Split (1)

train · 1.47k rows

filename stringlengths 13 34	markdown stringlengths 0 359k
10.1186_s12885-020-6609-x	He et al. BMC Cancer (2020) 20:116 https://doi.org/10.1186/s12885-020-6609-x R E S E A R C H A R T I C L E Open Access Early onset neutropenia: a useful predictor of chemosensitivity and favorable prognosis in patients with serous ovarian cancer Yijing He1, Ting Li1, Jue Liu1, Qiong Ou1 and Junlin Zhou2* Abstract Background: Epithelial ovarian cancer (EOC) is the leading cause of gynecological cancer-associated deaths and a majority of its histological type is manifested as serous ovarian cancer (SOC). In this study, we investigated whether the timing of onset of chemotherapy-induced neutropenia (CIN) is related to chemotherapeutic response and disease outcome of SOC. Methods: One hundred sixty-nine primary SOC patients receiving six doses of carboplatin plus paclitaxel adjuvant chemotherapy following cytoreductive surgery were retrospectively included in this research. CIN was grouped as early onset and late onset neutropenia depending on the timing of development. Development of CIN prior to or with administration of 3rd cycle of chemotherapy was listed as early onset neutropenia, while those CIN due to later stage chemotherapy were grouped into non-early type. The relevance of time of CIN onset with the clinical characteristics, chemotherapeutic response, progression free survival (PFS) and overall survival (OS) were determined and analyzed by using Kaplan–Meier curves, Logistic regression method, Cox proportional hazards models, and Chi-square tests. Results: The age distribution of the patients was between 27 to 77 years. Fifty years was the median. No statistical significances of difference in age, FIGO stage, histological grade, tumor residual and lymph node invasion, as well as CA125 level in each CIN group were found (all P>0.05). The patients from non-early onset group showed higher chemoresistance rates (78.33%) compared to those from early onset group (9.17%). Additionally, patients in early onset group showed improved median PFS (23 vs. 9 months; P<0.001) and median OS (55 vs.24 months; P<0.001). Conclusions: Early onset neutropenia may be potentially used as a potential indicator for chemosensitivity and favorable prognosis of SOC in patients who underwent six cycles of carboplatin plus paclitaxel adjuvant chemotherapy following primary cytoreductive surgery. Keywords: Timing of onset of chemotherapy-induced neutropenia (CIN), Chemotherapeutic response, Prognosis, Serous ovarian cancer * Correspondence: zjl18627660255@163.com 2Clinical Research Institute, The First Affiliated Hospital of University of South China, Hengyang, China Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. He et al. BMC Cancer (2020) 20:116 Page 2 of 8 Background Epithelial ovarian cancer (EOC) is the leading cause of gynecological cancer-associated deaths and a majority of its histological type is manifested as SOC [1]. Despite high clinical response rate, recurrences of the SOC post pri- mary combined surgery and chemotherapy are common. Majority of the relapses accompany non-responsiveness to further chemotherapy which eventually leads to death [2, 3]. Although, in recent years, some studies have attempted to reveal the prognostic factors and biomarkers for predic- tion of responses to chemotherapy and survival, the appli- limited. cation of such prediction parameters are still Therefore, identification of an easy and reliable prognostic biomarker for disease surveillance and stratification of ovarian cancer is essential. Neutropenia is a frequent adverse reaction following chemotherapy. The risk of developing neutropenia fol- lowing the standard chemotherapy for EOC with carbo- is approximately 30–90% [4]. platin and paclitaxel Despite being an adverse effect of chemotherapy, several researches have reported that CIN can be used for pre- diction of a favorable prognosis in different carcinomas of the breast [5], non-small cell lung [6], gastric [7], pan- creas [8], and colorectum carcinoma [9, 10]. The associ- ation between CIN and the progress of ovarian cancer has been controversial. While, Kim et al. [11] suggested CIN as non-significant prognostic indicator in ovarian cancer, studies by Tewari et al. [4], the indicate im- proved survival rate in patients with CIN as compared to those patients who do not develop CIN. Recently, sev- eral studies reported that timing of CIN may predict chemotherapeutic response or survival [12–16]. How- ever, the role of timing of CIN onset for predicting che- motherapeutic response and clinical outcome has not been evaluated for SOC patients. Therefore, this research aims to determine the correl- ation of CIN onset and the response to chemotherapy, with carboplatin and paclitaxel, in terms of chemosensi- tivity and survival. Methods Patients and data collection This retrospective study comprises of patients diagnosed with SOC and were admitted in the Second Affiliated Hospital of University of South China during the period between January, 2011 to June, 2013. The approval for the research was provided by the ethical committee of Second Affiliated Hospital of University of South China. Before study, written informed consents were obtained from the patients. All treatments and blood tests were performed according to institutional guidelines. The clinical records were gathered from the database of Sec- ond Affiliated Hospital of University of South China. The criteria for inclusion in the study were as follows: 1) histological or cytological confirmation of developing SOC and without prior treatment, such as radiotherapy or chemotherapy; 2) patients underwent cytoreductive surgery followed by carboplatin plus paclitaxel adjuvant chemotherapy; 3) normal bone marrow profile; 4) nor- mally functioning liver and kidney. The exclusion cri- teria were: 1) incomplete record of toxicities; 2) lost follow-up; 3) second malignancies or multiple primary malignancies; 4) primary treatment in other hospital. One hundred sixty-nine SOC patients were found fit as per the criteria set for inclusion and exclusion in the present research. Dose intensity of chemotherapy Chemotherapy regimens for all the patients were initi- ated within 4 weeks after primary cytoreduction. Each dose of carboplatin and paclitaxel comprise of (AUC = 5) and (175 mg/m2) respectively and were administered intravenously six times with a gap of 3 weeks. Assessment of neutropenia Blood samples were collected both before (day 0 or day 1) and on every 7 days after initiation of chemotherapy. The development of CIN of the highest grade was used for analysis. CIN grading were carried out according to ruling of the National Cancer Institute (NCI) Common Terminology Criteria for Adverse Events (CTCAE, ver- sion 4.0). Grade 1, 2, 3 and 4 were assigned based on ab- solute neutrophil count (ANC) limit of 1.5 × 109/L to 2.0 × 109/L; 1.0 × 109/L to 1.5 × 109/L; 0.5 × 109/L to 1.0 × 109/L; less than 0.5 × 109/L respectively. Grade 1 and 2 represent mild neutropenia, while grade 3 and 4 are denoted to severe form of neutropenia. Moreover, depending on minimum number of chemotherapeutic dose for development of CIN, they were listed as early onset and late onset neutropenia. Early onset group ex- perience ANC fall less than2.0 × 109/L with chemother- apy cycle 1–3, while in non-early onset group ANC level did not fall below2.0 × 109/L until 4th cycle of chemo- therapy. The use of granulocyte colony-stimulating fac- tor (G-CSF) for prophylaxis was prohibited unless ANC reached below 0.5 × 109/L. Follow-up study were regularly All patients enrolled in this followed-up every 3 months until June 30, 2018 to obtain recurrence and survival information. Follow-up included a complete history of the disease, physical examination, blood tests, abdominal ultrasonography, CT scan of the chest and abdomen to exclude recurrence and metasta- sis. Recurrence was evaluated as per the guidelines of re- sponse evaluation criteria in solid tumors (RECIST) [17]. Development of progressive disease before 6 months of He et al. BMC Cancer (2020) 20:116 Page 3 of 8 initial treatment were grouped as chemoresistant; while the others were grouped as chemosensitive [18]. PFS is determined by the time from the surgery to disease pro- gression, while OS represents the time duration between cytoreductive surgery and death or, as the case may be, date of latest follow-up. Statistical analysis Statistical differences between groups were determined using Wilcoxon and Pearson’s Chi-Square tests. Logistic regression method was applied for prediction of inde- pendent risk factors of chemoresistance. Survival curves were analyzed by the Kaplan–Meier curves and the log- rank test. Analysis of multivariates were done by Cox proportional hazards regression models. If the value P was found to be less than 0.05, then difference in the groups were considered statistically significant. The SPSS, version 23.0 (Chicago, IL, USA) software tool was used for all the statistical analysis. Results Patient demographics A total of 169 patients with histologically identified SOC, who underwent cytoreductive surgery followed by carboplatin plus paclitaxel adjuvant chemotherapy, were eligible for this analysis. Table 1 showed clinical vari- ables and the timing of CIN of the 169 patients. The me- dian age of the patients was 50 years (range 25–77 years). Among 169 patients, 109 (64.50%) experienced early onset and 60 (35.50%) experienced non-early onset neutropenia. One hundred fifteen developed mild and the remaining 38 developed severe neutropenia. There were no significant differences in age, FIGO stage, histo- logical grade, tumor residual and lymph node invasion, as well as CA125 level among groups by timing of CIN (all P>0.05) (Table 1). The timing of CIN and chemoresistance Table 2 showed clinical characteristics predicting che- moresistance. In this study, 57 out of 169 patients (33.73%) were found to be chemoresistant. The patients in non-early onset group have higher chemoresistance rates (78.33%) compared to the early onset group (9.17%). Besides, histological grade, severity of CIN was associated with chemotherapeutic response. Furthermore, use Logistic analysis to assess the pre- dictive significance of timing of CIN (Table 3) revealed that the non-early onset CIN was an independent pre- dictor of chemoresistance [odds ratio (OR) 36.371, 95% confidence interval (CI) 12.364–106.993; P<0.001]. Survival analysis The patients had a median PFS of 19 months and a me- dian OS of 44 months. There was a significant association Table 1 Clinical characteristics of patients by timing of CIN in patients with serous ovarian cancer(n = 169) n Variables Non-early onset Early onset P value Age (year) <50 ≥ 50 FIGO stage I-II III-IV Histological grade Low High Lymph node invasion Negative Positive Tumor residual (cm) Optimal(≤1) Sub-optimal(>1) CA125 level (U/mL) ≤ 35 >35 Severity of CIN Absence Mild Sever 75 94 45 124 42 127 132 37 127 42 49 60 27 82 23 86 89 20 87 22 12 9 157 100 16 115 38 0 83 26 26 34 18 42 19 41 43 17 40 20 3 57 16 32 12 0.839 0.462 0.128 0.133 0.058 0.634 <0.001 between timing of CIN and survival using Kaplan−Meier analysis. As shown in Fig. 1, the early onset group showed significantly higher PFS and OS than the non-early onset group. The median PFS in early and non-early onset groups were 23 and 9 months, respectively (P<0.001), while the median OS were 55 and 24 months, respectively (P<0.001). To assess the prognostic significance of timing of CIN, we performed the univariate and multivariate Cox re- gression analysis. According to Table 4, univariate ana- lysis focused on several variables of survival, including tumor residual, age, FIGO stage, histological grade, lymph node invasion, CA125 level, severity of CIN, and the timing of CIN. FIGO stage (P<0.001), tumor residual (P = 0.003), lymph node invasion (P = 0.013), CA125 level (P<0.001), timing of CIN (P<0.001) were all signifi- cant in terms of effects on PFS. However, multivariate analysis revealed only advanced FIGO stage (HR 3.337, 95%CI 2.049–5.436; P<0.001) and non-early onset CIN (HR 5.098, 95%CI 3.389–7.669; P<0.001) as independent prognostic factors associated with poor PFS. Moreover, analysis for OS showed that age ≥ 50 years, advanced FIGO stage, high histological grade, lymph node involve- ment, sub-optimal tumor residual, CA125 level >35 U/ He et al. BMC Cancer (2020) 20:116 Page 4 of 8 Table 2 Association between chemotherapeutic response and clinical characteristics Variables Chemotherapeutic response P value Chemosensitive (n = 112) Chemoresistance (n = 57) Age (year) <50 ≥ 50 FIGO stage I-II III-IV Histological grade Low High Lymph node invasion Negative Positive Tumor residual (cm) Optimal(≤1) Sub-optimal(>1) CA125 level (U/mL) ≤ 35 >35 Severity of CIN Absence Mild Sever Timing of CIN Early onset Non-early onset 50 62 31 81 19 93 89 23 87 25 9 103 1 81 30 99 13 25 32 14 43 23 34 43 14 40 17 3 54 15 34 8 10 47 0.923 0.665 0.001 0.550 0.286 0.729 <0.001 <0.001 mL, non-early onset of CIN were risk factors for OS in univariate analysis, but only advanced FIGO stage (HR 5.004, 95%CI 2.951–8.485; P<0.001), sub-optimal tumor residual (HR 3.182, 95%CI 1.970–5.140; P<0.001) and timing of CIN (HR 6.713, 95%CI 4.295–10.492; P<0.001) were independent prognosis factors for OS in multivari- ate analysis (Table 5). However, there was no correlation between severity of CIN and PFS or OS. Table 3 Logistic analysis of the association between chemoresistance and timing of CIN Variable OR Histological grade (high vs low) 0.188 Timing of CIN (Non-early vs Early) 36.371 95% CI 0.062–0.567 12.364–106.993 Severity of CIN Sever vs Absence Mild vs Absence Mild vs Sever 0.085 0.223 2.606 0.009–0.852 0.025–2.010 0.784–8.667 P value 0.003 <0.001 0.074 0.036 0.181 0.118 Discussion Patients who undergo carboplatin plus paclitaxel adju- vant chemotherapy experience different levels and types of adverse effects. Neutropenia is the most evident ad- verse effects of chemotherapy. Since 2013, several inves- tigations represented that timing of CIN may predict [12–16]. The chemotherapeutic response or survival present investigation, to our best knowledge, is the first report on the association between timing of CIN and chemotherapeutic response or survival in SOC patients. A significantly better chemotherapeutic response and survival outcomes were observed in patients who had early onset CIN as compared to that of non-early onset. Consistent with previous researches, our study provides evidences that the timing of CIN onset can be exploited for prediction of chemotherapeutic response and sur- vival. For example, the chemoresistance incident was more likely to occur in non-early onset neutropenia (78.33% vs. 9.17%; P<0.001). In addition, early onset of CIN leads to significantly improved PFS as well as OS than the non-early onset group. The median PFS in early onset neutropenia group was 23 months as compared to 9 months in case of non-early onset group(P<0.001), and similarly the median OS were 55 and 24 months, in the respective groups (P<0.001). Several studies with different types of cancer have demonstrated the effect of CIN on the improved survival of patients. Rocconi et al. [19] first reported the associ- ation of CIN and survival in 255 primary EOC patients treated with 6 cycles of platinum plus taxane regimen. However, Kim et al. [11] reported that CIN as a non- significant prognostic indicator in ovarian cancer pa- tients treated with carboplatin plus paclitaxel. In 2013, Jang SH et al. [12] have provided the viewpoint that the timing of CIN onset following chemotherapy can be a determinant of survival against metastatic non-small cell lung cancer. Similar relations were also found in pancre- atic [14], gastric [15], and metastatic colon cancer [16]. This study demonstrates that early onset CIN is a pre- dictor of better survival outcomes against SOC. This may be a due to chemotherapy induced effective killing of residual as well as cancer stem cells. It suggested that CIN reflects the pharmacokinetics of cytotoxic drugs, the genetic predisposition of the patients, and inflamma- tion in the tumor microenvironment, which are the common factors related to survival outcomes. First, CIN reflects the dose and pharmacokinetics of chemotherapy regimen. In practice, the cytotoxic drugs dosing is based on body-surface area (BSA). Several re- ports have showed that this method of selecting dose may be insufficient or suboptimal in some patients due to the uncertain correlation between the pharmacokinet- ics of many cytotoxic drugs and BSA [20]. Differences in metabolisms, drug distribution, and catabolism affects He et al. BMC Cancer (2020) 20:116 Page 5 of 8 Fig. 1 Kaplan–Meier survival curves demonstrating relationships between timing of CIN and PFS (a) and OS (b) of patients with SOC Table 4 Univariate and multivariate analysis for the association between clinical characteristics and progression-free survival Variable Age (year) <50 ≥ 50 FIGO stage I-II III-IV Histological grade Low High Lymph node invasion Negative Positive Tumor residual (cm) Optimal(≤1) Sub-optimal(>1) CA125 level (U/mL) ≤ 35 >35 Severity of CIN Mild versus Absence Sever versus Absence Mild versus Sever Timing of CIN Early onset Non-early onset Univariate HR(95% CI) 1 1.352(0.980–1.866) 1 2.577(1.717–3.868) 1 1.250(0.852–1.835) 1 1.608(1.107–2.336) 1 1.732(1.204–2.466) 1 3.156(1.468–6.785) 0.241(0.139–0.419) 0.275(0.150–0.506) 0.876(0.598–1.282) 1 3.803(2.687–5.383) P value 0.066 <0.001 0.254 0.013 0.003 0.003 <0.001 <0.001 0.495 <0.001 Multivariate HR(95% CI) P value 1 3.337(2.049–5.436) 1 1.069(0.692–1.653) 1 1.314(0.864–1.997) 1 1.454(0.621–3.407) 0.593(0.322–1.092) 0.615(0.318–1.888) 0.965(0.649–1.435) 1 5.098(3.389–7.669) <0.001 0.763 0.202 0.388 0.093 0.148 0.860 <0.001 He et al. BMC Cancer (2020) 20:116 Page 6 of 8 Table 5 Univariate and multivariate analysis for the association between clinical characteristics and overall survival Variable Age (year) <50 ≥ 50 FIGO stage I-II III-IV Histological grade Low High Lymph node invasion Negative Positive Tumor residual (cm) Optimal(≤1) Sub-optimal(>1) CA125 level (U/mL) ≤ 35 >35 Severity of CIN Mild versus Absence Sever versus Absence Mild versus Sever Timing of CIN Early onset Non-early onset Univariate HR(95% CI) 1 1.598(1.147–2.226) 1 3.794(2.454–5.864) 1 1.613(1.073–2.425) 1 2.583(1.759–3.795) 1 4.183(2.845–6.149) 1 4.360(1.779–10.681) 0.224(0.128–0.393) 0.211(0.112–0.397) 1.060(0.708–1.588) 1 3.696(2.593–5.268) P value 0.006 <0.001 0.022 <0.001 <0.001 0.001 <0.001 <0.001 0.776 <0.001 Multivariate HR(95% CI) 1 1.264(0.887–1.802) 1 5.004(2.951–8.485) 1. 1.302(0.823–2.059) 1 1.042(0.642–1.694) 1 3.182(1.970–5.140) 1 1.142(0.411–3.168) 0.512(0.269–0.975) 0.493(0.246–0.989) 1.039(0.671–1.608) 1 6.713(4.295–10.492) P value 0.195 <0.001 0.259 0.867 <0.001 0.799 0.042 0.046 0.864 <0.001 the plasma concentration of cytotoxic drugs which may lead to variation in therapeutic effectiveness due to under-dosing with standard chemotherapy [21]. How- ever, it is evident that at least the cornerstone of the medical treatment of ovarian cancer patients, the carbo- platin, is not dosed based on the BSA but on AUC. Therefore, lack of prognostic value of CIN in ovarian cancer might be explained by the fact that AUC dosing of carboplatin, the cornerstone of chemotherapy in this disease, prevents underdosing more than dosing strat- egies based on BSA [22]. Moreover, it is too expensive and not practical to assess drug plasma concentration in each patient. Therefore, based on our findings, the early onset of CIN may be a biomarker of pharmacokinetic changes, and can be used by physicians for adjustment of drug dose. Second, patient’s genetic predisposition may determine tumor chemosensitivity. Theoretically all cells of a pa- tient (including healthy cells, particularly hemopoietic cells) have similar pharmacokinetics characteristics [23]. In other words, we believe that the sensitivity to the che- motherapeutic drug in tumor cells is similar to the neu- trophils in an individual patient. Our research showed that the chemoresistance incident was more likely to occur in non-early onset neutropenia, suggesting that patients with early onset CIN are chemosensitive to car- boplatin and paclitaxel. On the other hand, the efficiency of cancer chemotherapy is determined by intrinsic and acquired chemoresistance [24]. The patients who do not develop neutropenia within 6 cycles in our research might be resistant to carboplatin plus paclitaxel regimen intrinsically. Furthermore, inflammation at the tumor site are crucial for regulation of tumor development and progression [25–27]. Elevated blood neutrophil could suppress the anti-tumor immune response and promote tumor angio- genesis, resulting in speeding up tumor proliferation. Therefore, early onset CIN may slow tumor progression by releasing immune suppression and disrupting angio- genesis, resulting in better survival. He et al. BMC Cancer (2020) 20:116 Page 7 of 8 Based on the above three possible mechanisms, it is evident that early onset CIN may be a factor for predict- ing chemosensitivity and favorable prognosis. However, there are several limitations concerning the present re- search. Firstly, it was retrospective in design, with a lim- ited sample size. Secondly, the patients enrolled in the study belong to same ethnicity and received a single carboplatin plus paclitaxel. chemotherapy Despite these drawbacks, the study forms the lead for an accurate and easily measurable surrogate marker for pre- dicting chemotherapeutic response and prognosis of ovarian cancer. regimen, Conclusion The findings of the research suggest that early onset CIN may be used to predict chemosensitivity and favor- able prognosis in SOC patients receiving carboplatin plus paclitaxel adjuvant chemotherapy post cytoreduc- tive surgery. However, a large-scale multicentric study would be essential to fully elucidate the association of timing of CIN onset and effective chemotherapy. Abbreviations ANC: Absolute neutrophil count; BSA: Body-surface area; CI: Confidence interval; CIN: Chemotherapy-induced neutropenia; EOC: Epithelial ovarian cancer; HR: Hazard ratio; OR: Odds ratio; OS: Overall survival; PFS: Progression free survival; SOC: Serous ovarian cancer Acknowledgements None. Authors’ contributions JLZ designed the study; YJH, TL, JL, QO enrolled patients; YJH, TL analyzed the data; YJH wrote the article; JLZ made the final approval. All authors have read and approved the final version for publication. Funding No funding. Availability of data and materials The dataset used and analysed during the present study is available from the corresponding author upon reasonable request. Ethics approval and consent to participate The study was approved by the ethics committee of the Second Affiliated Hospital of University of South China. All participants signed informed consent forms. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1Department of Obstetrics and Gynecology, The Second Affiliated Hospital of University of South China, Hengyang, China. 2Clinical Research Institute, The First Affiliated Hospital of University of South China, Hengyang, China. Received: 3 April 2019 Accepted: 6 February 2020 References 1. Jayson GC, Kohn EC, Kitchener HC, Ledermann JA. Ovarian cancer. Lancet. 2014;384(9951):1376–88. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. Alkema NG, Tomar T, van der Zee AG, Everts M, Meersma GJ, Hollema H, et al. Checkpoint kinase 2 (Chk2) supports sensitivity to platinum-based treatment in high grade serous ovarian cancer. Gynecol Oncol. 2014;133(3):591–8. Tate TJ. Contemporary phase III clinical trial endpoints in advanced ovarian cancer: assessing the pros and cons of objective response rate, progression- free survival, and overall survival. Gynecol Oncol. 2015;136(1):121–9. Tewari KS, Java JJ, Gatcliffe TA, Bookman MA, Monk BJ. Chemotherapy- induced neutropenia as a biomarker of survival in advanced ovarian carcinoma: an exploratory study of the gynecologic oncology group. Gynecol Oncol. 2014;133(3):439–45. Han Y, Yu Z, Wen S, Zhang B, Cao X, Wang X. Prognostic value of chemotherapy-induced neutropenia in early-stage breast cancer. Breast Cancer Res Treat. 2012;131(2):483–90. Huang CS, Liu L, Liu J, Chen Z, Guo J, Li CZ, et al. Association of chemotherapy-induced leucopenia with treatment outcomes in advanced non-small-cell lung cancer cases receiving the NP regimen. Asian Pac J Cancer Prev. 2012;13(9):4481–5. Chen Z, Chen W, Wang J, Zhu M, Zhuang Z. Pretreated baseline neutrophil count and chemotherapy-induced neutropenia may be conveniently available as prognostic biomarkers in advanced gastric cancer. Intern Med J. 2015;45(8):854–9. Kurihara T, Kogo M, Ishii M, Shimada K, Yoneyama K, Kitamura K, et al. Chemotherapy-induced neutropenia as a prognostic factor in patients with unresectable pancreatic cancer. Cancer Chemother Pharmacol. 2015;76(6): 1217–24. Rambach L, Bertaut A, Vincent J, Lorgis V, Ladoire S, Ghiringhelli F. Prognostic value of chemotherapy-induced hematological toxicity in metastatic colorectal cancer patients. World J Gastroenterol. 2014;20(6): 1565–73. Kasi PM, Kotani D, Cecchini M, Shitara K, Ohtsu A, Ramanathan RK, et al. Chemotherapy induced neutropenia at 1-month mark is a predictor of overall survival in patients receiving TAS-102 for refractory metastatic colorectal cancer: a cohort study. BMC Cancer. 2016;16:467. Kim JJ, Park JY, Kim DY, Kim JH, Kim YM, Nam JH, et al. Is chemotherapy- induced neutropenia a prognostic factor in patients with ovarian cancer? Acta Obstet Gynecol Scand. 2010;89(5):623–8. Jang SH, Kim SY, Kim JH, Park S, Hwang YI, Kim DG, et al. Timing of chemotherapy-induced neutropenia is a prognostic factor in patients with metastatic non-small-cell lung cancer: a retrospective analysis in gemcitabine-plus-platinum-treated patients. J Cancer Res Clin Oncol. 2013; 139(3):409–17. Lee CY, Park SY, Shin TR, Park YB, Kim CH, Jang SH, et al. Early-onset neutropenia during perioperative chemotherapy is predictive of increased survival in patients with completely resected non-small cell lung cancer: a retrospective analysis. Anticancer Res. 2013;33(6):2755–61. 14. Chen Y, Shi Y, Yan H, Wang YR, Dai GH. Timing of chemotherapy-induced neutropenia: the prognostic factor in advanced pancreatic cancer patients treated with gemcitabine / gemcitabine-based chemotherapy. Oncotarget. 2017;8(39):66593–600. 15. Chen Y, Wang Y, Shi Y, Dai G. Timing of chemotherapy-induced neutropenia predicts prognosis in metastatic colon cancer patients: a retrospective study in mFOLFOX6 -treated patients. BMC Cancer. 2017;17(1):242. 16. Wang Y, Chen Y, Yin H, Gu X, Shi Y, Dai G. Timing of chemotherapy-induced neutropenia is a prognostic factor in patients with advanced gastric cancer undergoing first-line chemotherapy with oxaliplatin and capecitabine: a retrospective study. Cancer Med. 2018;7(4):997–1005. Therasse P, Arbuck SG, Eisenhauer EA, Wanders J, Kaplan RS, Rubinstein L, et al. New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. J Natl Cancer Inst. 2000;92(3):205–16. 17. 18. Berek JS, Kehoe ST, Kumar L, Friedlander M. Cancer of the ovary, fallopian tube, and peritoneum. Int J Gynaecol Obstet. 2018;143(Suppl 2):59–78. 19. Rocconi RP, Matthews KS, Kemper MK, Hoskins KE, Barnes MN. Chemotherapy-related myelosuppression as a marker of survival in epithelial ovarian cancer patients. Gynecol Oncol. 2008;108(2):336–41. 20. Michael M, Doherty MM. Tumoral drug metabolism: overview and its implications for cancer therapy. J Clin Oncol. 2005;23(1):205–29. 21. Gurney H. Dose calculation of anticancer drugs: a review of the current practice and introduction of an alternative. J Clin Oncol. 1996;14(9): 2590–611. He et al. BMC Cancer (2020) 20:116 Page 8 of 8 22. Daniele G, Arenare L, Scambia G, Pisano C, Sorio R, Breda E, et al. Prognostic role of chemotherapy-induced neutropenia in first-line treatment of advanced ovarian cancer. A pooled analysis of MITO2 and MITO7 trials. Gynecol Oncol. 2019;154(1):83–8. Kvinnsland S. The leucocyte nadir, a predictor of chemotherapy efficacy? Br J Cancer. 1999;80(11):1681. Fraczek N, Bronisz I, Pietryka M, Kepinska D, Strzala P, Mielnicka K, et al. An outline of main factors of drug resistance influencing cancer therapy. J Chemother. 2016;28(6):457–64. 23. 24. 25. Diakos CI, Charles KA, McMillan DC, Clarke SJ. Cancer-related inflammation and treatment effectiveness. Lancet Oncol. 2014;15(11):e493–503. 26. He JR, Shen GP, Ren ZF, Qin H, Cui C, Zhang Y, et al. Pretreatment levels of peripheral neutrophils and lymphocytes as independent prognostic factors in patients with nasopharyngeal carcinoma. Head Neck. 2012;34(12):1769–76. Taniguchi K, Karin M. NF-kappaB, inflammation, immunity and cancer: coming of age. Nat Rev Immunol. 2018;18(5):309–24. 27. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12879-023-08167-2	Sadio et al. BMC Infectious Diseases (2023) 23:200 https://doi.org/10.1186/s12879-023-08167-2 BMC Infectious Diseases High SARS-CoV-2 seroprevalence among street adolescents in Lomé, Togo, 2021 Arnold Junior Sadio1,2,3, Valentine Marie Ferré4,5, Rodion Yao Konu1,2,3, Anoumou Claver Dagnra6, Diane Descamps4,5, Didier Koumavi Ekouevi1,2,3* and Charlotte Charpentier4,5 Abstract Background There is almost no data on the circulation of SARS-CoV-2 among street adolescents. We conducted a study to document the immunization status of street adolescents in Togo against different variants of SARS-CoV-2. Methods A cross-sectional study was carried out in 2021 in Lomé, the city with the highest number of COVID 19 cases in Togo (60%). Adolescents aged 13- and 19 years old living on the street were eligible for inclusion. A standardized questionnaire was administered face-to-face to adolescents. A sample of blood was taken and aliquots of plasma were transported to the virology laboratory of the Hôpital Bichat-Claude Bernard (Paris, France). SARS- CoV-2 anti-S and anti-N IgG were measured using chemiluminescent microparticle immunoassay. A quantitative miniaturized and parallel-arranged ELISA assay was used to detect IgG antibodies specifically directed against the different SARS-CoV-2 Variants of Concern (VOC). Results A total of 299 street adolescents (5.2% female), median age 15 years, interquartile range (14-17 years), were included in this study. The prevalence of SARS-CoV-2 infection was 63.5% (95%CI: 57.8–69.0). Specific-IgG against the ancestral Wuhan strain was developed by 92.0% of subjects. The proportion of patients being immunized against each VOC was 86.8%, 51.1%, 56.3%, 60.0, and 30.5% for the Alpha, Beta, Gamma, Delta, and Omicron VOCs, respectively. Conclusion This study showed a very high prevalence with approximately 2/3 of Togolese street adolescents having antibodies to SARS-CoV-2 due to a previous infection. These results confirm an under-reporting of COVID-19 cases in Togo, questioning the hypothesis of low virus circulation in Togo and even in Africa. Keywords SARS-CoV-2 seroprevalence, Street adolescents, Togo *Correspondence: Didier Koumavi Ekouevi didier.ekouevi@gmail.fr 1Faculty of Health Sciences, Department of Public Health, University of Lomé, Center for Training and Research in Public Health, Lomé, Togo 2African Center for Research in Epidemiology and Public Health (CARESP), Lomé, Togo 3Research Institute for Sustainable Development (IRD), University of Bordeaux, National Institute for Health and Medical Research (INSERM), Bordeaux Population Health Centre, UMR 1219, Bordeaux, France 4Paris Cité University and Sorbonne Paris Nord University, IAME, Inserm, Paris F-75018, France 5Virology Unit, AP-HP, Hôpital Bichat-Claude Bernard, Paris F-75018, France 6Laboratory of Molecular Biology and Immunology, University of Lomé, Lomé, Togo © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessPage 2 of 6 Introduction The true magnitude of the SARS-CoV-2 epidemic is unknown in sub-Saharan Africa. According to existing data, only 1.7% of Covid-19 cases have been reported in Africa out of more than 517 million cases worldwide [1]. One way to estimate the true extent of the epidemic is to conduct seroprevalence surveys [2]. The last study conducted in Togo in 2021, in twelve health districts, with the inclusion of more than 7000 people, reported a seroprevalence of 65.5% [3]. These studies are essentially household surveys, which are costly and time-consuming and therefore difficult to repeat over time. Epidemiologi- cal surveillance surveys can also be carried out in spe- cific populations, as is the case in France where a study was carried out among homeless people and showed a circulation of the virus among this population after the first epidemic wave [4]. Homeless people are particularly vulnerable to SARS-CoV-2 infection. Indeed, several risk factors for SARS-CoV-2 infection are found in these populations, including promiscuity, multiple residence in often poorly ventilated dwellings, and frequent con- tact with many people in community support services. In addition, they are at increased risk for severe COVID- 19, being exposed to a high prevalence of comorbidities [5–7]. In Africa, there are populations that live in conditions as precarious as the homeless and are just as vulner- able: street adolescents. This is a very mobile population that is often found in city markets, border crossings and roadside pay stations. This population of street adoles- cents was on the almost excluded of all risk mitigation measures taken by governments. In 2021, we conducted a study among street adolescents to assess the feasibil- ity of HIV self-testing. As this study was conducted in the midst of the COVID-19 pandemic, and consider- ing the lack of epidemiological data in the literature on SARS-CoV-2 infection in this population, we conducted an ancillary study to document the immunization status of street adolescents in Togo against different variants of SARS-CoV-2. Methods Study design and setting This study was part of a survey aimed at describing the acceptability and feasibility of HIV self-testing among street adolescents in Lomé (the capital city of Togo). This was a cross-sectional study conducted from june 26th to july 3rd, 2021 after the second wave of pandemics. Togo is a West African country covering a 56,800 km² area with an average density of 145 inhabitants per square kilometer [8]. The population was 8.08 million in 2019, of which 50.2% were women [9]. Most of the population is under 25 years of age (60%), and lives in rural areas (62%) [9]. Togo’s health system has a pyramidal structure with three levels: central, intermediate and peripheral. Each level has administrative and health care delivery com- ponents. Lomé is the largest urban center in the country and at the time of the survey, the city alone accounted for more than 60% of reported COVID-19 cases in Togo [10]. Study population and sample size All adolescents aged between 13- and 19-years old liv- ing in the street were eligible for inclusion following informed consent. After listing all the places where street adolescents gather in Lomé, a team of investigators went to each site accompanied by local NGOs working with the street child population. After explaining the study to the adolescents, those who agreed to participate were asked to sign a consent form if they were over 18 years old. Adolescents under the age of 18 were asked to give their assent to participate in the study, and then a mem- ber of the NGO involved in the care of street adolescents was asked to sign a consent form. Data collection A standardized questionnaire was administered face-to- face to adolescents. It included socio-demographic char- acteristics, sexual practices, and history of HIV testing. Then, a sample of 04 ml of venous blood was taken. Laboratory procedures Aliquots of plasma were taken to the laboratory of molec- ular biology and immunology of the University of Lomé (Lomé, Togo) and transported frozen to the virology laboratory of the Hôpital Bichat-Claude Bernard (Paris, France), for the search for anti-SARS-CoV-2 antibodies. SARS-CoV-2 anti-S and anti-N IgG were measured using the automated Abbott SARS-CoV-2 IgG kit (che- miluminescent microparticle immunoassay, CLIA) (Abbott, IL, USA) using the Alinity platform according to the manufacturer’s instructions. In addition, the CoViD- iag kit (SirYus CoViDiag+, Innobiochips®, Loos, France), a quantitative miniaturized and parallel-arranged ELISA assay [11], was used to detect IgG antibodies specifi- cally directed against the different SARS-CoV-2 Variants Of Concern (VOC) including BA.1 Omicron sublineage with a threshold at 18 BAU/mL. CoViDiag microtitration plates wells are coated with 4 different SARS-CoV-2 anti- gens and 4 RBD domains from different VOC separately. They are dedicated to bind specific antibodies in the tested samples and therefore deliver different responses in one single assay. Case definition A sample was considered positive if any of the following conditions were met: (i) the presence of anti-protein S antibodies; (ii) the presence of anti-N antibodies; or (iii) a result in the greyzone of the assays. An humoral response Sadio et al. BMC Infectious Diseases (2023) 23:200 Page 3 of 6 was defined as having a positive titer against the VOC using the COVIDIAG assay. Statistical analysis We performed descriptive statistics, and the results were presented using frequency tabulations and percentages for categorical variables. Quantitative variables were pre- sented as medians with their interquartile range (IQR). Kruskal-Wallis rank sum test or Wilcoxon test were used for comparison when appropriate. Prevalences of SARS- CoV-2 antibodies were estimated with their 95% confi- dence interval (95%CI). Results A total of 299 street adolescents, median age 15 years, interquartile range (IQR) (14-17 years), of which 5.2% (n = 16) were female were included in this study. Among these, 246 out of 299 (82.3%) are of Togolese nationality (Table 1). Anti-S IgG serology was positive for 190 of the 299 tested samples leading to a prevalence of SARS-CoV-2 infection of 63.5% (CI 95%: 57.8–69.0). Anti-N IgG anti- bodies, marker of a more recent infection, were positive for 125 samples (41.8%; CI 95%: 36.2–47.6) (Table 1). VOC-specific anti-S IgG titers were assessed for the 190 samples with positive anti-S IgG. 175/190 (92.0%) sub- jects developed specific-IgG against the ancestral Wuhan strain. The 8% of subjects showing no anti-S IgG against the ancestral Wuhan strain had anti-S IgG titers ≤ 18 BAU/mL, which is the LOQ of the multiplex ELISA assay. The proportion of patients with presence of anti-S IgG against each VOC was 86.8% (n = 165/190), 51.1% (n = 97/190), 56.3% (n = 107/190), 60.0% (n = 114/190), and 30.5% (n = 58/190) for the Alpha, Beta, Gamma, Delta and Omicron VOCs, respectively (Fig. 1). The proportion of the population harbouring a humoral response against the ancestral strain and the Alpha VOC were significantly higher than against all other VOCs (p < 0.0001 for Beta, Gamma, Delta and Omicron VOCs). Among subjects who developed an immunity against one or several VOCs, the median titers were 68 BAU/mL (IQR = 28–162), 81 (IQR = 34–182), 62 (IQR = 26–175), 53 (IQR = 24–155) and 33 (IQR = 21–54) for the Alpha, Beta, Gamma, Delta and Omicron VOCs, respectively (Fig. 2). Among participants who showed an humoral response, Anti-S IgG median titers were significantly lower against Delta and Omicron variants compared to the ancestral Wuhan strain (p = 0.0003 and p < 0.0001, respectively). Regarding a potential humoral response against the Omicron variant not yet encountered by this population at the time of sampling, anti-S1 RBD IgG median titers were significantly lower against this future new VOC at that time compared to all the variants studied (p < 0.0001, p < 0.0001, p = 0.0002 and p = 0.0023 for Alpha, Beta, Gamma and Delta VOCs, respectively) (Fig. 2). Discussion A seroprevalence study was conducted in a Togolese vulnerable population at high risk of infection because of non-compliance with government-mandated barrier measures to mitigate the risk of SARS-CoV-2 infection. A total of 2 out of 3 adolescents had anti-S antibodies Table 1 Seroprevalence of SARS-CoV-2 antibodies by sociodemographic characteristics Overall Age (years) Median [IQR] 10–14 15–19 Sex Female Male Nationality Other Togolese Education level No academic Primary Secondary Superior HIV serology Negative Positive N 299 15 105 194 16 283 53 246 23 121 121 34 296 3 Anti-S antibody n 190 [14–17] 63 127 9 181 36 154 13 75 81 21 189 1 % 63.5 60.0 65.5 56.2 64.0 67.9 62.6 56.5 62.0 66.9 61.8 63.9 33.3 95%CI 57.8–69.0 50.0–69.3 58.3–72.0 30.6–79.2 58.0–69.5 53.6–79.7 56.2–68.6 34.9–76.1 52.6–70.5 57.7–75.1 43.6–77.3 58.1–69.3 1.8–87.5 Anti-N antibody n 125 % 41.8 40 85 4 121 22 103 13 40 55 17 125 0 38.1 43.8 25.0 42.8 41.5 41.9 56.5 33.1 45.5 50.0 42.2 0.0 95%CI 36.2–47.6 28.9–48.1 36.8–51.1 8.3–52.6 36.9–48.8 28.4–55.8 35.7–48.3 34.9–76.1 24.9–42.3 36.5–54.7 34.1–65.9 36.6–48.1 0.0–69.0 Sadio et al. BMC Infectious Diseases (2023) 23:200 Page 4 of 6 Fig. 1 Proportion of population harbouring humoral response against the SARS-CoV-2 Spike RBD of the different strains studied among the participants tested positive for anti-S SARS-CoV-2 IgG in ELISA (n = 190) Fig. 2 Titers of IgG antibodies anti-SARS-CoV-2 Spike RBD of different variants using the COVIDIAG multiplex ELISA assay measured in the population harbouring a positive SARS-CoV-2 anti-S IgG serology (n = 190) demonstrating significant virus circulation as of June 2021. The results observed are comparable to those observed in the Togolese general population in April 2021 (65.5%) [3], and prove an under-reporting of SARS- CoV-2 infection in the country. Indeed, as of July 30, 2021, less than 40,000 cases (0.5% of the Togolese popu- lation) of SARS-CoV-2 infections were officially reported in Togo [10]. This study, which is the first seroprevalence study among the street’s adolescents, also documented recent infections with the identification of Anti-N IgG antibod- ies, which is not the case in the majority of SARS-CoV-2 seroprevalence surveys that document only antibodies targeting the Spike protein. The results show that 42% of street adolescents had a recent SARS-CoV-2 infec- tion. Differential reactivity of S- and N-specific anti- bodies can be used to help differentiate prior infection Sadio et al. BMC Infectious Diseases (2023) 23:200 from vaccination in serological studies, particularly for vaccines that produce antibodies only to the S protein [12–14]. The presence of isolated anti-S antibodies can also result from a past COVID-19 infection, after the dis- appearance of the anti-N antibodies, known to decrease more rapidly than the anti-S antibodies [15, 16]. How- ever, for the present study, given the start date of vacci- nation in Togo and the target chosen at the beginning, it can be said that surveyed street adolescents were not vac- cinated. Indeed, vaccination in Togo began on March 10, 2021, and was limited to health professionals and people aged 20 years and older at the end of the study. Regarding the research of SARS-CoV-2 variants, this is not systematically carried out in Togo. It is only per- formed during epidemic peaks for surveillance purposes and to adjust the response measures against the pan- demic. So, no naso-pharyngeal samples were available in this study, preventing to have a description of the circu- lating SARS-CoV-2 variants in this adolescent’s popula- tion. As of August 4, 2021, official genotyping data from samples collected in Togo reported 90% of Delta and 1.3% of Eta variant [17]. In December 2021 the same data reported 73.6% of Delta and 24.5% of Omicron [18]. The results of the COVIDIAG assay confirmed the capacity of immune escape of the Delta and Omicron VOCs with a lower proportion of patients immunized and if is the case in a lower level [19]. This assay was indeed proven to show a strong correlation between specific RBD antibod- ies titers and live viral neutralization ability of sera [19]. Thus, even if some serological cross-reactivity has been previously described with other serological assays on African samples [16], no such cross-reactivity is expected with the assays used in this study. A major limitation of this study is that it was not ini- tially designed for an SARS-CoV-2 investigation, which did not allow us to collect specific information about SARS-CoV-2 infection such as clinical manifestations of SARS-CoV-2 or history of screening or hospitalization related to COVID-19. Conclusion This is the first seroprevalence study of SARS-CoV-2 con- ducted in sub-Saharan Africa among street adolescents in mid-2021. This study showed a very high prevalence with approximately 2/3 of Togolese street adolescents having antibodies to SARS-CoV-2 due to a previous infection. The seroprevalence data from this survey in a population with limited access to the health system and non-compli- ance with government measures confirm those observed in the general population. These results confirm an under-reporting of COVID-19 cases in Togo, questioning the hypothesis of a low circulation of the virus in Togo and even in Africa. Page 5 of 6 Acknowledgements We are thankful to the members of the NGOs (JADE pour la Vie, AGOPODE, and ANGE) who helped to mobilize the street adolescents. We are also thankful to ANRS\|MIE for funding the salary of Rodion Konu for his thesis. Authors’ contributions AJS and DKE designed the study. AJS, ACD and DKE supervised the study implementation and data collection. VMF, CC and DD performed, verified and supervised biological procedures for SARS-CoV-2 antibodies detection and quantification. Statistical analysis was performed by VMF, AJS and RYK. Data interpretation and first draft of manuscript was done by AJS, RYK, VMF, CC and DKE. Critical revision of the manuscript for important intellectual content was provided by all the coauthors who have read and commented on the original manuscript and all agreed on the version finalized for submission. Funding This study was supported by the Center for Training and Research in Public Health, Lomé, Togo. Data availability The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate All methods were carried out in accordance with relevant guidelines and regulations. This study has been approved by the National ethic Committee in Togo (No. 012/2020/CBRS). Informed consent was sought from participants through an informational note detailing the objectives of the study and an introductory consent to participate question. Adolescents over 18 years old were asked to sign a consent form. Adolescents under the age of 18 were asked to give their assent to participate in the study, and then a member of the NGO involved in the care of street adolescents was asked to sign a consent form. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Received: 16 September 2022 / Accepted: 15 March 2023 References 1. World Health Organization (WHO). WHO Coronavirus (COVID-19) Dashboard. 2. 3. 4. 5. 6. 7. Available at: https://covid19.who.int/. Accessed May 10, 2022. Bergeri I, Whelan M, Ware H et al. Global epidemiology of SARS-CoV-2 infection: a systematic review and meta-analysis of standardized population- based seroprevalence studies, Jan 2020-Oct 2021. medRxiv; 2021. https://doi. org/10.1101/2021.12.14.21267791. Ekouevi DK, Gbeasor-Komlanvi F, Malou-Adom V et al. Seroprevalence of SARS-CoV-2 infection in the general population in Togo in 2021 (1.0) [Data set]. Zenodo. 2022; https://doi.org/10.5281/zenodo.6519821. Loubiere S, Monfardini E, Allaria C, et al. Seroprevalence of SARS-CoV-2 anti- bodies among homeless people living rough, in shelters and squats: a large population-based study in France. PLoS ONE. 2021;16(9):e0255498. https:// doi.org/10.1371/journal.pone.0255498. Tsai J, Wilson M. COVID-19: a potential public health problem for homeless populations. Lancet Public Health. 2020;5:e186–7. https://doi.org/10.1016/ S2468-2667(20)30053-0. Leung CS, Ho MM, Kiss A, et al. Homelessness and the response to emerg- ing Infectious Disease Outbreaks: Lessons from SARS. J Urban Health. 2008;85:402–10. https://doi.org/10.1007/s11524-008-9270-2. Culhane DP, Kane V, Johnston M. Homelessness Research: shaping policy and practice, now and into the future. Am J Public Health. 2013;103:181–S182. https://doi.org/10.2105/AJPH.2013.301728. Sadio et al. BMC Infectious Diseases (2023) 23:200 Page 6 of 6 8. World Bank, Data Bank. Country Profile. Available at: https://cutt.ly/zHBWUXB. 16. Woodford J, Sagara I, Dicko A, et al. Severe Acute Respiratory Syndrome 9. Accessed March 26, 2022 République, Togolaise. Direction Générale de la Statistique et de la Compt- abilité Nationale du Togo (DGSCN). Résultats définitifs du 4ième Recense- ment Général de la Population et de l’Habitat (4ième RGPH4). Lomé: DGSCN ; 2010, 66p. Disponible sur : https://inseed.tg/download/2958/. 10. République, Togolaise. COVID 19: situation au Togo. Available from: https:// covid19.gouv.tg/situation-au-togo/. Accessed May 8, 2022. 11. Gillot C, Douxfils J, Cadrobbi J, et al. An original ELISA-Based multiplex method for the simultaneous detection of 5 SARS-CoV-2 IgG antibodies Directed against different antigens. J Clin Med. 2020;9(11):3752. 12. Hall VJ, Foulkes S, Charlett A, et al. SARS-CoV-2 infection rates of antibody- positive compared with antibody-negative health-care workers in England: a large, multicentre, prospective cohort study (SIREN). Lancet. 2021;397(10283):1459–69. 13. Harvey RA, Rassen JA, Kabelac CA, et al. Association of SARS-CoV-2 seropositive antibody test with risk of future infection. JAMA Intern Med. 2021;181(5):672–9. 14. Fotis C, Meimetis N, Tsolakos N, et al. Accurate SARS-CoV-2 seroprevalence surveys require robust multi-antigen assays. Sci Rep. 2021;11(1):6614. 15. Wheatley AK, Juno JA, Wang JJ, et al. Evolution of immune responses to SARS-CoV-2 in mild-moderate COVID-19. Nat Commun. 2021;12(1):1162. https://doi.org/10.1038/s41467-021-21444-5. Coronavirus 2 Seroassay performance and optimization in a Population with high background reactivity in Mali. J Infect Dis. 2021;224(12):2001–9. 17. World Health Organization (WHO)-Africa Centres for Disease Control and Prevention. (Africa CDC) regional SARS-CoV-2 sequencing lab. SARS-CoV-2 genomic Sequencing results for Togo #4. Available at: https://cutt.ly/THCT- BSo. Accessed May 10, 2022. 18. World Health Organization (WHO)-Africa Centres for Disease Control and Prevention. (Africa CDC) regional SARS-CoV-2 sequencing lab. SARS-CoV-2 genomic Sequencing results for Togo #7. Available at: https://cutt.ly/FHCYi7a. Accessed May 10, 2022. 19. Ferré VM, Lebourgeois S, Chenane HR, et al. Vaccine ab neutralization against Omicron and SARS-CoV-2 variants using neutralization and specific ELISA assays. J Infect. 2022;84(6):834–72. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Sadio et al. BMC Infectious Diseases (2023) 23:200
10.1186_s12870-020-02783-9	Li et al. BMC Plant Biology (2020) 20:558 https://doi.org/10.1186/s12870-020-02783-9 R E S E A R C H A R T I C L E Open Access Genome-wide analysis of the AREB/ABF gene lineage in land plants and functional analysis of TaABF3 in Arabidopsis Fangfang Li†, Fangming Mei†, Yifang Zhang, Shumin Li, Zhensheng Kang* and Hude Mao* Abstract Background: Previous studies have shown that ABFs (abscisic acid-responsive transcription factors) are important ABA-signaling components that participate in abiotic stress response. However, little is known about the function of ABFs in Triticum aestivum. In addition, although various ABFs have been identified in other species, the phylogenetic relationship between ABF transcription factors has not been systemically investigated in land plants. Results: In this study, we systemically collected ABFs from land plants and analyzed the phylogenetic relationship of these ABF genes. The ABF genes are present in all the land plants we investigated, including moss, lycophyte, monocots, and eudicots. Furthermore, these ABF genes are phylogenetically divided into seven subgroups, differentiations that are supported by variation in the gene structure, protein properties, and motif patterns. We further demonstrated that the expression of ABF genes varies among different tissues and developmental stages, and are induced by one or more environmental stresses. Furthermore, we found that three wheat ABFs (TaABF1, TaABF2, and TaABF3) were significantly induced by drought stress. Compared with wild-type (WT) plants, transgenic Arabidopsis plants overexpressing TaABF3 displayed enhanced drought tolerance. Conclusions: These results provide important ground work for understanding the phylogenetic relationships between plant ABF genes. Our results also indicate that TaABFs may participate in regulating plant response to abiotic stresses. Keywords: ABFs, Land plants, Phylogenetic relationship, Expression analysis, TaABFs, Drought stress Background Drought is a major environmental stressor that affects plant growth, survival, distribution, and productivity. Plants have evolved complex mechanisms in molecular, cellular, and physiological processes to respond to envir- onmental stresses in order to survive [1]. Stressful condi- tions induce the production of stress response genes in plants [2, 3]. The phytohormone abscisic acid (ABA) is regulates some critical an important hormone that * Correspondence: kangzs@nwsuaf.edu.cn; mhd163com@163.com †Fangfang Li and Fangming Mei contributed equally to this work. State Key Laboratory of Crop Stress Biology for Arid Areas and College of Plant Protection, Northwest A&F University, Yangling, Shaanxi 712100, People’s Republic of China biological processes in plants, such as stomatal move- ment, adaptation to drought stress, and seed germin- ation [4–7]. The endogenous ABA is produced when plants encounter adverse environmental stresses such as prolonged periods of osmotic stress. Several stress- responsive genes were expressed due to these increased ABA levels. Additional research indicates that many stress-responsive genes can also be induced by the ex- ogenous application of ABA [2, 7–9]. ABA detects stress in a unique way and acts as an en- dogenous messenger in plant cells by inducing a double- negative regulatory pathway where ABA is bound to the ABA receptors RCARs/PYR1/PYLs, forming the com- plex that provides an active site for the PP2Cs. This © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Li et al. BMC Plant Biology (2020) 20:558 Page 2 of 15 inhibits the ability of PP2C to act as a negative regulator of the pathway, leading to the induction of SnRK2 as a positive regulator of downstream signalling and subse- quent phosphorylation of the target proteins [10, 11]. Thus, in the presence of ABA, the PP2Cs are inactivated to repress SnRK2 phosphatase activity. SnRK2 could then initiate the ABA-responsive regulation pathway and activate the most significant cis-element ABA-responsive element (ABRE) to regulate the expression of many genes under osmotic stress conditions. Subsequently, through the yeast one-hybrid method, a subgroup of bZIP transcription factors was isolated by using ABREs as bait [12, 13]. Genes of this subgroup of bZIP tran- scription factors primarily participated in osmotic stress response by regulating stress-related genes. In Arabidop- sis, nine group-A bZIP proteins were found as homologs of AREB/ABFs, and phylogenetically divided into two subfamilies, the AREB/ABF subfamily (ABF1, ABF2/ AREB1, ABF3 and ABF4/AREB2) and the ABI5/AtDPBF subfamily (AtDPBF1/ABI5, AtDPBF2, AtDPBF3/AREB3 and AtDPBF4/EEL) [14]. ABF/AREB family members have four conserved domains, including two located in the C-terminus (which includes a highly conserved bZIP domain and a C4 domain) and three located in the N- terminus (which include C1, C2 and C3 domains) [15]. To date, all of these AREB/ABF genes in Arabidopsis have been functionally characterized. These four genes (ABF1, ABF2/AREB1, ABF3 and ABF4/AREB2) are primar- ily expressed in vegetative tissues [12, 13, 16, 17]. In addition, the induced ABF1 expression changes in response to abiotic stress are minimal [18], while ABF2/AREB1, ABF3 and ABF4/AREB2 are significantly up-regulated under ABA and osmotic stresses [12–14, 17–20]. Ectopic expression of these four genes in Arabidopsis showed that ABF1 is a functional homolog of ABF2/AREB1, ABF3 and ABF4/AREB2; and ABF2/AREB1, ABF3 and ABF4/AREB2 are the core ABA signaling components responding to abi- otic stresses [16–18, 20, 21]. Moreover, the areb1areb2abf3 triple mutant and areb1areb2abf3abf1 quadruple mutant showed increased drought sensitivity and decreased ABA sensitivity by impairing the expression of ABA and osmotic stress-responsive genes [18, 22]. Additionally, the overexpres- sion of many AREB/ABFs in various species have been shown to confer increased tolerance to osmotic stress [23–27]. Sev- eral studies have reported that AREB/ABF transgenic agricul- tural plants showed substantial increases in drought tolerance with little or no effect on growth [23, 25, 26]. Bread wheat (Triticum aestivum L.) is the most widely cultivated crop on earth, accounting for approximately one-fifth of the total calories consumed by humans [28]. Consequently, wheat yields and production affect the global economy. However, its productivity is frequently hampered by water scarcity, making improved drought tolerance an important goal of many breeding programs. Although several studies have demonstrated the import- ance of ABFs in response to abiotic stresses, our knowledge of ABFs in wheat is still very limited. In this study, we sys- temically described the characteristics of plant ABFs, in- cluding gene members, phylogenetic relationships, gene structures, protein structural similarities and differences, and gene expression. We performed additional functional analyses of the wheat ABF gene TaABF3 by investigating drought tolerance in transgenic Arabidopsis plants. Our re- sults provide an important framework for understanding the phylogenetic relationship between plant ABF genes and deepens our understanding of the function and mechanism of wheat ABF genes in responses to drought stress. Results Identification and analysis of AREB/ABF family in plants Based on the 34 genomes listed in the Phytozome data- base, we performed a genome-wide BLAST search using Arabidopsis ABF1, AREB1/ABF2, AREB2/ABF4, and ABF3 amino acid sequences. We found the candidate ABFs in only 29 land plants, including moss, lycophyte, monocots, and eudicots. Among the ABF sequences we identified, some proteins had shorter amino acid residues (fewer than 200 amino acids). These short sequences were eliminated from subsequent analyses. In the end, 190 ABF-like sequences were collected for further analysis. We subjected these 190 protein sequences to SMART and Pfam analyses, and all of them were classified into the pro- tein family containing bZIP domains (Pfam: 00170). Previous studies have reported that the plant group-A bZIP family proteins can be phylogenetically clustered into two major groups, the AREB/ABF and the ABI5/ AtDPBF subfamilies [14]. As such, we constructed a maximum likelihood (ML) tree, using 190 full-length ABF-like gene sequences (Additional file 1: Figure S1). Our results show that these ABF-like sequences are di- vided into two major clades, designated as group A and B, each having 95 identified sequences. Group A con- tains all experimentally characterized AREB/ABFs, in- cluding Arabidopsis, Thellungiella salsuginea, and rice ABFs [14, 15]. According to previously characterized such as ABI5/AtDPBF1, AtDPBF2, AREB3/ genes, AtDPBF3 and EEL/AtDPBF4 [14], group B was classified as ABI5/AtDPBF subfamilies. Therefore, group A se- quences are designated ABF and were included for fur- ther analyses (Additional file 2: Table S1). The number of ABFs in each species is shown in Fig. 1. In summary, the moss Physcomitrella patens and the lycophyte Sela- ginella moellendorffii have two copies of the ABFs. In monocots, all species contain only four copies of ABFs, with the exception of wheat, maize and Panicum virga- tum. The ABF copy number differed, from one to seven, in eudicots. This indicates that several duplication inci- dents took place. The quantity of ABF paralogs in rice, Li et al. BMC Plant Biology (2020) 20:558 Page 3 of 15 Fig. 1 Number of ABF paralogs in each species and their clade distributions. The species tree is based on information in Phytozome (http://www. phytozome.net). The star on the branch point within eudicot species indicates the divergence point between a basal eudicot (A. coerulea) and the core eudicots Arabidopsis, and Thellungiella observed by this study are in line with previous research [14, 15]. We further analyzed protein length, molecular mass, and the pI values of 95 ABF proteins (Additional file 2: Table S1). According to our results, the length and molecular mass of ABFs ranged from 254 to 485 amino acid residues, and 27.81 to 52.95 kD, with a mean of 389 amino acid residues and 42.16 kD. Aquilegia coerulea_ABF1 is the longest and largest ABF (485 amino acid residues and 52.95 kD), while Citrus sinensis_ABF3 is the shortest and smallest ABF (254 amino acid residues and 27.81 kD). Zea mays_ABF5 has the lowest pI value, with 5.44, while Citrus sinensis_ABF3 has the highest value, with 10.42. ABFs in clades III, VI and VII have very close pI values, while the pI values of ABFs in clades I, II, IV, and V varied widely. Interestingly, ABFs from clade V dis- played a tendency to maintain acidic pI values, with an aver- age of 6.97, while more alkaline pI values (greater than 7) were observed in 83 out of 95 ABFs belonging to other clades (Table 1; Additional file 2: Table S1). Phylogenetic and structural analysis of plant ABFs In order to better understand the evolutionary relation- ship of AREB/ABF members in land plants, we further constructed an ML tree using full-length protein se- quences of 95 ABFs. According to support values (85% or greater) of the phylogenetic tree, ABFs can be divided into seven clades (clades I to VII) (Fig. 1; Fig. 2). Inside the phylogenetic tree, ABFs from the moss Physcomitrella patens and the lycophyte Selaginella moellendorffii form two independent clades, assigned as clades I (P.patens_ ABFs) to II (S.moellendorffii_ABFs). The monocots can be placed into the next two clades, IV and V. The eudicots can be divided into three clades: III, VI, and VII. It is worth mentioning that the phylogenetic tree (Fig. 2) aligns with the species tree shown in Phytozome (Fig. 1) with the ABFs from moss (Physcomitrella patens) and lycophyte (Selaginella moellendorffii) form- ing the two basal lineages of land plants. Monocot and eudicot ABFs are closer on the phylogenetic tree and form two monophyletic clades. To further investigate the accuracy of the ABF phylogenetic tree, we analyzed the exon/intron organization for each individual gene (Additional file 3: Figure S2). Of the 95ABFs, one has one exon; three have two exons; four have three exons; 71 have four exons, 13 have five exons, two have six exons, and two have seven exons. Within each clade, the Li et al. BMC Plant Biology (2020) 20:558 Page 4 of 15 Table 1 Summary of ABF protein properties Clade Total No. of ABFs Protein length (aa) Molecular Mass (KD) Theoretical pI I II III IV V VI VII Minimum Maximum 2 2 23 14 15 19 21 – – 407 ± 60.81 328 ± 101.82 394 ± 55.16 343 ± 13.91 344 ± 20.08 417 ± 32.06 421 ± 24.87 254 485 42.45 ± 6.76 35.39 ± 7.72 42.51 ± 5.82 36.99 ± 1.51 36.67 ± 2.21 45.62 ± 3.48 46.01 ± 2.68 27.81 52.95 8.60 ± 3.03 8.05 ± 1.61 9.63 ± 0.34 8.69 ± 1.32 6.97 ± 1.03 9.10 ± 0.66 9.40 ± 0.63 5.44 10.42 No. of ABFs with PI > 7 2 2 23 12 5 18 21 – – gene structure of ABFs is relatively conserved, and the adjacent ABFs have a similar exon/intron structure. We then investigated the intron phases of all ABF gene struc- tures. There are three categories of intron phase: phase 0 intron, phase 1 intron, and phase 2 intron. Our analysis indicated that the intron phase patterns (0, 0, 0) and (0, 0, 0, 0) are the predominant patterns across 95 land plant ABFs (Additional file 3: Figure S2). This analysis indicated that we have constructed a phylogenetic tree of the ABF genes in land plants that is highly accurate. Motif composition and arrangement of plant ABFs In order to better understand the phylogenetic relation- ships between plant ABFs, we aligned all of the ABF se- quences to better identify the conserved amino acid residues. Based on the alignment, 35 amino acid residues are completely conserved in 88 ABFs (except for eight shorter ABFs). We further identified the conserved motifs in 95 plant ABFs using the SMART program. Finally, we found fiveconserved protein motifs in all ABFs, which are BRLZ domain and the other four low complexity regions (LCR 1–4; Fig. 3a; Additional file 4: Figure S3). ABFs be- long to the basic-leucine zipper (bZIP) domain transcrip- tion factor family, and we found that the BRLZ domains are highly conserved in all plant sequences (Fig. 4). However, the ability of SMART to comprehensively identify the motifs present in ABFs is limited, so we used the MEME program to identify conservation and vari- ation in the motif arrangements among ABFs. We iden- tified 20 distinct motifs in ABFs. The occurrences and arrangements of the motifs in ABFs from seven major clades are shown in Fig. 3b and Additional file 5: Figure S4. Among 20 motifs, 8 motifs are shared by all ABFs, which are components of the BRLZ domain (motif 1 and 2) and the other four conserved low complexity regions (motif 3 and 6 for LCR1, motif 5 for LCR2, motif 4 for LCR3, and motif 7 for LCR4). Next, we examined the non-conserved motif composition in land plant ABFs. We then split the ABFs into four regions, based on the location of the LCR motifs and the BRLZ domain (Fig. 3b): Region 1 is the part before the LCR1, Region 2 is the part between LCR1 and LCR2, Region 3 is the part between LCR3 and the BRLZ domains (there were no motifs between LCR2 and LCR3), and Region 4 is the part between the BRLZ domain and LCR4. Of these four, Regions 2 and 4 are highly conserved in plants on land (they are mainly comprised of motifs 15 and 16). Less conserved is Region 1, which is primarily comprised of motif 11 in clades III, VI, and VII. Region 3 is the most divergent region: motif 8 was observed in clades I, III, IV, V, VI, and VII; motifs 9 and 10 were found in clades III, IV, VI, and VII; motifs 12 and 17 were found in clades III, V, VI, and VII; motifs 13 and 14 were found in clades III, VI,and VII; motifs 18 were found in clades III and VII; motif 19 was found in clades IV and VI; and motif 20 was found in clade V (Fig. 3b). Taken together, the conserved and non-conserved motif pat- terns of plant ABFs that we identified match the pattern of clades in the phylogenetic tree. Expression analysis of plant ABF genes To obtain the expression profiles of Arabidopsis ABFs, we extracted the expression data from the Arabidopsis eFP Browser (http://bar.utoronto.ca/efp/cgi-bin/efpWeb. cgi). We found that the expression of Arabidopsis ABF paralogs displayed tissue differentiation. For example, A.thaliana_ABF1 displayed significantly higher expres- sion in roots, and A.thaliana_ABF2 displayed signifi- cantly higher expression in seeds, indicating that ABF paralogs have followed the trend of tissue subfunctiona- lization. We found that ABF paralogs in clade III (A.thaliana_ABF2) have higher expression levels than clade VI paralogs (A.thaliana_ABF1, A.thaliana_ABF3; Additional file 6: Figure S5A). We next investigated the expression profiles of other plant ABF genes. Our results demonstrated that soybean (Glycine max) and common bean (Phaseolus vulgaris) ABF paralogs are expressed more in leaves, roots, and flowers than in other tissues, Li et al. BMC Plant Biology (2020) 20:558 Page 5 of 15 Fig. 2 Phylogenetic relationship of plant ABF genes. ABFs are grouped into seven distinct clades (I-VII). Numbers above branches represent the support values and that ABF paralogs in clade III (G.max_ABF3, P.vul- garis_ABF2) have higher expression levels than clade VII paralogs (G.max_ABF1, G.max_ABF1, P.vulgaris_ABF1) (Additional file 6: Figure S5B and C). Within monocots, we studied the expression of the four ABF paralogs from rice (Oryza sativa), and found that O.sativa_OsABF2 (clade III) had higher expression levels than O.sativa_ and O.sativa_OsAREB2 TRAB1, O.sativa_OsAREB1, (Additional file 6: Figure S5D). The five ABFs in maize (Zea mays) display similar expression patterns among tissues, except for Z.mays_ABF3, which is less expressed among all tissues (Additional file 6: Figure S5E). The Li et al. BMC Plant Biology (2020) 20:558 Page 6 of 15 Fig. 3 Sequemce alignment and motif composition of plant ABFs. a Alignment of protein sequences in representative plant ABFs. The conserved domains are annotated as red lines (LCR1-LCR4, BRLZ domain). The alignment was generated using ClustalW implemented in the Geneious software and represented as thick lines (aligned characters) and thin lines (gaps). b Motif patterns in representative ABFs. Motif occurrences were predicted using the MEME program, and the different colored and numbered boxes represent separate and distinct motifs. Stars indicate the BRLZ domain and four different colored arrows indicate the four conserved motifs (LCR1-LCR4) expression divergence of plant ABFs indicated the func- tional differentiation of ABFs. We also investigated the expression of Arabidopsis ABFs under abiotic stresses using microarray expression data. The results showed that the expression of all Arabidopsis ABFs was induced by ABA, cold temperatures, drought conditions, and high salinity, but the degrees of induction differed. A.thaliana_ABF1 was significantly induced by cold tempera- tures; A.thaliana_ABF2 was significantly induced by drought conditions; A.thaliana_ABF3 was significantly induced by ABA, drought conditions and salt; and A.thaliana_ABF4 was significantly induced by drought conditions and salt (Add- itional file 7: Figure S6). We further investigated the expres- sion of other plant ABFs to abiotic stresses (ABA, drought and highly salinity) using quantitative real-time PCR (qRT- PCR). From the heatmap, we found that the expression of most ABFs was induced by ABA, drought conditions, and high salinity. Except for B.rapa_ABF7, G.max_ABF2, and Z.mays_ABF3, all ABFs were significantly induced by drought conditions. ABFs are known for their importance in ABA-mediated abiotic stress responses, meaning a significant induction in the ABF genes might play a crucial role in plant adaptation to environmental stresses (Fig. 5). Molecular characterization and expression analysis of TaABFs Phylogenetic analyses suggest that TaABFs might serve a role in regulating abiotic stress response in wheat. We Li et al. BMC Plant Biology (2020) 20:558 Page 7 of 15 Fig. 4 The sequences of the BRLZ domain from ABFs are conserved across land plants. a A heat map diagram of the protein sequence alignment of BRLZ domain from land plants. b ICE LOGO of the BRLZ domain in the N-terminus of ABF shows that the motif is strictly conserved in land plants Fig. 5 Expression profiles of representative plant ABF genes under different abiotic stresses. The relative expression levels of ABFs were determined using quantitative real-time RT-PCR in the leaves of three-leaf-stage seedlings treated with drought, NaCl, and ABA, compared with the control Li et al. BMC Plant Biology (2020) 20:558 Page 8 of 15 cloned three TaABF genes from the wheat cv. Chinese Spring. Each gene had three homologous components in the A, B, and D genomes of wheat; we named them TaABF1-5A/B/D, TaABF2-7A/B/D, and TaABF3-6A/B/ D. Additional phylogenetic analyses indicated that TaABF1 was most closely related to the rice OsAREB2, TaABF2 was most closely related to the rice OsABF2, and TaABF3 was most closely related to the rice OsAREB1 (Fig. 6a). An analysis of the protein sequence revealed that TaABFs displayed 55–98% sequence simi- larity (Fig. 6b). We then analyzed the subcellular localization of TaABF3, first constructing the expression cassette and fusing TaABF3 with the GFP protein. The fused proteins were then transiently expressed in Arabi- dopsis protoplasts. We used fluorescence microscopy to analyze and reveal that the TaABF3-GFP fusion proteins were exclusively localized in the nucleus in the trans- formed cells, while the control GFP was uniformly dis- tributed throughout (Fig. 7a). These results the cell confirmed that TaABF3 is a nuclear-localized protein. To examine the expression pattern of TaABFs, we first identified the cis-element in its region of promotion, which was ~ 2 kb upstream of the transcription initiation codon, finding a number of cis-acting elements related to stress response in the promoter of TaABFs. This in- cludes LTR (low temperature-responsive element), MYB (MYB recognition site), MYC (MYC recognition site), drought- site MBS inducibility), ABRE (ABA-responsive element), and DRE (Dehydration-responsive element) (Fig. 7b). In order to better understand the role that TaABFs play in response to drought conditions, we executed quantitative real- (MYB binding involved in Fig. 6 Phylogeny, subcellular localization, and expression of TaABFs. a Phylogenetic relationship between TaABFs and ABF members from other plant species. The phylogenetic tree was constructed by MEGA6.0 using the neighbor-joining method. The numbers at each node indicate the percentage of bootstrap values from 1000 replicates. b Protein sequence alignment of TaABFs. The locations of the highly conserved BRLZ domain was indicated by black lines Li et al. BMC Plant Biology (2020) 20:558 Page 9 of 15 Fig. 7 Molecular characterization and expression analysis of TaABFs. a Subcellular localization of TaABF3 protein. 35S:GFP was used as positive control. b Distribution of several stress-related cis-elements in the promoter region (~ 2.0 kb) of TaABFs. LTR, low temperature responsive element; ABRE, ABA-responsive element; DRE, dehydration-responsive element; MBS, MYB binding site involved in drought-inducibility; MYB, MYB recognition site; MYC, MYC recognition site. c The expression profiles of TaABFs in different tissues. R, root of wheat seedling at five-leaf stage; S, stem of wheat seedling at five-leaf stage; L, leaf of wheat seedling at five-leaf stage; FL, flag leaf at heading stage; YS5, young spike at early booting stage; GR5, grain of 5 days post-anthesis; GR15, grain of 15 days post-anthesis. d The expression pattern of TaABFs under drought stress treatment. The numbers on X axis indicate the time point subject to drought stress. The error bars indicate standard deviations derived from three independent biological experiments time PCR (qRT-PCR) on RNA taken from various tis- sues and conditions of drought. Considering the highly sequence similarity of wheat homeologous genes, the PCR primers were designed to amplify the conserved locus of three TaABF homeologs; for example, the rela- tive expression level of TaABF1 represents the combined expression of all three TaABF homeologs (TaABF1-5A, TaABF1-5B and TaABF1-5D). The results demonstrated that TaABFs were found in higher levels in the leaves of the seedlings (Fig. 7c) and that under drought stress conditions, all TaABFs leaves were up- regulated (Fig. 7d). in wheat Overexpression of TaABF3 confers drought tolerance in Arabidopsis To better understand how TaABFs function in plant abi- otic stress tolerance, we generated 35S::TaABF3-GFP transgenic Arabidopsis lines. We then selected three Li et al. BMC Plant Biology (2020) 20:558 Page 10 of 15 independent transgenic lines for 35S::TaABF3-GFP trans- genic Arabidopsis that exhibited higher expression levels of TaABF3 in order to further analyze their response to drought stress (Additional file 8: Figure S7). We then compared the drought tolerance of transgenic and vector- transformed (WT) plants. We grew WT and each 35S:: TaABF3-GFP transgenic plants for 3 weeks in soil before withholding water for ~14d. After the drought treatment and 6 days of re-watering, ~ 65–75% of the transgenic plants survived, while only ~ 8% of the WT plants sur- vived (Fig. 8a and b). We next assayed the proline contents, malondialde- hyde (MDA) contents, and the soluble sugar contents in 35S::TaABF3-GFP transgenic and WT plants (Fig. 8c-e). Our results showed that in transgenic lines the proline contents and the soluble sugar contents were signifi- cantly higher and the MDA contents were significantly lower than in WT under both well-watered and drought conditions. We also detected the expression of several well-known drought-responsive genes in the transgenic including Arabidopsis-homologous LEA14 [29], lines, RD29A [30], DREB2A [31], RAB18 [32], RD20 [33], and GolS2 [34]. These results showed that all of these genes were up-regulated in 35S::TaABF3-GFP transgenic lines (Fig. 8f). Collectively, these findings indicate that the overexpression of TaABF3 in Arabidopsis could enhance the drought tolerance of transgenic plants. Discussion Transcription factors (TFs) are a group of regulatory proteins that regulate gene expression by binding to spe- cific cis-acting elements in the promoters of target genes [35]. Despite the fact that many studies have revealed the crucial role of AREB/ABF TFs in response to abiotic Fig. 8 Drought tolerance assay of TaABF3 overexpression transgenic Arabidopsis plants. a Performance of TaABF3 overexpression Arabidopsis plants under drought stress. b Statistical analysis of survival rates after the drought stress treatment. The average survival rates and standard errors were calculated based on data obtained from three independent experiments. c-e Proline contents (c), malondialdehyde (MDA) contents (d) and soluble sugar contents (e) of the WT and transgenic lines before and after drought treatment. F qRT-PCR analysis of the selected marker genes involved in water response. TaActin was used as the internal control. WW, well-watered; DT, drought stress Li et al. BMC Plant Biology (2020) 20:558 Page 11 of 15 stresses [16–18], our knowledge of ABFs is still limited. Previous studies have primarily focused on studying the function of ABF/AREB proteins, whereas phylogenetic studies of ABFs are restricted to some model plants, such as Physcomitrella patens, Selaginella moellendorffii, Arabidopsis, and rice [14]. To advance our understand- ing of the involvement of ABFs in stress response and other biological processes, it is essential to first under- stand their evolution and diversity. In this study, we col- lected most land plant group-A bZIP TFs from available genome databases (Fig. 1; Additional file 2: Table S1) and performed phylogenetic analyses with full coding se- quences. This allowed us to identify the ABF clade within the bZIP TFs (Fig. 2), the intron/exon structure of genes (Additional file 3: Figure S2), and the character- istic protein domains (Fig. 3; Fig. 4; Additional file 4: Figure S3; Additional file 5: Figure S4). We next extracted expression profiles of selected plants from a public expres- sion database and explored the functional differences of paralog genes during land plant evolution (Fig. 5; Add- itional file 6: Figure S5; Additional file 7: Figure S6). In addition, we systemically investigated the function of target genes of TaABF3 (Fig. 6; Fig. 7; Fig. 8). The goal of our study was to provide an overall picture of plant ABFs and deepen our knowledge of the function and mechanism of wheat ABF genes when responding to abiotic stresses. At the protein level, ABFs in land plants share many of the same structural features, all ABFs have four conserved LCR motifs, and one BRLZ domain (Fig. 3a; Additional file 4: Figure S3). However, the differences of ABF proteins are also existed. For example, the protein structure between the LCR3 motif and the BRLZ domain exhibit the highly variable (Fig. 3b). This region requires additional research to further elucidate the differences in structure and func- tion between the ABFs in land plants from various clades. A robust phylogenetic tree is essential for tracing the evolutionary history of ABF genes. As sequencing tech- niques have advanced, increasing amounts of plant ge- nomes have been sequenced and released. In this study, we surveyed 34 different plant genomes and collected 95 ABF genes. With the exception of algae, ABF candidates exist in all land plants, including lowland plants (a moss and a lycophyte) and highland plants (monocots and eudicots). It is increasingly apparent that gene families present in embryophytes (land plants) and absent from se- quenced chlorophyte genomes have their origins in the kind of algae from which the ancestral land plant evolved. This indicates that the ABF gene family originated during the evolution of the algal to land plants. Previous studies favor the single-origin theory of land plants, originating from charophycean green algae [36, 37]. Moving from an aqueous to a gaseous environment subjects various plants to different physical conditions, which results in particular changes to their structure and physiology. Significant involving flavenoids, metabolic pathways, lignins, plant hormones, and cutins from vascular plants come from existing structures of the primary metabolism in charo- phycean algae [36]. During this process, various families of genes evolved and helped land plants to adapt to challen- ging new environmental conditions, which included abi- otic stressors. It is possible that ABF genes could be induced by several abiotic stresses, participating in stress response to abiotic factors [16–18]. This evolution of ABFs could have played an important role in allowing plants to adapt to conditions on land. Our analysis of the phylogenetic relationship demon- strated that the ABF gene family underwent two changes that led to the seven distinct subfamilies (Fig. 2). The first instance happened after Selaginella moellendorffii and Physcomitrella patens diverged from a common an- cestor, that of seed plants. This occurrence is consistent with the known patterns of divergence in land plants, where Selaginella moellendorffii and Physcomitrella patens are the precursors of the seed plants. Following this event, the family of ABF genes could have been lim- ited to their historical functions. However, our phylogen- etic tree shows that the second instance of duplication that resulted in lineages similar to ABF happened in seed plants. Prior research has found that a whole-genome duplication (WGD) event that occurred in an ancestor of extant angiosperms produced exact copies of each gene [38, 39]. Monocots have seen many instances of WGD throughout their history, which are surely respon- sible for the high instance of ABF genes (Fig. 1) [40]. All of the ABFs in monocots were found in clades IV and V, which is evidence of a duplication event early in the evo- lutionary history of monocots. In contrast, eudicot ABFs were all found in clades III, VI, and VII. Members of clade III are paralogs of Arabidopsis ABF2, while mem- bers of clade VI are paralogs of ABF1, ABF3, and ABF4. Members of clade VII are more similar to members of clade VI but are not found in the paralogs of Arabidopsis. All of this is evidence of the functional similarity of ABFs, while the differences between the clades are indicative of functional differentiation between the clades. Recent re- search has shown that ABFs are involved in ABA signaling when responding to abiotic stressors [16–27], while the functional differences between ABF genes remain scarce. One study found that ABF2, ABF3, and ABF4 play im- portant roles when regulating the mediation of ABA- triggered Chl degradation as well as leaf senescence in Arabidopsis [41]. This demonstrates evolutionary diver- gence in the functionality of ABFs,but in order to under- stand the practical differences within the lineage of ABFs, further research is required. The responsiveness of ABF genes to abiotic stress strongly suggests that they serve roles in adapting to changing envir- further onmental conditions. Our qRT-PCR analyses Li et al. BMC Plant Biology (2020) 20:558 Page 12 of 15 revealed that all TaABF genes were induced by drought stress (Fig. 7). To investigate the role of TaABF genes in the abiotic stress response, TaABF3 was transformed into Arabi- dopsis, and its overexpression was confirmed by RT-PCR (Additional file 8: Figure S7). The transgenic plants showed significantly improved drought and salt tolerance compared to WT plants (Fig. 8). Consistently, several stress-responsive including LEA14, RD29A, DREB2A, RAB18,RD20, genes, and GolS2 were found to be significantly up-regulated in TaABF3 transgenic Arabidopsis under drought stress (Fig. 8). This research strongly indicates that TaABF3 increases the tolerance of transgenic Arabidopsis plants to drought condi- tions. Prior research has shown that an overexpression of TF genes can slow the growth of transgenic plants [42–45]. We also closely monitored the growth and morphological fea- tures of TaABF3 transgenic Arabidopsis plants, finding that transgenic plants exhibited a slight reduction in the size of rosette leaves (Fig. 8). Conclusions In summary, our study provides a comprehensive analysis of the plant ABF genes, include phylogenetic relationships, gene structures, protein structures and properties, and ex- pression profiles. Phylogenetic analysis combined with gene structure and motif composition clustered the plant ABFs into seven distinct clades. In addition, expression analyses demonstrate that plant ABFs have extensively in- duced by abiotic stress. Further functional analysis of TaABF3 trangenic Arabidopsis showed that they could confers drought tolerance in plants. Our results will help elucidate the functions of the AREB/ABF lineage in plants, and providing clues for the identification of candidate genes involved in abiotic stress responses in plants. Methods Plant materials and stress treatments After subjecting four plant species (Brassica rapa, Gly- cine max, Oryza sativa, and Zea mays) to different stress conditions, we assayed the expression of ABFs. We ob- tained Brassica rapa cv. ZS11, Glycine max cv. Jidou-7, Oryza sativa cv. Nipponbare, and Zea mays cv. B73 from Northwest A&F University, though these strains also could have been acquired from the Chinese Crop Germplasm Resources Information System (http://www. cgris.net/zhongzhidinggou/index.php). Growth condi- tions and the application of stress conditions proceeded according to the following: we germinated 1‰ (v/v) Topsin-M sterilized seeds at 25 °C for 3 days on wet fil- ter paper. The germinated seeds were then grown hydro- ponically, with Hoagland nutrient solution, under a 16 h light/8 h dark photoperiod in an artificially controlled climate chamber at 25 °C. The abscisic acid (ABA) was then applied to the cultivated seedlings, followed by high salinity conditions, and finally, drought conditions. The three-week-old seedlings were placed into a 200 mmol/L NaCl solution for the high-salinity treatment, into a 100 μmol/L ABA culture solution for the ABA treat- ment, and on a clean bench for the drought treatment (where they were dehydrated at 25 °C and relative hu- midity of 40–60%). The whole seedlings were collected 0, 2, and 10 h after subjecting them to stress conditions. We collected a minimum of five seedlings from each plant species at each time point, while each experiment was performed three times. All samples were subse- quently frozen in liquid nitrogen and refrigerated at − 80 °C prior to RNA extraction. ABF genes identification For Arabidopsis, ABFs (AREB1/ABF2, AREB2/ABF4, ABF1 and ABF3) were used to conduct a TBLASTN query in the Phytozome databases (http://www.phyto- zome.net/). We used 34 plant species, including algae (Chlamydomonas reinhardtii, Coccomyxa subellipsoidea C-169, Micromonas pusilla CCMP1545, Ostreococcus lucimarinus, and Volvox carteri), moss (Physcomitrella patens), lycophyte (Selaginella moellendorffii), monocots (Triticum aestivum, Oryza sativa, Panicum virgatum, Sorghum bicolor, Setaria italica, and Zea mays), and eudicots (Aquilegia coerulea, Arabidopsis lyrata, Arabi- dopsis thaliana, Brassica rapa, Capsella rubella, Carica papaya, Citrus clementina, Citrus sinensis, Cucumis sati- vus, Eucalyptus grandis, Fragaria vesca, Glycine max, Gossypium raimondii, Manihot esculenta, Mimulus gut- tatus, Phaseolus vulgaris, Prunus persica, Ricinus com- munis, Solanum lycopersicum, Thellungiella salsuginea, Theobroma cacao, and Vitis vinifera). The amino acid sequences, cDNA, and genomic DNA associated with each putative ABF or ABF were obtained from the Phy- tozome database, while we used the Simple Modular Architecture Research Tool (SMART; http://smart.embl- to heidelberg. identify ABFs with protein structures containing bZIP and other common domains. The Compute pI/Mw tool, from ExPASy (http://web.expasy.org/compute_pi/), was used to generate the theoretical molecular mass and Pi (isoelectric point) values. de/smart/set_mode.cgi?NORMAL = 1) Phylogenetic tree construction The TranslatorX server (http://translatorx.co.uk/) [46] was used to align the coding sequences (CDS), while we conducted a jModelTest analysis [47] to identify the model with the best fit. We created a ML (maximum tree using the online program RAxML likelihood) (http://www.trex.uqam.ca/index.php?action=raxml&pro- ject=trex) [48], via the best-fit model with 100 bootstrap samples. FigTree (http://tree.bio.ed.ac.uk/software/ fig- tree/) was used to visualize the phylogenetic tree. Li et al. BMC Plant Biology (2020) 20:558 Page 13 of 15 Analysis of gene structure analysis and conserved motif detection The online Gene Structure Display Server (GSDS; http:// gsds.cbi.pku.edu.cn) was used to assess the distribution of introns and exons and intron phase patterns. The Multiple Expectation Maximization for Motif Elicitation program (MEME; http://meme.nbcr.net/meme/cgi-bin/ meme.cgi) was used to obtain the functional motifs of ABF proteins, using these parameters: maximum num- ber of motifs = 20, optimum motif width = 6 to 100 resi- dues, distribution of motifs = any number of repetitions. ABF gene expression profile qRT-PCR was used to assess ABF expression patterns under different stress conditions, and TRIZOL reagent (Biotopped) was used to isolate total RNA using at least five seedlings from the three separate experiments. Total RNA was treated with Rnase-free DNAse (Takara) to re- move genetic contamination. A Nanodrop1000 (Thermo Scientific product, USA) was used to measure the total RNA levels, while 5 μg of total RNA was run on 0.8% agar- ose gel from each sample to validate the number and in- tegrity of RNA. The cDNAs were synthesized using recombinant M-MLV reverse transcriptase and total RNA (1 μg) mixed with 1 μg Oligo (dT)23 (Promega). The PCR conditions involved a preliminary denaturation for 10 min at 95 °C, 40 cycles of 15 s at 95 °C, and 40 cycles of 30 s at 60 °C. The internal control was TaActin (TraesC- S1A01G274400). We applied the quantification method (2-ΔCt) and approximated the expression variation using three biological replicates [49]. Additional file 9: Table S2 outlines the primers used in this study. Subcellular localization of TaABF3-GFP fusion proteins Prof. Zhensheng Kang’s Lab (Northwest A&F University, China) provided the Triticum aestivum cv. Chinese spring, which was used in the functional analysis of TaABF3. The full-length CDS sequence of TaABF3 was amplified using PCR from the wheat cv. Chinese Spring with specific primers for the subcellular localization assay of TaABF3. This was then placed into the binary vector pCAMV35S::GFP, in between BamH I and Xba I, to find the subcellular localization of TaABF3. Sequen- cing was used to obtain positive clones, and the wheat mesophyll protoplasts were obtained from the constructs using the methods previously described [50]. A confocal microscope (Olympus, FluoViewTM FV300, Japan) was used to assess GFP fluorescence. Transformation of Arabidopsis and isolation of TaABF3 Prof. Zhensheng Kang’s Lab (Northwest A&F University, China) provided the Arabidopsis ecotype Columbia, which was used to transform TaABF3. We amplified the full-length opening reading frame of TaABF3 from the 35S into virus (CaMV) promoter wheat cv. Chinese Spring, using gene-specific primers which were subsequently cloned with the cauliflower mosaic the pGreen0029-GFP vector. We then introduced the re- combinant vector (35S::TaABF3-GFP) into Agrobacter- ium tumefaciens, and used the floral dip method [51] to transform it into Arabidopsis (Arabidopsis thaliana; eco- type Columbia). T1 seeds were placed on a MS medium containing 2% sucrose and 50 mg/mL kanamycin to identify the transformants. Phenotypic analyses were performed using homozygous T3 plants. Drought tolerance assay We placed germinated, seven-day-old transgenic Arabi- dopsis plants on an MS medium into pots with a 130 g mix of 2:1 mixture of Jiffy mix and vermiculite to per- form the drought tolerance assays. The 32-day-old plants grown under optimal conditions (22 °C, relative humidity of 60%, 16/8 h light/dark photoperiod) were subjected to drought stress conditions by withholding water from the plant for 14 days, after which they were watered and allowed to recover. We then counted how many plants survived after 6 days. A minimum of 48 plants from each line were analyzed against wild-type (WT) plants for each test. Statistical data shown is based on data obtained from the three independent experi- ments. We used a student’s t-test to analyze the differ- ences between transgenic and WT plants. Measuring proline, MDA, and soluble sugar levels We measured the proline contents, MDA levels, and sol- uble sugar levels of the transgenic and WT plants sub- jected to 10 days of drought stress, at which most leaves began to wilt, using detection kits (Solarbio) according to the manufacturer’s instructions. Statistical analyses Each experiment was conducted a minimum of three times. Data shown are the mean ± standard deviation (SD) of the three independent replicates. A Student’s t- test was used to perform the statistical analysis, while P < 0.05 was considered statistically significant and P < 0.01 was considered extremely significant. Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12870-020-02783-9. Additional file 1: Figure S1. Phylogenetic relationship of group-A bZIP TFs from 29 plant species. The land plant group-A bZIP TFs are grouped into two major clades, designated as AREB/ABF and ABI5/AtDPBF subfamilies. Additional file 2: Table S1. ABF protein properties and their clade- distributions in land plants. Gene identifiers are obtained from Phytozome database. Li et al. BMC Plant Biology (2020) 20:558 Page 14 of 15 Additional file 3: Figure S2. Schematic diagram of gene structures of 95 plant ABFs. The thin lines represent introns and thick bars represent exons. The numbers above the gene structure indicate intron phases. A scale bar with a unit of base pair (bp) is graphed on the bottom. Additional file 4: Figure S3. Alignment of 95 plant ABF protein sequences. The alignment was generated using ClustalW implemented in Geneious software and represented as thick lines (aligned characters) and thin lines (gaps). Overall alignment identity and a scale bar indicating the numbers of amino acid residues are graphed on the top. Additional file 5: Figure S4. Combined motif diagram of 95 ABF proteins. Thick lines represent the ABF proteins. Different colored boxes represent separate and distinct motifs identified using MEME program. A scale bar indicating the numbers of amino acid residues is shown on the top. Motifs are drawn approximately to scale as boxes. Additional file 6: Figure S5. Gene expression profile of ABF paralogs in plants. Gene expression data was extracted from Arabidopsis thaliana (http://jsp.weigelworld. org/expviz/expviz.jsp), soybean (Glycine max, http:// soybase.org/soyseq/), common bean (Phaseolus vulgaris, http://plantgrn. noble.org/PvGEA/SearchVisual.jsp), maize (Zea mays, http://www.plexdb. org/index.php), and rice (Oryza sativa, http://www.plexdb.org/index.php). Additional file 7: Figure S6. Gene expression profile of ABF paralogs in Arabidopsis thaliana under different abiotic stresses. The mean- normalized expression values were obtained from the AtGenExpress microarray database via the web http://jsp.weigelworld.org/ expviz/ expviz.jsp. Additional file 8: Figure S7. RT-PCR analysis of TaABF3 transcription levels in the transgenic Arabidopsis lines. Additional file 9: Table S2. Primers used in this research. Abbreviations ABA: Abscisic acid; ABFs: Abscisic acid-responsive transcription factors; ABRE: ABA-responsive element; BLAST: Basic local alignment search tool; CDS: Coding sequences; DRE: dehydration-responsive element; GEO: Gene Expression Omnibus; GO: Gene Ontology; GSDS: Gene Structure Display Server; LCR: low complexity region; LTR: low temperature responsive element; MBS: MYB binding site involved in drought-inducibility; MEME: Multiple Expectation Maximization for Motif Elicitation; ML: Maximum likelihood; MYB: MYB recognition site; MYC: MYC recognition site; Pi: isoelectric point; SMART: Simple Modular Architecture Research Tool; TFs: Transcription factors Acknowledgements We thank reviewers for checking our manuscript and the editors for editing the paper. We would like to thank the members of the Bioinformatics Center of Northwest A & F University for their useful input. Authors’ contributions HM conceived and initiated the research; HM designed the experiments; FL, FM, YZ, and SL carried out the experiments. HM analyzed the data and wrote the manuscript. All authors have read and approved the final manuscript. Funding This research was supported by Talent Fund of Northwest A&F University (grant no. Z111021602). The funders had no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript. Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare that they have no conflict of interest. Received: 25 January 2020 Accepted: 3 December 2020 References 1. Nakashima K, Takasaki H, Mizoi J, Shinozaki K, Yamaguchi-Shinozaki K. NAC transcription factors in plant abiotic stress responses. Biochim Biophys Acta. 2012;1819(2):97–103. Yamaguchi-Shinozaki K, Shinozaki K. Transcriptional regulatory networks in cellular responses and tolerance to dehydration and cold stresses. Annu Rev Plant Biol. 2006;57:781–803. Zhu JK. Abiotic stress signaling and responses in plants. Cell. 2016;167(2): 313–24. Finkelstein RR, Gampala SS, Rock CD. Abscisic acid signaling in seeds and seedlings. Plant Cell. 2002;14(Suppl):S15–45. Himmelbach A, Yang Y, Grill E. Relay and control of abscisic acid signaling. Curr Opin Plant Biol. 2003;6(5):470–9. 2. 3. 4. 5. 7. 6. Munemasa S, Hauser F, Park J, Waadt R, Brandt B, Schroeder JI. Mechanisms of abscisic acid-mediated control of stomatal aperture. Curr Opin Plant Biol. 2015;28:154–62. Hauser F, Li Z, Waadt R, Schroeder JI. SnapShot: Abscisic acid signaling. Cell. 2017;171(7):1708. Zhu JK. Salt and drought stress signal transduction in plants. Annu Rev Plant Biol. 2002;53:247–73. Fujita Y, Fujita M, Shinozaki K, Yamaguchi-Shinozaki K. ABA-mediated transcriptional regulation in response to osmotic stress in plants. J Plant Res. 2011;124(4):509–25. 8. 9. 10. Ma Y, Szostkiewicz I, Korte A, Moes D, Yang Y, Christmann A, Grill E. Regulators of PP2C phosphatase activity function as abscisic acid sensors. Science. 2009;324(5930):1064–8. 11. Park SY, Fung P, Nishimura N, Jensen DR, Fujii H, Zhao Y, Lumba S, Santiago J, Rodrigues A, Chow TF, Alfred SE, Bonetta D, Finkelstein R, Provart NJ, Desveaux D, Rodriguez PL, McCourt P, Zhu JK, Schroeder JI, Volkman BF, Cutler SR. Abscisic acid inhibits type 2C protein phosphatases via the PYR/ PYL family of START proteins. Science. 2009;324(5930):1068–71. 12. Choi H, Hong J, Ha J, Kang J, Kim SY. ABFs, a family of ABA responsive element binding factors. J Biol Chem. 2000;275(3):1723–30. 13. Uno Y, Furihata T, Abe H, Yoshida R, Shinozaki K, Yamaguchi-Shinozaki K. Arabidopsis basic leucine zipper transcription factors involved in an abscisic acid-dependent signal transduction pathway under drought and high- salinity conditions. Proc Natl Acad Sci U S A. 2000;97(21):11632–7. Fujita Y, Yoshida T, Yamaguchi-Shinozaki K. Pivotal role of the AREB/ABF- SnRK2 pathway in ABRE-mediated transcription in response to osmotic stress in plants. Physiol Plant. 2013;147(1):15–27. 14. 16. 15. Vysotskii DA, de Vries-van Leeuwen IJ, Souer E, Babakov AV, de Boer AH. ABF transcription factors of Thellungiella salsuginea: structure, expression profiles and interaction with 14-3-3 regulatory proteins. Plant Signal Behav. 2013;8(1):e22672. Kang JY, Choi HI, Im MY, Kim SY. Arabidopsis basic leucine zipper proteins that mediate stress-responsive abscisic acid signaling. Plant Cell. 2002;14(2):343–57. Fujita Y, Fujita M, Satoh R, Maruyama K, Parvez MM, Seki M, Hiratsu K, Ohme- Takagi M, Shinozaki K, Yamaguchi-Shinozaki K. AREB1 is a transcription activator of novel ABRE-dependent ABA signaling that enhances drought stress tolerance in Arabidopsis. Plant Cell. 2005;17(12):3470–88. 18. Yoshida T, Fujita Y, Maruyama K, Mogami J, Todaka D, Shinozaki K, 17. Yamaguchi-Shinozaki K. Four Arabidopsis AREB/ABF transcription factors function predominantly in gene expression downstream of SnRK2 kinases in abscisic acid signalling in response to osmotic stress. Plant Cell Environ. 2015;38(1):35–49. Fujii H, Chinnusamy V, Rodrigues A, Rubio S, Antoni R, Park SY, Cutler SR, Sheen J, Rodriguez PL, Zhu JK. In vitro reconstitution of an abscisic acid signalling pathway. Nature. 2009;462(7273):660–4. Furihata T, Maruyama K, Fujita Y, Umezawa T, Yoshida R, Shinozaki K, Yamaguchi-Shinozaki K. Abscisic acid-dependent multisite phosphorylation regulates the activity of a transcription activator AREB1. Proc Natl Acad Sci U S A. 2006;103:1988–93. 19. 20. 21. Wang Z, Su G, Li M, Ke Q, Kim SY, Li H, Huang J, Xu B, Deng XP, Kwak SS. Overexpressing Arabidopsis ABF3 increases tolerance to multiple abiotic Li et al. BMC Plant Biology (2020) 20:558 Page 15 of 15 42. Sun Y, Fan XY, Cao DM, Tang W, He K, Zhu JY, He JX, Bai MY, Zhu S, Oh E, Patil S, Kim TW, Ji H, Wong WH, Rhee SY, Wang ZY. Integration of brassinosteroid signal transduction with the transcription network for plant growth regulation in Arabidopsis. Dev Cell. 2010;19(5):765–77. 43. Zhang W, Ruan J, Ho TH, You Y, Yu T, Quatrano RS. Cis-regulatory element based targeted genefinding: genome-wide identification of abscisic acid- and abiotic stress-responsive genes in Arabidopsis thaliana. Bioinformatics. 2005;21(14):3074–81. Lu Z, Yu H, Xiong G, Wang J, Jiao Y, Liu G, Jing Y, Meng X, Hu X, Qian Q, Fu X, Wang Y, Li J. Genome-wide binding analysis of the transcription activator ideal plant architecture1 reveals a complex network regulating rice plant architecture. Plant Cell. 2013;25(10):3743–59. 44. 45. Birkenbihl RP, Kracher B, Somssich IE. Induced genome-wide binding of three Arabidopsis WRKY transcription factors during early MAMP-triggered immunity. Plant Cell. 2017;29(1):20–38. 46. Abascal F, Zardoya R, Telford MJ. TranslatorX: multiple alignment of nucleotide sequences guided by amino acid translations. Nucleic Acids Res. 2010;38:W7–13. 47. Posada D. jModelTest: phylogenetic model averaging. Mol Biol Evol. 2008; 48. 25(7):1253–6. Stamatakis A. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics. 2006; 22(21):2688–90. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real- time quantitative PCR and the 2−ΔΔCt method. Methods. 2001;25(4):402–8. 50. Yoo SD, Cho YH, Sheen J. Arabidopsis mesophyll protoplasts: a versatile cell system for transient gene expression analysis. Nat Protoc. 2007;2(7):1565–72. 49. 51. Clough SJ, Bent AF. Floral dip: a simplified method for agrobacterium- mediated transformation of Arabidopsis thaliana. Plant J. 1998;16(6):735–43. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. stresses and reduces leaf size in alfalfa. Plant Physiol Biochem. 2016;109: 199–208. 22. Yoshida T, Fujita Y, Sayama H, Kidokoro S, Maruyama K, Mizoi J, Shinozaki K, Yamaguchi-Shinozaki K. AREB1, AREB2, and ABF3 are master transcription factors that cooperatively regulate ABRE-dependent ABA signaling involved in drought stress tolerance and require ABA for full activation. Plant J. 2010; 61(4):672–85. 23. Oh SJ, Song SI, Kim YS, Jang HJ, Kim SY, Kim M, Kim YK, Nahm BH, Kim JK. Arabidopsis CBF3/DREB1A and ABF3 in transgenic rice increased tolerance to abiotic stress without stunting growth. Plant Physiol. 2005;138(1):341–51. 24. Chinnusamy V, Zhu JK, Sunkar R. Gene regulation during cold stress acclimation in plants. Methods Mol Biol. 2010;639:39–55. 25. Orellana S, Yañez M, Espinoza A, Verdugo I, González E, Ruiz-Lara S, Casaretto JA. The transcription factor SlAREB1 confers drought, salt stress tolerance and regulates biotic and abiotic stress-related genes in tomato. Plant Cell Environ. 2010;33(12):2191–208. Li C, Yue J, Wu X, Xu C, Yu J. An ABA-responsive DRE-binding protein gene from Setaria italica, SiARDP, the target gene of SiAREB, plays a critical role under drought stress. J Exp Bot. 2014;65(18):5415–27. Kerr TC, Abdel-Mageed H, Aleman L, Lee J, Payton P, Cryer D, Allen RD. Ectopic expression of two AREB/ABF orthologs increases drought tolerance in cotton (Gossypium hirsutum). Plant Cell Environ. 2018;41(5):898–907. IWGSC. Shifting the limits in wheat research and breeding using a fully annotated reference genome. Science. 2018;361(6403). https://pubmed.ncbi. nlm.nih.gov/30115783/. 26. 27. 28. 29. Hundertmark M, Hincha DK. LEA (late embryogenesis abundant) proteins and their encoding genes in Arabidopsis thaliana. BMC Genomics. 2008;9:118. 30. Nakashima K, Fujita Y, Katsura K, Maruyama K, Narusaka Y, Seki M, Shinozaki K, Yamaguchi-Shinozaki K. Transcriptional regulation of ABI3- and ABA- responsive genes including RD29B and RD29A in seeds, germinating embryos, and seedlings of Arabidopsis. Plant Mol Biol. 2006;60(1):51–68. Sakuma Y, Maruyama K, Osakabe Y, Qin F, Seki M, Shinozaki K, Yamaguchi- Shinozaki K. Functional analysis of an Arabidopsis transcription factor, DREB2A, involved in drought-responsive gene expression. Plant Cell. 2006; 18(5):1292–309. 31. 32. Nylander M, Svensson J, Palva ET, Welin BV. Stress-induced accumulation and tissue-specific localization of dehydrins in Arabidopsis thaliana. Plant Mol Biol. 2001;45(3):263–79. 33. Aubert Y, Vile D, Pervent M, Aldon D, Ranty B, Simonneau T, Vavasseur A, 34. Galaud JP. RD20, a stress-inducible caleosin, participates in stomatal control, transpiration and drought tolerance in Arabidopsis thaliana. Plant Cell Physiol. 2010;51(12):1975–87. Selvaraj MG, Ishizaki T, Valencia M, Ogawa S, Dedicova B, Ogata T, Yoshiwara K, Maruyama K, Kusano M, Saito K, Takahashi F, Shinozaki K, Nakashima K, Ishitani M. Overexpression of an Arabidopsis thaliana galactinol synthase gene improves drought tolerance in transgenic rice and increased grain yield in the field. Plant Biotechnol J. 2017;15(11):1465–77. 35. Badis G, Berger MF, Philippakis AA, Talukder S, Gehrke AR, Jaeger SA, Chan ET, Metzler G, Vedenko A, Chen X, Kuznetsov H, Wang CF, Coburn D, Newburger DE, Morris Q, Hughes TR, Bulyk ML. Diversity and complexity in DNA recognition by transcription factors. Science. 2009;324(5935):1720–3. Kenrick P, Crane PR. The origin and early evolution of plants on land. Nature. 1997;389:33–9. 36. 38. 37. Morris JL, Puttick MN, Clark JW, Edwards D, Kenrick P, Pressel S, Wellman CH, Yang Z, Schneider H, Donoghue PCJ. The timescale of early land plant evolution. Proc Natl Acad Sci U S A. 2018;115(10):E2274–83. Jiao YN, Wickett NJ, Ayyampalayam S, Chanderbali AS, Landherr L, Ralph PE, Tomsho LP, Hu Y, Liang HY, Soltis PS, Clifton SW, Schlarbaum SE, Schuster SC, Ma H, Leebens-Mack J, dePamphilis CW. Ancestral polyploidy in seed plants and angiosperms. Nature 2011;473:97–100. Jiao YN, Paterson AH. Polyploidy-associated genome modifications during land plant evolution. Philos Transact Royal Soc London Series B. 2014;369: 20130355. Jiao YN, Li JP, Tang HB, Paterson AH. Integrated syntenic and phylogenetic analyses reveal and ancient genome duplication in monocots. Plant Cell. 2014;26:2792–802. 40. 39. 41. Gao S, Gao J, Zhu X, Song Y, Li Z, Ren G, Zhou X, Kuai B. ABF2, ABF3, and ABF4 promote ABA-mediated chlorophyll degradation and leaf senescence by transcriptional activation of chlorophyll catabolic genes and senescence- associated genes in Arabidopsis. Mol Plant. 2016;9(9):1272–85.
10.1155_2019_6891831	Hindawi Occupational erapy International Volume 2019, Article ID 6891831, 12 pages https://doi.org/10.1155/2019/6891831 Research Article The Child Evaluation Checklist (CHECK): A Screening Questionnaire for Detecting Daily Functional “Red Flags” of Underrecognized Neurodevelopmental Disorders among Preschool Children Sara Rosenblum ,1 Irit Ezra Zandani,1,2 Tsofia Deutsch-Castel,3 and Sonya Meyer 1 1Laboratory of Complex Human Activity and Participation (CHAP), Department of Occupational Therapy, Faculty of Social Welfare & Health Sciences, University of Haifa, Mount Carmel, Israel 2Child Development Center, Maccabi Healthcare Services, Ashkelon, Southern District, Israel 3Neurodevelopmental Service, Maccabi Healthcare Services, Haifa, Northern District, Israel Correspondence should be addressed to Sara Rosenblum; rosens@research.haifa.ac.il Received 2 July 2019; Revised 23 September 2019; Accepted 11 November 2019; Published 1 December 2019 Academic Editor: Erna I. Blanche Copyright © 2019 Sara Rosenblum et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Background. Early identiﬁcation of invisible comorbid neurodevelopmental disorders, such as speciﬁc learning disorders, attention deﬁcit hyperactive disorders, and developmental coordination disorders, is crucial to improving children’s daily functional deﬁcits related to executive functions. However, a practical questionnaire to address parents’ concerns is lacking. Aims. To develop a reliable and valid assessment tool that can identify young children at risk for invisible underrecognized neurodevelopmental disorders. This article describes the development and standardization of the Child Evaluation Checklist (CHECK). Methods and Procedures. Participants were 186 children aged 3 to 6 years: 91 with suspected invisible neurodevelopmental disorders, and 95 controls with typical development. Parents completed a demographic questionnaire, the CHECK, and the Behavior Rating Inventory of Executive Function-Preschool Version (BRIEF-P). Outcomes and Results. The CHECK’s construct validity indicated high internal consistency for each part (Part A: α = :94; Part B: α = :90) and moderate-to-high consistency for each of Part A’s four factors. Signiﬁcant correlations, as well as signiﬁcant group diﬀerences, were found between the CHECK factors and BRIEF-P scores. Conclusions and Implications. Use of the CHECK allows for timely identiﬁcation of suspicious (“red ﬂags”) invisible neurodevelopmental disorders. It may support parents’ suﬃcient awareness and knowledge to refer their children for comprehensive evaluation and intervention. 1. Introduction Neurodevelopmental disorders are a group of developmental conditions characterized by developmental deﬁcits in per- sonal, social, academic, or occupational functioning [1]. Among these disorders are three diagnoses that are underre- cognized and have high comorbidity with each other [2]. Prevalent among 3% to 20% of children [1], these diagnoses include speciﬁc learning disorders (SLD) [3, 4]), attention deﬁcit hyperactive disorders (ADHD) [5], and developmen- tal coordination disorders (DCD). This paper describes the development of the Child Evaluation Checklist (CHECK), a short screening tool aimed at identifying children who are at risk for these underrecognised invisible neurodevelopmen- tal conditions. CHECK focuses on the small nuances of children’s daily functional activity performance features as related to their executive functions. According to the Diag- nostic and Statistical Manual of Mental Disorders (DSM-5) [1], SLD refers to deﬁciencies in reading, writing, and math- ematics [6]. There is a high occurrence (40%–70%) of SLD with other developmental deﬁcits [7], and as many as 30% to 50% of children with SLD also have attention deﬁcits [8, 9]. DSM-5 deﬁnes ADHD as attention deﬁcit and dis- ruptive behavior disorders characterized by three sets of 2 Occupational Therapy International symptoms—inattention, hyperactivity, and impulsivity—with each set comprising a list of nine observable behavioral symptoms [1, 10]. Developmental coordination disorder or “clumsiness” is deﬁned as a substantially below age-expected acquisition and execution of coordinated motor skills that sig- niﬁcantly and persistently interferes with the activities of daily living (ADL) and aﬀects academic/school productivity, prevo- cational and vocational activities, leisure, and play [11]. The DCD diagnosis is phenotypically heterogeneous, with up to 70% of children meeting criteria for at least one other neuro- developmental disorder [12]. Despite the high commonality (5% to 10%) of DCD, and it being a signiﬁcant risk factor in the long-term development of children and adolescents, it is underrecognized and underdiagnosed [11]. Although high comorbidity has been found between the abovementioned diagnoses and sensory processing deﬁcits, this diagnosis is not yet included in DSM-5 [12–14], and therefore sensory processing deﬁcits are not included in this study. Regardless of the high percentage of comorbidity or cooccurrence among these three diagnoses that appear in DSM-5 (from 50% to 70%) or the complexity of their clinical manifestation [15], the general trend is to refer to each dis- ability as a separate diagnosis. Usually, children are diagnosed at school, around the age of 8 years or older [1, 16]. In fact, young children with SLD, ADHD, and DCD exhibit day-to-day functional deﬁcits even before going to school, for example, in their personal hygiene, ADL, interpersonal relationships, communication, ﬁne and gross motor activities, organization in space and time, and play and leisure [17–21]. Such deﬁcits may concern both children and adults and negatively impact the child and their whole family [18, 22, 23]. Previous literature indicated that people with invisible neurodevelopmental disabilities feel already from very early that “Something with me or my child is not the same as others” in reference to their daily func- tional capabilities. Nevertheless, a tool that can combine all the pieces of daily functional evidence to one whole picture of functional deﬁcits tied to invisible neurodevelopmental disabilities is lacking [17–23]. Theoretical models and research ﬁndings have previously connected these daily functional deﬁcits with deﬁcient executive functions (e.g., [24, 25]). The term “executive function” refers to a neuro- psychological process that enables physical, cognitive, and emotional self-control [26]. The hierarchical hybrid model of executive functions developed by Barkley [27] places inhibition at the top of the hierarchy. Other executive functions including nonverbal working memory; internali- zation of speech (verbal working memory); self-regulation of aﬀect, motivation, and arousal; and reconstitution are placed at the lower level [27]. Together, these skills impact the functional cognition performance used to accomplish essential activities in daily life [27, 28]. Eﬃcient executive functions provide children eﬃcient day-to-day functioning while managing age-related daily tasks, coping with chal- lenges, and solving problems [29]. Executive functions are also described as body functions in the International Classiﬁcation of Functioning, Disability, and Health (ICF) [30], a conceptual and operational framework published by the World Health Organization. The ICF describes executive multidirectional interactions between body functions and structures, activities, and participation, while considering both environmental and personal factors. Deﬁcient execu- tive functions have been described among children and adults with SLD [31], ADHD [32], and DCD [33]. Fur- function-related deﬁciencies were thermore, described in these populations related to their body and motor control [34], verbal abilities [35], self-regulation and control, and social abilities [36–38]. Such deﬁciencies may appear very early in life in the way children engage in daily functional activities. Furthermore, they predict functional impairments later in life, limiting participation in various life areas [39–41]. Children suspected with neurodevelopmental invisible disabilities are at signiﬁcant developmental risk in the long term. Their performance abilities, self-esteem, and wellbeing may be negatively aﬀected [25, 42]. Despite knowing the importance of the early detection of executive function deﬁ- ciency in daily functioning among younger children, most daily functional delays among children suspected for neuro- developmental invisible disabilities are diagnosed (if at all) only when the child reaches school age. Thus, this phenome- non remains underrecognized [1, 16]. Knowledge is scarce about how deﬁcient executive functions can impact these children’s performance in daily activities, especially in pre- school ages. Furthermore, it is important to consider gender diﬀerences. Especially at young ages, boys are generally more extroverted than girls and, thus, may reﬂect signs of impaired daily function abilities in their behavior, whereas girls are more introverted and may express their frustration less often and more verbally [43]. Early identiﬁcation is therefore important to prevent future emotional problems, such as reduced self-esteem, anxiety, and depression, as well as social and behavioral problems [44, 45]. There are several screening question- naires for detecting developmental delay among children aged from birth to ﬁve that focus on physical, cognitive, linguistic, and social-emotional growth and development [46]. One parent questionnaire is the Behavior Rating Inven- tory of Executive Function: Preschool Version (BRIEF-P) that assesses executive functions [47]. The Child Evaluation Checklist (CHECK) adds to those existing tools by providing the option of identifying children ages 3 to 6 years at risk for invisible neurodevelopmental disorders through a short, easy to complete parent report about their children’s day-to-day functioning, with speciﬁc emphasis on executive functions. Both the complexity and latency were considered for enabling identiﬁcation, even before a physician provides a speciﬁc diagnosis according to the DSM-5 criteria [48]. As such, this current study’s research hypotheses are as follows: (1) The CHECK’s construct validity will be estab- lished by the factor analysis for Parts A and B, and each factor will show adequate internal reliability (α > :70). (2) Signiﬁ- cant correlations will be found between the CHECK scores and all ﬁve BRIEF-P subscale scores (inhibition, shift, emo- tional control, working memory, and planning), thus estab- lishing concurrent validity. (3) Signiﬁcant diﬀerences will be found in the CHECK scores, beyond gender and age, between children diagnosed by a pediatrician as suspected Occupational Therapy International 3 for invisible neurodevelopmental disorders and those with typical development. 2. Materials and Methods 2.1. Questionnaire Development and Content Validity Determination. The CHECK is a one-page questionnaire designed for use by parents to provide information about their children’s ability to function within the context of their natural environments during the previous three months. The questionnaire was developed based on a number of resources that established the tool’s content validity: (a) the DSM-5 deﬁnitions of SLD, ADHD, and DCD [1], (b) the ICF frame- work, which served as the basis for understanding interac- tions between components that reﬂect functioning [30], (c) Barkley’s hybrid model of executive functions [27], (d) the current literature about daily functional challenges of chil- dren with these conditions, (e) previously designed screening questionnaires (e.g., [49, 50]), and (f) analysis of interviews with parents of children and adults with SLD, ADHD, and DCD about their children’s daily functioning experiences, confrontations, and challenges (e.g., [17, 51–54]). Initially, based on these resources and the researchers’ extensive clin- ical experience, including observations on young children, 48 statements were formulated by the ﬁrst author to reﬂect daily routine functions that challenge children aged 3 to 6 years with suspected invisible disabilities because of the need for executive function involvement in their performance (i.e., ADL, communication, inhibition and self-regulation, and organization in space and time). Each statement was associ- ated with at least one of the main concepts of the ICF [30]. For example, “Can solve problems created in play\wardrobe” classiﬁed as body functions, and “Does his\her needs inde- pendently, in comparison with what’s expected of children of his\her age” classiﬁed as activities and participation. The 48-statement questionnaire was sent to three physicians and ﬁve senior occupational therapists, all experts in child development. To establish content validity, the experts were asked to comment whether (1) each item is appropriate for detecting functional deﬁciency related to executive functions among children with suspected invisible neurodevelopmen- tal disabilities or whether (2) the items were clearly worded. There was 100% expert agreement for 40 items but only 60% agreement items that were consequently deleted. Additionally, the wording of ﬁve items was improved following expert recommendations. For example, in Item 15, “Organizes body for activity,” the following examples were added in parenthesis for clarity (i.e., jumping, skipping, and throwing a ball). for eight The 40-item questionnaire was then sent to two other expert pediatric occupational therapists with 20 years clinical experience, and three occupational therapy researchers expe- rienced in populations with invisible disabilities across the lifespan. A second round of content expert validity was per- formed based on their feedback. Following the expert input, the wording of three items was again improved and 100% agreement was achieved for the ﬁnal 40 items. The CHECK’s ﬁnal version is divided into two parts. Part A includes 30 items that entail the domains previously described that are related to various daily activities (e.g., Item 7, “Eats in a manner suiting his\her counterparts, such as cleanliness, tidiness, control over utensils” or Item 2, “Under- stands instructions he\she is provided”). Parents are asked to score the frequency in which the item describes their child: always (4), often (3), rarely (2), or never (1). Part B includes 10 sentences about the child’s global performance level related to various executive function outcomes reﬂected in daily function. Items are rated on a scale from 1 (very low performance level) to 5 (high performance level) compared with that of the child’s typically developing peers. A higher score represents better performance. Examples from Part B include Item 2, “In comparison with other children, the child’s attention and concentration ability is …” or Item 6, “The child’s adaptation ability to changes in routine is …” 2.2. Participants. A required sample size of 135 partici- pants was calculated with the G∗Power statistical program, based on a moderate eﬀect size (f 2ðVÞ = :0625, alpha = :05, Power = :80). Children previously diagnosed with an intel- lectual, physical, or neurological disability were excluded. Initially, 96 children diagnosed with suspected invisible neurodevelopmental disorders participated in the study. Those children were referred by their family physician or pediatrician to a child developmental center because their parents or teachers were concerned that the children were not performing like other children. A developmental pedi- atrician conﬁrmed the parents’ concerns and deﬁned the children as suspected for invisible neurodevelopmental dis- orders (SLD, ADHD, and DCD symptoms) based on the DSM-5 criteria [1]. This developmental pediatrician did not determine a speciﬁc diagnosis but recommended follow-up with or intervention by a pediatrician or occu- pational therapist. Another group of 95 children with typical development were recruited from the same kindergartens or communities as the children with invisible disabilities through a chain-referral sampling method. Parents com- pleted a demographic questionnaire and reported neither dif- ﬁculties in daily functioning nor the need to be referred to a health or educational professional because of any develop- mental functional concerns. Participants were then matched for age, gender, and socioeconomic level as reﬂected in the mothers’ years of education (ranged from 9 to 20 years). Pre- vious research has found that the level of mother’s education and socioeconomic status can impact their child’s develop- ment because of the learning opportunities and possibilities that are directed by the mother-child interaction [55]. The results presented hereafter refer to 186 children (91 children with suspected invisible neurodevelopmental disor- ders and 95 with typical development) aged 3 to 6 years—141 (75.8%) boys and 45 (24%) girls. 2.3. Instruments 2.3.1. BRIEF-P. The BRIEF-P [47] consists of 63 items related to behavioral manifestations of executive functions, rated on a 3-point scale indicating whether the behavior occurs never (1), sometimes (2), or often (3). The items are divided into ﬁve scales of executive functions (inhibitory control, shifting, 4 Occupational Therapy International emotional control, working memory, and planning and orga- nization) that produce three index scores (inhibitory self- control, ﬂexibility, and emergent metacognition). The sum of the clinical scales reveals the global executive composite. In addition, two validity scales were obtained. The inconsis- tency scale aims to determine if the respondent has answered in an especially conﬂicting manner, and the negativity scale measures whether the respondent answered in an unusually pessimistic manner. Higher scores indicate more dysregula- tion in behaviors associated with executive functions. 2.4. Procedure. The Health Care Service Human Research Ethics Committee (Helsinki approval No. 2009087), as well as the Israeli Ministry of Education (No. 506/7902) and the Institutional Ethics Committee (No. 320/13), authorized the study. All parents signed informed consent forms and were then asked to complete a demographic questionnaire, the BRIEF-P, and the CHECK. 2.5. Data Analysis. Data were analyzed using SPSS version 22 and descriptive statistics to describe the participants. To verify the CHECK’s construction and dimensions based on the theoretical and clinical experience of the CHECK’s developer, exploratory factor analysis was con- ducted using principle components to ﬁnd the factors of Parts A and B. The number of extracted factors in each part was chosen based on a screen plot of the eigenvalues and on factor interpretability. The resulting factor solution was subsequently rotated by means of an oblique (Oblimin) rotation procedure due to the possible correlation of the factors which all represent functional reﬂections of executive function deﬁciency. Item- factor loading with values of at least .35 were deemed salient. All items that did not meet this criterion were dropped, as were all items that loaded highly on multiple factors. Internal consistency reliability was evaluated using Cronbach’s alpha coeﬃcient. After conﬁrming the CHECK’s ﬁnal format, Pearson’s correlation analyses were performed on the entire sample to better understand the relationship between the CHECK fac- tors and BRIEF-P subscale scores and to establish concurrent validity. Consequently, gender diﬀerences and diﬀerences between children with invisible disabilities and those with typical development were analyzed across the CHECK fac- tors via MANCOVA analysis, holding age as the covariate. Univariate ANCOVA analyses were used to determine the source of the group diﬀerences. The ANCOVA was per- formed to check for group diﬀerentiation of the ﬁnal CHECK scores while holding age as covariate. 3. Results and Discussion 3.1. Examination of the Questionnaire Validity and Reliability 3.1.1. Construct Validity 1: Exploratory Factor Analysis for CHECK Part (A and B) (1) CHECK Part A: Child’s Daily Function Nuances as Reﬂectors of Executive Function Abilities. Analysis of Part A revealed four distinct factors, comprised of 30 items, with eigenvalues greater than 1 (Table 1). The four factors yielded a cumulative variance percentage of 54.05%, with an internal consistency of α = :94. The four factors, as well as the internal consistency reliability measured by the coeﬃcient alpha of each factor are as follows: (1) The ﬁrst factor, expression/performance manage- ment, included 12 items and accounted for 37.23% of the variance with α = :91 (2) The second factor, included six items and accounted for 6.31% of the variance with α = :79 self-regulation, (3) The third factor, organization—body, essentials, social, included nine items and accounted for 5.89% of the variance with α = :88 (4) The fourth factor, ADL, included three items and accounted for 4.67% of the variance with α = :68 (2) CHECK Part B: Child’s General Daily Function Compared to Others. The analysis revealed one distinct factor with eigenvalues > 1, comprised of 10 items (Table 2). The cumu- lative percentage of this one factor was 59.33% with an inter- nal consistency of α = :92. 3.1.2. Internal Consistency Reliability. As presented in Tables 1 and 2, Cronbach’s alpha coeﬃcient—calculated separately for each part—indicated excellent internal con- sistency for each (Part A, 30 items: α = :94; Part B, 10 items: α = :92). Suﬃcient-to-high levels of internal consis- tency were achieved for each factor in Part A, ranging from α = :68 to α = :91. 3.1.3. Concurrent Validity of the CHECK with the BRIEF- P: Entire Sample (N = 186). As shown in Table 3, signif- icant negative low-to-moderate correlations (r = −:23 to r = −:62, p < :001) were found between the four factors of Part A (cognitive-language, self-regulation, organiza- tion, and ADL), the CHECK Parts A and B ﬁnal scores with the three BRIEP-P indexes (inhibitory self-control, ﬂexibility, and emergent metacognition), and the global executive composite and each subscale (inhibition, shift, control, memory, and planning). 3.1.4. Construct Validity 2. In this phase, construct validity of the CHECK factors was examined by analyzing gender diﬀer- ences and group diﬀerences (children with suspected invisi- ble neurodevelopmental disorders and controls). Although no signiﬁcant diﬀerence was found in age between groups, age was held as the covariate because the signiﬁcance level was borderline (p = :062). Pearson’s correlations were con- ducted to test the correlation between age and CHECK fac- tors showing a weak positive correlation (r = −:25 to r = −:30, p < :001). Occupational Therapy International 5 Table 1: CHECK factor loading of questionnaire items and internal consistency: Part A. Item Description Expression and performance management Factor Self-regulation Organization ADLa 10 3 4 2 9 18 25 1 30 8 19 11 27 29 26 16 24 28 15 21 20 22 23 13 14 12 17 6 5 7 Expresses thoughts Remembers stories Verbalizes him/herself Understands instructions Talks about school Takes responsibility Corrects him/herself when wrong Attentive Correctly estimates diﬃculty of task Solves problems in play/dressing Completes tasks Does activities in reasonable time Is generally calm Calms him/herself down Gets out of bed willingly Deals with changes Is not impulsive Sleeps well at night Organizes body for activity Has good balance Has good coordination Has good in-hand control of small objects Has good pencil control; writes and draws what he/she wants Cooperates with friends Gets ready for a game Communicates properly to get what he or she wants Is likable among friends Is independent in the lavatory Is independent dressing Eats cleanly and in order Eigenvalue % of variance Internal consistency (α): entire sample (N = 186) α-Children with suspected disorders (n = 91) α-Children with typical development (n = 95) Note. N = 186. aActivities of daily living. .777 .772 .758 .715 .699 .645 .564 .520 .520 .514 .473 .359 11.16 37.23 .91 .91 .73 .683 .641 .618 .497 .445 .355 1.89 6.31 .79 .67 .68 .842 .828 .727 .627 .551 .524 .505 .486 .435 1.77 5.89 .88 .81 .83 .762 .732 .703 1.38 4.89 .68 .73 .69 As presented in Table 4, initial analysis indicated no signiﬁcant group diﬀerences in participants’ sociodemo- graphic variables, such as children’s age and gender or mother’s education. The MANCOVA analysis indicated signiﬁcance diﬀer- ences across age (Fð4,178Þ = 8:63, p < :001, partial η2 = :162), no signiﬁcance across gender (Fð4,178Þ = 1:88, p = :12, partial η2 = :040), and signiﬁcant group diﬀerences (Fð4,178Þ = :11:40, p < :001, partial η2 = :204) in Part A. Furthermore, no signiﬁ- cant gender or group interaction was found (Fð4,178Þ = 1:37, p = :246, partial η2 = :03). The subsequent ANCOVA analy- sis for the four CHECK factors indicated signiﬁcant age con- language and cognition (Fð1,181Þ = 19:05, tribution for p < :001, partial η2 = :10), organization (Fð1,181Þ = 15:48, p < :001, partial η2 = :08), and ADL (Fð1,181Þ = 19:01, p < :001, partial η2 = :10). No signiﬁcant contribution was found for emotional regulation (Fð1,181Þ = 1:43, p = :234, partial η2 = :01). Results of the ANCOVA analysis for the four CHECK factors across gender and groups are presented in Table 5. In addition, ANCOVA analysis for the ﬁnal scores of CHECK Parts A and B showed a signiﬁcant eﬀect of age to the ﬁnal CHECK A and B scores (Fð1,181Þ = 20:81, 6 Occupational Therapy International Table 2: CHECK factor loading of questionnaire items and internal consistency: Part B. Item (compared to other children): Loading General functioning Attentional capacity Work habits Initiation capability Inhibition Emotional domain Verbal communication Social domain Memory capacity Adapt to changes Eigenvalue % of variance Internal consistency (α): entire sample (N = 186) α-Children suspected disorders (n = 91) α-Children with typical development (n = 95) Note. N = 186. .854 .850 .819 .794 .792 .774 .752 .751 .691 .588 5.930 59.330 .920 .91 0 .76 0 p < :001, partial η2 = :10; Fð1,181Þ = 21:54, p < :001, partial η2 = :10, respectively). 4. Discussion This study describes the development of CHECK—a quick, easy-to-use, and practical screening tool focused on chil- dren’s daily functional characteristics. As such, it is suitable for use with parents. CHECK responds to the need for assessment tools that address the concepts of activities and participation, deﬁned by the International Classiﬁca- tion of Functioning Disability and Health [30] as central concepts related to each individual’s health. The four factors achieved for Part A of CHECK support the study’s results presented in the Introduction and reﬂect the children’s daily challenges. The items in the ﬁrst factor reﬂect how the children express themselves and how they manage their daily performance. Domains required at this phase are attention (Item 1), understanding instructions (Item 2), memory (Item 3, remembering stories), and correctly estimating the diﬃculty of a task (Item 25). The children then perform the daily tasks while managing their performance, thus self-verbalizing (Item 4), talking about school (Item 9), expressing thoughts (Item 10), and doing the activities in a reasonable time (Item 11), but also correct- ing themselves when wrong (Item 25), solving problems if they occur (Item 8), taking responsibility (Item 18), and completing the task (Item 19). Such a process of expression, performance, and monitoring is based on a combination of executive function abilities such as attention, inhibition, ini- tiation, working memory, shifting, and planning and organi- zation [40]. Among the items in this factor are verbal skills that allow children to learn, understand, and interpret their physical, social, and conceptual worlds [56]. Indeed, verbal delay may be one of the ﬁrst reasons for parents’ concern about their children’s development and for seeking profes- sional help [56, 57]. After preparation and performance management, the second factor reﬂects the self-regulation and control abilities required for success in daily demands. The literature has described varied self-regulation and adaptive functioning aspects, such as cognitive skills, exec- utive functions, emotions, and motivation [58]. Thus, the second factor includes items that reﬂect the children’s abil- ity to self-regulate, not be impulsive, be generally calm, calm themselves, cope with changes, and get out of bed willingly. At preschool age, children’s ability to regulate their behavior relates to school readiness [59, 60]. Because deﬁcits in self-regulation, self-control, and self-monitoring were described among children and adults with ADHD [61–63] and SLD [64], identifying the deﬁciency early in the child’s development may lead to appropriate strategies to prevent future failure. The third factor uniquely combines body organization (including pencil control) and social behavior under the umbrella term of organizational abilities. This ability, described by Godefroy [65] as an executive function domain, is the ability to organize thoughts to eﬃciently plan and carry out activities in the correct sequence and tempo within the given time range and space [66, 67]. Thus, organizational abilities are required to eﬃciently execute body motion (Items 15, 20, and 21; [33]), converse a message (Item 12), and play and communicate with friends (Items 13, 14, and 17), as well as control a pen or small object (Items 22–23). For example, controlling small objects requires gentle organi- zation and motor adjustments of the small intrinsic hand muscles in space and time, deﬁned as in-hand manipulation [68], for eﬃcient skilled performance. Because organizational abilities such as dysgraphia are signiﬁcantly inferior among children with SLD, ADHD, and DCD and signiﬁcantly corre- lated with their handwriting proﬁciency [69], more emphasis needs to be given to this skill in the preschool years. Thus, deﬁcient organization abilities may be reﬂected in how children organize their bodies and control objects in their play and social behavior. In all, these three CHECK factors successfully represent the action control necessary for task execution [58]. The ﬁnal CHECK factor consists of ADL items, which are the basic tasks children learn to perform by themselves as they develop and become more independent. Performing these skills provides children with self-competence and is important for both the children and the family atmosphere, especially in the morning when getting organized to leave home for school and work [70]. The medium-to-high internal reliability values found for these factors (range α = :68‐:91), as well as for the entire Part A (α = :94), indicates that these items successfully reﬂect daily functional challenges among children aged 3 to 6 years with suspected invisible neurodevelopmental disorders. Fur- thermore, it indicates that parents are able to report success- fully those items as reﬂecting their children’s functional abilities. As shown in the high internal reliability achieved for Part B (α = :92), parents are also able to rank their Occupational Therapy International 7 Table 3: Correlations between four CHECK factors: Parts A and B ﬁnal scores, entire sample. CHECK Factor Part total Self-regulation Organization ADL A B BRIEF Shifting Inhibitory control Working memory Emotional control Expression and performance management -.427∗∗ -.276∗∗ -.360∗∗ -.669∗∗ -.548∗∗ -.433∗∗ -.656∗∗ -.356∗∗ -.419∗∗ -.349∗∗ -.356∗∗ -.587∗∗ -.502∗∗ Planning and organization -.425∗∗ ISCIa -.584∗∗ FIb -.393∗∗ EMIc Note. N = 186. aInhibitory self-control index; bﬂexibility index; cemergent metacognition index. ∗Correlation is signiﬁcant at the 0.05 level (2-tailed). ∗∗Correlation is signiﬁcant at the 0.01 level (2-tailed). -.529∗∗ -.367∗∗ -.457∗∗ -.668∗∗ -.583∗∗ -.542∗∗ -.669∗∗ -.464∗∗ -.624∗∗ -.408∗∗ -.495∗∗ -.497∗∗ -.425∗∗ -.621∗∗ -.493∗∗ -.510∗∗ -.431∗∗ -.345∗∗ -.397∗∗ -.574∗∗ -.540∗∗ -.453∗∗ -.590∗∗ -.419∗∗ -.241∗∗ -.160∗ -.253∗∗ -.252∗∗ -.252∗∗ -.265∗∗ -.264∗∗ -.229∗∗ Table 4: Comparison of demographic characteristics of children suspected for neurodevelopmental disabilities and those with typical development. Variable Child’s age (months) Mother’s education (years) Children with suspected neurodevelopmental disabilities (n = 91) M (SD) 53.88 (9.27) 14.35 (2.41) Children with typical development (n = 95) M (SD) 54.06 (9.24) 14.94 (2.07) Gender Boys Girls 72 (79.1%) 19 (20.9%) 69 (72.6%) 26 (27.4%) t p .89 1.79 χ2 .062 .886 p 1.07 .301 children’s function well as compared to other children based on the 10 items the authors chose to include in this part. Those results support previous ﬁndings about the accuracy of the information parents supply when asked appropriate questions to identify “red ﬂags” in their chil- dren’s daily functional abilities [71, 72]. However, previous studies have indicated that 3 to 6 years pass between the time when parents sense that something about their child is not the same as other children and the time a diagnosis is given [73]. Thus, it is important to provide opportuni- ties to identify invisible-disability developmental delays in the preschool years. During the preidentiﬁcation period, negative inﬂuences may appear in the children’s and the parent’s functional and emotional experiences, as well as in the interactions between parents and their children [51]. Children’s frustra- tion and sense of failure, accompanied by family strains, can cause secondary socioemotional, social, and family prob- lems and deﬁcient participation and self-perception [74–77]. Hence, obtaining knowledge from parents by means of a standardized early screening tool based on the children’s abilities, as reﬂected in daily activity performance, can be valuable towards this essential identiﬁcation of “red ﬂags.” Further evidence for the beneﬁts of CHECK in detecting functional deﬁcits related to executive functions was achieved through the concurrent validity results. Signiﬁcant negative low-to-medium correlations (r = −:27 to −:57) were found between CHECK’s three action control factors and executive function domains (i.e., inhibitory control, shifting, emotional control, working memory, and planning and organization). Speciﬁcally, the high correlation level between both CHECK parts and the BRIEF-P working memory and planning abili- ties (r = −:50 to −:67) align with previous literature stating that deﬁcits in these executive functions may cause deﬁcient daily functioning among children [70]. Interestingly, the low- est signiﬁcant correlation was found with ADL performance (i.e., independent lavatory, independent dressing, and clean eating; r = :16‐:26). Indeed, ADL performance at a young age requires learning the order and sequence of the activities to be performed, and there is a need to implement executive functions while using the lavatory, dressing, or eating. However, those speciﬁc activities all have less complex demands than other life tasks (such as those described in the other three factors). The signiﬁcant group diﬀerences found for all CHECK factors and the ﬁnal scores while considering age indicate the sensitivity of the scale in distinguishing between children at risk for invisible neurodevelopmental disorders and those not at risk. In fact, deﬁcient abilities in speciﬁc tasks were reported in this population concerning how they manage their actual performance self-regulation and self-control [56], and organization abilities [2, 33, 68]. [77–79], 8 Occupational Therapy International Table 5: Gender and group diﬀerences across four CHECK factors and Parts A and B ﬁnal scores. Factor Gender Control group (n = 95) Suspected invisible disabilities group (n = 91) Total gender Gender F ð (ηp 2) 1,181 Þ Group F ð (ηp 2) 1,181 Þ Gender ∗ group F 1,181 ð (ηp 2) Þ Expression and performance management Self-regulation Organization ADLd CHECK A ﬁnal score CHECK B ﬁnal score Boysa (n = 141) Girlsb (n = 45) Totalc (N = 186) Boysa Girlsb Totalc Boysa Girlsb Totalc Boysa Girlsb Totalc Boysa Girlsb Totalc Boysa Girlsb Totalc 3.54 (.31) 3.04 (.52) 3.29 (.49) 3.58 (.27) 3.26 (.52) 3.45 (.42) 3.55 (.30) 3.09 (.52) 3.47 (.39) 3.49 (.45) 3.48 (.40) 3.73 (.31) 3.79 (.26) 3.75 (.29) 3.66 (.34) 3.64 (.51) 3.65 (.39) 3.60 (.27) 3.63 (.24) 3.61 (.26) 4.17 (.62) 4.25 (.53) 4.19 (.60) 3.16 (.44) 3.39 (.46) 3.21 (.46) 3.18 (.50) 3.52 (.45) 3.25 (.51) 3.34 (.62) 3.47 (.60) 3.37 (.61) 3.13 (.40) 3.39 (.42) 3.19 (.42) 3.41 (.71) 3.62 (.84) 3.46 (.74) 3.31 (.45) 3.44 (.45) 3.45 (.50) 3.67 (.37) 3.50 (.53) 3.57 (.55) 3.36 (.41) 3.53 (.35) 3.78 (.77) 3.98 (.74) 2.49 .01 2.37 .01 7.15∗∗ .04 0.12 .00 4.86∗ .03 0.89 .01 34.45∗∗∗ .016 7.88∗∗ .04 36.60∗∗∗ .17 7.80∗∗ .04 39.05∗∗∗ .18 38.40∗∗∗ .17 1.97 .01 2.20 .01 5.17∗ .03 0.99 .01 4.24∗ .02 0.49 .00 Note. an = 141; bn = 45; cN = 186. dADL = activities of daily living. Furthermore, signiﬁcant relationships were found between children’s executive function abilities and their instrumental ADL [77], play [21], and physical activity performance [80]. The results of this study also indicate the need to consider gender diﬀerences when attempting to identify children sus- pected for neurodevelopmental invisible disorders. Signiﬁ- cant diﬀerences were found between boys and girls in their organization abilities for their bodies and objects, as well as social organization, in favor of girls. On one hand, this sug- gests that boys have more deﬁciencies in this area; on the other hand, it challenges identiﬁcation of girls who, although they tend to be more organized, may confront invisible disabilities. More studies are needed to further establish the CHECK reliability and validity; however, these primary results enable its practical clinical use as a screening tool. Considering the emotional and social consequences of “invisible disabilities” for children, parents, and families [81], early identiﬁcation is crucial. Parents are in a position to observe their children over time and across settings and are the best source of infor- mation about their children [80]. They are often the ﬁrst to recognize something concerning about their child’s develop- ment and therefore seek further professional services from pediatricians. However, although screening questionnaires can provide important information, they serve only as a basis for a more comprehensive evaluation by proﬁcient clinicians [82]. Nevertheless, this study’s ﬁndings provide encouraging indications that CHECK may be a useful screening tool to identify “red ﬂags” for children aged 3 to 6 years with neuro- developmental invisible disability characteristics through parents’ reports about their children’s daily functioning. Using CHECK can provide parents and the education team with a tool to see each child and his or her needs. It not only highlights diﬀerences in the child’s daily perfor- mance compared to other children, but also more speciﬁ- cally deﬁnes the nature of the diﬃculty. Furthermore, because of the complexity and comorbidity of invisible neu- rodevelopmental disorders, CHECK may aid in observing individual diﬀerences, strengths, and weakness throughout development, while identifying not only disabilities but also abilities [15, 81, 83]. 5. Conclusions The current study conﬁrms that the CHECK scale, which is eﬃcient, short, and easy to use, can reveal reliable and valid information concerning the presence of subtle early signs of possible future SLD, ADHD, or DCD at the ages of 3 to 6 years. Therefore, the use of CHECK can increase early detec- tion of such neurodevelopmental disorders among children and help professionals refer children for comprehensive eval- uation and further intervention with caution, while consider- ing normal variation [52]. Occupational Therapy International 9 6. Limitations and Future Research The children in this study were all from the northern region of the country and were referred to a regional child develop- mental center. The control group of children with typical development were recruited by a chain-referral sampling method. Although parents reported no developmental con- cerns, these children did not undergo a developmental assessment. Given that Part B items achieved the highest loading values, it is questionable whether asking worried par- ents speciﬁcally about their child’s general function, attention capacity, and work habits in comparison to those of peers will lead to more a comprehensive evaluation. Although CHECK exhibited a good level of internal construct and concurrent validity, it is important to emphasize that the clinical validity of any measure requires testing over time and implementa- tion across a variety of larger sample groups. In addition, a longitudinal study is required to discover whether the chil- dren suspected for invisible neurodevelopmental disabilities (as identiﬁed by CHECK) were indeed diagnosed with ADHD, SLD, or DCD later in childhood. Data Availability The data used to support the ﬁndings of this study have not been made available because they are restricted by the The Health Care Service Human Research Ethics Committee (Helsinki approval No. 2009087), as well as the Israeli Minis- try of Education (No. 506/7902) in order to protect patient privacy. Data are available from Dr. Tsoﬁa Deutsch-Castel, deutsh_t@mac.org.il for researchers who meet the criteria for access to conﬁdential data. Conflicts of Interest The authors declare that there is no conﬂict of interest regarding the publication of this paper. Acknowledgments The authors thank Efrat Ben-Nevat and Wafaa Yassin Khar- boush for their contributions to data collection in the process of their master’s theses. References [1] American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, American Psychiatric Associa- tion, Arlington, VA, 5th ed. edition, 2013. [2] N. Josman and S. Rosenblum, “A metacognitive model for children with neurodevelopmental disorders,” in Cognition, Occupation, and Participation across the Life Span: Neurosci- ence, Neurorehabilitation and Models for Intervention in Occu- pational Therapy, N. Katz and J. Toglia, Eds., pp. 273–294, AOTA Press, Bethesda, MD, 2018. [3] I. S. Fortes, C. S. Paula, M. C. Oliveira, I. A. Bordin, J. de Jesus Mari, and L. A. Rohde, “A cross-sectional study to assess the prevalence of DSM-5 speciﬁc learning disorders in representa- tive school samples from the second to sixth grade in Brazil,” European Child & Adolescent Psychiatry, vol. 25, no. 2, pp. 195–207, 2016. [4] K. A. Kavale and L. S. Spaulding, “Is response to intervention good policy for speciﬁc learning disability?,” Learning Disabil- ities Research & Practice, vol. 23, no. 4, pp. 169–179, 2008. [5] B. Pajo and P. H. Stuart, “A comparative review of "how to" books for parents of ADHD children and "how to" books for parents of typical children,” Children and Youth Services Review, vol. 34, no. 4, pp. 826–833, 2012. [6] D. Scanlon, “Speciﬁc learning disability and its newest deﬁni- tion: which is comprehensive? And which is insuﬃcient?,” Journal of Learning Disabilities, vol. 46, no. 1, pp. 26–33, 2013. [7] S. Vaughn, “Social skills enhancement in students with learn- ing disabilities,” in Learning about Learning Disabilities, B. Y. L. Wong, Ed., pp. 407–436, Elsevier academic press, San Diego, CA, 3rd ed edition, 2014, http://search.ebscohost.com/login .aspx?direct=true&db=nlebk&AN=195174&site=eds-live. [8] C. Cortiella and S. H. Horowitz, The State of Learning Disabil- ities: Facts, Trends and Emerging Issues, National Center for Learning Disabilities, New York, NY, 3rd ed. edition, 2014. [9] G. J. DuPaul, M. J. Gormley, and S. D. Laracy, “Comorbidity of LD and ADHD: implications of DSM-5 for assessment and treatment,” Journal of Learning Disabilities, vol. 46, no. 1, pp. 43–51, 2012. [10] R. Tannock, “Rethinking ADHD and LD in DSM-5: proposed changes in diagnostic criteria,” Journal of Learning Disabilities, vol. 46, no. 1, pp. 5–25, 2013. [11] R. Blank, A. L. Barnett, J. Cairney et al., “International clinical practice recommendations on the deﬁnition, diagnosis, assess- ment, intervention, and psychosocial aspects of developmental coordination disorder,” Developmental Medicine & Child Neu- rology, vol. 61, no. 3, pp. 242–285, 2019. [12] B. Elbasan, H. Kayıhan, and I. Duzgun, “Sensory integration and activities of daily living in children with developmental coordination disorder,” Italian Journal of Pediatrics, vol. 38, p. 14, 2012, http://www.ijponline.net/38/1/14. [13] A. Ghanizadeh, “Sensory processing problems in children with ADHD, a systematic review,” Psychiatry Investigation, vol. 8, no. 2, pp. 89–94, 2011. [14] X. Tong, X. Tong, and F. King Yiu, “Beyond auditory sensory processing deﬁcits: lexical tone perception deﬁcits in Chinese children with developmental dyslexia,” Journal of Learning Disabilities, vol. 51, no. 3, pp. 293–301, 2017. [15] D. Dewey and F. P. Bernier, “The concept of atypical brain development in developmental coordination disorder (DCD): a new look,” Current Developmental Disorders Reports, vol. 3, no. 2, pp. 161–169, 2016. [16] C. Gillberg, “The ESSENCE in child psychiatry: early symp- tomatic syndromes eliciting neurodevelopmental clinical examinations,” Research in Developmental Disabilities, vol. 31, no. 6, pp. 1543–1551, 2010. [17] A. Aviran, S. Berman, C. Frisch, and S. Rosenblum, “How par- ents with symptoms of attention deﬁcit/hyperactivity disorder (ADHD) experience the parental occupational executive training (POET) intervention process: initial impressions,” Israeli Journal of Occupational Therapy, vol. 26, no. 3, p. H234, 2017. [18] C. Dunford, C. Missiuna, E. Street, and J. Sibert, “Children’s perceptions of the impact of developmental coordination dis- order on activities of daily living,” British Journal of Occupa- tional Therapy, vol. 68, no. 5, pp. 207–214, 2005. [19] E. Perez, M. Turner, A. Fisher, J. Lockwood, and D. Daley, the preschool ﬁve minute speech “Linguistic analysis of 10 Occupational Therapy International sample: what the parents of preschool children with early signs of ADHD say and how they say it?,” PLoS One, vol. 9, no. 9, article e106231, 2014. [20] S. Rosenblum, “The development and standardization of the children activity scales (ChAS-P/T) for the early identiﬁcation of children with developmental coordination disorders,” Child: Care, Health and Development, vol. 32, no. 6, pp. 619– 632, 2006. [21] S. Rosenblum, P. Waissman, and G. W. Diamond, “Identifying play characteristics of pre-school children with developmental coordination disorder via parental questionnaires,” Human Movement Science, vol. 53, pp. 5–15, 2017. [22] C. K. Whalen, B. Henker, L. D. Jamner et al., “Toward map- ping daily challenges of living with ADHD: maternal and child perspectives using electronic diaries,” Journal of Abnormal Child Psychology, vol. 34, no. 1, pp. 111–126, 2006. [23] B. P. White and S. E. Mulligan, “Behavioral and physiologic response measures of occupational task performance: a pre- liminary comparison between typical children and children with attention disorder,” American Journal of Occupational Therapy, vol. 59, no. 4, pp. 426–436, 2005. [24] C. Frisch and S. Rosenblum, “Reliability and validity of the executive scale (EFORTS),” Research in Developmental Disabilities, vol. 35, no. 9, pp. 2148–2157, 2014. occupational function routines and [25] S. Rosenblum, “Handwriting measures as reﬂectors of execu- tive functions among adults with developmental coordina- tion disorders (DCD),” Frontiers in Psychology, vol. 4, p. 357, 2013. [26] B. A. Corbett, L. J. Constantine, R. Hendren, D. Rocke, and S. Ozonoﬀ, “Examining executive functioning in children with autism spectrum disorder, attention deﬁcit hyperactivity dis- and typical development,” Psychiatry Research, order vol. 166, no. 2–3, pp. 210–222, 2009. [27] R. A. Barkley, “Behavioral inhibition, sustained attention and executive functions: constructing a unifying theory of ADHD,” Psychological Bulletin, vol. 121, no. 1, pp. 65–94, 1997. [28] G. M. Giles, “Neurocognitive rehabilitation: skills or strate- gies?,” American Journal of Occupational Therapy, vol. 72, no. 6, pp. 7206150010p1–7206150010p16, 2018. [29] P. Dawson and R. Guare, Executive Skills in Children and Ado- lescents: A Practical Guide to Assessment and Intervention, Guilford Press, New York, NY, 3rd ed. edition, 2018. [30] World Health Organization, ICF: International Classiﬁcation of Functioning, Disability, and Health, World Health Organi- zation, Geneva, Switzerland, 2001. [31] L. Meltzer, Executive Function in Education: From Theory to Practice, Guilford Press, New York, NY, 2011. [32] S. Rosenblum, C. Frisch, T. Deutsh-Castel, and N. Josman, “Daily functioning proﬁle of children with attention deﬁcit hyperactive disorder: a pilot study using an ecological assess- ment,” Neuropsychological Rehabilitation, vol. 25, no. 3, pp. 402–418, 2015. [33] S. Rosenblum, “Do motor ability and handwriting kinematic measures predict organizational ability among children with developmental coordination disorders?,” Human Movement Science, vol. 43, pp. 201–215, 2015. [34] C. Gillberg, “Deﬁcits in attention, motor control, and percep- tion: a brief review,” Archives of Disease in Childhood, vol. 88, no. 10, pp. 904–910, 2003. [35] R. Nathanson, J. N. Crank, K. J. Saywitz, and E. Ruegg, “Enhancing the oral narratives of children with learning dis- abilities,” Reading & Writing Quarterly, vol. 23, no. 4, pp. 315–331, 2007. [36] N. Eisenberg, T. L. Spinrad, and N. D. Eggum, “Emotion- related self-regulation and its relation to children’s maladjust- ment,” Annual Review of Clinical Psychology, vol. 6, pp. 495– 525, 2010. [37] M. M. Martel and J. T. Nigg, “Child ADHD and personality/- temperament traits of reactive and eﬀortful control, resiliency, and emotionality,” Journal of Child Psychology and Psychiatry, vol. 47, no. 11, pp. 1175–1183, 2006. [38] P. M. Wehmeier, A. Schacht, and R. A. Barkley, “Social and emotional in children and adolescents with ADHD and the impact on quality of life,” Journal of Adolescent Health, vol. 46, no. 3, pp. 209–217, 2010. impairment [39] B. Engel-Yeger, N. Josman, and S. Rosenblum, “Behavioural Assessment of the Dysexecutive Syndrome for Children (BADS-C): an examination of construct validity,” Neuropsy- chological Rehabilitation, vol. 19, no. 5, pp. 662–676, 2009. [40] M. Miller, A. J. Nevado-Montenegro, and S. P. Hinshaw, “Childhood executive function continues to predict outcomes in young adult females with and without childhood- diagnosed ADHD,” Journal of Abnormal Child Psychology, vol. 40, no. 5, pp. 657–668, 2012. [41] K. Sharﬁ and S. Rosenblum, “Executive functions, time organi- zation and quality of life among adults with learning disabil- ities,” PLoS One, vol. 11, no. 12, article e0166939, 2016. [42] A. Kirby, L. Edwards, D. Sugden, and S. Rosenblum, “The development and standardization of the adult developmental co-ordination disorders/dyspraxia checklist (ADC),” Research in Developmental Disabilities, vol. 31, no. 1, pp. 131–139, 2010. [43] C. S. Tamis-LeMonda, R. D. Briggs, S. G. McClowry, and D. L. Snow, “Maternal control and sensitivity, child gender, and maternal education in relation to children's behavioral out- comes in African American families,” Journal of Applied Developmental Psychology, vol. 30, no. 3, pp. 321–331, 2009. [44] L. Erlenmeyer-Kimling, D. Rock, S. A. Roberts et al., “Atten- tion, memory, and motor skills as childhood predictors of schizophrenia-related psychoses: the New York High-Risk Project,” American Journal of Psychiatry, vol. 157, no. 9, pp. 1416–1422, 2000. [45] E. Swanson and S. Vaughn, “Learning disabilities: academic and mental health needs,” in Critical Issues in School-Based Mental Health: Evidence-Based Research, Practice, and Inter- ventions, M. K. Holt and A. E. Grills, Eds., Routledge, New York, NY, 2015. [46] S. Moodie, P. Daneri, S. Goldhagen, T. Halle, K. Green, and L. LaMonte, Early Childhood Developmental Screening: A Compendium of Measures for Children Ages Birth to Five. (OPRE Report 2014-11), Oﬃce of Planning, Research and Eval- uation, Administration for Children and Families, U.S. Department of Health and Human Services, Washington, DC, 2014. [47] G. A. Gioia, K. A. Espy, and P. K. Isquith, BRIEF-P: Behavior Rating Inventory of Executive Function: Preschool Version, Psychological Assessment Resources, Lutz, FL, 2003. [48] E. R. Blanche and L. D. Parham, “Praxis and organization of behavior in time and space,” in The Nature of Sensory Integra- tion with Diverse Populations, S. Smith Roley, E. I. Blanche, and R. C. Schaaf, Eds., pp. 183–200, Therapy Skill Builders, Tucson, AZ, 2001. Occupational Therapy International 11 [49] J. Squires, D. Bricker, and L. Potter, “Revision of a parent- completed development screening tool: ages and stages ques- tionnaires,” Journal of Pediatric Psychology, vol. 22, no. 3, pp. 313–328, 1997. [50] F. Glascoe, Collaborating with Parents: Using Parents’ Evalua- tion of Developmental Status (PEDS) to Detect and Address Developmental and Behavioral Problems, Ellsworth & Van- dermeer, Nashville, TN, 2nd edn edition, 2013. [51] C. Frisch, E. Tirosh, T. Deutsh-Castel, and S. Rosenblum, Parental Occupational Executive Training (POET): A Unique Intervention Model for Improving Daily Functioning of Young Children with ADHD Symptoms, Paper presented at the World Federation of Occupational Therapy Congress, Cape Town, South Africa, 2018. [52] S. Rosenblum, “Where from, where to, why, and how? The daily functional journey as a source of meaning and self- creation,” Israeli Journal of Occupational Therapy, vol. 27, pp. 8–28, 2018. [53] S. Rosenblum and N. Weintraub, “Learning disabilities and occupational therapy: review of research and practice as reﬂected in the IJOT,” Israeli Journal of Occupational Therapy, vol. 16, pp. H137–H158, 2007. [54] K. Sharﬁ and S. Rosenblum, “Learning disabilities among adults: an application of the ICF model as a key for the under- standing of individual needs; a position paper,” Mifgash: Jour- nal of Social-Educational Work, vol. 39, pp. 205–220, 2014, http://www.jstor.org/stable/24524238. [55] K. S. Buckles, “Maternal socio-economic status and the well- being of the next generation(s),” The Oxford Handbook of Women and the Economy, 2017, http://ftp.iza.org/dp10714.pdf. [56] S. M. Sheridan, L. L. Knoche, K. A. Kupzyk, C. P. Edwards, and C. A. Marvin, “A randomized trial examining the eﬀects of parent engagement on early language and literacy: the getting ready intervention,” Journal of School Psychology, vol. 49, no. 3, pp. 361–383, 2011. [57] C. F. Norbury and A. Sparks, “Diﬀerence or disorder? Cultural in understanding neurodevelopmental disorders,” issues Developmental Psychology, vol. 49, no. 1, pp. 45–58, 2013. [58] G. J. Geldhof and T. D. Little, “Inﬂuences of children’s and adolescents’ action-control processes on school achievement, peer relationships, and coping with challenging life events,” New Directions for Child and Adolescent Development, vol. 2011, no. 133, pp. 45–59, 2011. [59] C. Blair and A. Diamond, “Biological processes in prevention and intervention: the promotion of self-regulation as a means of preventing school failure,” Development and Psychopathol- ogy, vol. 20, no. 3, pp. 899–911, 2008. [60] E. M. Lewit and L. S. Baker, “School readiness,” The Future of Children, vol. 5, no. 2, pp. 128–139, 1995, https://www.jstor .org/stable/1602361. [61] R. A. Barkley, Executive Functions: What They Are, How They Work, and Why They Evolved, Guilford Press, New York, NY, 2012. [62] R. A. Barkley, “Diﬀerential diagnosis of adults with ADHD: the role of executive function and self-regulation,” Journal of Clin- ical Psychiatry, vol. 71, no. 7, article e17, 2010. [63] K. Shiels and L. W. Hawk Jr., “Self-regulation in ADHD: the role of error processing,” Clinical Psychology Review, vol. 30, no. 8, pp. 951–961, 2010. [64] M. Montague, “Self-regulation strategies to improve mathe- matical problem solving for students with learning disabil- ities,” Learning Disability Quarterly, vol. 31, no. 1, pp. 37–44, 2008. [65] O. Godefroy, “Frontal syndrome and disorders of executive functions,” Journal of Neurology, vol. 250, no. 1, pp. 1–6, 2003. [66] G. Árnadóttir, The Brain and Behavior: Assessing Cortical Dys- function through Activities of Daily Living (ADL), Mosby, St. Louis, MO, 1990. [67] S. S. Zentall, G. W. Harper, and M. Stormont-Spurgin, “Chil- dren with hyperactivity and their organizational abilities,” Journal of Educational Research, vol. 87, no. 2, pp. 112–117, 1993. [68] J. R. Flanagan, M. C. Bowman, and R. S. Johansson, “Control strategies in object manipulation tasks,” Current Opinion in Neurobiology, vol. 16, no. 6, pp. 650–659, 2006. [69] S. Rosenblum, T. Aloni, and N. Josman, “Relationships between handwriting performance and organizational abilities among children with and without dysgraphia: a preliminary study,” Research in Developmental Disabilities, vol. 31, no. 2, pp. 502–509, 2010. [70] P. M. Wehmeier, R. W. Dittmann, A. Schacht, K. Helsberg, and G. Lehmkuhl, “Morning and evening behavior in children and adolescents treated with atomoxetine once daily for atten- tion-deﬁcit/hyperactivity disorder (ADHD): ﬁndings from two 24-week, open-label studies,” Child and Adolescent Psychi- atry and Mental Health, vol. 3, no. 1, 2009. [71] F. P. Glascoe, “Developmental and behavioral screening,” in Handbook of Intellectual and Developmental Disabilities, J. W. Jacobson, J. A. Mulick, and J. Rojahn, Eds., pp. 353–372, Springer, New York, NY, 2007. [72] R. C. Tervo, “Parent’s reports predict their child’s develop- mental problems,” Clinical Pediatrics, vol. 44, no. 7, pp. 601– 611, 2005. [73] S. Rosenblum and E. Yom-Tov, “Seeking web-based informa- tion about attention deﬁcit hyperactivity disorder: where, what, and when,” Journal of Medical Internet Research, vol. 19, no. 4, p. e126, 2017. [74] L. L. Dyson, “The experiences of families of children with learning disabilities: parental stress, family functioning, and sibling self-concept,” Journal of Learning Disabilities, vol. 29, no. 3, pp. 280–286, 1996. [75] S. Rosenblum and B. Engel-Yeger, “Predicting participation in children with DCD,” Current Developmental Disorders Reports, vol. 1, no. 2, pp. 109–117, 2014. [76] L. B. Silver, The Misunderstood Child: Understanding and Cop- ing with Your Child’s Learning Disabilities, Harmony, New York, NY, 4th ed. edition, 2006. [77] A. Thornton, M. Licari, S. Reid, J. Armstrong, R. Fallows, and C. Elliott, “Cognitive orientation to (daily) occupational per- formance intervention leads to improvements in impairments, activity and participation in children with developmental coordination disorder,” Disability and Rehabilitation, vol. 38, no. 10, pp. 979–986, 2016. [78] N. Josman, A. Goﬀer, and S. Rosenblum, “Development and standardization of a “Do-Eat” activity of daily living perfor- mance test for children,” American Journal of Occupational Therapy, vol. 64, no. 1, pp. 47–58, 2010. [79] S. Rosenblum and M. Livneh-Zirinski, “Do relationships exist between brain-hand language and daily function characteris- tics of children with a hidden disability?,” in Recent Advances of Neural Network Models and Applications, S. Bassis, A. 12 Occupational Therapy International Esposito, and F. C. Morabito, Eds., pp. 269–281, Springer, Cham, Switzerland, 2014. [80] B. Engel-Yeger, A. Hanna-Kassis, and S. Rosenblum, “Can gym- nastic teacher predict leisure activity preference among chil- dren with developmental coordination disorders (DCD)?,” Research in Developmental Disabilities, vol. 33, no. 4, pp. 1006–1013, 2012. [81] J. W. Gilger and B. J. Kaplan, “Atypical brain development: a framework for understanding developmental disabilities,” Developmental Neuropsychology, conceptual learning vol. 20, no. 2, pp. 465–481, 2001. [82] A. M. Leon, S. Holliker, and J. Pepe, “The importance of the ﬁrst 5 years: pediatrician identiﬁcation of developmental delays and other related concerns,” Journal of Social Service Research, vol. 41, no. 4, pp. 425–444, 2015. [83] J. W. Gilger and B. J. Kaplan, “The concept of atypical brain development (ABD) as applied to developmental learning disorders,” in Adult Learning Disorders: Contempo- rary Issues, E. E. Wolf, H. E. Schrieber, and J. W. Wasser- stein, Eds., pp. 55–79, Psychology Press, New York, NY, 2008.
10.1186_s12879-022-07139-2	Ebinger et al. BMC Infectious Diseases (2022) 22:178 https://doi.org/10.1186/s12879-022-07139-2 RESEARCH ARTICLE Open Access Seasonal COVID-19 surge related hospital volumes and case fatality rates Joseph E. Ebinger1,2† Margo B. Minissian5, Bernice Coleman5, Richard Riggs6, Pamela Roberts6,7 and Susan Cheng1,2 , Roy Lan3†, Matthew Driver2, Nancy Sun2, Patrick Botting2, Eunice Park4, Tod Davis4, Abstract Background: Seasonal and regional surges in COVID-19 have imposed substantial strain on healthcare systems. Whereas sharp inclines in hospital volume were accompanied by overt increases in case fatality rates during the very early phases of the pandemic, the relative impact during later phases of the pandemic are less clear. We sought to characterize how the 2020 winter surge in COVID-19 volumes impacted case fatality in an adequately-resourced health system. Methods: We performed a retrospective cohort study of all adult diagnosed with COVID-19 in a large academic healthcare system between August 25, 2020 to May 8, 2021, using multivariable logistic regression to examine case fatality rates across 3 sequential time periods around the 2020 winter surge: pre-surge, surge, and post-surge. Sub- group analyses of patients admitted to the hospital and those receiving ICU-level care were also performed. Addition- ally, we used multivariable logistic regression to examine risk factors for mortality during the surge period. Results: We studied 7388 patients (aged 52.8 COVID-19 during the study period. Patients treated during surge (N period had 2.64 greater odds (95% CI 1.46–5.27) of mortality after adjusting for sociodemographic and clinical factors. Adjusted mortality risk returned to pre-surge levels during the post-surge period. Notably, first-encounter patient- level measures of illness severity appeared higher during surge compared to non-surge periods. 19.6 years, 48% male) who received outpatient or inpatient care for 6372) compared to the pre-surge (N 536) = ± = Conclusions: We observed excess mortality risk during a recent winter COVID-19 surge that was not explained by conventional risk factors or easily measurable variables, although recovered rapidly in the setting of targeted facility resources. These findings point to how complex interrelations of population- and patient-level pandemic factors can profoundly augment health system strain and drive dynamic, if short-lived, changes in outcomes. Keywords: COVID-19, Surge, Case fatality Background Adverse clinical outcomes, particularly case fatality, are known to increase during periods of strain on health- care systems caused by excess patient volume [1, 2]. The COVID-19 pandemic has led to especially profound Correspondence: joseph.ebinger@csmc.edu †Joseph E. Ebinger and Roy Lan contributed equally to this work 1 Department of Cardiology, Cedars-Sinai Medical Center, Los Angeles, CA, USA Full list of author information is available at the end of the article challenges, many related to the uniquely evolving fea- tures of SARS-CoV-2 infection, with numerous hospi- tals having experienced substantially greater COVID-19 case fatality during periods of regional surges. However, the vast majority of published reports on the relationship between hospital volume and excess mortality risk have been focused almost exclusively on data collected during the initial months of the pandemic—prior to the imple- mentation of more developed standards of care [3–6]. The earlier reports also tended to highlight data from © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Ebinger et al. BMC Infectious Diseases (2022) 22:178 Page 2 of 8 facilities with limited staff and operational resources; these factors are likely to have contributed to greater increases in mortality during periods of surge [7], lead- ing to potentially extreme estimates of excess mortality associated with rapid increases in patient volume for a given health system [8, 9]. Reports from the initial phase of the pandemic are also limited to the effects of the ear- lier SARS-CoV-2 variants, and more recently emerged variants are known to have differential impacts on clini- cal outcomes [10]. Amidst ongoing regional surges of COVID-19, related in part to recently emerged SARS-CoV-2 variants, more uptodate information is needed regarding how the pres- sures of COVID-19 surges on health systems can impact outcomes—especially during the winter season, when colder weather tends to increase both viral transmissibil- ity and patient-level susceptibility to more severe types of illness [11]. Most hospitals have adopted more advanced SARS-CoV-2 therapies, developed standards of care for more severely ill patients, and developed protocols for anticipating rapid increases in patient volume. However, the more transmissible SARS-CoV-2 variants and the overall epidemiologic persistence of COVID-19 across all communities have led to surges that continue to impose dynamic challenges for all health systems. Methods Study design and sampling To investigate the nature and correlates of COVID-19 associated outcomes before, during, and after the Winter 2020 surge, we performed a retrospective cohort study of all adult patients (age ≥ 18 years) treated for confirmed COVID-19 infection in our large multisite healthcare system based in Los Angeles, California (Cedars-Sinai Health System), from August 25, 2020 through May 8, 2021. Cedars-Sinai Medical Center is the largest non- profit hospital in the western United States, with a total of 886 hospital beds, 96 of which are in intensive care units (ICU). In addition, it has a catchment area of > 1.8 mil- lion people, with over a quarter million inpatient hospital days for admitted patients, > 90,000 emergency depart- ment visits, and nearly 800,000 outpatient appointments annually. All laboratory testing for COVID-19 were per- formed using reverse transcriptase polymerase chain reaction (PCR) of extracted RNA from nasopharyngeal swabs. Data collection We obtained demographic, clinical, and outcomes data from the Cedars-Sinai electronic health record (EHR) and manually confirmed key clinical and outcomes vari- ables. We defined race/ethnicity membership as follows: Asian, Hispanic/Latinx ethnicity (all races), non-Hispanic Black, non-Hispanic White, and other (including indi- viduals with multiple races listed). To estimate rela- tive comorbid status, the Elixhauser Comorbidity Index score was calculated with van Walraven weighting, using the International Classification of Diseases-10 (ICD-10) codes present at the time of COVID-19 presentation [12]. Specific clinical characteristics were identified for each patient using ICD-10 diagnoses at presentation, including: obesity, hypertension, diabetes mellitus, prior myocardial infarction (MI) or heart failure (HF), and prior chronic obstructive pulmonary disease (COPD) or asthma. Laboratory values from the time of admission were also obtained from the EHR. Exposures and outcomes Our primary exposure was the receipt of care for COVID-19 during three distinct time periods: pre-surge (August 25, 2020–November 7, 2020), surge (Novem- ber 8, 2020–February 22, 2021), and post-surge (Febru- ary 23, 2021–May 8, 2021). The start of the surge period was declared by hospital capacity management based on trends in internal and regional case volumes. The end of the surge period was calculated as the date at which the 7-day rolling average of newly diagnosed COVID- 19 cases dropped below the 7-day rolling average at the beginning of the surge. The pre-surge period and post- surge period were defined as the 75 days before and after the surge period, respectively, as a larger observation window would include cases from a prior surges. Our primary outcome was COVID-19 case fatality. For patients admitted to the hospital, we defined case fatal- ity as a death during hospitalization or up to 30 days from the time of discharge as documented in the EHR. For patients not requiring admission, case fatality was defined as death within 30 days from the date of initial COVID-19 diagnosis. Statistical analyses Demographic and clinical characteristics were summa- rized using mean and standard deviation (SD) for con- tinuous variables and as counts with percentages for all categorical variables. We compared demographic, clinical and laboratory characteristics across time peri- ods using analysis of variance (ANOVA) for continuous measures and Chi-squared tests for categorical measures. We conducted multivariable logistic regression to exam- ine the association between time period and case fatality, both overall and by subgroups of patients admitted to the hospital and those receiving ICU-level care. Additionally, we used multivariable logistic regression to examine risk factors for mortality during the surge period. All analy- ses were adjusted for age, sex, race/ethnicity, Elixhauser score, hypertension, and diabetes mellitus. A two-tailed Ebinger et al. BMC Infectious Diseases (2022) 22:178 Page 3 of 8 P-value of < 0.05 were considered significant. All analyses were conducted using R V.4.0.2 (R Foundation for Statis- tical Computing, Vienna, Austria). Results Cohort characteristics A total of 7,388 patients with COVID-19 were iden- tified during the study period with a mean age of 52.8 ± 19.6 years, 47.8% of whom were male. A total of 536 patients were diagnosed during the pre-surge period, 6372 during the surge, and 480 during the post- surge period (Fig. 1). Overall, patients in the surge period were on average older (53.3 ± 19.5) than those in the pre-surge (50.3 ± 19.8) and post-surge periods (48.5 ± 19.5; p < 0.001). Patients during the surge period also exhibited greater rates of obesity, hypertension and diabetes mellitus when compared to the pre-surge and post-surge periods. The average mean daily COVID-19 case count increased (p < 0.001) during the surge period (59.6 ± 41.6), compared to both the pre-surge period (7.1 ± 3.7) and the post-surge period (6.4 ± 3.1) (Table 1). Transfers from other acute care hospitals ranged between 2.5% and 4.9% of all COVID-19 cases across the 3 periods (Additional file 1: Table S1). Multivariable analysis During the study period there were 412 deaths (case fatality rate 5.6%), with 11 (2.1%) during the pre-surge period, 385 (6.0%) during the surge period, and 16 (3.3%) during the post-surge period. Following multi- variable adjustment for demographic and clinical char- acteristics, patients diagnosed with COVID-19 during the surge period experienced higher odds of death (OR: 2.64, 95% CI 1.46–5.27) compared to patients diagnosed in the pre-surge time period (Table 2). Odds of death were also higher during the surge period for Cases and Deaths per Week 150 100 y c n e u q e r F 50 0 0 30 60 90 120 150 180 210 240 Days since index date Cases Deaths Fig. 1 COVID-19 cases and deaths per week Ebinger et al. BMC Infectious Diseases (2022) 22:178 Page 4 of 8 Table 1 Demographic and clinical characteristics of COVID-19 patients, 8/25/2020 to 5/8/2021 Overall (n = 7388) Pre‑surge period (n 536) = Surge period (n = 6372) Post‑surge period (n 480) = p‑value 28.7 (37.5) 7.1 ( 3.7) 59.6 (41.6) 6.4 ( 3.1) < 0.001 52.76 (19.60) 3529 (47.8) 50.26 (19.78) 53.29 (19.53) 260 (48.5) 3054 (47.9) 48.48 (19.52) 215 (44.8) Average daily cases, mean (SD) Demographic characteristics Age, mean (SD), years Male sex, n (%) Race/ethnicity, n (%) Asian Hispanic/Latinx Non-Hispanic Black Non-Hispanic White Other Clinical characteristics Elixhauser comorbidity scorea, mean Obesity, n (%) ± SD Hypertension, n (%) Diabetes mellitus, n (%) Prior myocardial infarction or heart failure, n (%) 1230 (16.6) Prior COPD or asthma, n (%) 1326 (17.9) COPD, chronic obstructive pulmonary disease; SD, standard deviation a Elixhauser comorbidity score calculated using the van Walraven method 624 (8.4) 2423 (32.8) 1212 (16.4) 2416 (32.7) 368 (5.0) 31 (5.8) 158 (29.5) 97 (18.1) 193 (36.0) 21 (3.9) 564 (8.9) 2136 (33.5) 1034 (16.2) 2031 (31.9) 324 (5.1) 7.44 (11.84) 6.31 (10.24) 7.50 (11.95) 1697 (23.0) 3001 (40.6) 1804 (24.4) 109 (20.3) 176 (32.8) 105 (19.6) 70 (13.1) 78 (14.6) 1498 (23.5) 2648 (41.6) 1597 (25.1) 1073 (16.8) 1174 (18.4) < 0.001 0.389 < 0.001 0.060 0.019 < 0.001 0.004 0.052 0.026 29 (6.0) 129 (26.9) 81 (16.9) 192 (40.0) 23 (4.8) 7.82 (11.98) 90 (18.8) 177 (36.9) 102 (21.2) 87 (18.1) 74 (15.4) hospitalized patients (3.20, 1.76–6.43), and those admitted to the ICU, (2.81, 1.20–7.29) (Additional file 1: Table S2). No statistically significant differences in case fatality were observed between the pre-surge and post-surge periods in the overall, hospitalized, and ICU groups. In the fully adjusted model, during the surge period patients over age 65 (5.76, 4.29–7.81), males (1.55, 1.23–1.96), Hispanic/Latinx patients (1.64, 1.05–2.50), and Asian patients (1.64, 1.05–2.50) were more likely to experience death. Increasing comorbidity burden, as assessed by Elixhauser score, was also positively asso- ciated with risk of death (1.06, 1.05–1.07) (Table 3). First‑encounter measures of illness severity A total of 2537 patients were hospitalized during the study period. To assess the severity of illness at the time of initial clinical presentation (i.e. first encoun- ter), we examined the presenting vital signs and labora- tory values at the time of hospital admission. Clinically modest but statistically significant differences were appreciated among patients across time periods, includ- ing for the average mean C-Reactive Protein (in mg/L 110.4 ± 85.4 surge, 84.8 ± 79.1 pre-surge, 101.0 ± 100.3 post-surge; p = 0.019), serum Creatinine (in mg/dL 1.6 ± 2.4, ± 1.2 ± 1.4, ± 1.7 ± 2.6, p = 0.031), serum HCO3 (in mmol/L 22.8 ± 5.7, 25.2 ± 6.0, 22.9 ± 8.7; p = 0.046), mean systolic blood pressure (in mmHg 124.8 ± 19.7, 123.1 ± 18.9, ± 121.3 ± 20.6, p = 0.039), mean respira- tory rate (20.2 ± 4.1, 19.3 ± 3.7, 18.7 ± 3.4; p < 0.001), SPO2 (95.3 ± 3.4, 96.0 ± 2.6, 96.2 ± 3.9; p < 0.001), and (degrees Fahrenheit 99.6 ± 1.4, mean (Additional file 1: 99.8 ± 1.5, ± 99.2 ± 1.4; p < 0.001) Table S3). temperature Table 2 Odds of death, by time period, among patients with COVID-19 Time period Pre-surge (8/25/2020–11/7/2020) Surge (11/8/2020–2/22/2021) Post-surge (2/23/2021–5/8/2021) Unadjusted OR (95% CI) Adjusted OR (95% CI)a Ref. 3.07 (1.76, 5.98) 1.65 (0.76, 3.68) Ref. 2.64 (1.46, 5.27) 1.63 (0.72, 3.81) Bold value indicate odds ratios who’s 95% CI does not cross unity, indicating statistical significance CI, Confidence Interval; OR, Odds Ratio a Model adjusted for age, sex, race/ethnicity, Elixhauser score, hypertension, diabetes, obesity, chronic obstructive pulmonary disease or asthma, and prior myocardial infarction or heart failure Ebinger et al. BMC Infectious Diseases (2022) 22:178 Page 5 of 8 Table 3 Risk factors for death during surge among patients with COVID-19 Unadjusted OR (95% CI) Adjusted OR (95% CI)a Age Below 65 Above 65 Sex Female Male Race/ethnicity Non-Hispanic white Non-Hispanic Black Hispanic/Latinx Asian Other Elixhauser comorbidity score Diabetes No Yes Hypertension No Yes Ref. 10.91 (8.47, 14.22) Ref. 1.84 (1.49, 2.28) Ref. 0.57 (0.41, 0.79) 0.64 (0.50, 0.82) 0.74 (0.50, 1.07) 0.76 (0.46, 1.20) 1.08 (1.08, 1.09) Ref. 3.55 (2.88, 4.38) Ref. 4.25 (3.38, 5.38) Ref. 5.76 (4.29, 7.81) Ref. 1.55 (1.23, 1.96) Ref. 0.72 (0.50, 1.02) 1.53 (1.15, 2.05) 1.64 (1.05, 2.50) 1.17 (0.67, 1.96) 1.06 (1.05, 1.07) Ref. 1.19 (0.93, 1.53) Ref. 0.96 (0.72, 1.29) Bold values indicate odds ratios who’s 95% CI does not cross unity, indicating statistical significance CI, Confidence Interval; OR, Odds Ratio a Model adjusted for age, sex, race/ethnicity, Elixhauser score, hypertension, diabetes, obesity, chronic obstructive pulmonary disease or asthma, and prior myocardial infarction or heart failure Discussion In this study of over 7000 patients who received outpa- tient or inpatient care for COVID-19 between August 2020 and May 2021, adjusted mortality risk increased significantly from the Fall pre-surge period to the Win- ter surge period—corresponding with the very rapid rise in patient volume. Adjusted mortality risk then returned to pre-surge levels during the Spring post-surge period— corresponding to a subsequent similarly rapid decline in patient volume. The excess mortality risk during the winter COVID-19 surge was not adequately explained by conventional sociodemographic or pre-existing risk traits or easily measurable variables. However, we did observe that the excess risk recovered rapidly in the setting of tar- geted facility resources. Notably, however, first-encounter patient-level measures of illness severity appeared higher during surge compared to non-surge periods—sug- gesting that timing of patient presentation, as related to timing illness onset, may have contributed along with external socioeconomic or other epidemiological fac- tors to augmenting risk for adverse outcomes during the surge period. Our findings extend from numerous earlier scientific and lay reports that have chronicled the overwhelm- ing nature of the initial COVID-19 surge that began the United States in March of 2020 [13–16]. This first wave was compounded by multiple factors including lack of knowledge around appropriate treatment of SARS-CoV-2 infection, unprepared resource supply chains, and a lack of adequately trained personnel in highly impacted com- munities. Advances in standards of care including the use of monoclonal antibodies, steroids and Remdesivir [17], among others, as well as more robust supply chains [18, 19] were present during the Winter surge period evalu- ated in the current study—allowing for a more focused evaluation of the excess patient volume effect on COVID- 19 case outcomes at a health system level. Further, the presence of ‘valleys’ in patient volume during non-surge time periods allowed for comparison of surge case fatal- ity rates to those when excess patient volume was not a predominant factor. Expanding longitudinally from the earlier reports, our analysis from the Winter 2020 surge found that patients treated for COVID-19 during the surge period had higher odds of death, both overall and when stratified by maximum level of care required (outpatient, inpa- tient, and ICU). Importantly, we observed that patients presenting for care during the surge were more likely to be older, male, Hispanic/Latinx, and with a greater bur- den of comorbidities than those presenting during the Ebinger et al. BMC Infectious Diseases (2022) 22:178 Page 6 of 8 non-surge periods; all these factors have been linked to greater severity of COVID-19 illness [4, 20–22]. None- theless, odds of death remained elevated during the surge period even when adjusting for these risk factors. We also observed apparently modest but statistically significant differences in laboratory characteristics among patients hospitalized during the surge. It is well described that during periods of high COVID-19 activity in the commu- nity, patients delay seeking care due to fear of becoming ill or spreading the virus themselves [23–27]. As such, delayed presentations, particularly among vulnerable patient populations, with subsequent late initiation of COVID-19 specific therapies, may well have contributed to at least a portion of the observed excess mortality risk. Previous studies have also found that discernible non- patient factors contribute to measurable variation in COVID-19 outcomes. Increased COVID-19 case rates [3, 28, 29], ICU strain [5, 12], and limited hospital resource availability, including number of hospital beds and staff [7], have been linked to increased case fatality, though these studies examine outcomes solely during the ini- tial stages of the pandemic. Nonetheless, these phe- nomena are known to continue to impact COVID-19 outcomes across in at-risk regions and communities. For- tunately, although surges in COVID-19 patient volume required the transformation of previously non-critical care environments into advanced care locations within our health system, we were ultimately able to house and medically accommodate all patients requiring advanced care including intubation, mechanical ventilation, and mechanical circulatory support. These advanced care needs were met through redirecting staffing support from non-critical care to critical care settings. Beyond facility-level factors, lack of statistically significant dif- ferences in case fatality between pre- and post-surge periods suggest that the increase in case fatality was not related to secular trends in COVID-19 outcomes during the study period, such as improvements in the standard of care and the circulation of regional COVID-19 vari- ants linked to increased mortality [9, 10]. In fact, while not statistically significant, we observed a trend towards slightly higher mortality during the post-surge compared to pre-surge period. This finding could have been related to a bias towards more severely ill patients presenting for medical encounters over time, or a residual excess in hospitalized patients; further studies using more detailed data are needed to clarify the factors contributing to vari- ations in post-surge recovery periods. Several limitations of this study merit consideration. Our data were derived from a single healthcare system, and thus our findings may not be generalizable to other populations, especially those outside the United States. However, our patient cohort was found to be diverse, both demographically as well as clinically, and our insti- tution is a high-volume center serving a large and diverse urban population. Reliance on EHR data to identify deaths may result in misclassification, particularly by undercounting deaths occurring outside of the hospital, though we would expect this to attenuate rather than confound our results. We were unable to systematically capture data on timing of illness onset, which precluding assessment of symptom duration prior to presentation to medical care. We recognize that all the factors driving as well as correlated with delayed patient presentations (i.e. delays in patients seeking or receiving medical attention), especially during COVID-19 surge periods, are critically important to identify and yet not easily measured in the real-world community setting. Detailed data on temporal trends in hospital occupancy, medical care staffing (e.g. nurse-to-patient ratios), and medical care supplies and other resources were not available for the current analysis and will be important for future investigations of excess mortality during surge periods. Finally, we were unable to control for COVID-19 vaccination status as vaccines were not available for the majority of the cohort until the post-surge period, and vaccine uptake in the post-surge period may have lowered risk for severe outcomes. How- ever, given that all patients in our cohort were COVID- positive and that reported breakthrough infection rates are relatively low [30], it is unlikely that enough patients in the post-surge period were vaccinated to have biased our results. Conclusions In summary, our study highlights the reality of excess mortality risk seen during the last Winter surge of COVID-19 experienced by a high-volume healthcare sys- tem serving a diverse and large metropolitan region. The excess mortality risk was not explained by conventional risk factors or easily measurable variables, although recovered rapidly in the setting of targeted facility resources. These findings point to how complex interre- lations of population- and patient-level pandemic factors can profoundly augment health system strain and drive dynamic, if short-lived, changes in outcomes. Abbreviations ANOVA: Analysis of variance; COPD: Chronic obstructive pulmonary disease; EHR: Electronic health record; ICD-10: International Classification of Dis- eases-10; ICU: Intensive care units; HF: Heart failure; MI: Myocardial infarction; PCR: Polymerase chain reaction; SD: Standard deviation. Ebinger et al. BMC Infectious Diseases (2022) 22:178 Page 7 of 8 Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12879- 022- 07139-2. USA. 7 Department of Biomedical Sciences, Division of Informatics, Cedars-Sinai Medical Center, Los Angeles, CA, USA. Received: 21 December 2021 Accepted: 9 February 2022 Additional file 1: Table S1. COVID-19 patients transferred from other acute care facilities and associated mortality, by time period. Table S2. Odds of death, by time period, among patients with COVID-19 among those admitted to the hospital and to those admitted to the ICU. Table S3. Diagnostic values at admission among patients hospitalized for COVID-19, by time period. Acknowledgements We are grateful to all the front-line healthcare workers in our healthcare sys- tem who continue to be dedicated to delivering the highest quality care for all patients, as well as the invaluable contributions of the CORALE and EMBARC study investigators and staff. Authors’ contributions JEE took part in conceptualization, methodology, validation, formal analysis, investigation, resources, data curation, writing—original draft, writing—review & editing, visualization, and funding acquisition. RL took part in conceptual- ization, formal analysis, data curation, writing—original draft, and writing— review & editing. MD took part in methodology, validation, formal analysis, writing—review & editing, and visualization. NS took part in Validation, formal analysis, and writing—review & editing. PB took part in methodology, validation, and writing—review & editing. EP took part in validation, formal analysis, and writing—review & editing. TD, MBM, RR, and PR took part in writing—review & editing, and supervision. BC took part in conceptualiza- tion, methodology, writing—review & editing, and supervision. SC took part in conceptualization, methodology, validation, formal analysis, investigation, resources, data curation, writing—original draft, writing—review & editing, visualization, supervision, project administration, and funding acquisition. All authors read and approved the manuscript. Funding This work was supported in part by Cedars-Sinai Medical Center, the Erika J Glazer Family Foundation, and NIH grants U54-CA260591 and K23-HL153888. Funding sources had no role in the design of the study and collection, analy- sis, and interpretation of data and in writing the manuscript. Availability of data and materials Due to their sensitive nature, restrictions apply to the availability of the data that support the findings, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of Cedars-Sinai Medical Center. Declarations Ethics approval and consent to participate This study and access to the data were approved by the Cedars-Sinai Institutional Review Board (CORALE_EHR: Study 00000603), with a waiver for informed consent. Data was anonymized before use. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1 Department of Cardiology, Cedars-Sinai Medical Center, Los Angeles, CA, USA. 2 Smidt Heart Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA. 3 College of Medicine, University of Tennessee Health Science Center, Memphis, TN, USA. 4 Enterprise Data Intelligence, Cedars-Sinai Medical Center, Los Angeles, California, USA. 5 Brawerman Nursing Institute and Nursing Research Department, Cedars-Sinai Medical Center, Los Angeles, CA, USA. 6 Department of Medical Affairs, Cedars-Sinai Medical Center, Los Angeles, CA, References 1. Anesi GL, Liu VX, Gabler NB, Delgado MK, Kohn R, Weissman GE, Bayes B, Escobar GJ, Halpern SD. Associations of intensive care unit capacity strain with disposition and outcomes of patients with sepsis presenting to the emergency department. Ann Am Thorac Soc. 2018;15(11):1328–35. Eriksson CO, Stoner RC, Eden KB, Newgard CD, Guise JM. The Association between hospital capacity strain and inpatient outcomes in highly devel- oped countries: a systematic review. J Gen Intern Med. 2017;32(6):686–96. 2. 3. Block BL, Martin TM, Boscardin WJ, Covinsky KE, Mourad M, Hu LL, Smith AK. Variation in COVID-19 mortality across 117 US hospitals in high- and low-burden settings. J Hosp Med. 2021;16(4):215–8. 4. Gupta S, Hayek SS, Wang W, Chan L, Mathews KS, Melamed ML, Brenner SK, Leonberg-Yoo A, Schenck EJ, Radbel J, et al. Factors associated with death in critically ill patients with coronavirus disease 2019 in the US. JAMA Intern Med. 2020;180(11):1436–47. 5. Karaca-Mandic P, Sen S, Georgiou A, Zhu Y, Basu A: Association of COVID- 19-Related Hospital Use and Overall COVID-19 Mortality in the USA. J Gen Intern Med. 2020; 1–3. Statistics and research: mortality risk of COVID-19, Coronavirus Pandemic (COVID-19). https:// ourwo rldin data. org/ morta lity- risk- covid. Janke AT, Mei H, Rothenberg C, Becher RD, Lin Z, Venkatesh AK. Analysis of hospital resource availability and COVID-19 mortality across the United States. J Hosp Med. 2021;16(4):211–4. Ledford H. Why do COVID death rates seem to be falling? Nature. 2020;587(7833):190–2. 6. 7. 8. 9. NIH COVID-19 treatment guidelines: guidelines archives. https:// www. covid 19tre atmen tguid elines. nih. gov/ about- the- guide lines/ guide lines- archi ve/. 10. SeyedAlinaghi S, Mirzapour P, Dadras O, Pashaei Z, Karimi A, Mohs- seniPour M, Soleymanzadeh M, Barzegary A, Afsahi AM, Vahedi F, et al. Characterization of SARS-CoV-2 different variants and related morbidity and mortality: a systematic review. Eur J Med Res. 2021;26(1):51. 11. Moriyama M, Hugentobler WJ, Iwasaki A. Seasonality of respiratory viral infections. Ann Rev Virol. 2020;7(1):83–101. 12. van Walraven C, Austin PC, Jennings A, Quan H, Forster AJ. A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care. 2009;47(6):626–33. 13. Kleinpell R, Ferraro DM, Maves RC, Kane Gill SL, Branson R, Greenberg S, Doersam JK, Raman R, Kaplan LJ. Coronavirus disease 2019 pandemic measures: reports from a national survey of 9,120 ICU clinicians. Crit Care Med. 2020;48(10):e846–55. 14. Chokshi DA, Katz MH. Emerging lessons from COVID-19 response in New York City. JAMA. 2020;323(20):1996–7. 15. Keeley C, Jimenez J, Jackson H, Boudourakis L, Salway RJ, Cineas N, Vil- lanueva Y, Bell D, Wallach AB, Boyle Schwartz D, et al. Staffing up for the surge: expanding the New York City Public Hospital workforce during the COVID-19 pandemic. Health Aff (Millwood). 2020;39(8):1426–30. 16. Argenziano MG, Bruce SL, Slater CL, Tiao JR, Baldwin MR, Barr RG, Chang BP, Chau KH, Choi JJ, Gavin N, et al. Characterization and clinical course of 1000 patients with coronavirus disease 2019 in New York: retrospective case series. Bmj. 1996;2020:m369. 17. Beigel JH, Tomashek KM, Dodd LE, Mehta AK, Zingman BS, Kalil AC, Hohmann E, Chu HY, Luetkemeyer A, Kline S, et al. Remdesivir for the treatment of Covid-19—final report. N Engl J Med. 2020;383(19):1813–26. 18. Rebmann T, Alvino RT, Holdsworth JE. Availability and crisis standards of care for personal protective equipment during fall 2020 of the COVID-19 pandemic: a national study by the APIC COVID-19 task force. Am J Infect Control. 2021;49(6):657–62. 19. Rebmann T, Vassallo A, Holdsworth JE. Availability of personal protective equipment and infection prevention supplies during the first month of the COVID-19 pandemic: a national study by the APIC COVID-19 task force. Am J Infect Control. 2021;49(4):434–7. Ebinger et al. BMC Infectious Diseases (2022) 22:178 Page 8 of 8 20. Rosenthal N, Cao Z, Gundrum J, Sianis J, Safo S. Risk factors associated with in-hospital mortality in a US national sample of patients with COVID- 19. JAMA Netw Open. 2020;3(12):e2029058. 21. Cardemil CV, Dahl R, Prill MM, Cates J, Brown S, Perea A, Marconi V, Bell L, Rodriguez-Barradas MC, Rivera-Dominguez G, et al. COVID-19-related hospitalization rates and severe outcomes among veterans from 5 veter- ans affairs medical centers: hospital-based surveillance study. JMIR Public Health Surveill. 2021;7(1):e24502. 22. Bassett MT, Chen JT, Krieger N. Variation in racial/ethnic disparities in COVID-19 mortality by age in the United States: a cross-sectional study. PLoS Med. 2020;17(10):e1003402. 23. Roth GA, Emmons-Bell S, Alger HM, Bradley SM, Das SR, de Lemos JA, Gakidou E, Elkind MSV, Hay S, Hall JL, et al. Trends in patient characteris- tics and COVID-19 in-hospital mortality in the United States during the COVID-19 pandemic. JAMA Netw Open. 2021;4(5):e218828–e218828. 24. Vahidy FS, Drews AL, Masud FN, Schwartz RL, Askary BB, Boom ML, Phillips RA. Characteristics and outcomes of COVID-19 patients during initial peak and resurgence in the houston metropolitan area. JAMA. 2020;324(10):998–1000. 25. Bansal S, Roy M, Chatterjee T, Roy AK. Deaths due to delayed presenta- tion to the hospital from fear of contracting COVID-19 during lockdown period: a tertiary care center experience. J Commun Hosp Intern Med Perspect. 2021;11(3):299–301. 26. Sürme Y, Özmen N, Ertürk Arik B. Fear of COVID-19 and related factors in emergency department patients. Int J Mental Health Addiction. 2021. https:// doi. org/ 10. 1007/ s11469- 021- 00575-2. 27. Baugh JJ, White BA, McEvoy D, Yun BJ, Brown DFM, Raja AS, Dutta S. The cases not seen: patterns of emergency department visits and procedures in the era of COVID-19. Am J Emerg Med. 2021;46:476–81. 28. Asch DA, Sheils NE, Islam MN, Chen Y, Werner RM, Buresh J, Doshi JA. Variation in US hospital mortality rates for patients admitted with COVID-19 during the first 6 months of the pandemic. JAMA Intern Med. 2021;181(4):471–8. 29. Association Between Caseload Surge and COVID-19 Survival in 558 U.S. Hospitals, March to August 2020. Ann Intern Med. 2021; 174(9):1240–1251. 30. Washington State Department of Health. SARS-CoV-2 Vaccine Break- through Surveillance and Case Information Resource. https:// www. doh. wa. gov/ Porta ls/1/ Docum ents/ 1600/ coron avirus/ data- tables/ 420- 339- Vacci neBre akthr oughR eport. pdf. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12874-023-01902-y	Klatte et al. BMC Medical Research Methodology (2023) 23:84 https://doi.org/10.1186/s12874-023-01902-y BMC Medical Research Methodology Development of a risk-tailored approach and dashboard for efficient management and monitoring of investigator-initiated trials Katharina Klatte1, Suvitha Subramaniam1, Pascal Benkert1, Alexandra Schulz1, Klaus Ehrlich1, Astrid Rösler1, Mieke Deschodt2,3, Thomas Fabbro1, Christiane Pauli-Magnus1† and Matthias Briel1,4† Abstract Background Most randomized controlled trials (RCTs) in the academic setting have limited resources for clinical trial management and monitoring. Inefficient conduct of trials was identified as an important source of waste even in well-designed studies. Thoroughly identifying trial-specific risks to enable focussing of monitoring and management efforts on these critical areas during trial conduct may allow for the timely initiation of corrective action and to improve the efficiency of trial conduct. We developed a risk-tailored approach with an initial risk assessment of an individual trial that informs the compilation of monitoring and management procedures in a trial dashboard. Methods We performed a literature review to identify risk indicators and trial monitoring approaches followed by a contextual analysis involving local, national and international stakeholders. Based on this work we developed a risk-tailored management approach with integrated monitoring for RCTs and including a visualizing trial dashboard. We piloted the approach and refined it in an iterative process based on feedback from stakeholders and performed formal user testing with investigators and staff of two clinical trials. Results The developed risk assessment comprises four domains (patient safety and rights, overall trial management, intervention management, trial data). An accompanying manual provides rationales and detailed instructions for the risk assessment. We programmed two trial dashboards tailored to one medical and one surgical RCT to manage identified trial risks based on daily exports of accumulating trial data. We made the code for a generic dashboard available on GitHub that can be adapted to individual trials. Conclusions The presented trial management approach with integrated monitoring enables user-friendly, continuous checking of critical elements of trial conduct to support trial teams in the academic setting. Further work is needed in order to show effectiveness of the dashboard in terms of safe trial conduct and successful completion of clinical trials. Keywords Clinical trial, Trial management, Risk-tailored monitoring, Trial dashboard †Shared senior authorship. Correspondence: Katharina Klatte Katiklatte@icloud.com 1Department of Clinical Research, University Hospital Basel and University of Basel, Spitalstrasse 12, Basel CH- 4031, Switzerland 2Department of Public Health & Primary Care, KU Leuven, Leuven, Belgium 3Competence Centre of Nursing, University Hospitals Leuven, Leuven, Belgium 4Department of Health Research Methods, Evidence, and Impact, McMaster University, Hamilton, ON, Canada © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessPage 2 of 11 Introduction Randomized controlled trials (RCTs) are the gold stan- dard for assessing the effects of medical interventions. However, they are typically resource intense and pose various organisational challenges [1–3]. Inefficient man- agement and monitoring of RCTs have been identified as an important source of waste [1–5]. Monitoring efforts are traditionally quite generic and extensive, [6–8] but problems such as slow participant recruitment, con- siderable losses to follow-up, or poor data quality are often recognized too late during trial conduct delaying necessary adjustments of processes or the protocol. In addition, resources for clinical trial monitoring and man- agement are usually scarce in the academic setting and sophisticated commercial solutions can be costly [9, 10]. Organisational challenges and critical factors jeopar- dizing trial integrity and quality may vary considerably across trials; therefore, a risk assessment conducted prior to trial initiation or at certain intervals during trial con- duct may yield different risk profiles for individual trials. Trial monitoring protects the safety and rights of partici- pants, ensures data are accurate, complete and verifiable, and that the trial follows the principles of good clinical practice [11, 12]. Currently recommended risk-based trial monitoring allows for an adaptation of the monitor- ing intensity according to an initial risk assessment of a trial and has been developed to reduce resource intense onsite visits with source data verification for non-high- risk trials [1–3, 13–15, 16, 17. However, this approach typically does not consider individual risk profiles of RCTs, but rather classifies trials by generic risk catego- ries [16]. To accommodate individual trial risks, a moni- toring strategy may include several components such as centralized monitoring (evaluation of accumulated trial data performed in a timely manner at a central location), onsite monitoring (performed at investigator sites with source data verification and review of protocol-specified processes), or remote monitoring (same tasks as onsite monitoring but performed away from investigator sites) [17, 18, 19]. Trial management should provide for smooth and reli- able trial procedures including participant recruitment, randomisation, intervention application, data collection, and data cleaning [20, 21]. Data cleaning and checking of recruitment and retention rates, for instance, need to be performed in a timely fashion, so that corrective mea- sures can be taken early on and detrimental effects on the trial can be avoided [22]. Trial monitoring is most effec- tive when performed on cleaned data, because incorrect processes may be missed due to poor data quality and monitoring efforts are wasted on individual data errors. Therefore, trial management and monitoring ideally are integrated tasks that make use of accumulating data dur- ing trial conduct, i.e. continuously keeping oversight of complex study processes and performing centralized data monitoring [23–25]. The objective of this project was to develop a risk- tailored approach that integrated trial management and monitoring in investigator-initiated RCTs. We closely collaborated with relevant stakeholders (trial coordina- tors, principal investigators, data managers, trial moni- tors, statisticians) to create a user-friendly dashboard that efficiently visualizes data on critical processes of individual trials. Methods Overview of research process In the first phase of this user-centred project, [26] we developed a concept of a risk-tailored trial monitor- ing and management approach with corresponding trial dashboard (Fig. 1). We anticipated users to be primarily trial managers, principal investigators, and trial moni- tors. The development involved relevant stakeholder groups and was based on the results of systematic litera- ture reviews on existing monitoring strategies, [17] and a contextual analysis to identify current practices and needs of anticipated users. The concept and dashboard were piloted and refined in an iterative process involving different end users and other stakeholder groups. In the second phase, we performed formal user testing of the developed risk assessment and dashboard. Experiences of investigators and trial staff of one medical and one surgi- cal investigator-initiated RCT were gathered using semi- structured interviews to further refine the concept and dashboard. Setting Before the introduction of the new concept, a risk assess- ment was routinely performed by the monitoring team to assess the extent of the monitoring needed for the trial according to the ADAMON criteria. This approach allowed the rough classification of trials into the catego- ries low, medium, or high risk [27]. The new risk assess- ment incorporates many more factors related to the study specific conduct including challenges in the study management. It is not meant to categorize trials and adjust the extent of monitoring based on the category. The trial teams included in our project were not involved in other pre-trial risk assessments. Both trial teams assessing the benefits of the risk assessment and dash- board tool had started participant recruitment and data collection before the implementation of the new tool and, thus, compared it to the situation without structured risk assessment and tool support.” Systematic literature review To identify and structure components for the initial risk assessment of individual trials, we systematically Klatte et al. BMC Medical Research Methodology (2023) 23:84 Page 3 of 11 Fig. 1 Overview of the two phases of the development and user-testing of the risk-tailored approach and trial dashboard searched for published risk assessment approaches and risk indicators used to support trial oversight and to identify centres in need for support. We considered dif- ferent components and qualitative evidence from process evaluations of tested monitoring strategies summarized in a previously conducted systematic review [17]. We further considered the guideline of the European Clini- cal Research Infrastructure Network (Ecrin) [16] and the risk assessment guideline developed by the Swiss Clinical Trial Organization [28], TransCelerate metrics [29, 30], Whitham metrics [31], and the trial specific metrics used by the Medical Research Council (MRC) Clinical Trials Unit (CTU) at University College London (UCL) Trial specific metrics [32]. Results from this literature review are summarized in Supplementary Table 1. groups provided an additional opportunity for feedback and exchange of information on the risk assessment and dashboard development as well as on the application strategy. In order to get input from a national group of stakeholders in Switzerland, we contacted the national platform of the Swiss Clinical Trial Organisation for trial monitoring. Finally, we gathered experiences from inter- national methodological research groups and UK-based CTUs using risk-based approaches or study dashboards to support trial conduct. The different activities with stakeholders at all levels are summarized in Supplemen- tary Table 2. We extracted information from protocols of meetings and interviews and summarized the output in Supplementary Table 3. Contextual analysis Stakeholder involvement We set up a local, multidisciplinary working group including end users and representatives of different stakeholder groups within the Department of Clinical Research (DKF) and associated research groups at the University Hospital Basel. At this local level, we involved members from the Data Science and Data Management Teams of the DKF experienced in central monitoring, R shiny applications, dashboard development, database structures and exports; we involved trial monitors with experience in on-site and remote monitoring, knowl- edge of study site structures and processes; study coor- dinators and investigators experienced in managing RCTs. Stakeholder meetings with all members of these Gathering contextual input from various end users and the above-mentioned stakeholders guided the devel- opment of the risk-tailored approach and helped to determine relevant domains and applications to be con- sidered in the initial risk assessment. We structured the identified stakeholder needs into content related fac- tors such as the inclusion of the follow-up visits into the risk assessment, and design related factors such as the suggested separation of severity and likelihood in the assessment or the colour code for the status of queries visualized in the dashboard (Supplementary Table 3). In terms of content of the risk assessment, it became clear, for instance, that the assessment covers a wide spectrum of risks applicable to a large variety of RCTs. The design Klatte et al. BMC Medical Research Methodology (2023) 23:84 of the risk assessment guide should support the intuitive assessment by different end user groups (monitors, study managers, principal investigators). The study dashboard should reflect the outcome of the risk assessment and the design of the dashboard should enable an efficient navi- gation within the routine study procedure by end-users. The findings of the contextual analysis are summarized in Supplementary Table 3. Development and piloting of the concept and dashboard Based on the systematically reviewed literature, our contextual analysis and stakeholder input, we drafted a generic risk-assessment template. We then created trial- specific dashboards for a medical and a surgical mul- ticentre trial that differed in their risk profile, but both comprised complex study procedures and data collec- tion. The risk-tailored approach continued to evolve as we gathered contextual information, detected gaps in the assessment procedure, and identified critical components of study management. We developed R code to extract data values from exported data tables of the trial database secuTrial and summarized, compared, and calculated rel- evant information to create pathways for the identified risks. The output of these operations was then visual- ized in the trial dashboard. The piloting and refinement was an iterative process incorporating repeated feedback from the end-users and the stakeholder representatives in the project group on dashboard content, structure, user-friendly interface, and visualization of critical study data. User testing The aim of the user testing was to identify challenges in the routine use of the dashboard experienced by different user groups. Each of the six users (i.e. 2 trial managers, 2 monitors, 2 principal investigators) received a detailed manual of the features and operation mode of the study dashboard. Table 1 Domains and their attributed risk elements Domain Participant Safety and Rights Overall Study Management Device/ Medication Management Study Data Risk Elements Informed consent AE/SAE reporting and documentation Inclusion/exclusion Recruitment Retention Study procedures and endpoint assessment (e.g. bio sampling, imaging quality) Participant schedule (e.g. timeframe of visits) AE/SAE management Administration Accountability/ storage Data quality – completeness, consistency, timeliness Documentation/ storage Abbreviations: AE, adverse event; SAE, serious adverse event Page 4 of 11 We interviewed users 6–12 weeks after using the study dashboard in daily trial routine. We followed a semi- structured interview guide, which allowed for expan- sion on topics that emerged during the interview. All interviews took approximately 30 min. The interviewer (KK) transcribed the recorded interviews and extracted suggestions for improvement. We then updated the trial dashboard based on the feedback of the users and pro- vided the adapted version for further use and evaluation. Results The final concept consisted of the following three steps: trial-specific risk assessment prior to study start, selec- tion and development of data-based pathways to address identified risks, and visualization of pathways output in a trial dashboard. Trial-specific risk assessment The four trial-specific risk assessment comprised domains (participant safety and rights, overall study management, device/medication management, study data), and each domain contained several risk elements (Table 1). To better assess if these elements are critical for a specific trial and which trial components are at par- ticular risk, we determined trial assets and corresponding risk scenarios. Trial assets are conditions essential for the successful and proper conduct of a trial, e.g. visits must be scheduled and take place in the required timeframe, Serious Adverse Events (SAEs) have to be reported on time and need to be closely followed over the whole study conduct. If a trial includes many follow-up visits over a long follow-up time and assessments have to take place in a very narrow time window, this asset would be con- sidered at risk (example shown in Table 2, Part A). Other assets, for example SAE reporting and oversight, are essential for all clinical trials and, thus, are considered as a risk that applies to all trials (marked in red, Example shown in Table 2, Part B). The identified risks are then analysed in terms of severity and likelihood. For exam- ple, if many follow-up visits need to be coordinated but the time window of the endpoint assessment is wide the severity is rated as less critical. The likelihood is highly influenced by the experience of the trial team and partici- pating centres with similar trials, training and experience of all involved staff members, and the resources available for the study. The complete list of assets, as well as the corresponding risk scenarios, is provided in the full risk assessment in Supplementary Table 4. We suggest that the risk assess- ment is done by an experienced trial manager (e.g. from a trials support unit) supported by a trial monitor, a clini- cal expert, and the principal investigator. The first risk assessment should be performed before the start of the trial based on the study protocol, Case Report Forms Klatte et al. BMC Medical Research Methodology (2023) 23:84 Table 2 Example of assets and risk scenarios for risk elements in the domain Overall Study Management (Part A) and Participant Safety and Rights (Part B). Assets that apply to all trials are marked in red A) Domain Overall Study Management Risk element Participant Schedule Asset Visits/Phone calls must be within the given Timeframe Risk scenario (A) Time point of visit is critical for the endpoint assess- ment of the study (B) Large number of visits are difficult to organize and coordinate between centres and patients B) Domain Participant Safety and Rights Risk element SAE/AE Asset SAE have to be re- ported and documented correctly in the required timeframe Risk scenario Complexity of CRF or missing SOPs for SAE Reporting leads to (A) Incorrect docu- mentation and (B) Delayed report- ing of SAEs Abbreviations: CRF, case report form; SOPs, standard operating procedures; SAE, serious adverse events (CRFs), the planned and actual budget of the study, expected recruitment rates for all participating centres, information on the trial intervention, and information about planned study staff (see Appendix for detailed Manual). Pathways to manage identified risks In order to continuously manage identified risks, we cre- ated pathways that eventually allowed for tailored visu- alization of accumulating trial data and implemented action at suitable time intervals (e.g., email reminders, staff overviews) in a study dashboard. The operations applied to the exported data tables via R code are depen- dent on the specific information needed to provide a clear oversight on identified risk elements. The code is structured into modules that contain the operations of all pathways visualized in one dashboard tab (e.g. SAE management). For example, the module SAE contains operations that count the number of SAEs, determine the number of patients with SAE and calculate the ratio SAEs per patient randomized. In addition, information like severity, causality and outcome are extracted from the SAE form data table and percentages of value options (e.g. SAE outcome: Continuing, Resolved without sequel, Resolved with sequel, others) are calculated and graphi- cally displayed (Fig. 2, Panel A and B). The developed study dashboards contain tabs that visualize the output of created pathways reflecting identified study-specific risks. These tabs are based on the R modules contain- ing the pathways as well as the code required for a clear Page 5 of 11 visual presentation (value boxes, graphs, lists). When pilot testing our risk assessment guide, it became appar- ent that some risks apply to almost all trials (marked in red in the full risk assessment Supplementary Table 4). The management of these risks is, thus, based on tabs classified as “generic” in the study dashboard, while other, more seldom and study-specific risks are considered in “optional” tabs (Table 3). The content of generic tabs can also be adapted depending on, for instance, the complex- ity or time point of outcome assessment in a trial. The generic dashboard template is freely available on GitHub (https://github.com/CTU-Basel/viewTrial). Visualization of data based pathways The output of the pathways is visualized in the corre- sponding tabs in the study dashboard. The arrangement of the tabs within the study dashboard can be determined by study teams; a division into study management related tabs and oversight/study progress tabs may provide a better overview for the different user groups (principal investigator, study manager, and trial monitor). The main tabs can also contain sub-tabs. For example, the num- ber of due visits is displayed under the visits tab in the sub-category “due visits”. In this context, the definitions of due, overdue, and missed visits are dependent on the specific timeframes of the study protocol. Total num- bers are provided as well as a list of the patient ID and a direct link to the corresponding eCRF in the database (Fig. 2, Panel A). Each tab or sub-tab can represent sev- eral pathway outputs displayed in form of value boxes, graphical presentations, or lists of relevant patients. For example, the SAE management tab provides an overview on SAE prevalence in boxes, and in additional panels the user can switch between the graphical representation of SAE severity, causality, and outcome. Additionally, a list of patients with SAE is provided below, displaying infor- mation on SAE status (e.g. ongoing/closed) and a short description of the event (Fig. 2, Panel B). The informa- tion is provided for the overall study, including all ran- domized patients as numbers and percentages in boxes, while graphs differentiating between centres are provided to better assess which centres are in need for support in a certain aspect of the study conduct. In addition, the dashboard allows filtering for specific centres and time ranges of interest or choosing particular study visits from drop down menus to provide users with more detailed information (see Supplementary Fig. 1 for an example). The output of the pathways visualized in the dash- board is based on a daily export of trial data and, thus, includes up-to-date information on randomised patients and entered data. The generic and some of the optional tabs are listed in Table 3. Examples of the tabs from the two study dashboards are provided in Supplementary Figs. 2–5. The generic dashboard is accessible via GitHub Klatte et al. BMC Medical Research Methodology (2023) 23:84 Page 6 of 11 Fig. 2 Dashboard screenshots of the Visits tab, sub-tab “Due visits” (Panel A), and the Safety management tab, sub-tab “Serious adverse events” (Panel B) and generic data is provided to test the different code modules behind each tab (examples provided in Supple- mentary Figs. 6 and 7). suggestions for further elements to be included in the dashboard. A detailed summary of the results from the user testing is provided in Supplementary Table 5. User testing The user testing of our study dashboards provided posi- tive feedback in terms of improved study oversight and facilitated conduct. Trial monitors and study staff agreed that the initial risk assessment was beneficial, because it increased the awareness of critical processes in the col- lection of outcome data, enabling corrective measures at an early time point, e.g. adaptation of database struc- tures. A clear benefit perceived by all user groups was the more frequent and improved communication with trial sites; sites were better prepared for remote or on-site monitoring visits, because many issues were recognized and solved in advance. In addition, users made several Discussion Using a systematic approach involving relevant stake- holder groups, we developed a concept of risk-tailored trial monitoring and management that focuses on the identification and control of trial specific risks during trial conduct. The continuous evaluation of most impor- tant risks provides important information about the study progress, e.g. in terms of recruitment, endpoint assessment, as well as in terms of data management and data quality, e.g. CRF completion, timeliness of follow- up visits. Completeness of essential data points as the basis for analysable patient data is continuously evalu- ated and trial monitors and study managers maintain an Klatte et al. BMC Medical Research Methodology (2023) 23:84 Table 3 Structure and content of dashboard tabs Domain Participant Safety and Rights Risk Elements Informed consent Example Tabs Informed consent AE/SAE reporting and documentation AE/SAE Inclusion/exclusion Safety Overall Study Management Recruitment Recruitment Patient Characteristics Retention Retention Study procedures and endpoint assessment Bio sampling (e.g. blood samples) Imaging quality Content of Tab In case of a re-consent this tab can provide an overview of patients patients who have previously not been able to give consent themselves Provides an overview of timeliness and completeness of AE/SAE entries In case of safety-relevant inclusion or exclusion criteria, a verification of relevant information available in the database can provide ad- ditional security (e.g. blood pressure has to be within a certain range – check for the entry of blood pressure in the database) Recruitment trajectories for expected and actual recruit- ment in total and per centre (Supplementary Fig. 2) Relevant patient character- istics are summarized and presented (e.g. gender, age, background of treatment) Patients who have ended the study resulting in missing outcome data, reasons for leaving the study, kind of data collected before study end (Primary outcome data avail- able) (Supplementary Fig. 3) Overview of samples taken and availability of sample results Automated and visual verifica- tion of imaging data quality, e.g., for MRI or CT Participant schedule: Follow-up visits Overview of follow-up visits AE/SAE management Safety manage- ment (SAEs, AEs) with a particular focus on visits where primary outcome data is collected. (Fig. 2, Panel A) The Safety tab provides an overview of SAEs and AEs that have been reported in the study and information on severity and outcome of SAEs/AEs (Fig. 2, Panel B) Page 7 of 11 Generic/Optional Optional Generic Optional Functionality/Purpose To ensure patient rights and support of re-consent process through site-specific reminders, list of patients that still need a re-consent. To ensure that all AE/SAE forms are complete and that the date of first entry is within the required reporting timeframe To provide the option for addi- tional checks for inclusion/ exclu- sion criteria besides the marked list of criteria in the eCRF To monitor the progress of partici- pant recruitment enabling early action in case of slow recruitment. Generic Generic Generic Optional Optional Optional Generic To inform the study team on the accuracy of inclusion/exclusion criteria and provide an overview of the sample population in terms of relevant characteristics To monitor the progress of partici- pant retention, consider reasons for ending study in recruitment. Time point of ending the study important for amount of data analysable. To support sample management in terms of localization and status of bio sample. Important for biomarker determination. To enable early adjustments in case of low quality imaging data and ensure that the imaging data is analysable. To assist in integrating follow-up visits on time into the daily clinical routine might be difficult for trial sites. Support through remind- ers for due visits can be initiated through the dashboard. To estimate potential safety issues (e.g. SAEs occurring more often in one study arm, number of SAEs in total, number of patients with SAE) Klatte et al. BMC Medical Research Methodology (2023) 23:84 Table 3 (continued) Domain Device/ Medication Management Risk Elements Administration Accountability/ storage Example Tabs Medication Study Data Data Quality Data quality – com- pleteness, consis- tency, timeliness Documentation/ storage Content of Tab Overview of medication con- sumption based on number of patients and their current position in the medication plan per protocol and com- parison with IMP stock at sites Completeness of forms (Primary end point, secondary endpoint, SAE/AE forms) Timeliness of data entry, Number of queries, status of queries (open, resolved) (Supplementary Figs. 4,5) Page 8 of 11 Functionality/Purpose To assist in the managing of IMP stock overview and enable reminders for restocking Generic/Optional Optional Generic To increase awareness of items missing in the database Trial sites may have different chal- lenges when integrating a trial in their daily clinical routine and therefore need support in differ- ent aspects of the study conduct. Completeness and timeliness of data entry as well as query man- agement constitute indicators for need of support. Query status helps the study monitor to decide which centre needs more assistance/ on-site visit. Abbreviations: AEs, adverse events; CT, computerized tomography ;IMP, investigational medicinal product; MRI, magnetic resonance imaging; SAEs, serious adverse events overview of visit timeframes, SAE reporting, and query management. Strengths and limitations Strengths of our study are the systematic and structured process of development of the risk assessment and the trial dashboard, which included the involvement of all local stakeholder groups and the performance of a com- prehensive contextual analysis. In addition, the devel- opment was based on prior evidence gathered through systematic literature searches and exchange with interna- tional stakeholder groups. Directly involving end users in developing and evaluating the usability of our tool may facilitate the implementation process, promote wider adoption, maintain involvement, and increase user satis- faction with the concept as well as the tool [33]. Providing an R code repository for other study teams that can be adapted and applied to differently structured databases, constitutes a software-independent, affordable approach for the limited budget of investigator-initiated trials. Our study has the following limitations: First, we per- formed user testing in two ongoing RCTs only, and, thus, the spectrum of feedback may have been limited and may compromise the extrapolation of mentioned ben- efits and disadvantages to other trials. Both RCTs had already started participant recruitment when the dash- board was implemented. This allowed for a qualitative comparison of management and monitoring processes without and with the dashboard tool in place. However, it will be crucial to subsequently evaluate the impact and value of the study dashboard during the entire course of a clinical trial. Since both RCTs are still ongoing, we could not evaluate the impact of the tool on participant safety and overall trial success, including the percentage of analysable data, at the end of a trial. Lastly, we have not yet evaluated any cost-effectiveness of our developed approach, e.g. assessing whether the dashboard has the potential to reduce monitoring and management hours needed to ensure a safe and successful trial conduct. While some users felt that our dashboard would only be worthwhile for multicentre trials, others found that the costs of providing a study dashboard will always depend on the needs and preferences of the study team and the complexity of the study. Comparison with similar studies and frameworks Following the recommendations of the Clinical Trials Transformation Initiative (CTTI), effective and efficient monitoring and management needs to first determine what matters for a specific trial and focus on areas of highest risk for generating errors that matter [34, 35]. With our risk assessment guide and the study dashboard we address the need for this focus and provide a tool that supports the continuous oversight of the quality of the trial conduct. Dashboards that visualize time-dependent parameters have recently met a growing acceptance in medical and administrative health care settings [36–43]. Dashboards have been introduced to support various aspects of clinical trials, including web applications for eligibility screening and overview of the enrolment progress [41], web-based support of recruitment management and Klatte et al. BMC Medical Research Methodology (2023) 23:84 Page 9 of 11 communication; [42] graphical summaries and diagrams of the progress of patient accrual and form completion [43], feedback on data completeness by using a traffic light system [44], and automated reports of data compli- ance, protocol adherence and safety [45]. These available dashboards typically focus on specific elements of trial conduct and communication with trial sites; however, our dashboard provides a comprehensive overview of all elements of a trial identified as critical. In addition, tables and graphical representations are often limited to certain time intervals [41]. The daily export of trial data providing up-to-date trial information is part of the core idea of our approach as it enables immediate actions and improves communication with site staff. Various methods for assessing the risk of non-conform trial conduct at trial sites including central statistical monitoring have been introduced in the academic set- ting with increasing prevalence [46]. Most methods use statistical testing of all or a subset of trial data items to compare sites and identify atypical trial centres. While many methods focus on the detection of data errors and fraud, [47] triggered monitoring is frequently used to direct on-site monitoring to atypical trial sites [46]. In our approach components of central data evaluations are used to assess whether actions are required constituting some sort of triggered intervention. However, the data evaluation is not based on statistical testing, it is rather an assessment of trial progress (recruitment, retention), management challenges, and conform data collection progress. It is also not intended to categorize trials and predetermine the extent of on-site monitoring [48]. Our concept focuses on directing attention to the most criti- cal areas of a trial and should help to minimize and tailor on-site monitoring. Several commercial solutions supporting the over- all trial conduct in various aspects are readily available [9, 49–53], but for investigator-initiated trials with tight budgets such software packages typically remain unaf- fordable. We wanted to provide a comprehensive and affordable option for investigator-initiated trials that can be adapted to individual needs and preferences and fur- ther developed by the research community. Therefore, we transparently present all details of the structured risk assessment and manual as well as the generic code for our dashboard in publicly accessible repositories via GitHub. We invite users to report difficulties or sugges- tions for improvement for consideration in future modifi- cations of the generic dashboard via GitHub. Implications Besides the emphasis on the feasibility and design of clin- ical trials, measures to increase the efficiency of clinical trial conduct are needed [54]. Current challenges include premature discontinuation of a significant proportion of clinical trials, and inflated costs mainly due to delayed recruitment and organisational issues [54]. We propose a comprehensive approach integrating management and monitoring of a clinical trial into one risk management tool supporting the conduct of investigator-initiated trials. Overseeing the progress of a trial in each centre based on up-to-date information, provides the opportunity for trial monitors to prioritize centres for on-site visits or remote interactions, tailor their action to the specific issues of a centre, and guide decisions on where resources and training is needed the most. In addition, providing automated reminders for upcoming visits or sampling, overview of investigational medicinal product supply, overview of patients who need a re-consent, overview of ongoing SAEs, etc. could increase the efficiency of the trial management processes. The tool further provides the opportunity to improve the overall communication between the study team and trial sites and may increase motivation through the involvement of sites in the trial progress and the option to compliment active partici- pation in the trial. The dashboard tool is intended to address site-level monitoring, trial-wide monitoring, and finding per-patient issues. Feedback from the user testing also revealed a positive perception of study managers and investigators to improved data quality visible in the dash- board: “If incomplete is empty, I am at ease. The impact of this tool is largely dependent on the suc- cessful implementation into clinical trial practice. The perception of benefits and opportunities by stakeholders and end-users have been collected while the effectiveness of the tool in terms of analysable data collected, timeline of recruitment, conformity of SAE/Adverse Event (AE) reporting and documentation, support of the overall study management still have to be evaluated. The next step is now to implement the risk assessment as a routine step in the joint planning of clinical trials with the respective study teams. The timely generation of a dashboard on the basis of the generic template and further study-specific risks has to be organized. Strate- gies to further evaluate this implementation process as well as the effectiveness of this new approach in studies of different design and structure have to be developed. As an implementation outcome, the amount of studies tak- ing advantage of the study dashboard in relation to the studies for which a dashboard was recommended could be assessed along with the frequency of risk assessments performed per trial. The effectiveness of the concept of risk assessment and dashboard tool will be evaluated based on structured feedback from study teams on their experience and quantitative measures of the trial, e.g. proportion of analysable patients/data at the end of the trial. These evaluations will provide more information on the feasibility of study-specific dashboards supporting Klatte et al. BMC Medical Research Methodology (2023) 23:84 trial monitoring and management in the heterogeneous field of clinical trials. Conclusion In summary, the presented risk-assessment guide and dashboard tool provide a systematically developed and user-tested instrument for the risk-tailored support of trial monitoring and trial management. Feedback from the user testing of the instrument revealed many benefits for the involved stakeholder groups. However, the effec- tiveness of the dashboard in terms of a safe trial conduct and overall support for a successful completion of clinical trials needs to be further evaluated. Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12874-023-01902-y. Supplementary Material 1 Supplementary Material 2 Acknowledgements We would like to thank all stakeholders that provided feedback in the development process of our risk assessment and the study dashboard. Author contributions K.K., M.B. and C.P.M. wrote the main manuscript text and K.K. prepared the figures. P.B., A.S., K.E., A.R. and T.F. were involved in the development of the concept, the risk assessment and dashboard content. S.S. and K.K. developed the study-specific dashboards and the generic dashboard. K.K. conducted the interviews for the user testing. All authors reviewed the manuscript. Funding Open access funding provided by University of Basel Data Availability We provide R modules as the basis for the dashboard development on GitHub (https://github.com/CTU-Basel/viewTrial). Qualitative data that supported the development of the risk assessment and study dashboard is provided in the supplementary material. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing Interest The authors declare no competing interests. Received: 10 September 2022 / Accepted: 23 March 2023 References 1. Yusuf S. Randomized clinical trials: slow death by a thousand unnecessary policies? CMAJ 2004;171(8):889 – 92; discussion 92 – 3. doi: https://doi. org/10.1503/cmaj.1040884 [published Online First: 2004/10/13] Page 10 of 11 4. 3. 2. Eisenstein EL, Lemons PW 2nd, Tardiff BE, et al. Reducing the costs of phase III cardiovascular clinical trials. Am Heart J. 2005;149(3):482–8. https://doi. org/10.1016/j.ahj.2004.04.049. [published Online First: 2005/05/03]. Eisenstein EL, Collins R, Cracknell BS, et al. Sensible approaches for reducing clinical trial costs. Clin Trials. 2008;5(1):75–84. https://doi. org/10.1177/1740774507087551. [published Online First: 2008/02/20]. Kasenda B, von Elm E, You J, et al. Prevalence, characteristics, and publication of discontinued randomized trials. JAMA. 2014;311(10):1045–51. https://doi. org/10.1001/jama.2014.1361. [published Online First: 2014/03/13]. Al-Shahi Salman R, Beller E, Kagan J, et al. Increasing value and reduc- ing waste in biomedical research regulation and management. Lancet. 2014;383(9912):176–85. https://doi.org/10.1016/S0140-6736(13)62297-7. [published Online First: 2014/01/15]. Tudur Smith C, Stocken DD, Dunn J, et al. The value of source data verifica- tion in a cancer clinical trial. PLoS ONE. 2012;7(12):e51623. https://doi. org/10.1371/journal.pone.0051623. [published Online First: 2012/12/20]. 7. Hearn J, Sullivan R. The impact of the ‘clinical trials’ directive on the cost 6. 5. and conduct of non-commercial cancer trials in the UK. Eur J Cancer. 2007;43(1):8–13. https://doi.org/10.1016/j.ejca.2006.09.016. [published Online First: 2006/11/23]. 8. Duley L, Antman K, Arena J, et al. Specific barriers to the con- duct of randomized trials. Clin Trials. 2008;5(1):40–8. https://doi. org/10.1177/1740774507087704. [published Online First: 2008/02/20]. CluePoints.Risk-basedmonitoring. http://cluepoints.com/risk-based-monitor- ing. 2016 [accessed 31st July 2022. 9. 10. Medidata.Strategic monitoring https://www.medidata.com/wp-content/ 11. uploads/2018/12/Strategic-Monitoring_2018Medidata_Fact-Sheet.pdf2018 [accessed 31st July 2022. International Conference on Harmonisation of technical requirements forpharmaceuticals for human use (ICH). Guideline for good clinical practice E6(R2) Available from: https://www.ema.europa.eu/en/documents/scientific- guideline/ich-e-6-r2-guideline-good-clinical-practice-step-5_en.pdf2018 [accessed 21st July 2022. 12. Medicines. Healthcare products Regulatory Agency. Good clinicalpractice guide. The Stationery Office; 2012. 13. Califf RM. Clinical trials bureaucracy: unintended consequences of well-inten- tioned policy. Clin Trials. 2006;3(6):496–502. doi: 10.1177/1740774506073173 [published Online First: 2006/12/16]. 14. Sertkaya A, Wong HH, Jessup A, et al. Key cost drivers of pharmaceutical clinical trials in the United States. Clin Trials. 2016;13(2):117–26. https://doi. org/10.1177/1740774515625964. [published Online First: 2016/02/26]. 15. Funning SGA, Eriksson K, Kettis-Lindblad. A Quality assurance within the scope of good clinical practice (GCP) - what is the cost of GCP-related activi- ties? A survey within the Swedish Association of the Pharmaceutical industry (LIF)’s members. Qual Assur J. 2009;12:3–7. 16. Ecrin. Risk-Based Monitoring Toolbox Available from: https://ecrin.org/tools/ risk-based-monitoring-toolbox2021 [ 17. Klatte K, Pauli-Magnus C, Love SB, et al. Monitoring strategies for clinical inter- vention studies. Cochrane Database Syst Rev. 2021;12:MR000051. https://doi. org/10.1002/14651858.MR000051.pub2. [published Online First: 2021/12/09]. 18. Baigent C, Harrell FE, Buyse M, et al. Ensuring trial validity by data quality assurance and diversification of monitoring methods. Clin Trials. 2008;5(1):49– 55. https://doi.org/10.1177/1740774507087554. [published Online First: 2008/02/20]. 19. Love SB, Armstrong E, Bayliss C, et al. Monitoring advances including consent: learning from COVID-19 trials and other trials running in UKCRC registered clinical trials units during the pandemic. Trials. 2021;22(1):279. https://doi. org/10.1186/s13063-021-05225-5. [published Online First: 2021/04/16]. 20. Farrell B, Kenyon S, Shakur H. Managing clinical trials. Trials. 2010;11:78. https://doi.org/10.1186/1745-6215-11-78. [published Online First: 2010/07/16]. 21. Medical Research Council. Clinical trials for tomorrow: an MRC Review of Randomised Control trials. Medical Research Council; 2003. 22. Love SB, Yorke-Edwards V, Diaz-Montana C, et al. Making a distinction between data cleaning and central monitoring in clinical trials. Clin Trials. 2021;18(3):386–88. https://doi.org/10.1177/1740774520976617. [published Online First: 2021/03/04]. 23. Swiss Clinical Trial Organisation (SCTO). F dated. Fact sheet: Central Data Monitoring in Clinical Trials – V 1.0. available from www.scto.ch/monitoring SCTO; [accessed May 3rd 2022. 24. Venet D, Doffagne E, Burzykowski T, et al. A statistical approach to central monitoring of data quality in clinical trials. Clin Trials. 2012;9(6):705–13. Klatte et al. BMC Medical Research Methodology (2023) 23:84 https://doi.org/10.1177/1740774512447898. [published Online First: 2012/06/12]. 25. Agrafiotis DK, Lobanov VS, Farnum MA, et al. Risk-based monitoring of clinical trials: an Integrative Approach. Clin Ther. 2018;40(7):1204–12. https://doi. org/10.1016/j.clinthera.2018.04.020. [published Online First: 2018/08/14]. 26. Walden A, Garvin L, Smerek M, et al. User-centered design principles in the development of clinical research tools. Clin Trials. 2020;17(6):703–11. https:// doi.org/10.1177/1740774520946314. [published Online First: 2020/08/21]. 27. von Niederhausern B, Orleth A, Schadelin S, et al. Generating evidence on a risk-based monitoring approach in the academic setting - lessons learned. BMC Med Res Methodol. 2017;17(1):26. https://doi.org/10.1186/s12874-017- 0308-6. [published Online First: 2017/02/15]. 28. Swiss Clinical Trial Organisation SCTO. Risk assessment Form for Clinical Research Projects available from: https://www.sctoplatforms.ch/en/tools/risk- assessment-form-for-clinical-research-projects-30.html2018 [accessed May 3rd 2022. 29. Wilson B, Provencher T, Gough J, et al. Defining a central monitoring capability:sharing the experience of TransCelerate BioPharma’s Approach, Part 1. Therapeutic Innov Regul Sci. 2014;48(5):529–35. https://doi. org/10.1177/2168479014546335. 30. Gough J, Wilson B, Zerola M, et al. Defining a central monitoring capability:sharing the experience of TransCelerate BioPharma’s Approach, Part 2. Therapeutic Innov Regul Sci. 2016;50(1):8–14. https://doi. org/10.1177/2168479015618696. 31. Whitham D, Turzanski J, Bradshaw L, et al. Development of a standardised set of metrics for monitoring site performance in multicentre randomised trials: a Delphi study. Trials. 2018;19(1):557. https://doi.org/10.1186/s13063-018-2940- 9. [published Online First: 2018/10/18]. 32. Diaz-Montana C, Cragg WJ, Choudhury R, et al. Implementing monitoring triggers and matching of triggered and control sites in the TEMPER study: a description and evaluation of a triggered monitoring management system. Trials. 2019;20(1):227. https://doi.org/10.1186/s13063-019-3301-z. [published Online First: 2019/04/19]. 33. Goodman MS, Sanders Thompson VL. The science of stakeholder engage- ment in research: classification, implementation, and evaluation. Transl Behav Med. 2017;7(3):486–91. https://doi.org/10.1007/s13142-017-0495-z. [published Online First: 2017/04/12]. 34. EFFECTIVE AND EFFICIENT CTTIRECOMMENDATIONS, MONITORING AS A COMPONENT OF QUALITY ASSURANCE IN, THE CONDUCT OF CLINICAL TRI- ALS Available at. : https://ctti-clinicaltrials.org/wp-content/uploads/2021/06/ CTTI_Monitoring_Recs.pdf2021 [accessed 21st July 2022. 35. CTTI. QUALITY BY DESIGN PROJECT - CRITICAL TO QUALITY (CTQ) FACTORS PRINCIPLES DOCUMENT Available at. : https://ctti-clinicaltrials.org/wp-con- tent/uploads/2021/07/CTTI_QbD_Workshop_Principles_Document.pdf 2021 [accessed 21st July 2022. 36. Simpao AF, Ahumada LM, Desai BR, et al. Optimization of drug-drug interac- tion alert rules in a pediatric hospital’s electronic health record system using a visual analytics dashboard. J Am Med Inform Assoc. 2015;22(2):361–9. https://doi.org/10.1136/amiajnl-2013-002538. [published Online First: 2014/10/17]. 37. Hartzler AL, Izard JP, Dalkin BL, et al. Design and feasibility of integrating per- sonalized PRO dashboards into prostate cancer care. J Am Med Inform Assoc. 2016;23(1):38–47. https://doi.org/10.1093/jamia/ocv101. [published Online First: 2015/08/12]. 38. Dowding D, Randell R, Gardner P, et al. Dashboards for improving patient care: review of the literature. Int J Med Inform. 2015;84(2):87–100. https://doi. org/10.1016/j.ijmedinf.2014.10.001. [published Online First: 2014/12/03]. Page 11 of 11 39. Waitman LR, Phillips IE, McCoy AB, et al. Adopting real-time surveillance dashboards as a component of an enterprisewide medication safety strategy. Jt Comm J Qual Patient Saf. 2011;37(7):326–32. https://doi.org/10.1016/ s1553-7250(11)37041-9. [published Online First: 2011/08/09]. 40. Nelson SD, Del Fiol G, Hanseler H, et al. Software Prototyping: a Case Report of Refining user requirements for a Health Information Exchange Dash- board. Appl Clin Inform. 2016;7(1):22–32. https://doi.org/10.4338/ACI-2015- 07-CR-0091. [published Online First: 2016/04/16]. 41. Arab L, Hahn H, Henry J, et al. Using the web for recruitment, screen, tracking, data management, and quality control in a dietary assessment clinical valida- tion trial. Contemp Clin Trials. 2010;31(2):138–46. [published Online First: 2009/11/21]. 42. Mattingly WA, Kelley RR, Wiemken TL, et al. Real-time enrollment dashboard for Multisite clinical trials. Contemp Clin Trials Commun. 2015;1:17–21. https://doi.org/10.1016/j.conctc.2015.09.001. [published Online First: 2016/02/16]. 43. Toddenroth D, Sivagnanasundaram J, Prokosch HU, et al. Concept and imple- mentation of a study dashboard module for a continuous monitoring of trial recruitment and documentation. J Biomed Inform. 2016;64:222–31. https:// doi.org/10.1016/j.jbi.2016.10.010. [published Online First: 2016/10/23]. Improving data entry and study compliance efficiently using immediate audit and feedback tools 5th International Clinical Trials Methodology Con- ference (ICTMC); 2019. Trials. 44. 45. Automation of clinical trial statistical monitoring. 5th International Clinical Trials Methodology Conference (ICTMC); 2019. Trials. 46. Cragg WJ, Hurley C, Yorke-Edwards V, et al. Dynamic methods for ongoing assessment of site-level risk in risk-based monitoring of clinical trials: a scoping review. Clin Trials. 2021;18(2):245–59. https://doi. org/10.1177/1740774520976561. [published Online First: 2021/02/23]. 47. Kirkwood AA, Cox T, Hackshaw A. Application of methods for central statisti- cal monitoring in clinical trials. Clin Trials. 2013;10(5):783–806. https://doi. org/10.1177/1740774513494504. [published Online First: 2013/10/17]. 48. Hurley C, Shiely F, Power J, et al. Risk based monitoring (RBM) tools for clinical trials: a systematic review. Contemp Clin Trials. 2016;51:15–27. https://doi. org/10.1016/j.cct.2016.09.003. [published Online First: 2016/10/30]. 49. PerkinElmer. Risk-based monitoring.: Available at: http://www.perkinelmer. 50. 51. com/product/risk-based-monitoring-rbm.; [accessed August 18th 2022. IQVIA. The next wave of central monitoring. Available at: https://www. iqvia.com/library/white-papers/the-next-wave-of-centralized-monitoring. [accessed August 18th 2022. Intelligence TR. Ttirials: a complete risk-based monitoring solution Available at: https://tritrials.com [accessed August 18th 2022. 52. Cyntegrity. Risk-based monitoring solutions Available at: https://cyntegrity. com [accessed August 18th 2022. 53. Plc ICON. Patient centric monitoring: preventing and learning from mistakes Available at: http://www.iconplc.com/icon-views/blog/2015/05/21/patient- centric-monitorin [accessed August 18th 2022. 54. Swiss Academy of Medical Sciences. White Paper: Clinical Research. Swiss Academies Communications 2021; 16 (4) Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Klatte et al. BMC Medical Research Methodology (2023) 23:84
10.1186_s12879-021-06753-w	Huang et al. BMC Infect Dis (2021) 21:1068 https://doi.org/10.1186/s12879-021-06753-w RESEARCH ARTICLE Open Access Surgical management of tuberculous epididymo-orchitis: a retrospective study of 81 cases with long-term follow-up Yin Huang1,2†, Bo Chen1†, Dehong Cao1†, Zeyu Chen1, Jin Li1, Jianbing Guo1, Qiang Dong1, Qiang Wei1* and Liangren Liu1* Abstract Background: Nowadays, most studies of tuberculous epididymo-orchitis (TBEO) are case reports or small sample cohort series. Our study is aimed to present the largest series of TBEO with our management experiences and long- term follow-up outcomes. Methods: Patients diagnosed with TBEO after surgical procedures at Department of Urology, West China Hospital from 2008 to 2019 were included. All clinical features, auxiliary examination results, treatment and histopathological findings were extracted if available. ± = 16.1 years) were included. Scrotal swelling (N 7.0 months. TBEO was considered in 30 (37.0%), tumors in 28 (34.6%) and nonspecific bacterial epididymo- 29, 35.8%) were the most common presenting complaint. Pyuria and microscopic hematuria were observed Results: Eighty-one patients (mean age 50.77 (N in twenty-two (27.2%) and eight patients (9.9%), respectively. Urine acid fast bacilli cultures were available in 16 patients and all were negative. The mean duration between the onset of symptoms and the definite diagnosis was 6.42 orchitis in 23 (28.4%) patients. All patients received triple therapy of chemotherapy-surgery-pharmacotherapy and definite diagnosis was confirmed through histopathology of surgical specimens. Fifty-five patients were followed up regularly (mean follow-up 82.35 36.6 months). One patient (1.2%) died from liver cirrhosis and no recurrence was observed. Postoperative complications included erectile dysfunction in 4 patients (4.9%), premature ejaculation in 5 patients (6.2%) and sterility in 7 patients (8.6%). 47, 58.0%) and pain ± = ± Conclusions: We recommend patients with advanced TBEO to receive triple therapy of chemotherapy-surgery- pharmacotherapy. Physicians should pay more attention to patients’ sexual function and fertility during follow up after treatment completed. Keywords: Tuberculosis, Epididymo-orchitis, Chemotherapy-surgery-pharmacotherapy, Follow up, Sexual function, Fertility *Correspondence: weiqiang933@126.com; liuliangren@scu.edu.cn †Yin Huang, Bo Chen and Dehong Cao contributed equally to this study 1 Department of Urology, Institute of Urology, West China Hospital, Sichuan University, Guoxue Alley, No. 37, Chengdu 610041, Sichuan, People’s Republic of China Full list of author information is available at the end of the article Background According to the 2019 World Health Organization global tuberculosis (TB) report, about 10 million (range, 9.0– 11.1 million) new cases of TB were reported worldwide in 2018. There were an estimated 1.2 million TB deaths among HIV-negative people and an additional 251,000 deaths among HIV positive people in 2018 [1]. TB can © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecom- mons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Huang et al. BMC Infect Dis (2021) 21:1068 Page 2 of 8 affect people of both sexes in all age groups but the high- est burden is in men (aged ≥ 15 years), who accounted for 57% of all TB cases, while women and children (aged < 15 years) accounted for 32% and 11%, respectively. China is the second highest TB-burden country after India, accounting for 9% of global TB cases in 2018 [1]. In addition to lymphatic involvement, urogenital TB is the most common manifestation of extrapulmonary TB and is more frequent in middle-aged men, which accounts for 33.7–45.5% of the extrapulmonary TB worldwide [2]. Compared with renal TB, male genital TB is a rare sub- type of the urogenital TB, which can be classified as TB epididymitis, TB orchitis, TB of the prostate, TB of the seminal vesicles, and TB of the penis [2, 3]. Since the lack of preoperative diagnostic methods with high sensitivity and specificity, tuberculous epididymo- orchitis (TBEO) with nonspecific clinical signs is often misdiagnosed with bacterial infection or tumor [4]. Standard anti-tuberculosis chemotherapy is the first-line therapy for TBEO. However, surgical intervention may be unavoidable in cases of hard to diagnose, or poorly responding to chemotherapy [5–7]. Patients with TBEO received regular anti-tuberculosis chemotherapy plus surgery intervention are seldom to recur [4, 8]. Nowa- days, the majority of studies of TBEO are case reports or small sample cohort series including less than 50 cases. Furthermore, long-term follow-up data after treatment are deficient [4–6, 8]. Therefore, our study is aimed to present our experiences on the clinicopathological char- acteristics, management and 11-year follow-up outcomes of TBEO at a large medical center in west China. Methods Setting and study design From January 2008 to May 2019, patients diagnosed with TBEO after surgical procedures at Department of Urol- ogy, West China Hospital were included in our retrospec- tive observational study. All clinical features (symptoms, physical signs, duration of disease, past medical history, comorbidities and organ involvement), auxiliary exami- nation results (hemogram, erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), urinalysis, blood biochemistry, tumor markers, urine culture, ultrasonog- raphy and radiology), treatment (drug therapy and surgery) and histopathological findings (microscopy, Ziehl–Neelsen acid fast stain and polymerase chain reac- tion (PCR)) were retrieved from medical records if avail- able. Patients who only received drug therapy without surgery were excluded. Diagnosis TBEO was definitely diagnosed in the presence of clinical findings combined with one of the following criteria: (1) Positivity of acid fast bacilli (AFB) in urine. (2) Positive urine culture for M. tuberculosis. (3) Positivity of PCR for M. tuberculosis in urine. (4) Typical granulomatous inflammation with caseous necrosis in microscopy plus any positivity of Ziehl–Neelsen acid fast stain or PCR for M. tuberculosis in any relevant tissue specimen [9]. In our institution, histopathological evidence of TBEO was only obtained from postoperative histopathological find- ings of surgical specimens, no preoperative histopatho- logical examination (e.g. fine-needle aspiration cytology (FNAC)) was performed. According to the criteria of our institution: ESR > 20 mm/1 h, CRP > 5 mg/L, serum beta- human chorionic gonadotropin (β-HCG) ≥ 3.81 mIU/ml, serum alpha-fetoprotein (AFP) ≥ 8 ng/ml and serum lac- tate dehydrogenase (LDH) ≥ 220 U/L were considered to be elevated. Treatment and follow up Surgical indications for patients with TBEO in our center including: (1) Regular anti-tuberculosis pharmacotherapy for 1–2 months was completed but tuberculous lesions were still not controlled. (2) TBEO was diagnosed in advanced stage with widely spread of M. tuberculosis. (3) Tuberculous complications including hydrocele, abscess, sinus or fistula were observed. (4) Clinical diagnosis of scrotal tumors was suspected. (5) Clinical diagnosis was nonspecific bacterial epididymo-orchitis but the efficacy of antibiotic treatment was limited. Patients with clini- cal diagnosis of TBEO received anti-tuberculosis chemo- therapy for 2–4 weeks before surgery. Surgical procedure (orchiectomy, epididymectomy or epididymo-orchidec- tomy) and postoperative anti-tuberculosis therapy with 4 drugs (rifampicin, isoniazid, ethambutol and pyrazina- mide) for 6–12 months were performed in all patients. After completion of the therapy, patients were followed up for symptoms, physical examination, scrotal ultra- sound, sexual function and fertility until June, 2020. Statistical analysis The data were analyzed by the IBM SPSS Statistics (version 25). Continuous variables were expressed as mean ± standard deviation (SD) or median with inter- quartile range (IQR). P value < 0.05 was considered as sta- tistically significant. Results Clinical characteristics During the 11 years period of the study, a total of 98 cases were recorded as TBEO. After exclusion of those not meeting the diagnostic criteria (N = 11), repeating records (N = 2) and patients who did not undergo surgery (N = 4), 81 separate patients with TBEO were included in our study. Clinical characteristics including symptoms, Huang et al. BMC Infect Dis (2021) 21:1068 Page 3 of 8 physical examination, urinalysis, laboratory findings and history of TB of 81 patients are shown in Table 1. The average age of the patients was 50.77 ± 16.1 years (range, 13–90 years). Scrotal swelling (N = 47, 58.0%) was the most common presenting complaint, followed by scrotal pain (N = 29, 35.8%). Unilateral and bilateral scrotal mass were detected through physical examina- tion in 43 (53.1%) and 4 (4.9%) patients, respectively. The most positive findings of urinalysis were pyuria (N = 22, 27.2%) and microscopic hematuria (N = 8, 9.9%). AFB culture results were available in only 16 Table 1 Clinical characteristics of 81 patients Characteristics Ages (years) < 60 60 ≥ Symptoms Scrotal swelling Scrotal pain Fever Night sweats Weight loss Frequency / urgency Dysuria Physical examination Left scrotal mass Right scrotal mass Bilateral scrotal mass Abscess Sinus Urinalysis Pyuria Microscopic hematuria Positive urine AFB culture Laboratory findings Increased ESR Increased CRP Increased WBC Increased serum AFP Increased serum β-HCG Increased serum LDH Anemia Hypoalbuminemia History of TB Lung Kidney Bone No. Pts/total (%) 62/81 (76.5) 19/81 (23.5) 47/81 (58.0) 29/81 (35.8) 6/81 (7.4) 1/81 (1.2) 11/81 (13.6) 2/81 (2.5) 3/81 (3.7) 16/81 (19.8) 27/81 (33.3) 4/81 (4.9) 2/81 (2.5) 1/81 (1.2) 22/81 (27.2) 8/81 (9.9) 0/16 (0) 7/14 (50.0) 5/8 (62.5) 4/81 (4.9) 3/30 (10.0) 1/30 (3.3) 10/81 (12.3) 8/81 (9.9) 12/81 (14.8) 10/81 (12.3) 8/81 (9.9) 1/81 (1.2) 1/81 (1.2) AFB acid fast bacilli, ESR erythrocyte sedimentation rate, CRP C-reactive protein, WBC white blood cell, AFP alpha-fetoprotein, β-HCG beta-human chorionic gonadotropin, LDH lactate dehydrogenase, TB tuberculosis patients and all were negative. The duration between the onset of symptoms and the definite diagnosis var- ied from 0.3 to 36 months, with an average duration of 6.42 ± 7.0 months. Furthermore, TB history of other organs was reported in ten patients (12.3%). Besides, fifteen patients (18.5%) were diagnosed with TBEO in other medical centers before registered in our hospital and had received regular anti-tuberculosis pharmaco- therapy for 1–12 months, but the scrotal tuberculosis lesions of these patients were not controlled effectively. Abnormal rates of imaging findings in 81 patients are shown in Table 2. Scrotal ultrasound detected the signs of epididymo-orchitis in 30 of 39 patients. Forty- one patients received computed tomography (CT) scan and infectious disease was observed in 36 patients. Magnetic resonance imaging (MRI) was performed in 2 patients and both found an abnormality. Chest radio- logical evidence of pulmonary infection was found in 43 of 81 patients, and 6 patients (7.4%) had and evi- dence of active pulmonary TB. In addition, ten cases (12.3%) of hydrocele and four cases (4.9%) of varicocele were confirmed by scrotal imaging. Clinical diagnosis and treatment Before surgery, thirty patients (37.0%) were clinically diagnosed with TBEO, while 28 patients (34.6%) were diagnosed with tumors. Twenty-three patients (28.4%) were preoperatively diagnosed with nonspecific bac- terial epididymo-orchitis but routine antibiotic treat- ment was noneffective (Table 3). Surgical procedure was performed for all patients. Thirty patients with clinical diagnosis of TBEO received preoperative anti- tuberculosis chemotherapy for 2–4 week. Orchiec- tomy, epididymectomy and epididymo-orchidectomy were performed in two (2.5%), twenty-seven (33.3%) and fifty-two (64.2%) patients, respectively. Seventy- five patients (92.6%) received unilateral orchiectomy or epididymectomy. Bilateral surgical procedure was per- formed in 6 patients (7.4%) (Table 3). Scrotal masses were found in 43 patients (53.1%) in surgery, with the mean diameter of 3.18 ± 1.6 cm (range, 0.2–7.0 cm). Table 2 Abnormal rates of imaging findings in 81 patients Imaging modality No. Pts (total) Percentage (%) Ultrasonography Computed tomography Magnetic resonance Chest radiography 30 (39) 36 (41) 2 (2) 43 (81) 76.9 87.8 100 53.1 Huang et al. BMC Infect Dis (2021) 21:1068 Page 4 of 8 Table 3 Clinical diagnosis, treatment and histopathology characteristics of 81 patients Characteristics Clinical diagnosis TBEO Tumor Nonspecific bacterial epididymo-orchitis Surgical procedure Unilateral Bilaterala Orchiectomy Epididymectomy Epididymo-orchidectomy Histopathology Positive acid fast stain Positive PCR Any of the above Granulomatous inflammation with caseous necrosis Isolated testicular TB Isolated epididymal TB Testicular and epididymal TB No. Pts (total) Percentage (%) 30 (81) 28 (81) 23 (81) 75 (81) 6 (81) 2 (81) 27 (81) 52 (81) 62 (77) 24 (43) 81 (81) 81 (81) 23 (81) 31 (81) 27 (81) 37.0 34.6 28.4 92.6 7.4 2.5 33.3 64.2 80.5 55.8 100 100 28.4 38.3 33.3 TBEO tuberculous epididymo-orchitis, PCR polymerase chain reaction, TB tuberculosis a One patient underwent bilateral orchiectomy, one patient underwent bilateral epididymectomy, and four patients underwent orchiectomy and contralateral epididymectomy Histopathological findings Postoperative histopathology showed the typical gran- ulomatous inflammation with amorphous caseous necrosis in all surgical specimens (Fig. 1A). At high magnification, the amorphous caseous necrosis was surrounded by granulomas contain epithelioid histio- cytes, Langhans giant cells and lymphocytes (Fig. 1B). Ziehl–Neelsen acid fast stain was performed in 77 specimens and was positive in sixty-two (80.5%). PCR findings for M. tuberculosis identification were available in 43 surgical specimens and twenty-four cases (55.8%) showed a positive result. Results of histopathological findings are shown in Table 3. Prognosis All patients were followed up regularly until June, 2020 except 26 patients (32.1%) due to loss of contacts. Fol- low-up ranged from 14 months to 11.5 years, with an average follow-up of 82.35 ± 36.6 months. During fol- low up, only 1 patient (1.2%) died from liver cirrhosis and no recurrence was observed. Symptoms such as fever, scrotal pain and irritative urinary symptoms were under control in 47 cases (58.0%). Urinalysis param- eters and scrotal imaging were stable in 50 patients (61.7%). Postoperative complications included erectile dysfunction in 4 patients (4.9%), premature ejaculation in 5 patients (6.2%) and sterility in 7 patients (8.6%). Discussion Until now, M. tuberculosis was the most frequently iso- lated species in humans all over the world, followed by M. bovis [10]. However, the species M. tuberculosis has been inaccurately used to represent the Mycobacterium tuberculosis complex (MTBC), including M. tuberculosis, M. africanum, M. bovis, M. canettii and so on [10, 11]. As one of the most virulent pathogens for humans, M. tuberculosis has a slow replication rate, which accounts for the latent nature of the infection and its resistance to conventional antibiotics [3]. Despite the bacillus could stay dormant in the human body without any symptoms for a long time, injury of immune function may induce its reactivation [10]. Urogenital TB is the second common form of extrapul- monary TB which occurs in 15% to 20% cases of pulmo- nary TB with a prevalence of 400 per 100,000 population, mostly affecting middle-aged men [2, 12]. Given the high prevalence of TB worldwide, urogenital TB reflects a large burden of urogenital diseases, especially in coun- tries with a severe epidemic situation including China. Male genital TB is a rare subtype of urogenital TB, usu- ally occurring in men aged 30–50 years [5]. Similarly, our Huang et al. BMC Infect Dis (2021) 21:1068 Page 5 of 8 which probably attributed to the latent presentation and delayed diagnosis of TB. In addition, sexual transmission is thought to be possible since M. tuberculosis has been isolated form the ejaculate of men with prostatic TB [5, 13]. The onset of clinically evident TBEO is insidious, with variable clinical manifestations. Most studies found that a scrotal swelling, scrotal pain and irritative voiding symp- toms are the common initial symptoms of patients with TBEO, which is similar to our findings [4, 8, 12]. On phys- ical examination, the scrotal mass may be either painful or painless [6]. Nonspecific constitutional symptoms of TB such as fever, weight loss, fatigue and night sweats are uncommon [9, 13]. However, suspicion of concomitant TB outside the urogenital tract should arise when these constitutional symptoms are present, such as pulmonary TB [13]. Compared with other reports in the literature, the average duration between onset of the symptoms and the definite diagnosis (mean 6.42 ± 7.0 months, range 0.3–36 months) in our study was longer, probably due to the insidious and asymptomatic onset of TBEO [5, 14–16]. Urinalysis was reported abnormal in 77–90% of patients with urogenital TB [5]. Altiparmak et al.[9] reported that hematuria and pyuria were detected in 79.7% and 67.1% of patients, respectively. However, in our study, pyuria and hematuria were only observed in twenty-two (27.2%) and eight patients (9.9%), respec- tively. Similar to our findings, abnormal urinalysis (hema- turia and pyuria) was detected in 59.6% of patients in another cohort study with 47 cases of epididymal tuber- culosis [4]. This result may be because all patients in our study were isolated TBEO without renal involvement. Urine AFB culture was long considered the gold standard in diagnosis of urogenital TB. However, low sensitivity of culture was reported in the literature and the negative urine cultures do not rule out the possibility of TBEO. In addition, cultures may take several weeks to show a delayed result [5, 6, 17]. In our series, however, results of urine AFB cultures were only available in 16 patients since the retrospective nature of the study, and all were negative. On the other hand, patients might not receive the urine AFB cultures when the clinical diagnosis of tumor was considered before surgery. Given the small sample size, our results are not enough to suggest the low significance of urine AFB culture in diagnosis of TBEO. Recent years, as a rapid test for detecting M. tuberculo- sis DNA and rifampicin resistance, GenXpert MTB/RIF was identified to be the better choice in diagnosing uro- genital TB according to its higher sensitivity compared with urine microscopy and culture [13]. Furthermore, interferon-γ release assays (including QuantiFERON- TB Gold In-Tube and T-SPOT TB) were recommended Fig. 1 Microscopy Images of Tuberculous Granulomas in Hematoxylin and Eosin Stain. At low magnification (A), the typical granulomatous inflammation with amorphous caseous necrosis (arrows) were observed. At high magnification (B), the amorphous caseous necrosis was surrounded by granulomas contain epithelioid histiocytes (arrowheads), Langhans giant cells (arrows) and lymphocytes cohort was consisted of a wide range of ages with a mean age of 50.77 ± 16.1 years (range, 13–90 years). In our study, 10 patients (12.3%) reported a history of TB com- pared with 34% to 76% reported in the literature [5]. The low rate in our series was probably due to the retrospec- tive nature of our study. Some studies suggest that male genital TB often results from direct infection from urine, while other experts suggest that haematogenous and lymphatic spread are the most common pathways of initial infection in male genital TB [2, 5]. Muneer et al.[13] thought that TBEO is caused by direct spread from the lower urinary tract or retrograde spread of M. tuberculosis via the prostate and into the epididymis, and TB of the testis is always second- ary to infection of the epididymis. However, in our study, 23 patients (28.4%) were defined as isolated testicular TB and no patients were diagnosed with prostatic TB, Huang et al. BMC Infect Dis (2021) 21:1068 Page 6 of 8 to detect latent TB infection, especially for asympto- matic individuals with high risk of TB infection [13]. However, data of GenXpert MTB/RIF and interferon-γ release assays were not available in our study due to the retrospective design. In addition, some studies reported that nucleic acid amplification (NAA) tests of the urine are helpful adjunctive tools for rapid diagnosis of renal TB, with a specificity and sensitivity of 95.6% and 98.1%, respectively. But the value of NAA tests in diagnosis of male genital TB is still controversial [5, 6]. According to some reviews, FNAC can be used to diagnose TB of the external male genitals [2, 13]. In our opinion, how- ever, FNAC should not be used for diagnosis of TBEO or testicular tumor, given the risk of fistula formation and spread of M. tuberculosis or tumor cell. Therefore, none of our patients received FNAC. TBEO in the early stages always has no specific scro- tal imaging findings [5, 13]. Ultrasound of TBEO can show diffusely or nodular enlarged hypoechoic lesions. Other features including scrotal wall and tunica albug- inea thickening, hydrocele, varicocele and intratesticular abscesses can also be seen in scrotal ultrasound. On a contrast-enhanced CT scan, TBEO can be seen as heter- ogeneous or annular enhancement, cavitation lesions or irregular mass (Fig. 2). Calcification may also be observed in advanced TB. However, these findings in scrotal imag- ing are not TB-specific and cannot be distinguished from an abscess or malignancy [2, 5, 13]. In addition, the pro- portion of cases with active pulmonary TB in our study (7.4%) was consistent with the literature [9, 13]. Male TBEO can present as a nonspecific epididymo- orchitis and testicular mass in clinical features that is dif- ficult to differentiate from nonspecific infectious diseases and malignancy [5]. Similarly, scrotal tumors (34.6%) and nonspecific bacterial epididymo-orchitis (28.4%) were the most common misdiagnosis in our study. Given the deficiency of preoperative diagnostic method with high sensitivity and specificity, histopathology of surgical spec- imens including microscopy, acid fast stain and/or PCR for M. tuberculosis remains the gold standard for diagno- sis of TBEO, especially in isolated TBEO without renal and prostate involvement [2, 8]. Borges et al. reported a case of TBEO recently, in which a surgical intervention of right epididymo-orchidectomy through the inguinal canal was performed, given the possibility of malignant neoplasm of the epididymis [18]. As reported by a review of urogenital TB, in up to one-fifth of patients, TBEO is only diagnosed after epididymo-orchidectomy and histo- pathological examination [13]. In our study, all definitive diagnosis of TBEO was confirmed by histopathology of surgical specimens. Most studies recommended anti-tuberculosis chemo- therapy as the first-line treatment for TBEO. Surgery Fig. 2 Scrotal Contrast-enhanced CT Scan Images. A contrast-enhanced CT scan showed asymmetric enlargement of the scrotum, in which the irregular mass or nodules (A, arrow), cystic lesions (B, arrow) and heterogeneous or annular enhancement (arrowheads) were observed should only be considered for patients not respond to chemotherapy and for the correction of complications [2, 5, 8]. However, the value of anti-tuberculosis chemo- therapy alone for TBEO is limited in our study since most patients have developed with complications (hydrocele, abscess, sinus or fistula) or been advanced stages when diagnosis of TBEO was confirmed due to its latent pres- entation. Moreover, successful medical treatment of TB might be hampered by drug tolerance of M. tuberculosis [19]. Recently, Goossens et al. have proposed four pos- sible mechanisms for drug tolerance of M. tuberculosis, including metabolic slowdown through reducing the metabolism and growth rate, metabolic shifting, cell wall thickening, and the upregulation of efflux pumps [19]. In our series, 15 patients (18.5%) had received regular anti-tuberculosis pharmacotherapy for 1–12 months in other medical centers before surgery but the scrotal tuberculosis lesions were not controlled effectively, and Huang et al. BMC Infect Dis (2021) 21:1068 Page 7 of 8 tuberculous complications including hydrocele, abscess and sinus were observed in 13 patients (16.0%). In this situation, radical surgery is often unavoidable, which can be used for resection of the lesions and histopathologi- cal examination [4, 6, 13, 20]. However, it is worth not- ing that Bedi et al. has successfully treated a 38-year-old patient with TBEO by standard anti-tuberculous medica- tions [14]. Besides, Abraham et al. also reported an Ecua- dorian man with TBEO who was cured with 6 months of drug therapy and no surgery was required [21]. There- fore, for patients at early stages without server compli- cations, standard anti-tuberculosis chemotherapy is still necessary to avoid surgical resection [15, 16, 22, 23]. In our study, all patients clinically diagnosed with TBEO received triple therapy of preoperative anti-tuber- culosis chemotherapy for 2–4 weeks, radical surgery and postoperative anti-tuberculosis therapy with 4 drugs (rifampicin, isoniazid, ethambutol and pyrazinamide) for 6–12 months. The preoperative anti-tuberculosis chemo- therapy was aimed to control the quantity of M. tubercu- losis in tissues and blood to prepare for surgery, while the postoperative anti-tuberculosis drug therapy was used for eradicating the remaining M. tuberculosis to prevent recurrence. Given TB of the testis is always secondary to infection of the epididymis, epididymectomy (33.3%) and epididymo-orchidectomy (64.2%) were the main sur- gical procedure for TBEO in our study. But there were still two patients (2.5%) who received simple orchiec- tomy due to the preoperative imaging suggested that the lesion was confined in testis and no epididymal involve- ment was found during operation. After standard therapy was completed, all patients recovered well and no recur- rence was observed during follow up except one patient died from liver cirrhosis. Postoperative complications including sexual dysfunction and sterility were respec- tively reported in nine (11.1%) and seven (8.6%) patients, suggesting that physicians should pay more attention to the sexual function and fertility of TBEO patients who received radical surgery during follow up. We must acknowledge several limitations of our study. Firstly, patient data such as serum ESR, CRP and gonadal hormone levels, urine AFB cultures, PCR for M. tuber- culosis, GenXpert MTB/RIF and interferon-γ release assays were not completed because of the nature of ret- rospective study. Secondly, the data of patients success- fully treated with anti-tuberculosis drugs without surgery were not available for comparative analysis. Thirdly, our study was a single-center study without sufficient data from other medical centers, which may have resulted in a certain degree of selection bias. Finally, some patients lost to follow-up even if try to contact with their rela- tives and families, which may affect the accuracy of our findings. Conclusions Given the deficiency of diagnostic tools with high sen- sitivity and specificity, radical surgery followed by histopathological examination might be unavoidable for diagnosis and treatment. We recommend patients diagnosed with advanced TBEO to receive triple ther- apy of chemotherapy-surgery-pharmacotherapy to correct complications and minimize the risk of recur- rence especially in endemic area. After triple therapy is completed, physicians should pay more attention to patients’ sexual function and fertility during follow up. Abbreviations TBEO: Tuberculous epididymo-orchitis; TB: Tuberculosis; ESR: Erythrocyte sedi- mentation rate; CRP: C-reactive protein; PCR: Polymerase chain reaction; AFB: Acid fast bacilli; FNAC: Fine-needle aspiration cytology; β-HCG: Beta-human chorionic gonadotropin; AFP: Alpha-fetoprotein; LDH: Lactate dehydrogenase; SD: Standard deviation; IQR: Interquartile range; CT: Computed tomography; MRI: Magnetic resonance imaging; MTBC: Mycobacterium tuberculosis com- plex; NAA: Nucleic acid amplification. Acknowledgements Not applicable. Authors’ contributions YH and BC collected, analyzed clinical data and wrote the manuscript. LRL, DHC and QW designed the study, supervised the project, and revised the manuscript. DHC, JL and QD assisted with detailed statistical analysis. ZYC and JBG helped with patients follow-up and interpreted the clinical data. All authors read and approved the final manuscript. Funding The collection, analysis, and interpretation of data of this study was funded by the National Natural Science Foundation of China (Grant Number 82000721) and Program from the Department of Science and Technology of Sichuan Province (Grant Number 2020YJ0054). Availability of data and materials The datasets used and/or analyzed in the current study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate This study was approved by the Ethics Committee on Biomedical Research, West China Hospital of Sichuan University. All methods were carried out in this study in accordance with relevant guidelines and regulations. Individual written informed consent for participating this study was obtained from the patients or patients’ parents or guardians. Consent for publication Not applicable. Competing interests The authors declare that they have no conflicts of interest. Author details 1 Department of Urology, Institute of Urology, West China Hospital, Sichuan University, Guoxue Alley, No. 37, Chengdu 610041, Sichuan, People’s Republic of China. 2 West China School of Medicine, Sichuan University, Chengdu, China. Received: 14 December 2020 Accepted: 1 October 2021 Huang et al. BMC Infect Dis (2021) 21:1068 Page 8 of 8 References 1. Organization WH. Global tuberculosis report 2019. Geneva: World Health Organization. 2019. https:// apps. who. int/ iris/ bitst ream/ handle/ 10665/ 329368/ 97892 41565 714- eng. pdf? ua 1. Accessed 12 June 2020. 2. Kulchavenya E, Naber K, Bjerklund Johansen TE. Urogenital Tuber- = culosis: Classification, Diagnosis, and Treatment. Eur Urol Suppl. 2016:S1569905616300471. Figueiredo AA, Lucon AM, Srougi M. Urogenital tuberculosis. Microbiol Spectr. 2017;5:1. https:// doi. org/ 10. 1128/ micro biols pec. TNMI7- 0015- 2016. 3. 5. 4. Man J, Cao L, Dong Z, Tian J, Wang Z, Yang L. Diagnosis and treatment of epididymal tuberculosis: a review of 47 cases. Peer J. 2020;8: e8291. https:// doi. org/ 10. 7717/ peerj. 8291. Jacob JT, Nguyen TM, Ray SM. Male genital tuberculosis. Lancet Infect Dis. 2008;8(5):335–42. https:// doi. org/ 10. 1016/ s1473- 3099(08) 70101-4. Thompson RW, Yu H, Dahl DM, Hurtado RM, Sajed DP. Case 10–2018: an 84-year-old man with painless unilateral testicular swelling. N Engl J Med. 2018;378(13):1233–40. https:// doi. org/ 10. 1056/ NEJMc pc171 2224. 7. Paton NI, Borand L, Benedicto J, Kyi MM, Mahmud AM, Norazmi MN, 6. 8. et al. Diagnosis and management of latent tuberculosis infection in Asia: Review of current status and challenges. Int J Infect Dis. 2019;87:21–9. https:// doi. org/ 10. 1016/j. ijid. 2019. 07. 004. Suankwan U, Larbcharoensub N, Viseshsindh W, Wiratkapun C, Chalermsanyakorn P. A clinicopathologic study of tuberculous epididymo-orchitis in Thailand. Southeast Asian J Trop Med Public Health. 2012;43(4):951–8. 9. Altiparmak MR, Trabulus S, Balkan II, Yalin SF, Denizli N, Aslan G, et al. Uri- nary tuberculosis: a cohort of 79 adult cases. Ren Fail. 2015;37(7):1157–63. https:// doi. org/ 10. 3109/ 08860 22x. 2015. 10574 60. 10. Chang A, Blackburn B, Hsieh M. Tuberculosis and parasitic infections of the genitourinary tract Campbell-Walsh urology. 11th ed. New York: Elsevier; 2016. p. 421–46. 11. Yeboah-Manu D, de Jong BC, Gehre F. The Biology and Epidemiology of Mycobacterium africanum. Adv Exp Med Biol. 2017;1019:117–33. https:// doi. org/ 10. 1007/ 978-3- 319- 64371-7_6. 12. Gupta NP, Kumar R, Mundada OP, Aron M, Hemal AK, Dogra PN et al. Reconstructive surgery for the management of genitourinary tuberculo- sis: a single center experience. J Urol. 2006;175(6):2150–4. doi: https:// doi. org/ 10. 1016/ s0022- 5347(06) 00310-7 (discussion 4). 13. Muneer A, Macrae B, Krishnamoorthy S, Zumla A. Urogenital tuberculosis — epidemiology, pathogenesis and clinical features. Nat Rev Urol. 2019. 14. Bedi N, Rahimi MNC, Menzies S, Kalsi J. Atypical testicular pain. BMJ Case Rep. 2019;12:2. https:// doi. org/ 10. 1136/ bcr- 2018- 226697. 15. Hadadi A, Pourmand G, Mehdipour-Aghabagher B. Unilateral testicular tuberculosis: case report. Andrologia. 2012;44(1):70–2. https:// doi. org/ 10. 1111/j. 1439- 0272. 2010. 01097.x. 16. Kinnear N, Hoh I, Campillo P, Bolt J. Tuberculous epididymo-orchitis masquerading as acute scrotum. BMJ Case Rep. 2016. https:// doi. org/ 10. 1136/ bcr- 2015- 214060. 17. Altintepe L, Tonbul HZ, Ozbey I, Guney I, Odabas AR, Cetinkaya R, et al. Urinary tuberculosis: ten years’ experience. Ren Fail. 2005;27(6):657–61. https:// doi. org/ 10. 1080/ 08860 22050 02348 57. 18. Borges WM, Bechara GR, de Miranda MML, de Figueiredo GB, Venturini BA, Laghi CR. Epididymis tuberculosis: case report and brief review of the literature. Urol Case Rep. 2019;26: 100969. https:// doi. org/ 10. 1016/j. eucr. 2019. 100969. 19. Goossens SN, Sampson SL, Van Rie A. Mechanisms of Drug-Induced Tolerance in Mycobacterium tuberculosis. Clin Microbiol Rev. 2020;34:1. https:// doi. org/ 10. 1128/ cmr. 00141- 20. 20. Cek M, Lenk S, Naber KG, Bishop MC, Johansen TE, Botto H, et al. EAU guidelines for the management of genitourinary tuberculosis. Eur Urol. 2005;48(3):353–62. https:// doi. org/ 10. 1016/j. eururo. 2005. 03. 008. 21. Abraham S, Izaguirre Anariba DE, Dua K, Mir M, Ankireddypalli A. A case of testicular tuberculosis mimicking malignancy in a healthy young man. Ther Adv Infect Dis. 2016;3(3–4):110–3. https:// doi. org/ 10. 1177/ 20499 36116 655261. 22. Ravikanth R, Kamalasekar K, Patel N. Extensive primary male genital tuberculosis. J Human Reprod Sci. 2019;12(3):258–61. https:// doi. org/ 10. 4103/ jhrs. JHRS_3_ 19. 23. Klebanov N, Raghavan A. Tuberculous orchitis following intravesical Bacille Calmette-Guérin (BCG) therapy. Cureus. 2018;10(5): e2703. https:// doi. org/ 10. 7759/ cureus. 2703. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12879-021-05907-0	Johnson et al. BMC Infectious Diseases (2021) 21:215 https://doi.org/10.1186/s12879-021-05907-0 R E S E A R C H A R T I C L E Open Access Monitoring HIV testing and pre-exposure prophylaxis information seeking by combining digital and traditional data Derek C. Johnson1,2, Alicia L. Nobles1,2, Theodore L. Caputi2,3, Michael Liu4, Eric C. Leas2,5, Steffanie A. Strathdee1, Davey M. Smith1 and John W. Ayers1,2 Abstract Background: Public health is increasingly turning to non-traditional digital data to inform HIV prevention and control strategies. We demonstrate a parsimonious method using both traditional survey and internet search histories to provide new insights into HIV testing and pre-exposure prophylaxis (PrEP) information seeking that can be easily extended to other settings. Method: We modeled how US internet search volumes from 2019 for HIV testing and PrEP compared against expected search volumes for HIV testing and PrEP using state HIV prevalence and socioeconomic characteristics as predictors. States with search volumes outside the upper and lower bound confidence interval were labeled as either over or under performing. State performance was evaluated by (a) Centers for Disease Control and Prevention designation as a hotspot for new HIV diagnoses (b) expanding Medicaid coverage. Results: Ten states over-performed in models assessing information seeking for HIV testing, while eleven states under- performed. Thirteen states over-performed in models assessing internet searches for PrEP information, while thirteen states under-performed. States that expanded Medicaid coverage were more likely to over perform in PrEP models than states that did not expand Medicaid coverage. While states that were hotspots for new HIV diagnoses were more likely to over perform on HIV testing searches. Conclusion: Our study derived a method of measuring HIV and PrEP information seeking that is comparable across states. Several states exhibited information seeking for PrEP and HIV testing that deviated from model assessments. Statewide search volume for PrEP information was affected by a state’s decision to expand Medicaid coverage. Our research provides health officials with an innovative way to monitor statewide interest in PrEP and HIV testing using a metric for information- seeking that is comparable across states. Keywords: Google trends, HIV, PrEP, Internet, HIV testing Correspondence: dcjohnson@ucsd.edu 1Department of Medicine, Division of Infectious Diseases and Global Public Health, University of California San Diego, 9500 Gilman Dr, La Jolla, California 92093, USA 2The Center for Data Driven Health at the Qualcomm Institute, University of California San Diego, La Jolla, California, USA Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Johnson et al. BMC Infectious Diseases (2021) 21:215 Page 2 of 7 Background As people increasingly turn to digital sources of news and information, online activity has the potential to become a window into the public’s consciousness [1]. Measuring the public’s online information seeking has the potential to predict health behavior, as what people are searching the internet for can be predictive of what they intend to do in the future [2]. It is possible that seeking information about HIV testing and Pre-Exposure Prophylaxis (PrEP) online could be a new surveillance tool in the fight against HIV. Previous studies have shown a spike in internet searches for HIV testing has corresponded with increases in HIV testing, suggesting that seeking HIV testing information online could be predictive of testing behavior [3]. Utilizing internet searches could be a way to enhance the surveil- lance of the public’s interest in seeking information on HIV and HIV health seeking behavior. Past efforts to enhance HIV surveillance relied mostly on upscaling trad- itional data (e.g., clinical records or surveys) that have in- trinsic shortcomings, such as a limited ability to provide current information. For instance, the most recent data for HIV testing on AIDSvu.org is from 2016 and the most recent AIDSvu.org data on PrEP usage is from 2018 [4]. These limitations have driven public health to increasingly turn to digital data, such as news, social media, and inter- net searches, to learn how people seek HIV information [5–8]. For example, internet search trends can be used to investigate public interests as evident by actor Charlie Sheen’s HIV positive disclosure concurring with record levels of Google searches for HIV awareness, HIV testing, and condoms [3]. This finding was valid, as it was later confirmed by traditional data after a 16 month delay [7]. Internet search histories have potential utility for assessing both help-seeking behavior regarding public interest in PrEP for HIV prevention and for HIV testing. For ex- ample, one study conducted in Hong Kong found that a direct relation between HIV news trends and online search behavior for issues regarding HIV/AIDS and men who have sex with men (MSM) [9]. Other studies have found that areas with high levels of HIV prevalence have greater internet search volumes for HIV related terms then areas of low HIV prevalence [10]. These studies show that the use of internet search histories combined with traditional surveillance data has the potential to create synergies that can yield new insights into HIV related health behavior. Our study methods use both internet search histories and traditional survey data to provide new insights on information seeking for HIV testing and PrEP informa- tion that can be easily replicated and extended to other settings and outcomes. Specifically, we predicted ex- pected internet search volumes for HIV testing and PrEP based on statewide HIV prevalence and socioeconomic (SES) factors and compared them to observed search volumes in a model that allows us to identify if US states over or under perform against expectations. Moreover, we evaluated how state performance varied by (a) states that are designated as hotspots for new HIV diagnoses (b) states that received Medicaid expansion funding. Methods Our study used data from multiple sources. 1) We ob- tained the most current state-level prevalence of HIV from the Center for Disease Control and Prevention HIV surveillance report [11], which is from 2018. HIV prevalence was chosen over HIV incidence because we were interested in look at the association between the total number of HIV cases in a state and internet searches for HIV testing and PrEP 2) The following state-level socioeconomic attributes we obtained from the 2018 Center for Disease Control and Prevention’s Behavioral Risk Factor Surveillance System (BRFSS): proportion of males, white non-Hispanics, people aged 45 years or older, and people with household income over $50,000 [12]. 3) We obtained 2019 state-level an- nual internet search volumes for HIV testing and PrEP from using the Google Trends API. Information avail- able through this API includes the volume of searches for each term, the number of searches per unit of time, and the geographic location of the searches (country, re- gion, state, city, metropolitan area). Search volume data was calculated as a query fraction of the proportion of searches of a specific search term relative to all searches measured per 10 million searches. Standardizing search volumes was done in order to account for population sizes. We defined HIV testing searches as any query that included the terms “HIV” and “test,” “tests,” or “testing”, “AIDS test”, or “oraquick”. We defined PrEP searches as any query that included the terms “PrEP” and “HIV” or “pre-exposure prophylaxis HIV” or “Truvada” or “Descovy”. Internet search volumes are withheld by Google for states where searches do not achieve a minimum thresh- old of searches. As a result, we could not obtain search data for HIV testing for five states (Alaska, Montana, South Dakota, Vermont, and Wyoming). PrEP search data could not be obtained for two states (Vermont and Wyoming). 4) We obtained data on states that expanded Medicaid coverage from the Kaiser Family Foundation [13]. Our analysis followed a four-step process. First, we fit Poisson regression models with state-level HIV preva- lence data and state-level socioeconomic attributes to predict the expected internet search volumes of HIV testing and PrEP for each state. Second, we fit a centered least squares regression line of expected search volumes from our models versus observed search volumes from Google Trends. Third, we compared the expected search Johnson et al. BMC Infectious Diseases (2021) 21:215 Page 3 of 7 volumes from our models in step one with the observed search volumes from Google Trends for each state in order to assess the level of information seeking for HIV testing and PrEP by calculating the percent difference between the observed and expected values of the cen- tered least squares regression (i.e., (observed-fitted)/fit- ted * 100%). States with observed information seeking (measured by observed internet search volumes for HIV testing and PrEP) greater than their expected information seeking (predicted internet search volumes by the Poisson re- gression model) were considered to be over performing and exhibit greater information seeking for HIV testing and PrEP than expected given their prevalence of HIV. States that were over performing above the 95% confi- dence interval were highlighted in our results (see ex- ample of plotting observed vs. expected observations in Fig. 1). Similarly, states with observed information seek- ing less than their expected information seeking were considered to be underperforming and exhibit less infor- mation seeking for HIV testing and/or PrEP than ex- pected given their prevalence of HIV. States that were under performing below the 95% confidence interval were highlighted in our results To describe statistical uncertainty between expected and observed search vol- umes, we used bootstrap sampling to calculate the 95% confidence interval (CI) for the regression line and la- beled observations outside the confidence band as states that over or under performed. Fourth, to understand which states typically under or over performed we contrasted the deviations in expected searches against (a) What states were designated as hot- spots for new HIV diagnoses by the Centers for Disease Control and Prevention (CDC) [11], (b) What states re- ceived Medicaid expansion funding that covered HIV testing and PrEP [13]. Results We observed different levels of information seeking for HIV testing and PrEP across states (Fig. 1). Ten states over-performed for HIV testing searches. Georgia exhib- ited the greatest difference with 36.8% more searches than expected followed closely by Rhode Island (35.2%), then Indiana (28.8%), Pennsylvania (22.9%), Nevada (18.6%), Florida (18.0%), Louisiana (17.3%), Washington (15.3%), Iowa (13.4%), and Virginia (10.0%) (Table 1). Conversely, eleven states under-performed for HIV test- ing. New Hampshire exhibited the greatest difference with − 34.1% less searches than expected, followed by Maine (− 32.2%), Idaho (− 26.1%), Nebraska (− 21.9%), Oregon (− 20.5%), New Mexico (− 18.5%), Mississippi (− 16.8%), Arkansas (− 15.7%), Alabama (− 13.6%), Massa- chusetts (− 12.4%), Arizona (− 10.3%), Fig. 1 Observed vs. Expected HIV Testing and PrEP Internet Searches as Compared to a Hypothetical Perfectly Fitting Regression Line Johnson et al. BMC Infectious Diseases (2021) 21:215 Page 4 of 7 Table 1 Statewide Proportional Differences Between Observed and Expected Internet Search Volumes for Information Seeking About HIV testing and PrEP U.S State Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware District of Columbia Florida Georgia Hawaii Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland HIV prevalence per 100,000 individuals 394 136.6 326.6 278.7 451.9 305 371.9 461.4 2515.5 691.8 745.6 229.4 94.1 384.7 248.6 128.5 154.6 239.2 654.4 158 723.8 Massachusetts 385.2 Michigan Minnesota Mississippi Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico New York 223.1 209.9 456.9 281.9 83.7 161.3 501.4 119 511.3 239.9 822.7 North Carolina 411.4 North Dakota^ 101.0 Ohio Oklahoma Oregon 270.1 229.2 227 Pennsylvania 361.8 Rhode Island 318.4 South Carolina 480.8 HIV Search Differences −13.6* NA −10.3* −15.7* −9.3 7 −2.9 −5.6 −10.5 18# 36.9# −11.3 −26.1* −5.0 28.8# 13.5# −0.7 1.0 17.3# −32.2* 6.7 −12.4* 5.4 −3.5 −16.8* 1.7 NA −21.9* 18.6# −34.1* −2.6 −18.5* 0.5 −5.8 21.6 5.4 2.3 −20.5* 22.9# 35.2# 6.7 PrEP Search Differences −6.4 −2.3 14.1# −1.6 4.8 10.7# 4.1 −7.7* −1.9 −5.3 0.4 −1.1 − 31* 12.6# −5.0 −17.2* −13.3* 10.2# 3.9 11.5 −5.2 23.6# −9.4* −5.0 0.1 −6.1* −28.2* 16.9# 18# −23.6* −15.9* −1.6 12.1# −8.4* 12.8 0.1 −14.1* 12.4# 17.3# 29.2# −23.1* Expanded Medicaid Coverage@ Designated Hotspot for New HIV Infections Y Y Y Y Y Y Y Y Y N N Y N Y Y Y N Y Y N Y Y Y Y N N Y Y Y Y Y Y Y N Y Y N Y Y Y Y N N N N Y N N N Y Y Y N N Y Y N N Y Y N Y Y Y N N N N N N N Y N Y Y N Y N N Y N N Johnson et al. BMC Infectious Diseases (2021) 21:215 Page 5 of 7 Table 1 Statewide Proportional Differences Between Observed and Expected Internet Search Volumes for Information Seeking About HIV testing and PrEP (Continued) U.S State HIV prevalence per 100,000 individuals HIV Search Differences PrEP Search Differences Expanded Medicaid Coverage@ Designated Hotspot for New HIV Infections South Dakota 104.9 Tennessee Texas Utah Vermont Virginia 352.1 471.3 139.7 149.9 365.9 Washington 245 West Virginia 146.7 NA −4.3 7 −3.4 NA 10# 15.3# 8.4 −7.9 7.7 2.8 −9.3 −2.9 NA −6.4 19.9# 30.7# −27.9* N Y Y N Y N Y Y 83.5 148.3 Wisconsin Wyoming^ #Over performed in search models *Underperformed in search models @State decisions on the Affordable Care Act’s Medicaid expansion ^HIV prevalence for 2018 was missing for Wyoming and North Dakota and was substituted with HIV prevalence from 2017 NA NA N N N Y Y N N N Y N N N (10.2%). Conversely, Thirteen states over-performed for PrEP searches (Table 1). West Virginia exhibited the greatest difference with 30.7% more searches than expected, followed by Rhode Island (29.2%), Massachusetts (23.6%), Washing- ton (19.9%), Nevada (18.0%), Pennsylvania (17.3%), Neb- raska (16.9%), Arizona (14.1%), Illinois (12.6%), Oregon (12.4%), New York (12.1%), Colorado (10.7%), and Ken- tucky thirteen states under- performed for PrEP searches. Idaho exhibited the great- est difference with 30.9.% less searches than expected, followed by Montana (− 28.2%), Wisconsin (− 27.9%), New Hampshire (− 23.6%), South Carolina (− 23.1%), Iowa (− 17.2%), New Jersey (− 15.9%), Oklahoma (− 14.1%), Kansas (− 13.3%), Michigan (− 9.4%), North Car- olina (− 8.4%), Delaware (− 7.7%), and Missouri (− 6.1%). States that over or under performed on HIV testing searches did not necessarily do likewise for PrEP searches (r = 0.12). For instance, Nebraska ranked 7th for excess HIV testing searches, but then ranked 43rd for PrEP searches. Four states (Washington, Nevada, Pennsylvania, and Rhode Island) over-performed for both HIV testing and PrEP searches, while only 2 states (New Hampshire and Idaho) under-performed for both. States that expanded Medicaid coverage were more likely to over perform more on PrEP searches compared to states that did not expand Medicaid coverage (z = 2.04, p < 0.041). States that were hotspots for new HIV diagnoses were more likely to over perform on HIV test- ing searches than states that were not hotspots for new HIV diagnoses (z = 2.08, p < 0.037). Discussion Our study derived a method of measuring HIV testing and PrEP information seeking that is comparable across states. Several states exhibited information seeking for PrEP and HIV testing that deviated from what was ex- pected in our models. A state’s performance in our models was not affected by its designation as a hotspot for new HIV infections. However, performance for PrEP information seeking was associated with a state’s deci- sion to expand Medicaid coverage. By integrating inter- net search histories and traditional survey data, our results provide baseline benchmarks for monitoring statewide interest in seeking information on HIV testing and PrEP. Our research demonstrates a need for increased access to PrEP information, particularly among states that have not expanded their Medicaid coverage. Lower interest in seeking information on PrEP for states that did not ex- pand Medicaid coverage could be detrimental to increas- ing PrEP utilization given that insurance coverage affects PrEP uptake [14, 15]. Approximately 12% of PrEP users receive PrEP through Medicaid [16] and the refusal to extend coverage could deny people the ability to access PrEP. Our results, coupled with the inability to utilize PrEP due to a lack of health insurance, is a potentially disastrous combination that could result in an increase in HIV prevalence in states that underperformed in our PrEP models. Underperformance in PrEP models could be due to the unequal distribution of PrEP across different gen- ders, ages, and states. Our models control for age, sex, race, and income at the state level using BRFSS data. However, our models do not adjust for disparities in the distribution of PrEP. It is possible that PrEP interven- tions that do not specifically target key populations with indications for PrEP use could result in these neglected populations not searching for PrEP information on line, Johnson et al. BMC Infectious Diseases (2021) 21:215 Page 6 of 7 which would result in an underperformance in our PrEP models. For example, five states represented 50% of PrEP prescriptions and although women represent almost 20% of new HIV infections, they represented only 7% of PrEP prescriptions [16, 17]. These types of underlying dispar- ities in PrEP distribution could possibly be factors influ- encing how people look for information on PrEP. Our research suggest the possibility that increased at- tention to HIV testing, promoted by a state being listed as a CDC hotspot for new HIV diagnoses, does in fact result in increased public interest in seeking HIV testing information [11]. States that are listed as a hotspot for new HIV infections receive a rapid infusion of additional resources, expertise, and technology to develop and im- plement locally tailored HIV interventions [18]. It is pos- sible that the increased promotion of HIV interventions results in more public interest in seeking HIV testing in- formation. This would explain why states that were listed as hotspots for new HIV diagnoses were more likely to over perform on HIV testing searches than states that were not hotspots. Our results support pro- viding states with more resources to promote HIV test- ing, given that our models suggest increases in searches for HIV testing are correlated with more CDC support for HIV programs. Our study benefits from several strengths. We use a nationally representative survey to control for several SES covariates, ensuring that the US population is accur- ately represented. Because our methods adjusted for baseline state level SES characteristics, leaders in each individual state can use these methods to evaluate their state-specific progress. Our internet search volume data is measured in real-time, and while we used annual esti- mates, it is possible to use the same method to estimate weekly or monthly search volumes. Most importantly, our research presents a new method for surveillance and performance monitoring in HIV prevention. Our research is not without limitations. Internet search volume data is aggregated and is susceptible to ecological confounding. Additionally, it cannot be used to determine which racial/ethnic, gender, or age groups are or are not engaged with HIV testing or PrEP. While it is possible that adding more search terms could affect our results, the ef- fects of adding additional search terms to our models di- minishes after the most common search terms are added, as these terms make up the vast majority of search terms that the public uses. To insure we were using the most common search terms for HIV testing and PrEP, we con- sulted with HIV experts on HIV testing and PrEP nomen- clature. Using search data may be subject to selection bias, as not all people access the Internet equally and although some queries may reflect general curiosity rather than treatment-seeking, it is well known that internet search trends mirror many health-related behaviors [1]. Conclusions Our results are a call-to-action for underperforming states whose populations are not engaged in searching for information on HIV testing and PrEP. Our research provides health officials with an innovative way to moni- tor statewide interest in PrEP and HIV testing by highlighting the states that demonstrate the least online information seeking, which is critical for the promotion of HIV testing and PrEP as a way to help end the HIV epidemic. Further research should examine why certain states are deficient, and policy makers in deficient states should make efforts to expand HIV testing and PrEP promotion, perhaps by replicating the interventions and policies of better-performing states. Abbreviations AIDS: acquired immunodeficiency syndrome; CDC: Centers for Disease Control and Prevention; CI: confidence interval; HIV: human immunodeficiency virus; PrEP: pre-exposure prophylaxis; SES: socioeconomic Acknowledgements The content of this research is solely the responsibility of the authors and does not necessarily represent the official views of the California HIV/AIDS Research Program Office or National Institute of Drug Abuse. Authors’ contributions (last name of each author is listed under what they contributed to) Study Conception and Design Johnson, Ayers, Nobles, Leas Acquisition and Preparation of Data Johnson, Caputi, Liu Analysis and Interpretation of Data Johnson, Nobles, Strathdee, Smith Drafting of Manuscript Johnson, Nobles, Caputi, Liu, Leas, Strathdee, Smith, Ayers Critical Revisions/Revising Johnson, Ayers, Nobles, Leas All authors read and approved the final manuscript. Funding This research was supported by funds from the California HIV/AIDS Research Program Office of the University of California (OS17-SD-001) and the National Institute of Drug Abuse (T32 DA023356, R37 DA019829). Availability of data and materials All data is publically available through Google Trends and through The Behavioral Risk Factor Surveillance System (BRFSS). Ethics approval and consent to participate Our research was exempted from an ethics review by the University of California at San Diego Human Research Protections Program. Competing interests None of the authors declares any conflicts of interest. Author details 1Department of Medicine, Division of Infectious Diseases and Global Public Health, University of California San Diego, 9500 Gilman Dr, La Jolla, California 92093, USA. 2The Center for Data Driven Health at the Qualcomm Institute, University of California San Diego, La Jolla, California, USA. 3Department of Health Sciences, University of York, York, UK. 4University of Oxford, Oxford, UK. 5Department of Family Medicine and Public Health, Division of Health Policy, University of California San Diego, La Jolla, California, USA. Johnson et al. BMC Infectious Diseases (2021) 21:215 Page 7 of 7 Received: 8 October 2020 Accepted: 16 February 2021 References 1. Asur S, Huberman BA. Predicting the future with social media. In: Proceedings - 2010 IEEE/WIC/ACM International Conference on Web Intelligence, WI 2010. ; 2010. doi:https://doi.org/10.1109/WI-IAT.2010.63 Goel S, Hofman JM, Lahaie S, Pennock DM, Watts DJ. Predicting consumer behavior with web search. Proc Natl Acad Sci U S A. 2010. https://doi.org/1 0.1073/pnas.1005962107. Ayers JW, Althouse BM, Dredze M, Leas EC, Noar SM. News and internet searches about human immunodeficiency virus after Charlie Sheen’s disclosure. JAMA Intern Med. 2016. https://doi.org/10.1001/jamainternmed.2 016.0003. AIDSVu (aidsvu.org). Emory University, Rollins School of Public Health. Accessed 10 Mar 2020. Ayers JW, Althouse BM, Dredze M. Could behavioral medicine lead the web data revolution? JAMA. 2014. https://doi.org/10.1001/jama.2014.1505. Nobles AL, Leas EC, Latkin CA, Dredze M, Strathdee SA, Ayers JW. #HIV: Alignment of HIV-Related Visual Content on Instagram with Public Health Priorities in the US. AIDS Behav. 2020. doi:https://doi.org/10.1007/s10461-01 9-02765-5 Young SD, Yu W, Wang W. Toward automating HIV identification: machine learning for rapid identification of HIV-related social media data. J Acquir Immune Defic Syndr. 2017. https://doi.org/10.1097/QAI.0000000000001240. Allem JP, Leas EC, Caputi TL, et al. The Charlie sheen effect on rapid in- home human immunodeficiency virus test sales. Prev Sci. 2017. https://doi. org/10.1007/s11121-017-0792-2. Chiu, A. P. Y., Lin, Q., & He, D. (2017). News trends and web search query of HIV/AIDS in Hong Kong. PLoS One https://doi.org/https://doi.org/10.1371/ journal.pone.0185004. 2. 3. 4. 5. 6. 7. 8. 9. 10. Mahroum, N., Bragazzi, N. L., Brigo, F., Waknin, R., Sharif, K., Mahagna, H., Amital, H., & Watad, A. (2019). Capturing public interest toward new tools for controlling human immunodeficiency virus (HIV) infection exploiting data from Google trends. Health Informatics Journal https://doi.org/https:// doi.org/10.1177/1460458218766573. 11. Department of Health and Human Services, Centers for Disease Control and Prevention. HIV Surveillance Report 2020. https://www.cdc.gov/hiv/library/ reports/hiv-surveillance.html. 12. Centers for Disease Control and Prevention (CDC). Behavioral Risk Factor 13. Surveillance System Survey Data. Atlanta, Georgia: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, 2018. Kaiser Family Foundation. “Status of State Medicaid Expansion Decisions: Interactive Map”. https://www.kff.org/medicaid/issue-brief/status-of-state- medicaid-expansion-decisions-interactive-map/. Accessed April 1st, 2020. 14. Patel RR, Mena L, Nunn A, et al. Impact of insurance coverage on utilization of pre-exposure prophylaxis for HIV prevention. PLoS One. 2017. https://doi. org/10.1371/journal.pone.0178737. 15. Doblecki-Lewis S, Liu A, Feaster D, et al. Healthcare access and PrEP continuation in San Francisco and Miami after the US PrEP demo project. In: Journal of Acquired Immune Deficiency Syndromes. ; 2017. doi:https://doi. org/10.1097/QAI.0000000000001236. 16. Huang YLA, Zhu W, Smith DK, Harris N, Hoover KW. Hiv preexposure prophylaxis, by race and ethnicity — United States, 2014–2016. Morb Mortal Wkly Rep. 2018. doi:https://doi.org/10.15585/MMWR.MM6741A3 17. AIDSVu (aidsvu.org). Mapping PrEP, First Ever Data on PrEP Users Across the U. S. Emory University, Rollins School of Public Health. https://aidsvu.org/prep/. Accessed 10 Mar 2020. 18. U.S. Department of Health and Human Services. 2021. HIV National Strategic Plaen for the United States: a roadmap to end the epidemic 2021–2025. Washington, DC. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12877-021-02434-0	Backman et al. BMC Geriatrics (2021) 21:498 https://doi.org/10.1186/s12877-021-02434-0 R E S E A R C H A R T I C L E Open Access Characteristics of nursing home units with high versus low levels of person-centred care in relation to leadership, staff- resident- and facility factors: findings from SWENIS, a cross-sectional study in Sweden Annica Backman1* , Per-Olof Sandman1,2 and Anders Sköldunger1,3 Abstract Background: The context of care consists of factors that determines the extent to which staff can offer person- centred care. However, few studies have investigated factors that can explain variation in levels of person-centred care among nursing home units. The aim of this study was to explore factors characterizing nursing home units with high and low degree of person-centred care, with focus on leadership, staff, resident and facility factors. Methods: Cross-sectional data from residents, staff, and managers in 172 randomly selected nursing homes in Sweden were collected in 2014. Activities of Daily Living Index, Gottfries’ cognitive scale, Person-centred Care Assessment Tool together with demographic information and estimations of leadership engagement was used. Independent samples t-test and Chi2 test were conducted. Results: Highly person-centred units were characterised by leaders engaging in staff knowledge, professional development, team support and care quality. In highly person-centred units’ staff also received supervision of a nurse to a larger extent. Highly person-centred units were also characterised as dementia specific units, units with fewer beds and with a larger proportion of enrolled nurses. No differences in degree of person-centred care were seen between public or private providers. Conclusions: This study provides guidance for practitioners when designing, developing and adapting person- centred units in aged care contexts. Managers and leaders have an important role to promote the movement towards a person-centred practice of care, by supporting their staff in daily care, and engaging in staff knowledge and professional development. Targeting and adjusting environmental factors, such as provide small and dementia adapted environments to match the residents’ personal preferences and capacity are also important when striving towards person-centredness. Keywords: Person centred care, Physical environment, Leadership, Nursing management, Nursing homes, Organisation of care * Correspondence: annica.backman@umu.se 1Department of Nursing, Umeå University, SE-901 87 Umeå, Sweden Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Backman et al. BMC Geriatrics (2021) 21:498 Page 2 of 11 Background Moving away from task-oriented models of care, towards a more person-centred practice is now guiding the provision of care in nursing homes worldwide. The care context has been described as a decisive factor for the extent to which staff can provide person-centred care, but only a few studies have empirically investigated which factors define nursing home units as being more or less person-centred. To gain insight in determinants enhancing person-centred care with respect to personal factors (leadership, staffing and resident) as well as structural (facility) is essential. The concept person-centred care is commonly used to indicate a holistic view of the person in need of care, re- specting subjective experiences, values, needs and prefer- ences [1–3]. Studies has shown that residents may benefit from this approach. Among persons living with dementia, it has been shown that person-centred care has been associated to higher quality of life [4, 5]. A person-centred intervention including staff education, environmental adaptation and a variety of daily activities for residents was shown to improve well-being among residents [6]. Also, Dementia Care Mapping (DCM) in- creased well-being among residents and reduced depres- sive symptoms in a person-centred intervention by Rokstad et al. [7]. A cluster-randomized trial showed that DCM and staff training in dementia care were asso- ciated with reduced agitation among nursing home resi- dents [8]. However, opposite findings are also evident in previous literature as person-centred intervention stud- ies with no or negative effect also has been reported. An individualised tailored intervention did not significantly life among residents [9] and an change quality of activity-oriented intervention showed no reduction in anxiety [10]. An increased number of falls accidents among residents were also reported from a person- centred intervention, while DCM reduced resident falls [8]. Positive impact on staff health and work situation has also been reported in previous literature. Working in a person-centred manner has been associ- job satisfaction [11–14], ated with higher degree of lower degree of higher psychosocial climate [13] stress of conscience and lower degree of job strain [13, 15]. Taken together, available evidence indicates that person-centred care has predominantly beneficial outcomes wereas why person-centred practices have been declared by WHO as a global strategy to ad- dress the variety of care needs due to an aging popu- lation [1, is increasingly embraced and recommended by aged care providers, practitioners and the research society as the “gold standard” model of care [1, 10, 17, 18], there still remain challenges in putting a person- centred philosophy into practice [11]. 16] Although person-centred care Quality of leadership has been described as having po- tential to support or hinder person-centred practices [2, 19]. Until recently, a large amount of PCC intervention publications have, besides focusing on implementing PCC, and also highlighted the significance of leadership [20–23]. An intervention study using DCM have re- ported that managers who take an active part in the care practice, have clear visions, are supportive and act as role models with a leadership based on a person-centred philosophy, beneficially influenced the implementation of PCC in nursing homes [20]. Other studies have also reported that managers who promotes interpersonal re- lationships, communicating with staff with sensitivity, in- clusion and respect affect implementation of a Person- Centred Care program [21] and a DCM intervention in a positive way [22]. One study showed that leadership is positively associated to degree of PCC in Swedish nurs- ing homes [24]. Furthermore, staff in nursing home units in Sweden offering a high degree of PCC seem to be more satisfied with the leadership than units with low degree of PCC [25]. However, managerial obstacles have also been reported in previous studies where manager and leader resistance to change have been indicated as a barrier to enable person-centred practice [26]. A cluster randomized controlled trial of person-centred residential care by Chenoweth et al. [23] reported that the imple- mentation effectiveness was low when the manager fo- cused more on organisational than on enabling resident comfort and pleasure. In contrast, Jeon et al. [27] was not able to see any changes in terms of person-centred care after a 12-month leadership devel- opment programme, in their controlled trial. A system- atic literature review concluded that leadership is a vital part of the implementation process in nursing, but re- search has still not specified in what way, and therefore suggests that more research is needed to explore the possible role of the leader [28]. Thus, although the qual- ity of leadership has been described to have high poten- tial to support or hinder person-centred practices [2, 19], there seems to be a limited consensus on leadership determinants enabling successful delivery of PCC. efficiency factors are critical The theoretical discourse has postulated that the or- ganisational and contextual for person-centred processes [2, 19]. A factor that has been pointed out as an important prerequisite for PCC, is staff competence [2]. A Canadian study showed that demo- graphic characteristics of the staff (e.g., age, education, experience, job classification, ethnicity, work status) did not influence the provision of PCC, neither did facility characteristics (e.g., facility size, presence of a union, managers experience) [29]. In contrast, a Swedish study showed that units with high PCC had a more adapted environment and higher degree of staff educated for care of persons with dementia. Furthermore, staff in units Backman et al. BMC Geriatrics (2021) 21:498 Page 3 of 11 with higher degree of PCC to a larger extent received regular supervision as well as reported satisfaction with leadership compared to units with lower degree of PCC [25]. In terms of other contextual factors such as owner- ship of the facility, it has been shown that public nursing homes has been related to higher quality in terms of higher staffing levels and offering individual accommo- dation/kitchen, but scored lower in terms of processual quality such as user participation, updated care plans and medication reviews, when comparing with private nursing homes [30]. Although the way leadership is performed can play an important role in PCC processes, there seems to be a lack of scientific consensus underpinning this assump- tion. Although the context of care delivery is increas- ingly recognised, organisational and contextual actors facilitating person-centred care are mainly based on the- oretical models, with limited empirical knowledge on which factors actually characterise nursing home units offering different degree of PCC. Thus, it’s essential to gain insight in determinants enhancing person-centred care with respect to personal factors (leadership, staffing and resident) as well as structural therefor seems essential. The aim of this study was to explore factors characterizing high and low person-centred nurs- ing home units, with focus on leadership, staff, resident and facility variables. (facility) Methods Design The present study is part of the Swedish National Inven- tory of Care and Health in Residential Aged Care (SWENIS), a nationwide randomized longitudinal pro- ject with explorative design within The Umeå ageing and health research programme (U-AGE) [31]. The U- AGE research programme provides translational know- ledge on the structure, content and outcomes of person- centred care and health-promoting living conditions in nursing homes for older people and people with demen- tia. The cross-sectional data for this study was collected between November 2013 and September 2014. Sampling Out of 290 Swedish municipalities, a random selection of 60 municipalities was invited to participate in the pro- ject. Of the 60 municipalities 35 agreed to participate, contributing with data from 172 nursing homes. The final sample included data from staff n = 3605 (response rate 66.5%), residents n = 4831 (response rate 70%) and managers (n = 191). As this study is part of a large re- search programme [31], a more detailed description of sampling and data collection can found be in previous publications [24, 31]. Data collection Data was collected using a three-part survey that were sent out to the invited nursing homes. A self-reported staff survey comprising demographic information to- gether with estimations of leadership engagement and person-centred care. All direct care staff working day/ evening shifts with long-term employments were invited to participate. A resident survey collected demographic information about the residents together with assess- ments of the residents ADL and cognitive status. This survey was completed by the staff member who knew the resident best, their primary carer, through proxy- rating. Each primary carer commonly assessed one resi- dents each. The third survey was an organisational sur- vey leadership and organisational characteristics about the nursing home. This part was completed by the managers. consisting of questions about training/education (< 1 year) with one level Characteristics of workforce and study context In Sweden, nursing home managers have the operational responsibility for the care of residents, direct care staff and the work environment [32] and a qualification of so- cial work or nursing care seems to be most common al- though no formal education is required to hold a managerial position in Sweden [33]. Registered nurses are responsible for the nursing care provision and med- ical care [34]. The direct care staff consists primarily of enrolled nurses and nurse assistants [34] and they are re- sponsible for providing personal care and social services to residents [35]. Direct care staff consists primarily of enrolled nurses who have upper secondary level school- ing (up to 3 years of training), with a level 4 qualification in the European Qualifications Framework [36] and nurses’ assistant who have approximately, half the length of lower qualification in EQF [37–40]. The European Qualifica- tions Framework is a translation device explicating qualification requirements within different educations and training systems in Europe [36]. Regular tasks for nurse assistants includes care assistance, making beds, helping patients with nutrition and hygiene, while en- rolled nurses in addition also tests glucose, temperature, pulse, respiration and weight, carrying out simple chan- ging bandages, conducting simple laboratory tests and giving medication on delegation from reg. nurse [39]. Swedish nursing homes are defined as housing for people 65 years and older, who are no longer able to live at home [41]. About 82,000 residents resides in nursing homes due to extensive personal care needs and/or cog- nitive impairment [37, 42]. The number of beds in mu- nicipal aged care in Sweden has decreased from 120,000 since 2000 although the proportion of older persons is growing and in 2014 the beds were 108,835 [43]. Swed- ish aged care is mainly funded publicly and essentially Backman et al. BMC Geriatrics (2021) 21:498 Page 4 of 11 publicly produced, and the specification of the national policies postulates that older persons should have the possibility to live independently with high quality of life and furthermore that high-quality care should be pro- vided to older persons in need of care [44]. Study variables Demographic data including staff characteristics (sex, age, qualifications, work experience) and resident char- acteristics (sex, age) as well as organisational characteris- tics of the facility (number of beds, SCU/general unit, private/public provider). Leader engagement and support The extent to which staff perceived leader engagement and support were investigated by six study-specific single items inspired by Hällsten & Tengbland [45], and Beck [46],; To what extent is your manager engaged in issues related to your knowledge/skills at work? To what extent is your managers engaged in issues related to your profes- sional development at work? To what extent is your manager aware of the quality of the work you do? To what extent does your manager support you in providing care that is based on the individual older person’s needs? To what extent does your manager consciously work to improve the team spirit/mood of the staff group? To what extent do you get supervision of a reg. nurse in the direct care provision? The items are rated on a five-point Likert scale, ranging from 1 (to a very small extent) to 5 (to a very large extent). Higher scores implied that staff to a larger extent agreed with the statement. The study- specific items were treated as single items in the analyses. Activities of daily living Functional function of the residents was measured using a modified version of the Katz Activities of Daily Living Index (ADL) [47] previous published in K. Hulter Åsberg (1990) [48] which measures daily activities in six domains: eating, transferring, dressing, bathing, toileting, and continence. Each domain was scored dichotomously as dependent (0) or fully independent (1) to obtain a total score of 0–6 points. Higher score indicates a more independent functional function. Functional independ- ence was defined as dependent or independent where dependence included person’s dependent least three of six items on the Katz ADL index [49]. in at Gottfries’ cognitive scale (GCS) Cognitive function was assessed using the scale devel- oped by Gottfries and Gottfries [50], previously pub- lished in M. Gustafsson, U. Isaksson, S. Karlsson, PO. Sandman, H. Lovheim (2016). GCS consists of 27 items regarding ability to orientate. Statements are answered with a ‘yes’ (1 point) or ‘no’ (0 point). The range of the scale is 0–27 and high scores indicate a better orienta- tion ability. Scores < 24 indicate cognitive impairment. Cut-off and criterion validity have been established against the Mini-Mental State Examination and been confirmed by Lövheim [51]. Person-centred care assessment tool The Swedish version of the Person-centred Care Assess- ment Tool (P-CAT) used to assess the extent to which the staff perceived care as being person-centred [52, 53]. P-CAT was developed by D. Edvardsson, D. Fetherston- haugh, R. Nay and S. Gibson [52] and later translated to Swedish by K. Sjogren, M. Lindkvist, PO. Sandman, K. Zingmark and D. Edvardsson [53]. The Person-centred Care Assessment Tool includes 13 items rated on a five- point Likert scale, ranging from 1 (disagree completely) to 5 (agree completely). The total score is calculated with a possible range between 13 and 65 and higher scores indicate higher levels of person-centred care. As five items were negatively worded in P-CAT, (Items 7, 8, 9, 10, and 12) these were reversed before statistical ana- lysis. Permission to use P-CAT was obtained from Pro- fessor D. Edvardsson. Free of charge. Data analyses Data was analysed using SPSS statistics version 25. Nor- mality was tested using Kolmogorov-Smirnov and visual examination of the histogram. Up to two missing items in the P-CAT instrument were replaced with the mean value of the individual for the total scale (< 8% of scale missing) [54]. A sample size calculation has been con- ducted for the SWENIS project and reported in previous literature [55, 56], indicating that a sample of 4500 resi- dents would provide enough power to answer the U-Age SWENIS research questions at the 0.05 significance level. As a first step, staff-, resident- and facility charac- teristics were explored using descriptive statistics. Sec- ondly, P-CAT was aggregated on unit level (mean value for the unit), and divided into two groups; units with higher mean values of P-CAT than the mean for all in- cluded nursing homes (49.78 points) and a second co- hort including care units with a P-CAT score below the mean. As P-CAT was normally distributed a mean split was deemed appropriate. Thirdly, differences in leader engagement and support as well as staff-, resident- and facility characteristics between units with higher and lower levels of PCC were explored using independent samples t-test and Chi2 test. Results Direct care staff consisted of mostly women (95.3%) with a mean age of 46.6 years (SD 11.3) and enrolled nursing was the most common qualification (82.5%). Direct care Backman et al. BMC Geriatrics (2021) 21:498 Page 5 of 11 staff had 9.9 years (SD 8.0) as the average work ex- perience in the nursing home. Nursing home man- agers (n = 191) were mostly women (91.0%) with a mean age of 49.6 years (SD 9.0) and they had been working as managers for approximately 3.4 years (SD 3.4) in that specific nursing home. Among managers, a social work degree was the most common educa- followed by registered tional qualification (47.9%) nurse qualification (27.7%) and enrolled nursing (9.0%) qualifications. Approximately, 4.3% of the man- agers were human resource specialists and 11.2% had residents was other qualifications. The sample of comprised of mostly women (67.8%) and the mean age was 85.5 years. Their mean stay in the unit was the majority were ADL about 30 months, and dependent (84%) and cognitively impaired (66.6%) (see Tables 1 and 2). The participating nursing homes consisted of both regular units for older people (69.1%) and special care units for dementia (SCU) (30.9%). The number of beds varied between the nursing homes, 7–128 (mean 38) and most nursing homes were public (93.5%). Comparison of leader engagement and support in units with high and low scorers of PCC When comparing how leader engagement and support were experienced by staff in units with higher and lower scoring of PCC, all variables were significantly higher in units scoring higher in PCC (see Table 3). Table 1 Characteristics of staff (n = 3605) and managers (n = 191) Staff Age (Years) Sex Men Women Qualifications Registered nurses Enrolled nurses Nurse’s assistants No formal qualifications Other education Years of experience in aged care (mean ± SD) Years in this nursing homes (mean (±SD) Work shift Day shift Day and evening Day, evening, night shift Managers Age (Years; mean ± SD) Sex Men Women Qualifications Registered nurses Enrolled nurses Social work Human resource specialist Other education n 1(%) 167 (4.7) 3401 (95.3) 12 (0.3) 2918 (82,6) 463 (13.1) 82 (2.3) 60 (1.7) 80 (2.2) 3140 (88.2) 318 (8.9) n 2 (%) 17 (9.0) 172 (91.0) 52 (27.7) 17 (9.0) 90 (47.9) 8 (4.3) 21 (11.2) Years of experience in aged care (mean ± SD) Years of experience in this nursing home (mean ± SD) 1 n does not always add up to 3605 in all variables due to missing items varariables due to missing items 2 n does not always add up to 191 in all variables due to missing items m (SD) 46.6 (11.3) 17.9 (10.3) 9.9 (8.0) m (SD) 49.6 (9.0) 11.0 (8.7) 3.4 (3.4) Backman et al. BMC Geriatrics (2021) 21:498 Page 6 of 11 m (SD) 85.5 (7.8) Table 2 Characteristics of residents (n = 4831) n 1 (%) Age (Years) Sex Men Women ADL Capacity Independent Dependent Cognitive impairment Yes No Residing in SCU Residing in general units 1538 (32.2) 3239 (67.8) 716 (16) 3768 (84) 2827 (66.6) 1418 (33.4) 1778 (37.8) 2931 (62.2) Length of stay in months (mean ± SD 30.5 (33.1) 1 n does not always add up to 4831 in all variables due to missing items Comparison of staff, resident and facility characteristics in units with high and low scores of PCC There were no significant differences between staff in units with higher and lower scores of PCC in relation to age, sex and years of experience in aged care (see Table 4). In units with higher scores of PCC, a sig- nificantly higher proportion of staff (84.6%) were en- rolled nurses compared to units with lower scores of PCC (80.7%) (p = < 0.002). Units with higher scores of PCC had a significantly lower proportion of nurse’s assistants compared to units with lower scores of PCC (14.2%) (p = 0.021). Units with higher scores of PCC had a significantly lower proportion of other education (1.7%) compared to units with lower scores of PCC (2.2%) (p = 0.016). In units with higher scores of PCC, staff work experience in current nurs- ing home unit were significantly shorter (9.6 year), (11.8%) compared to units with lower scores of PCC (10.2 year) (p = 0.029) (see Table 4). There were no significant differences in terms of resi- dent age, sex and length of stay between units with higher and lower scores of PCC. When comparing facil- ity variables, a significantly larger proportion of demen- tia specific units (34.3%) were found in highly scored PCC units, compared to the units with low scoring of PCC (30.6%) (p = 0.019). Units with higher scores of PCC had significantly higher proportion of residents with ADL dependency (50.9%) compared to units with lower scoring of PCC (49.1%) (see Table 4). It was also found that the number of beds per units were signifi- cantly lower 12.4 (SD 5.8) in units with high PCC scor- ing compared to in units with low scoring of PCC 13.4 (SD 5.8). No significant differences were found related to ownership in terms of public or privately-operated nursing homes (see Table 4). Discussion This study aimed to explore factors characterizing high and low PCC nursing home units, with focus on leader- ship, staff, resident and facility variables. This study showed that leadership engagement and support were scored higher in highly person-centred units, which is in line with previous Swedish nursing home studies [24, 25]. More specifically, the findings showed that high PCC-units were characterised by having leaders en- gaging in staff knowledge, professional development and leaders supporting staff to provide care based on the in- dividual older person’s needs. Furthermore, working to improve the team spirit as a leader, as well as being aware of the quality provided by staff were scored higher in high PCC units. Other studies have described that de- veloping PCC requires leaders that acknowledge staffs’ unique competences and skills, are engaged in care Table 3 Comparison of leadership engagement in units with high versus low degree of PCC Low degree of PCC n 1 m (SD) n 1 m (SD) High degree of PCC p- value To what extent is your manager engaged in issues related to your knowledge / skills at work? 1817 3.3 1659 3.9 0.000 (1.04) (0.95) To what extent is your manager engaged in issues related to your professional development at work? 1820 3.3 1663 3.9 0.000 To what extent is your manager aware of the quality of the work you do? (1.05) (0.97) 1812 3.2 1661 3.8 0.000 (1.11) (1.01) To what extent does your manager support you in providing care that is based on the individual older person’s needs? 1816 3.4 1656 4.1 0.000 (1.06) (0.96) To what extent does your manager consciously work to improve the team spirit/mood of the staff group? 1815 3.1 1658 3.8 0.000 To what extent do you get supervision of a reg. Nurse in the direct care provision? 1 n = staff assessments. Does not always add up to 3605 staff all variables due to missing items (1.16) (1.07) 1820 3.3 1661 3.8 0.000 (1.13) (1.05) Backman et al. BMC Geriatrics (2021) 21:498 Page 7 of 11 Table 4 Comparison of staff, resident and facility characteristics’ between units with high and low degree of PCC Low degree of PCC n 1 (%) m (SD) High degree of PCC n 1 (%) m (SD) p- value Staff Age (Years) Sex Women Men Years of experience in aged care Years in this nursing home unit Education Nurse Enrolled nurse Nurse assistants No formal education Other Resident Age Sex Women Men Length of stay (months) Cognitive impairment ADL dependent Facility Number of beds per unit Public nursing home2 assessmetns Private nursing home2 Dementia specific unit2 General unit2 47 (11.2) 46.2 (11.4) 0.051 17.6 (10.4) 9.6 (7.8) 18.3 (10.3) 10.2 (8.2) 1600 (95.2) 81 (4.8) 7 (0.4) 1404 (84.6) 196 (11.8) 33 (2.0) 60 (1.7) 85.4 (7.9) 85.6 (7.8) 31.1 (31.3) 1576 (67.4) 764 (32.6) 1316 (63.1) 1195 (50.9) 30.2 (35.1) 0.582 0.074 0.029 0.327 0.002 0.021 0.144 0.016 0.454 0.746 0.460 0.000 0.008 13.4 (5.8) 12.4 (5.8) > 0.001 1439 (93.4) 102 (6.6) 571 (34.3) 1094 (65.7) 0.159 0.019 1748 (95.6) 81 (4.4) 5 (0.3) 1468 (80.7) 258 (14.2) 47 (2.6) 40 (2.2) 1428 (67.8) 678 (32.2) 1281 (68.7) 1154 (49.1) 1566 (94.6) 90 (5.4) 559 (30.6) 1269 (69.4) 1 n does not always add up to 3605 staff or 4831 residents in all variables due to missing item 2 n=PCC assessments by staff practices as well as promotes team performance [57– 60]. It has been shown that when staff feel supported, acknowledged and valued at work, optimal performance and commitment are likely to follow [57–60]. One inter- pretation of this study’s findings is, when staffs’ individ- ual knowledge, competence and the quality of their work is acknowledged and supported by their leader, a person-centred approach can be nurtured, both among the individual staff member and within the team. The results also show that staff in high PCC units scored that they received supervision of a registered nurse (RN) in the direct care provision to a larger extent that in units with low PCC. Escrig-Pinol, Corazzini, Blodgett, Chu, & McGilton [61], reports that effective nurse supervisors and support may improve work envi- ronments and staff’s ability to respond to residents’ needs in a timely, effective and compassionate manner. The expertise and clinical excellence a RN holds has shown to be crucial to build a person-centred approach [62]. Based on this, providing supervision to direct care staff appear to be an important focus of efforts when seeking to improve PCC. The results also indicate that high PCC-units are characterised by a significantly higher proportion of staff with higher educational qualification. This has been elaborated in person-centred theory [2]. They postulate that a prerequisite is that staff have compe- tence, being committed to the job, being able to dem- onstrate and also knowledge and skills to make decisions’ and prioritise care [2]. In Sweden, the education of enrolled nurses is twice as long as nurse’s assistants’ education, hence it seems reasonable that a longer education may have principles foundation contributed clarity of beliefs and values, the to of a Backman et al. BMC Geriatrics (2021) 21:498 Page 8 of 11 necessary for PCC. However, this finding is contrary to a previous Canadian study [29], where level of education did not influence the provision of PCC and it was reported that individual factors such as educa- tion exerted very little influence on staffs’ ability to provide PCC, whereas access to resources and seemed to be more of a predictor. This study adds new in- significance for sights on structural conditions of PCC, as de facto the proportion of staff with enrolled nursing education was significantly higher in high PCC units. This finding contributes to the literature as staff competence has been highlighted as a critical element in person-centred theory [2, 19]. Staff work experience was significantly shorter in high PCC units compared to units with low degree of PCC. It is well known that cultural values of conservative traditions can maintain a strong influence over long periods of time, with resistance to change traditional care to a more person-oriented care as a consequence [63–65]. One can interpret that staff with shorter work experi- ence more easily adapt to this person-centred culture shift, preferred by the national guidelines in Sweden [41] than staff with long experience of working with traditional care models. If this is the case here, subse- quent studies need to be explored. A larger proportion of dementia specific units were found in high PCC units, compared to units with low PCC. It was also found that the number of beds per units were lower in units with high PCC compared to in units with low PCC. This is consistent with previous findings from Swedish nursing home [25], where small, dementia-adapted environments characterised high PCC units in nursing homes. Previous research has reported that dementia-adapted environments may contribute to maintaining autonomy and independence and support social interactions and sense of self [66]. A reasonable implication is the need to tailor the physical environ- ment to meet the individual needs of the residents and create small, homelike environments that allow the resi- dents to be an active participant in everyday life rather than a passive recipient of care [66]. Swedish nursing home care has undergone a transformation under the last decade with a rapidly growing share of private ac- tors, from 1% in 1990 to 16% in 2010 [67], to approxi- mately 21% in 2016 [30]. The proponents argued that contracting nursing homes would lead the private actors to develop better ways to provide care with a quality im- provement as a result [68, 69]. However, this study’s finding did not show any care quality differences in terms of PCC provision, related to the ownership of the nursing homes. Previous intervention research from health care con- texts [70], showed that a leadership emphasised PCC values and working practice, and interprofessional team facilitators’ for working was implementing person- centred care in hospital contexts. This can be situated as this study findings as highly person-centred units were characterised by leader engaging and support staff in providing a care that is based on the individual older person’s needs and works to improve the team spirit. This implies that this current study’s finding can be used as managerial strategies when designing and de- veloping person-centred interventions in nursing homes care contexts as well. In summary, this study contributes to the existing research evidence, by sug- gesting that the support from managers and leaders is important for facilitating PCC, and highlights the im- portance of manager and leader support in daily care to enact staff to provide such care. However, further com- parative, longitudinal and interventional studies would be valuable to confirm or reject these findings on lead- ing towards PCC. Methodological considerations This cross-sectional study cannot answer questions of causal nature. All resident data is proxy-rated which may introduce rate bias when rating items concerning one’s own work. This has been addressed with written information to the raters. Proxy-rated resident data may introduce recall and/or observer bias; still, it may information due to the high be the best source of prevalence of cognitive impairment in the sample, and the time the proxies had known the person they assessed was fairly long, indicating good knowledge about the older person. As this study draws on exten- sive cross-sectional randomised data from a national in Swedish nursing sample of homes, this may serve as a means of avoiding system- atic bias. The results are from a Swedish context, and thus, may affect the findings generalizability. However, as this study draws on extensive cross-sectional ran- domised data from a national in Swedish nursing homes, it seems reasonable to argue that the findings could be applied across different contexts and settings with similar care structure. Dif- ferences between nursing homes were not explored, as this was not the aim of this study, still an import- ant research area subsequent studies can explore. staff and residents sample of staff Conclusions and recommendations This study provides information about leadership, staff, resident, facility determinants with capacity to enhancing person-centred care provision, were from leaders seems important when striving towards person- centred care in daily practice. The study also highlights several environmental factors associated with highly person-centred units. The findings suggest that factors of leadership, staff, resident and facility can be identified support Backman et al. BMC Geriatrics (2021) 21:498 Page 9 of 11 and targeted in efforts to facilitate PCC practice in nurs- ing home care. Addressing these gaps may provide im- portant insight into the factors that help or hinder the provision and development PCC. The study findings can be interpreted as predictors or facilitators for PCC and be used for leadership training and/or development ini- tiatives, or even as an empirical knowledge base for nursing curricula on nursing leadership and develop- ment. If nursing homes units or facilities struggles with implementing person-centred care, it seems that man- agers have an important role to promote the movement towards a person-centred practice of care, by supporting their staff in daily care, and engaging in staff knowledge and professional development. It also seems important to target and adjust environmental factors, such as pro- vide small and dementia adapted environments to match the residents’ personal preferences and capacity. It also seems reasonable to suggest that educational initiatives need to be contextually embedded and tailored to meet the person in need of care when seeking to develop and improve person-centred nursing home units. This study provides guidance for practitioners when designing, de- veloping and adapting person-centred units in aged care contexts. Abbreviations PCC: Person-centred care; WHO: World Health Organization; ADL: Activities in daily living; RN: Registered nurse Acknowledgements The authors wish to thank all participating staff and their managers for taking part in this study by completing the questionnaires. Authors’ contributions AB and AS: study design, analysis of data, literature review and drafting of manuscript. PO: study supervision and drafting of manuscript. All authors were involved in manuscript preparation, and read and approved the final manuscript. Funding The study was financed by the Swedish Research Council for Health, Working life and Welfare (2014–4016) and the Swedish Research Council (521-2014-2715). The funding bodies had no role in the design of the study, the data collection, data analysis or interpretation of data. Nor did they have any influence on the writing on the manuscript or decision to submit the paper. Open Access funding provided by Umea University. Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This study has received ethical approval from the Regional Ethical Review Board in Umeå, Sweden (Dnr 2013-269-31). The questionnaires were com- pleted by nursing home staff without the direct involvement of the resi- dents. Informed consent, written or verbal, was not obtained from all residents but an opt-out consent procedure approved by the ethics commit- tee, was used. Written information about the study was provided to all par- ticipating staff as well as on information posters in the entrances to the nursing homes, and residents as well as their relatives could decline partici- pation if they did not want to contribute their data. In agreement with the Swedish Ethical Review Board, a returned and completed survey was considered as a consent to participate. The ethical approval was obtained from the Regional Ethical Review Board in Umeå, Sweden (Dnr 2013-269-31). Competing interests The authors declare that they have no competing interests. Author details 1Department of Nursing, Umeå University, SE-901 87 Umeå, Sweden. 2NVS, Division of Nursing, Karolinska Institutet, Huddinge, Sweden. 3NVS, Division of Neurogeriatrics, Department of Nursing, Karolinska Institutet, Huddinge, Sweden. Received: 20 August 2020 Accepted: 24 August 2021 References 1. Edvardsson D, Winblad B, Sandman PO. Person-centred care of people with severe Alzheimer’s disease: current status and ways forward. Lancet Neurol. 2008;7(4):362–7. 2. McCormack B, McCance TV. Development of a framework for person- 3. 4. 5. 6. 7. 8. 9. centred nursing. J Adv Nurs. 2006;56(5):472–9. Brooker DL, I. Person-centred dementia care - making services better with the VIPS framework. London and Philadelphia: Jessica Kingsley Publishers; 2015. Sjogren K, Lindkvist M, Sandman PO, Zingmark K, Edvardsson D. Person- centredness and its association with resident well-being in dementia care units. J Adv Nurs. 2013;69(10):2196–205. Terada S, Oshima E, Yokota O, Ikeda C, Nagao S, Takeda N, et al. Person- centered care and quality of life of patients with dementia in long-term care facilities. Psychiatry Res. 2013;205(1–2):103–8. Bone C, Cheung G, Wade B. Evaluating person centred care and dementia care mapping in a psychogeriatric hospital in New Zealand: a pilot study. NZJOT. 2010;57(1):35–40. Rokstad AM, Rosvik J, Kirkevold O, Selbaek G, Saltyte Benth J, Engedal K. The effect of person-centred dementia care to prevent agitation and other neuropsychiatric symptoms and enhance quality of life in nursing home patients: a 10-month randomized controlled trial. Dement Geriatr Cogn Disord. 2013;36(5–6):340–53. Chenoweth L, King MT, Jeon YH, Brodaty H, Stein-Parbury J, Norman R, et al. Caring for Aged Dementia Care Resident Study (CADRES) of person-centred care, dementia-care mapping, and usual care in dementia: a cluster- randomised trial. Lancet Neurol. 2009;8(4):317–25. Fossey J, Ballard C, Juszczak E, James I, Alder N, Jacoby R, et al. Effect of enhanced psychosocial care on antipsychotic use in nursing home residents with severe dementia: cluster randomised trial. BMJ. 2006; 332(7544):756–61. 10. Brooker DJ, Woolley RJ. Enriching opportunities for people living with dementia: the development of a blueprint for a sustainable activity-based model. Aging Ment Health. 2007;11(4):371–83. 11. Wallin AO, Jakobsson U, Edberg AK. Job satisfaction and associated variables 12. 13. among nurse assistants working in residential care. Int Psychogeriatr. 2012; 24(12):1904–18. Edvardsson D, Fetherstonhaugh D, McAuliffe L, Nay R, Chenco C. Job satisfaction amongst aged care staff: exploring the influence of person- centered care provision. Int Psychogeriatr. 2011;23(8):1205–12. Sjogren K, Lindkvist M, Sandman PO, Zingmark K, Edvardsson D. To what extent is the work environment of staff related to person-centred care? A cross-sectional study of residential aged care. J Clin Nurs. 2015;24(9–10):1310–9. 14. Moyle W, Murfield JE, Griffiths SG, Venturato L. Care staff attitudes and experiences of working with older people with dementia. Australas J Ageing. 2011;30(4):186–90. Edvardsson D, Sandman PO, Nay R, Karlsson S. Predictors of job strain in residential dementia care nursing staff. J Nurs Manag. 2009;17(1):59–65. 16. WHO. Global strategy on people-centred and integrated health services: an 15. overview of evidence. 2015. 17. McCance TM, Dewing BJ. An exploration of person-centredness in practice. Faculty of Science, Medicine and Health - Papers: part A 2470; 2011. Epp T. Person-centred dementia care: a vision to be refined. Can Alzheimer Dis Rev. 2003;5(3):14–9. 18. 19. McCormack BM, T. Perso-centred nursing theory and practice. Chichester: Wiley-Blackwell Publishing; 2010. Backman et al. BMC Geriatrics (2021) 21:498 Page 10 of 11 20. Rokstad AM, Vatne S, Engedal K, Selbæk G. The role of leadership in the implementation of person-centred care using dementia care mapping: a study in three nursing homes. J Nurs Manag. 2015;23(1):15–26. 21. Rosemond CA, Hanson LC, Ennett ST, Schenck AP, Weiner BJ. Implementing person-centered care in nursing homes. Health Care Manag Rev. 2012;37(3): 257–66. Stein-Parbury J, Chenoweth L, Jeon YH, Brodaty H, Haas M, Norman R. Implementing person-centered care in residential dementia care. Clin Gerontol. 2012;35(5):404–24. 22. 23. Chenoweth L, Forbes I, Fleming R, King MT, Stein-Parbury J, Luscombe G, et al. PerCEN: a cluster randomized controlled trial of person-centered residential care and environment for people with dementia. Int Psychogeriatr. 2014;26(7):1147–60. 24. Backman A, Sjögren K, Lindkvist M, Lövheim H, Edvardsson D. Towards person-centredness in aged care – exploring the impact of leadership. J Nurs Manag. 2016;24(6):766–74. Sjögren K, Lindkvist M, Sandman P-O, Zingmark K, Edvardsson D. Organisational and environmental characteristics of residential aged care units providing highly person-centred care: a cross sectional study. BMC Nurs. 2017;16(1):44. 25. 26. Miller SC, Miller EA, Jung H-Y, Sterns S, Clark M, Mor V. Nursing home 27. 28. organizational change: the “culture change” movement as viewed by long- term care specialists. Med Care Res Rev. 2010;67(4_suppl):65S–81S. Jeon Y-H, Simpson JM, Li Z, Cunich MM, Thomas TH, Chenoweth L, et al. Cluster randomized controlled trial of an aged care specific leadership and management program to improve work environment, staff turnover, and care quality. J Am Med Dir Assoc. 2015;16(7):629.e19-.e28. Sandström B, Borglin G, Nilsson R, Willman A. Promoting the implementation of evidence-based practice: a literature review focusing on the role of nursing leadership. Worldviews Evid-Based Nurs. 2011;8(4):212–23. 29. Caspar S, Cooke HA, O’Rourke N, MacDonald SWS. Influence of individual and contextual characteristics on the provision of individualized care in long-term care facilities. The Gerontologist. 2013;53(5):790–800. 31. 30. Winblad U, Blomqvist P, Karlsson A. Do public nursing home care providers deliver higher quality than private providers? Evidence from Sweden. BMC Health Serv Res. 2017;17(1):487. Edvardsson D, Backman A, Bergland Å, Björk S, Bölenius K, Kirkevold M, Winblad B. The Umeå ageing and health research programme (U-Age): Exploring person-centred care and health-promoting living conditions for an ageing population. Nordic J Nurs Res. 2016;36(3):168–74. (NBHW) NBoHaW. Nationell ledarskapsutbildning för chefer inom äldreomsorg (in Swedish). 2011. 32. 33. Backman A, Sjögren K, Lindkvist M, Lövheim H, Edvardsson D. Characteristics of highly rated leadership in nursing homes using item response theory. J Adv Nurs. 2017;73(12):2903–13. 35. 34. National Board of Health and Welfare. Socialstyrelsen. Vård- och omsorgsutbildade- idag och i framtiden (in Swedish). 2015. Törnquist A. What to do and how to be : a study of the professional skills among unit managers and nursing staff within the institutions of eldercare [Doctoral thesis, monograph]. Stockholm: HLS Förlag; 2004. European qualifications framework. Initial vocational education and training: focus on qualifications at levels 3 and 4. Luxembourg: Publications Office of the European Union. Cedefop research paper; No 77. 2020. http://data. europa.eu/doi/10.2801/114528. 36. 37. National Board of Health and Welfare. Socialstyrelsen. Öppna jämförelser 38. 39. 40. 2018-Vård och omsorg om äldre. Jämförelser mellan kommuner och län (in Swedish). 2018. SOFS. Socialstyrelsens allmänna råd om grundläggande kunskaper hos personal som arbetar i socialtjänstens omsorg om äldre (In Swedish). In: National Board of Health and Welfare, editor; 2011. Eurodiaconia. The education, training and qualifications of nursing and care assistants across Europe. 2016. Education SCfH. Referensramar för svenska kvalifikationer och examina (In swedish). 2021. Available from: https://www.uhr.se/bedomning-av-utlandsk- utbildning/enic-naric-sverige/referensramar-for-svenska-kvalifikationer/?ram= 2&level=5. neuropsychiatric symptoms and ADL dependency among persons living in nursing homes; a cross-sectional study. BMC Geriatr. 2016;16:154. 43. NBoHaWN. The National Board of Health and Welfares statistics database. 44. 2020. https://sdb.socialstyrelsen.se/if_ald/resultat.aspx2020. Lagergren M. The systems of care for frail elderly persons: the case of Sweden. Aging Clin Exp Res. 2002;14(4):252–7. 45. Hällsten FTS. Personalansvar och medarbetarskap-om relationen mellan organisation och medarbetare. Gothenburg: Gothenburg University; 2002. 46. Beck I. Att fokusera på “varandet” i en värld av görande. Stöd till personalen i ett palliativt förhållningssätt vid vård-och omsorgsboende för äldre. Lund: Lund University; 2013. Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. Studies of illness in the aged. The index of ADL: a standardized measure of biological and psychosocial function. JAMA. 1963;185:914–9. 47. 48. Hulter Åsberg K. ADL-trappan. Studentlitteratur. Katz, S. et al. (1963). Studies 49. 51. of illness in the aged. The index of ADL: a standardized measure of biological and psychosocial function. JAMA. 1990;185(12):914–9. Skoldunger A, Wimo A, Sjogren K, Bjork S, Backman A, Sandman PO, et al. Resource use and its association to cognitive impairment, ADL functions, and behavior in residents of Swedish nursing homes: results from the U- Age program (SWENIS study). Int J Geriatr Psychiatry. 2019;34(1):130–6. 50. Adolfsson R, Gottfries CG, Nystrom L, Winblad B. Prevalence of dementia disorders in institutionalized Swedish old people. The work load imposed by caring for these patients. Acta Psychiatr Scand. 1981;63(3):225–44. Lövheim H. Psychotropic and analgesic drug use among old people: with special focus on people living in institutional geriatric care. Umeå: Umeå University; 2008. Edvardsson D, Fetherstonhaugh D, Nay R, Gibson S. Development and initial testing of the Person-Centered Care Assessment Tool (P-CAT). Int Psychogeriatr. 2010;22(1):101–8. Sjogren K, Lindkvist M, Sandman PO, Zingmark K, Edvardsson D. Psychometric evaluation of the Swedish version of the Person-Centered Care Assessment Tool (P-CAT). Int Psychogeriatr. 2012;24(3):406–15. Shrive FM, Stuart H, Quan H, Ghali WA. Dealing with missing data in a multi-question depression scale: a comparison of imputation methods. BMC Med Res Methodol. 2006;6(1):57. 54. 52. 53. 55. Björk S. Exploring resident thriving in Swedish nursing homes : the Umeå ageing and health research programme (U-Age) thesis I [Doctoral thesis, comprehensive summary]. Umeå: Umeå universitet; 2017. 56. Backman A. Leadership: person-centred care and the work situation of staff in Swedish nursing homes [Doctoral thesis, comprehensive summary]. Umeå: Umeå universitet; 2018. 57. McGilton KS, Sidani S, Boscart VM, Guruge S, Brown M. The relationship between care providers’ relational behaviors and residents mood and behavior in long-term care settings. Aging Ment Health. 2012;16(4):507–15. 58. Backman A, Ahnlund P, Sjögren K, Lövheim H, McGilton KS, Edvardsson D. 59. Embodying person-centred being and doing: leading towards person-centred care in nursing homes as narrated by managers. J Clin Nurs. 2020;29(1–2):172–83. Lynch BM, McCance T, McCormack B, Brown D. The development of the person-centred situational leadership framework: revealing the being of person-centredness in nursing homes. J Clin Nurs. 2018;27(1–2):427–40. 60. Cardiff S, McCormack B, McCance T. Person-centred leadership: a relational approach to leadership derived through action research. J Clin Nurs. 2018; 27(15–16):3056–69. Escrig-Pinol A, Corazzini KN, Blodgett MB, Chu CH, McGilton KS. Supervisory relationships in long-term care facilities: a comparative case study of two facilities using complexity science. J Nurs Manag. 2019;27(2):311–9. 62. McGilton KS, Heath H, Chu CH, Boström A-M, Mueller C, Boscart MV, et al. 61. Moving the agenda forward: a person-centred framework in long-term care. Int J Older People Nursing. 2012;7(4):303–9. Koren M. Person-centered care for nursing home residents: the culture- change movement. Health Aff. 2010;29(2):312–7. 63. 64. Carlström Eric D, Ekman I. Organisational culture and change: implementing person-centred care. J Health Organ Manag. 2012;26(2):175–91. 65. McGreevy J. Implementing culture change in long-term dementia care settings. Nurs Stand. 2016;30(19):44–50. 41. National Board of Health and Welfare. Socialstyrelsen. Nationella riktlinjer för vård 66. Davis S, Byers S, Nay R, Koch S. Guiding design of dementia friendly och omsorg vid demenssjukdom- Stöd för styrning och ledning-remissversion (in Swedish). 2017. http://www.socialstyrelsen.se/publikationer2017/2017-12-2. environments in residential care settings: considering the living experiences. Dementia. 2009;8(2):185–203. 42. Bjork S, Juthberg C, Lindkvist M, Wimo A, Sandman PO, Winblad B, et al. Exploring the prevalence and variance of cognitive impairment, pain, 67. Blomqvist. P WU. Privata aktörer och kontraktsstyrning i äldreomsorgen: förbättras kvaliteten? 2011. Backman et al. BMC Geriatrics (2021) 21:498 Page 11 of 11 68. 1992/93:43 P. Ökad konkurrens i kommunal verksamhet (in Swedish). 1992. 69. Civildepartementet S. SOU 1991:104. Konkurrensen inom den kommunala sektorn, Civildepartementet, editor. Stockholm; 1991. 70. Moore L, Britten N, Lydahl D, Naldemirci Ö, Elam M, Wolf A. Barriers and facilitators to the implementation of person-centred care in different healthcare contexts. Scand J Caring Sci. 2017;31(4):662–73. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12870-023-04147-5	Niu et al. BMC Plant Biology (2023) 23:179 https://doi.org/10.1186/s12870-023-04147-5 RESEARCH BMC Plant Biology Open Access Lint percentage and boll weight QTLs in three excellent upland cotton (Gossypium hirsutum): ZR014121, CCRI60, and EZ60 Hao Niu1, Meng Kuang1, Longyu Huang1, Haihong Shang1,2, Youlu Yuan1 and Qun Ge1* Abstract Background Upland cotton (Gossypium hirsutum L.) is the most economically important species in the cotton genus (Gossypium spp.). Enhancing the cotton yield is a major goal in cotton breeding programs. Lint percentage (LP) and boll weight (BW) are the two most important components of cotton lint yield. The identification of stable and effec- tive quantitative trait loci (QTLs) will aid the molecular breeding of cotton cultivars with high yield. Results Genotyping by target sequencing (GBTS) and genome-wide association study (GWAS) with 3VmrMLM were used to identify LP and BW related QTLs from two recombinant inbred line (RIL) populations derived from high lint yield and fiber quality lines (ZR014121, CCRI60 and EZ60). The average call rate of a single locus was 94.35%, and the average call rate of an individual was 92.10% in GBTS. A total of 100 QTLs were identified; 22 of them were overlap- ping with the reported QTLs, and 78 were novel QTLs. Of the 100 QTLs, 51 QTLs were for LP, and they explained 0.29–9.96% of the phenotypic variation; 49 QTLs were for BW, and they explained 0.41–6.31% of the phenotypic variation. One QTL (qBW-E-A10-1, qBW-C-A10-1) was identified in both populations. Six key QTLs were identified in multiple-environments; three were for LP, and three were for BW. A total of 108 candidate genes were identified in the regions of the six key QTLs. Several candidate genes were positively related to the developments of LP and BW, such as genes involved in gene transcription, protein synthesis, calcium signaling, carbon metabolism, and biosynthesis of secondary metabolites. Seven major candidate genes were predicted to form a co-expression network. Six signifi- cantly highly expressed candidate genes of the six QTLs after anthesis were the key genes regulating LP and BW and affecting cotton yield formation. Conclusions A total of 100 stable QTLs for LP and BW in upland cotton were identified in this study; these QTLs could be used in cotton molecular breeding programs. Putative candidate genes of the six key QTLs were identified; this result provided clues for future studies on the mechanisms of LP and BW developments. Keywords Upland cotton (Gossypium hirsutum L.), Lint percentage, Boll weight, Quantitative trait locus (QTL), Candidate gene Correspondence: Haihong Shang shanghaihong@caas.cn Youlu Yuan yuanyoulu@caas.cn Qun Ge gequn1989@126.com Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Niu et al. BMC Plant Biology (2023) 23:179 Page 2 of 18 Background Cotton (Gossypium) is an economically important natu- ral fiber plant. Upland cotton (Gossypium hirsutum) is the most widely cultivated cotton variety, accounting for approximately 95% of global cotton production [1, 2]. Increasing the yield of upland cotton remains the main objective of this important cash crop worldwide. Cotton yield is typically affected by several complex quantita- tive traits, including the boll number (BN), lint percent- age (LP), boll weight (BW), seed index (SI) and lint index (LI) [3]. These yield component traits are controlled by genetic factors and are affected by environmental fac- tors; they are also genetically related to each other [3–5]. LP is an economically important index for cotton culti- vars with the highest heritability [6]. Because LP is a key contributor to lint yield and is easy to measure, selection for increasing LP has become an important approach for enhancing lint yield [7, 8]. Numerous studies have shown that cotton yield mainly depends on LP, BW, and BN, and these traits have been positively selected in cultivated cotton throughout the domestication process [9–15]. Because cotton breeding requires excellent germplasm, a large amount of germplasm resources have been pre- served and improved in China, such as many high LP cultivars/lines [16–18]. Many interspecific introgressive lines (ILs) or chromosome segment introgression lines (CSILs) have been obtained by crosses between G. hirsu- tum and Gossypium barbadense [19, 20]; some of these lines have high LP and BW [21]. Many new germplasm resources and cultivars have been successfully bred [22– 26]. Our lab has also bred a set of advanced cotton lines/ cultivars, such as the parents used in this study. The identification of stable and effective quantitative trait loci (QTLs) is prerequisites for cotton molecular breeding. From 1998 to 2015, a total of 327 QTLs for LP and 170 QTLs for BW were identified on different chro- mosomes through meta-QTL analysis [27]. Following the release of the cotton genome sequence, the number of dis- covered QTLs is rapidly increasing via genome-wide asso- ciation study (GWAS) or linkage mapping [28–30]. For example, structural variations have been explored by rese- quencing 1,081 G. hirsutum accessions, and 446 structural variations are significantly associated with seven traits, including 21 with LP and 17 with BW [31]. Genetic link- age analysis and association analysis (AS, or GWAS) are the two major approaches for identifying QTLs in crops. Many high-density genetic linkage maps and association maps for cotton have been published. For example, more than 17 crosses or populations of upland cotton have been used to construct genetic maps, including crosses of Yumian1 × T586 [4, 32, 33], Yumian1 × Zhongmian- suo35 [1], NC05AZ06 × NC11-2091 [34], DH962 × Jim- ian5 [35–37], Zhongmiansuo12 (ZMS12) × 8891 [4], 43], [42, (Simian3 × Sumian12) × (Zhong4133 × 8891) [3], Baimian1 × TM-1 [38, 39], Xiangzamian2 [40, 41], and HS46 × MARCABUCAG8US-1–88 CCRI35 × Nan Dan Ba Di Da Hua (NH) [44]. One high- density bin linkage map contains 6,187 bin markers span- ning 4,478.98 cM with an average distance of 0.72 cM [18]. Different types of GWAS, including single-locus- GWAS (SL-GWAS), multi-locus GWAS (ML-GWAS), and restricted two-stage, multi-locus, and multi-allele GWAS (RTM-GWAS) approaches, have been used to identify quantitative trait nucleotides (QTNs) for LP and BW in several cotton accessions. More than 16 associa- tion maps and many candidate genes for agronomic traits have been reported [5, 8, 10, 12, 45–48]. For example, 86 single-nucleotide polymorphism linkage disequilibrium block (SNPLDB) loci for LP and 70 SNPLDB loci for BW have been identified from 315 cotton accessions using RTM-GWAS [12]. A total of 719 upland cotton accessions have been screened by GWAS using the cottonSNP63K array, and 62 identified single nucleotide polymorphism (SNP) loci were significantly associated with different traits; a total of 689 candidate genes were screened, and 27 of them contain at least one significant SNP, including three for LP and six for BW [5]. Although the inheritance, QTLs and candidate genes of LP and BW in upland cotton have been widely stud- ied, only a few of the studied QTLs have been used in the molecular breeding of cotton via marker-assisted selec- tion (MAS) [49, 50]. One of the reasons is that the iden- tified QTLs are unstable in multiple-environments and only explain little phenotypic variance. Consequently, mining stable, effective LP and BW-related QTLs or QTNs would greatly aid cotton molecular breeding. We have previously bred the excellent cotton lines ZR014121 and EZ60 and the cultivar CCRI60. Here, we identified stable, effective LP and BW-related QTLs to aid the uti- lization of the germplasm resources in cotton breeding. Results Phenotypic variation in LP and BW We evaluated two yield-related traits LP and BW, in the two recombinant inbred line (RIL) populations under four environments in 2020 and 2021. The LP and BW ranged from 32.56% to 48.26% and from 4.09 to 6.93 g in P-EZ60, respectively (Table 1); LP and BW ranged from 31.57% to 48.02% and from 3.68 to 6.83 g in P-CCRI60, respectively (Table 2). All of the absolute skewness values of LP and BW were less than 1.0. The distributions of the LP and BW in the four experimental environments were normal. This suggests that LP and BW are polygenic traits, and the data could be used to map QTLs (Fig. 1). LP and BW exhibited high degrees of phenotypic variation. The coef- ficient of variation for each trait was relatively consistent Niu et al. BMC Plant Biology (2023) 23:179 Page 3 of 18 n o i t a i r a v f o 2 8 5 . 9 6 7 . 3 4 5 . 8 6 . 3 8 5 . 3 4 7 . 9 6 5 . 3 7 6 . t n e i c ffi e o c s i s o t r u k s s e n w e k s e c n a i r a v d r a d n a t s n o i t a i v e d e u l a v n a e m e u l a v m u m i x a M e u l a v m u m n M i i e g n a r s L I R 0 6 Z E 1 2 1 4 1 0 R Z n o i t a l u p o p s t n e r a p t i a r t t n e m n o r i v n e 0 6 Z E - P n i P L d n a W B e h t f o s i s y a n a l l a c i t s i t a t S 1 e l b a T 1 6 1 0 . 6 9 1 0 - . 5 4 1 0 - . 6 1 2 0 - . 9 2 0 - . 3 2 1 0 . 4 8 0 0 - . 5 1 4 0 . 7 4 5 0 - . 2 5 1 0 . 8 9 1 0 - . 5 4 1 0 . 5 0 1 0 - . 7 0 2 0 . 4 1 0 0 - . 7 4 1 0 . 4 2 6 . 3 5 1 0 . 3 5 8 4 . 2 4 1 0 . 3 1 0 5 . 5 8 1 0 . 9 9 3 5 . 9 2 1 0 . 5 2 . 9 3 0 . 2 2 . 8 3 0 . 4 2 2 . 3 4 0 . 2 3 2 . 6 3 0 . 2 9 2 4 . 9 0 5 . 5 5 0 4 . 4 5 5 . 4 4 8 3 . 8 7 5 . 1 8 0 4 . 3 3 5 . 6 2 8 4 . 1 1 6 . 7 6 5 4 . 4 6 6 . 9 4 3 4 . 3 9 6 . 9 2 6 4 . 3 6 6 . 3 8 3 3 . 9 0 4 . 7 5 3 3 . 6 4 . 6 5 2 3 . 5 6 4 . 3 3 4 3 . 1 4 4 . 2 4 4 1 . 2 0 2 . 9 0 2 1 . 4 0 2 . 3 9 0 1 . 8 2 2 . 6 9 1 1 . 3 2 2 . 9 9 1 9 9 1 9 9 1 9 9 1 9 9 1 9 9 1 9 9 1 9 9 1 7 8 1 4 . 7 1 5 . 9 2 0 4 . 5 3 5 . 9 6 9 3 . 3 1 6 . 3 8 0 4 . 3 7 5 . 7 4 9 3 . 3 4 4 . 1 9 8 3 . 1 5 . 5 3 6 3 . 1 2 5 . 4 5 8 3 . 6 9 4 . P L W B P L W B P L W B P L W B Y A 0 2 X W 0 2 Y A 1 2 X W 1 2 Niu et al. BMC Plant Biology (2023) 23:179 Page 4 of 18 n o i t a i r a v f o 2 9 5 . 5 7 . 6 6 5 . 9 0 7 . 9 3 5 . 8 7 . 7 0 5 . 9 7 7 . t n e i c ffi e o c s i s o t r u k s s e n w e k s e c n a i r a v d r a d n a t s n o i t a i v e d e u l a v n a e m e u l a v m u m i x a M e u l a v m u m n M i i e g n a r s L I R I 0 6 R C C 1 2 1 4 1 0 R Z n o i t a l u p o p s t n e r a p t i a r t t n e m n o r i v n e I 0 6 R C C - P n i P L d n a W B e h t f o s i s y a n a l l a c i t s i t a t S 2 e l b a T 4 6 2 0 - . 2 3 7 0 . 5 9 0 0 . 5 2 3 0 - . 7 6 0 0 - . 8 7 4 0 . 1 9 3 0 . 5 2 3 0 . 8 6 4 0 - . 7 3 0 - . 5 0 3 0 - . 2 2 0 . 1 1 0 - . 2 7 1 0 - . 5 3 1 0 - . 3 9 0 0 . 8 1 3 6 . 5 1 0 . 2 7 9 4 . 8 5 1 0 . 7 3 0 4 . 8 9 1 0 . 2 9 9 3 . 8 6 1 0 . 1 5 2 . 9 3 0 . 3 2 2 . 4 0 . 1 0 2 . 4 4 0 . 1 4 0 . 2 4 4 2 4 . 6 1 5 . 7 3 9 3 . 1 6 5 . 4 2 7 3 . 7 5 . . 4 9 3 7 2 5 . 2 0 8 4 . 3 2 6 . 4 8 4 4 . 3 8 6 . 7 5 2 4 . 3 8 6 . 2 5 5 4 . 5 5 6 . 8 6 3 . 5 3 6 6 2 3 . 1 7 4 . 7 5 1 3 . 2 9 3 . 5 1 2 3 . 4 2 0 3 1 . 4 5 2 . 8 1 2 1 . 2 1 2 . 1 1 1 9 2 . 8 3 3 1 . 5 5 2 . 9 9 2 9 9 2 9 9 2 9 9 2 9 9 2 9 9 2 9 9 2 9 9 2 8 4 1 4 . 5 1 5 . 3 3 9 3 . 5 6 5 . 2 0 7 3 . 2 6 5 . 4 2 9 3 . 9 4 5 . 7 4 9 3 . 3 4 4 . 1 9 8 3 . 1 5 . 5 3 6 3 . 1 2 5 . 4 5 8 3 . 6 9 4 . P L W B P L W B P L W B P L W B Y A 0 2 X W 0 2 Y A 1 2 X W 1 2 Niu et al. BMC Plant Biology (2023) 23:179 Page 5 of 18 Fig. 1 The histograms of the LP and BW in P-EZ60 (EZ60) and P-CCRI60 (CCRI60) in Anyang and Weixian in 2020 and 2021 among the different environments, suggesting that LP and BW were significantly affected by the environment, and the effect on BW (average 7.16 in P-EZ60; 7.55 in P-CCRI60) was greater than that on LP (average 5.69 in P-EZ60; 5.51 in P-CCRI60) (Tables 1, and 2). The correlations between LP and BW of all the RILs in the four environments were analyzed separately. Generally, LP and BW were significantly negatively cor- related in P-EZ60 and P-CCRI60, and the coefficients ranged from -0.098 to -0.340, which suggested that it was difficult to improve LP and BW synchronously (Tables 3, and 4). Because the cotton field was water- logged in Anyang in 2021, the LP and BW were affected to some extent, but the phenotypic data met the requirements for GWAS (Fig. 1). Analysis of variance (ANOVA) showed that there were highly significant differences among the accessions and environments for the two traits of two populations (Table 5). It indicated that LP and BW were significantly influenced by the accessions and planting environments. SNP quality control and in silico mapping to According the high-throughput whole-genome sequencing data of upland cotton (Nanjing Agricul- tural University), a liquid SNP array with 10 K SNPs was developed. The two RIL populations of P-CCRI60 and P-EZ60, including their parents, were genotyped by genotyping by target sequencing (GBTS) (Table S1). The total number of samples was 500. The average call rate of a single locus was 94.35%, and the average call rate of an individual was 92.10%. The results of the genotype control are shown in supplementary table 2 (Table S2). The BLAST alignment tool was used to ana- lyze the probe sequences of SNPs against the G. hir- sutum TM-1 genome sequence [28, 51], and a total of 8,348 genotyped high-quality SNPs across the 500 sam- ples were used in association mapping. Genome‑wide association studies We used the genetic model of 3VmrMLM to detect interactions QTNs for LP and BW × environment Niu et al. BMC Plant Biology (2023) 23:179 Page 6 of 18 Table 3 Correlation analysis between BW and LP in P-EZ60 in Anyang in 2020 and 2021 20AYLP 20WXLP 21AYLP 21WXLP -0.107 -0.247 20AYBW 20WXBW 21AYBW 21WXBW -0.191 -0.340 Represents significance at the P < 0.01 level (two-tailed) * Represents significance at the P < 0.05 level (two-tailed) Table 4 Correlation analysis between BW and LP in P-CCRI60 in Anyang in 2020 and 2021 20AYLP 20WXLP 21AYLP 21WXLP -0.098* -0.247 20AYBW 20WXBW 21AYBW 21WXBW -0.185 -0.234 Represents significance at the P < 0.01 level (two-tailed) * Represents significance at the P < 0.05 level (two-tailed) (Fig. 2). A total of 104 stable quantitative trait nucleotides (QTNs) on 26 chromosomes were identified as signifi- cantly associated with LP and BW (Table S3). Following other similar studies [47], we defined the flanking 200-Kb regions of QTNs as an initial QTL and merged the over- lapping QTLs to obtain the final QTLs. In the end, 100 stable QTLs were detected; 51 of them were for LP and 49 were for BW, including three QEIs, one for LP and two for BW, which could be identified in the four envi- ronments (Table S4). A total of 20 stable QTLs, 14 for LP and 6 for BW, were identified in EZ60, including one QEI for BW that could be identified in the four environments; 33 stable QTLs, 18 for LP and 15 for BW, were identi- fied in CCRI60, including one QEI for LP that could be identified in the four environments; and 47 stable QTLs were identified in ZR014121, 19 for LP and 28 for BW, including one QEI for BW that could be identified in the Table 5 Analysis of variance for the two traits of two populations four environments (Table S4). One QTL in chromosome A10, qBW-E-A10-1, was identified in both populations. Among the 100 QTLs, 22 QTLs, 9 for LP and 13 for BW, were overlapping with the reported QTLs (Table S5); 78 QTLs, 42 for LP and 36 for BW, were novel (Table S6). The QTLs explained 0.29–9.96% of the phenotypic variations in LP or BW. In P-EZ60, the novel QTLs associated with LP explained 0.47–8.67% of the pheno- typic variation, and the novel QTLs associated with BW explained 0.91–6.31% of the phenotypic variation. In P-CCRI60, the novel QTLs associated with LP explained 0.29 –9.96% of the phenotypic variation, and the novel QTLs associated with BW explained 0.36–3.02% of the phenotypic variation. In sum, a total of 51 QTLs related to LP were detected in this study, including 14 in EZ60, 18 in CCRI60, and 19 in ZR014121; 28 QTLs were in the At subgenome, and 27 QTLs were in the Dt subgenome, indicating that LP-related QTLs were evenly distributed in the At and Dt subgenomes. A total of 49 QTLs related to BW were detected, including 6 in EZ60, 15 in CCRI60, and 28 in ZR014121; 34 QTLs were in the At subgenome, and 15 QTLs were in the Dt subgenome, indicating that the QTLs related to BW were mainly distributed in the At subgenome. There were two QEIs, which were located on chromosomes A02 and A10 (Fig. 3). Candidate genes in the regions of the six key QTLs To identify candidate genes of key QTLs, six QTLs were selected, including three QEIs, the common QTL qBW-E-A10-1 that was mapped in both populations and two important QTLs (qLP-E-D03-2 and qLP-C- D03-2). The three QEIs were QTLs that were stable in the four environments (Table S7). A total of 108 puta- tive candidate genes in the regions of the six key QTLs in multiple environments were identified, including genes that were positively related to LP and BW, such as the genes involved in gene transcription, protein synthesis, calcium signaling, phytohormone synthesis population trait Source P-EZ60 P-CCRI60 LP BW LP BW Accessions Environments Accessions Environments Accessions Environments Accessions Environments SS 6714.21 3856.09 151.92 110.12 7788.48 8105.79 216.49 124.62 df 190 3 190 3 290 3 290 3 MS 35.34 1285.36 0.80 36.71 26.86 2701.93 0.75 41.54 F 5.58 173.44 3.24 151.13 3.32 388.89 2.85 154.30 P‑value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Niu et al. BMC Plant Biology (2023) 23:179 Page 7 of 18 Fig. 2 Manhattan-plots of LP and BW using the genetic model 3VmrMLM. X-axes are cotton chromosomes. Y-axes on the left side report -log10 P-values of the main-effect QTNs, which were obtained from single-marker genome-wide scans for all the markers in the first step of 3VmrMLM; Y-axes on the right side report LOD scores, which were obtained from likelihood ratio tests for significant and suggested QTNs, with a threshold of LOD 3.0 (dashed line) in the second step of 3VmrMLM. These LOD scores are indicated by points with straight lines = Niu et al. BMC Plant Biology (2023) 23:179 Page 8 of 18 Fig. 3 A physical map of QTLs for LP and BW from the two RIL populations. The green letters are QTLs for LP, and the red letters are QTLs for BW. The scale on the left is in Mb and signaling, and fiber synthesis-related polysaccha- ride metabolism (Table S6). KEGG analysis showed that the 48 genes related to LP were mainly involved in “metabolic pathways” and “spliceosome” (Fig. 4). Eighteen metabolic path- ways such as “biosynthesis of secondary metabolites”, “microbial metabolism in diverse environments” and “DNA replication” were also detected. KEGG analysis showed that the 60 genes related to BW were mainly involved in “metabolic pathways” and “biosynthesis of secondary metabolites” (Fig. 5). “Microbial metabo- lism in diverse environments”, “carbon metabolism,” “glycolysis/gluconeogenesis,” and 19 other metabolic pathways were detected. Expression profiles of candidate genes during fiber development Most of the candidate genes associated with LP and BW were differentially expressed in cotton fiber at differ- ent developmental stages, and there were differences at expression levels between the high-LP parent EZ60 and the low-LP parent ZR014121 at the same stage (Fig. 6). Among the major candidate genes, Gh_A02G0096 was only expressed in the ovule developmental stage of EZ60. Gh_A02G0111 was mainly expressed in both EZ60 and ZR014121 at 0, 5, 10, and 20 days post-anthesis (DPA). Its expression levels were higher in ZR014121 than in EZ60 at 0, 5, and 25 DPA; its expression levels were higher in EZ60 than in ZR014121 at 10 DPA. Gh_D03G1064 was highly expressed in both EZ60 and ZR014121 at all stages. It was mainly expressed at 0, 5, and 10 DPA, and its expression level in ZR014121 was higher than that in EZ60 at 10 DPA. Gh_D03G1069 was expressed in both EZ60 and ZR014121 at all stages. Its expression levels were higher in ZR014121 than in EZ60 at 10 and 20 DPA; its expression levels were higher in EZ60 than in ZR014121 at 0, 5, 15, and 25 DPA. Gh_A02G0106 was significantly highly expressed during the ovule devel- opment stage in EZ60, highly expressed at 5 DPA, and weakly expressed at 10 DPA in ZR014121. Niu et al. BMC Plant Biology (2023) 23:179 Page 9 of 18 Fig. 4 A histogram of candidate genes enriched in different KEGG pathways for LP. The x-axis indicates the number of candidate genes. The y-axis represents biological processes. The details are listed in Table S9 Fig. 5 A histogram of candidate genes enriched in different KEGG pathways for BW. The x-axis indicates the number of candidate genes. The y-axis represents biological processes. The details are listed in Table S10 Co‑expression of candidate genes The interaction network of candidate genes associ- ated with LP and BW was investigated by construct- interaction (PPI) network ing the protein–protein using the STRING database [52] (Fig. 7). Correlations were observed in the expression of the following pro- teins that appear to comprise a co-expression network: Gh_A02G0111, Gh_D03G1056, Gh_D03G1134, Gh_ D03G1064, Gh_A02G0106, Gh_A10G1521, and Gh_ A10G1653. Network analysis of the major proteins was carried out using Cytoscape 3.7.2 (Fig. 8). Gh_D03G1056, Gh_D03G1064, Gh_D03G1134, and Gh_A02G0111 played important roles in the network. PPI analysis indicated that GAI interacted with six other proteins. GAI interacted with FRI; FRI interacted with FPA; FOA interacted with AT1G12775; AT1G12775 inter- acted with AT3G46960; and AT3G46960, AT3G06700, and AT1G80750 interacted with each other (Fig. 7). There were three groups of co-expressed genes, UBC32 and PCNA1; and CRT3 and ECA1; HCF107 and GOX1. Co-expression analysis of the 108 candidate genes of the six QTLs using Cytoscape 3.7.2 indicated that the seven genes (the same (See figure on next page.) Fig. 6 Gene expression profiles of the candidate genes of LP and BW QTLs during fiber development in EZ60 and ZR014121. Each column represents one sample, and rows represent candidate genes. The expression levels of the candidate genes (FPKM) were log2-normalized (i.e., log2(FPKM represents the ovule development stage. 5, 10, 15, 20, and 25 DPA represent the fiber development stages. Detailed information on gene expression is shown in Table S11 0.01)) and presented in different colours on the scale bar. ZR indicates cotton line ZR014121; DPA indicates days post-anthesis. 0 DPA + Niu et al. BMC Plant Biology (2023) 23:179 Page 10 of 18 Fig. 6 (See legend on previous page.) Niu et al. BMC Plant Biology (2023) 23:179 Page 11 of 18 Fig. 7 Protein–protein interaction of the candidate genes of the QTLs for LP and BW. Network nodes represent proteins with splice isoforms or post-translational modifications are collapsed, i.e. each node represents all the proteins produced by a single, protein-coding gene locus. Colored nodes: query proteins and first shell of interactors; white nodes: second shell of interactors; Empty nodes: unknown proteins. 3D structure filled nodes: some 3D structures are known or predicted. Edges represent protein–protein associations. Associations are meant to be specific and meaningful (i.e., proteins jointly contribute to a shared function); this does not necessarily mean that they physically bind to each other. Known Interactions, blue: from curated databases; purple: experimentally determined. Predicted Interactions, green: gene neighborhood, red: gene fusions; indigo: gene co-occurrence; Others, yellow: textmining, black: co-expression, light purple: protein homology as the result of PPI) were co-expressed, including Gh_ A02G0106 (GAI) (Fig. 8). Discussion A set of new major QTLs for LP and BW that could be used for MAS was obtained LP and BW are the most important traits in cotton breeding, and they have been widely studied. More than 417 unique QTLs for LP have been identified on 26 chromosomes, including 243 QTLs identified with LOD > 3. More than 60 were stable, major effective QTLs that could be used for MAS [50]. According to the CottonGen Database [53, 54], a total of 1,387 yield QTLs and four yield component trait QTLs have been identified. The numbers of these QTLs are increas- ing continually. Recently, 34 SNPs corresponding to 22 QTLs for LP, including 13 novel QTLs, were detected from 254 upland cotton accessions via GWAS [55]. Two stable LP QTLs and three BW QTLs were identi- fied in the RIL mapping population derived from the inter-specific cross between G. hirsutum cv DS-28 and G. barbadense cv SBYF-425 [56]. We also identified one QTL for LP, and nine QTLs for BW from a BC5F3:5 line population chromosome segment substitution derived from G. hirsutum CCRI36 and G. barbadense Hai1 [57]. Three QTLs for LP and one QTL for BW were identified from an F2 population derived from the G. hirsutum × G. barbadense cross [58]. Niu et al. BMC Plant Biology (2023) 23:179 Page 12 of 18 Fig. 8 Major gene coexpression network of the candidate genes of the QTLs for LP and BW. Lines indicate co-expression of two linked genes. Network nodes represent genes. The size of the circle shows the betweenness centrality points of the gene. The size of the circle indicates that the gene plays an important role in co-expression. In this graph, genes with higher betweenness centrality points are marked in green and placed in the outer circle, and genes with smaller BC values are marked in red and placed in the inner circle. The three genes in the outer ring, Gh_D03G1056, Gh_D03G1064, and Gh_D03G1134 were candidate genes for LP, and Gh_A02G0111 was a candidate gene for BW In this study, a total of 51 stable QTLs for LP and 49 stable QTLs for BW were identified from three upland cotton lines ZR014121, CCRI60, and EZ60; these QTLs could explain 0.29–9.96% of the phenotypic variation in LP and 0.41–6.31% of the phenotypic variation in BW. A total of 78 of these QTLs were novel. These findings enhance QTL resources that could be used to enhance the yield of cotton; this QTL information will also aid the molecular breeding of cotton cultivars with high yield. Many studies have shown that the heritability of LP is the highest among all yield component traits in cotton, and the heritability of BW was the lowest among all cot- ton yield components. Because the heritability of BW is low, environmental factors can have significant effects on BW [6, 59–61]. The results of this study also demonstrate that environmental factors have stronger effects on BW than on LP (Tables 1, and 2). Thus, selection for LP can achieve desired outcomes more efficiently than selec- tion for BW in cotton breeding. Correlations and path analysis among agronomic and technological traits of 16 upland cotton lines indicated that LP was negatively cor- related with BW (-0.2668) [62]. Generally, LP and BW are negatively related [50]. In our study, the correlation coef- ficients between LP and BW ranged from -0.098 to -0.340 (Tables 3, and 4). This indicates that increases in one of these traits limit increases in the other. LP may be the target of direct selection on cotton genotypes with high cotton fiber yield. Most QTLs for LP and BW explain less than 10% of the phenotypic variation. For example, one study indi- cates that nine QTLs for LP explain 1.84–13.50% of the observed phenotypic variation; two QTLs for BW explain 6.02–9.50% of the observed phenotypic variation [63]. The QTLs qLP-C13-1 and qLP-C25-1 for LP explain 5.77% and 8.87% of the phenotypic variation, respec- tively [64]. A GWAS of a set of 289 Gossypium arboreum chromosome segment ILs in G. hirsutum indicates that co-QTLs for LP explain 1.21–10.79% of the phenotypic variation, and co-QTLs for BW explain 1.17–11.56% of the phenotypic variation [65]. Some QTLs for LP identi- fied in this study explained nearly 10% of the phenotypic variation, and all QTLs for BW explained less than 10% of the phenotypic variation (Table S4). These QTLs, espe- cially the major effective QTLs, can be used to breed cot- ton plants with high yield via MAS. Niu et al. BMC Plant Biology (2023) 23:179 Page 13 of 18 Several putative candidates of the six QTLs for LP and BW were identified Understanding the molecular mechanisms of LP and BW developments is essential for the molecular breeding of cotton plants with high yield, especially via genetic engi- neering. Many candidate genes of the QTLs for LP and BW have been studied [48–50, 55]. The TIP41-like family protein (TIP41L) gene (GH_A12G0194) is thought to be the candidate gene of a stable major QTL (q(BW + SI)- A12-1) for BW [49]. One gene orthologous to the Arabidopsis receptor-like protein kinase gene HERK1 (GB_A07G1034) was predicted to be the candidate gene for LP in G. barbadense [48]. Two candidate genes (Gh_ D01G0162 and Gh_D07G0463) of QTLs for LP were identified. Gh_D01G0162 is a homolog of the auxin- responsive GH3 family protein gene, and Gh_D07G0463 is a homolog of the NADPH/respiratory burst oxidase protein D gene (RBOHD) in Arabidopsis [55]. A molecu- lar regulatory network for LP has been proposed based on the functions of the candidate genes of QTLs for LP [50]. In this study, the candidate genes of the six important QTLs for LP and BW were investigated. The QTLs for both traits have candidate genes involved in gene tran- scription, protein syntheses, signaling, calcium signaling, carbon metabolism, metabolic pathways, and biosynthe- sis of secondary metabolites, which demonstrates that there are several candidate genes of the QTLs for LP and BW (Figs. 4, and 5; Tables S8, S9, S10). This result is con- sistent with the findings of previous studies [48, 50, 55, 66, 67]. The difference is that a greater number of candi- date genes in QTLs for LP were involved in gene expres- sion processes, and a greater number of candidate genes in QTLs for BW were involved in metabolic pathways. Interaction network analysis of the candidate genes asso- ciated with LP and BW indicated that seven candidate genes could form a co-expression network. The candidate gene Gh_A02G0096 of qBW-E-A02-1 encodes a homolog of eukaryotic translation initiation factor 2A, and the candidate gene Gh_D03G1069 of qLP-E-D03-2 likely encodes a serine/threonine-protein kinase. Their inter- action suggests that LP and BW are closely related dur- ing development (Figs. 7, and 8). Additional studies are needed to clarify why LP and BW are negatively related. Many candidates of the six QTLs are involved in fiber development The MYB-bHLH-WD40 (including MYB-DEL-TTG and CPC-MYC-TTG) [33, 68] and TCP-HOX-HD [66, 69] regulatory complexes play key roles in cotton fiber development. Phytohormone balance, Ca2+ signaling, and ROS also play key roles regulating fiber develop- ment [50, 70, 71]. Many candidate genes of the QTLs for LP and BW are involved in various signaling pathways and metabolic processes in this study, such as the transcription factor bHLH113 gene (Gh_A02G0095); Ca2+ signaling genes (Gh_A10G1519, Gh_D03G1058, and Gh_D03G1266); protein kinase genes (Gh_D03G1144, Gh_D03G1264, and Gh_D03G1069); GA signaling genes (Gh_A02G0104 and Gh_A02G0106); and ROS metabolism-related genes (Gh_D03G1138, Gh_D03G1063, and Gh_D03G1062) [55] (Table S7). Gh_D03G1264 encodes a HERK1-like protein [48]. Gh_A02G0106 is a homolog of AT1G14920, that encodes a gibberellin insensitive protein (DELLA protein GAI), and plays a role in seed germination [72]. Gh_A02G0111 is a homolog of AT2G43410, which encodes a flowering time control protein FPA in Arabi- dopsis [73]. Gh_D03G1064 encodes a FRIGIDA-like pro- tein that can pleiotropically increase lint yield; it is also significantly associated with SI [5]. The homologous gene of Gh_D03G1064 in Arabidopsis is FRI (AT4G00650), which regulates flowering time in Arabidopsis [73–77]. GhFSN1 is a cotton NAC transcription factor that acts as a positive regulator to control secondary cell wall (SCW) formation in cotton fibers by activating down- stream SCW-related genes, including GhDUF231L1, GhKNL1, GhMYBL1, GhGUT1 and GhIRX12 [66]. The candidate gene Gh_A02G0101 also encodes a NAC (Table S7). The glucosyltransferases, Rab- protein like GTPase activators, and myotubularin (GRAM) domain gene GhGRAM31 (Ghir_D02G018120) regulate fiber elongation. GhGRAM31 directly interacts with GhGRAM5 and GhGRAM35. GhGRAM5 also interacts with the transcription factor GhTTG1, and GhGRAM35 interacts with the transcription factors GhHOX1 and GhHD1 [67]. The candidate gene Gh_A02G0094 also encodes the C2 and GRAM domain-containing protein At1g03370 (Table S7). The above data demonstrate that most of the putative candidates of the six QTLs for LP and BW identified in this study were involved in regulating cotton fiber devel- opment. Most of the data obtained in this research are consistent with the findings of other studies, indicating that our results were reliable. Candidate gene expression profiles determine LP and BW ZR014121 is an excellent high-yield but low-LP line. EZ60 is an early maturity line with high LP. The candidate gene expression profiles of the six QTLs for LP and BW in the two lines significantly differed (Fig. 6). Most can- didate genes were highly expressed at the ovule develop- mental stage (0 DPA) in both ZR014121 and EZ60. Four Niu et al. BMC Plant Biology (2023) 23:179 Page 14 of 18 key candidate genes were highly expressed at 5 DPA in ZR014121, including Gh_A02G0095 (BHLH113, which might be involved in MYB-bHLH-WD40 complexes [33, 68]), Gh_A02G0097 (RGA3), Gh_A10G1158 (CBDAS), and Gh_D03G1062 (RBOHC, which might be involved in ROS [70]). Gh_A02G0114 (ccdc94) was significantly highly expressed at 15 DPA in EZ60. Gh_A02G0101 (NAC014, which might be involved in SCW formation in cotton fibers [66]) was significantly highly expressed at 25 DPA in ZR014121. Most genes were highly expressed at the ovule devel- opmental stage, which demonstrates that these genes were highly active in this stage. The expression of four genes in ZR014121 after this stage was likely the main cause of high yield. These four genes, in addition to the other two highly expressed genes, Gh_A02G0114 and Gh_ A02G0101, were the key candidate genes of the six QTLs for LP and BW (Fig. 6). Although we were unable to deter- mine whether the six genes represent the six QTLs, our findings indicate that they are the key genes regulating LP and BW and thus affecting cotton yield. These genes provide important genetic resources for studies of the lint regulation mechanism and improvements in cotton yield. Conclusions Two RIL populations were constructed using the three excellent upland cotton lines ZR014121, CCRI60, and EZ60, which differ in fiber yield and quality traits. The RILs were genotyped by GBTS and phenotyped under four different environments; a GWAS was then conducted to identify useful yield-related QTLs. A total of 51 QTLs for LP and 49 QTLs for BW were identified, and these QTLs could explain 0.29–9.96% of the phenotypic varia- tion in LP and 0.41–6.31% of the phenotypic variation in BW. There were six major and effective QTLs, three for LP and three for BW, and these could be used to breed cotton with high yield via molecular breeding approaches. A total of 108 putative candidate genes were identified in the six key QTLs, including genes that were positively related to the development of LP and BW, such as genes involved in gene transcription, protein synthesis, calcium signaling, phytohormone synthesis and signaling, and fiber synthesis-related polysaccharide metabolism. Seven of the candidate genes form a co-expression network. Six significantly highly expressed candidate genes after anthe- sis were important factors regulating cotton yield. These candidate genes will help clarify the molecular mecha- nisms underlying variation in LP and BW. Methods Plant materials and growth conditions Three G. hirsutum lines ZR014121, CCRI60, and EZ60 were used as parents in this study, and they were bred at the Institute of Cotton Research, Chinese Academy of Agricultural Sciences. All of the three RIL lines we were authorized to use. EZ60 and ZR014121 were preserved in the National Germplasm Library (38 Huanghe Ave- nue, Anyang, Henan 455,000); their accession numbers were M116025 and ZM115357, respectively. CCRI60 is a variety. ZR014121 has high yield but low LP. EZ60 is an early maturity line with high LP. CCRI60 is an excellent cultivar with several desirable agronomic traits. Two RIL populations at the F6:8 generation in 2020 (at F6:9 in 2021), P-CCRI60 and P-EZ60 were constructed from crosses of ZR014121 × CCRI60 and ZR014121 × EZ60, respectively. P-CCRI60 consisted of 300 RILs, and P-EZ60 consisted of 200 RILs. There were four factors in the field experiment: two years (2020 and 2021) and two locations (Anyang (36°05′N, 114°29′E), Henan Province, and Weixian (37°58′N, 115°16′E), Hebei Province, China(both of them are our experimental field)); these were each referred to as 20AY, 20WX, 21AY, and 21WX. To eliminate field effects, the experiment was conducted in a randomized incomplete block design with two replicates of each envi- ronmental factor. The parents and RILs were planted in rows with lengths of 3 m and widths of 0.8 m; the one control, CCRI60, had 20 rows. The lines were planted in April and sampled in September each year. Field man- agement techniques followed those of regular breeding practices. Trait measurements Two yield-related traits LP and BW were evaluated at each field location. The samples were prepared around September 20 each year. Thirty naturally opened bolls from the central part of plants (two bolls on each plant) of each line were randomly hand-harvested to calculate the BW (g) and gin the fiber. Fiber samples were sepa- rately weighed to calculate the LP (%). All statistical anal- yses, including correlations between traits, analysis of variance and significance analyses were conducted using IBM SPSS 22.0 software. GBTS For genotyping, the young leaf tissues of the three par- ents ZR014121, CCRI60, and EZ60, and the RILs of the two populations, P-CCRI60 and P-EZ60, were sampled in July 2020. Genomic DNA was extracted from each sam- ple using a modified cetyltrimethylammonium bromide method [78]. For GBTS, we used the Allegro Targeted Genotyping of NuGEN Technologies; the stable markers covering whole cotton genomes were selected from known markers obtained from the high-throughput sequencing results. To prevent the 3′-ends of the probes from overlapping Niu et al. BMC Plant Biology (2023) 23:179 Page 15 of 18 with other known variable sites, the SNPs were tested in the parents and their F1 plants, and the polymorphic SNPs were used to design primers. DNA fragmentation, adapter ligation, target extension, and library amplifica- tion were performed following the instructions of vari- ous kits (NuGEN Technologies, San Carlos, California, USA). The libraries were tested using the most recently updated Illumina manufacturer’s instructions (Illumina, San Diego, CA, USA). Three replications of GBTS were performed on each sample. After the SNP data were generated by BCFtools, the raw SNPs and Indels were screened using three parameters QUAL, RPB, and AC [(-e ‘%QUAL < 100); (RPB < 0.1, %QUAL < 100); (AC < 2, %QUAL < 100)’)]. The cover rate of each sequenced SNP was statisti- cally analyzed using ‘samtools depth’. The SNPs with sequencing cover rates more than 10 times and without genotypes were considered to be genotypes consistent with those in the cotton reference genome; SNPs with sequencing cover rates less than one time and without genotypes were referred to as deletion genotypes. The two SNP quality control criteria were (1) call rate of a single locus and (2) call rate of an individual. The Perl soft program that we translated and edited was used to statistically analyze the quality control criteria. For the physical localizations of the SNP markers, the probe sequences of the SNPs were used to\| perform local BLAST [79] queries against the G. hirsutum TM-1 ref- erence genome [28, 52]. GWAS The high-quality SNPs determined from the whole study populations, P-CCRI60 and P-EZ60, were used to conduct a GWAS for LP and BW. Given the possibility of obtain- ing false-positive QTNs with low association frequencies, we selected QTNs with LOD > 3 as stable QTNs in subse- quent analyses. The software 3VmrMLM version 1.0 [80] was used to perform GWAS with the following settings: method = ‘Multi_env’; fileKin = NULL; filePS = NULL; PopStrType = ‘Q’; fileCov = NULL; SearchRadius = 20; svpal = 0.01; DrawPlot = TRUE; Plotformat = ‘pdf’; and Chr_name_com = NULL. We obtained significant and suggested main-effect QTNs, significant, as well as sug- gested QEIs. The significant QTNs were selected by Bon- ferroni correction, and the critical P-value was 0.05/m, where m is the number of tests or markers, and suggested QTNs were identified as those with LOD ≥ 3.0. Significant QEIs were selected by Bonferroni correction; the critical P-value was 0.05/m, where m is the number of tests or markers, and suggested QEIs were identified as those with LOD ≥ 3.0 using default parameters [80]. Prediction and identification of candidate genes We defined the flanking 200-Kb regions of the QTNs as the same QTL and merged the overlapping QTLs to confirm the number of QTLs [81]. Potential candidate genes were confirmed based on gene annotations in the G. hirsutum TM-1 genome [28, 52]. All the candidate genes were subjected to Gene Ontology [82] enrichment analysis and Kyoto Encyclopedia of Genes and Genomes [83–85] analysis. The interaction network of candidate genes was inferred by constructing a PPI network using the STRING database [52]. The network analysis was conducted using Cytoscape 3.7.2. RNA sequencing and gene expression profiles of the QTL candidates The ovules/fibers of EZ60 and ZR014121 were sampled at 0, 5, 10, 15, 20, and 25 days post-anthesis (DPA). The total RNAs were extracted using the mirVana™ miRNA Isola- tion Kit (Ambion) according to the manufacturer’s instruc- tions. Three biological replicates were performed for each sample. The Illumina PE libraries were sequenced on the HiSeqTM2500 (Illumina) platform. Raw reads were fil- tered using Trimmomatic-0.39 [86], and the clean reads were mapped to the reference genome [87] using STAR- 2.7.9a [88]; the abundances of transcripts were quantified using RSEM-1.2.26 [89]. Differentially expressed genes (DEGs) were identified using DESeq2-1.30.1 according (FoldChange) > 1 to the following criteria: padj < 0.05 and log2 DESeq2-1.30.1 [90]. Hierarchical cluster analysis of DEGs was conducted to measure expression levels. The expres- sion profiles of every candidate gene were used to prelimi- narily identify LP-related and BW-related genes. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12870- 023- 04147-5. Additional file 1:Table S1. The result of GBTS Additional file 2:Table S2. The results of sample genotyping Additional file 3:Table S3. The result of 3VmrMLM: QEI Additional file 4:Table S4. The identified QTLs Additional file 5:Table S5. The identified QTLs overlapped with the reported QTLs Additional file 6:Table S6. The identified new QTLs Additional file 7:Table S7. All candidate genes of the 6 key QTLs Additional file 8:Table S8. Annotations of the candidate genes of the six QTLs for BW and LP Additional file 9:Table S9. KEGG annotations of the candidate genes of the QTLs for LP Additional file 10:Table S10. KEGG annotations of the candidate genes of the QTLs for BW Additional file 11:Table S11. The expression levels of the candidate genes Niu et al. BMC Plant Biology (2023) 23:179 Page 16 of 18 Acknowledgements We thank the reviewers for comments and suggestions on improving the manuscript. Authors’ contributions H.N. performed the experiments, analyzed the data, and drafted the manu- script. M.K and L.H. helped with analysis of the data. H.S., Y.Y and Q.G designed the whole study, revised the manuscript and gave the final approval to the version of the manuscript that is being sent for consideration for publication. Funding This work was supported by funding from the National Key Research and Development Program (2021YFF1000100) and Agricultural Science and Tech- nology Innovation Program of Chinese Academy of Agricultural Sciences. Availability of data and materials The datasets generated and/or analyzed during the current study are available in the NCBI repository, [https:// www. ncbi. nlm. nih. gov/ biopr oject/ 906276] [Accession number: PRJNA906276]. Declarations Ethics approval and consent to participate We complied with all relevant institutional, national and international guide-lines with permissions from State Key Laboratory of Cotton Biology, Key Laboratory of Biological and Genetic Breeding of Cotton, The Ministry of Agriculture, Institute of Cotton Research, Chinese Academy of Agricultural Sciences. Consent for publication Not applicable. Competing interests The authors declare there are no competing interests. Author details 1 State Key Laboratory of Cotton Biology, Key Laboratory of Biological and Genetic Breeding of Cotton, Institute of Cotton Research, The Ministry of Agriculture, Chinese Academy of Agricultural Sciences, Anyang 455000, Henan, China. 2 Zhengzhou Research Base, State Key Laboratory of Cotton Biol- ogy, Zhengzhou University, Zhengzhou 450001, Henan, China. Received: 9 December 2022 Accepted: 1 March 2023 References 1. Chen ZJ, Scheffler BE, Dennis E, Triplett BA, Zhang T, Guo W, Chen X, Stelly DM, Rabinowicz PD, Town CD, Arioli T, Brubaker C, Cantrell RG, Lacape JM, Ulloa M, Chee P, Gingle AR, Haigler CH, Percy R, Saha S, Wilkins T, Wright RJ, Van Deynze A, Zhu Y, Yu S, Abdurakhmonov I, Katageri I, Kumar PA, Rahman, M-U, Zafar Y, Yu JZ, Kohel R J, Wendel JF, Paterson AH. Toward sequencing cotton (Gossypium) genomes. Plant Physiol. 2007;145:1303–10. Imran M, Shakeel A, Azhar FM, Farooq J, Saleem MF, Saeed A, Nazeer W, Riaz M, Naeem M, Javaid A. Combining ability analysis for within-boll yield components in upland cotton (Gossypium hirsutum L.). Genet Mol Res. 2012;11(3):2790–800. 2. 3. Qin H, Guo W, Zhang YM, Zhang T. QTL mapping of yield and fiber traits based on a four-way cross population in Gossypium hirsutum L. Theor Appl Genet. 2008;117:883–94. 4. Wang B, Guo W, Zhu X, Wu Y, Huang N, Zhang T. QTL Mapping of yield and yield components for elite hybrid derived-RILs in upland cotton. J Genet Genomics. 2007;34(1):35–45. Sun Z, Wang X, Liu Z, Gu Q, Zhang Y, Li Z, Ke H, Yang J, Wu Ji, Wu L, Zhang G, Zhang C, Ma Z. A genome wide association study uncovers novel genomic regions and candidate genes of yield related traits in upland cotton. Theor Appl Genet. 2018;131(11):2413–25. 5. 7. 6. Badigannavar A, Myers GO. Genetic diversity, population structure and marker trait associations for seed quality traits in cotton (Gossypium hirsutum). J Genet. 2015;94(1):87–94. Tang F, Xiao W. Genetic effects and heterosis of within-boll yield components in upland cotton (Gossypium hirsutum L.). Euphytica. 2013;194:45–51. Su J, Fan S, Li L, Wei H, Wang C, Wang H, Song M, Zhang C, Gu L, Zhao S, Mao G, Wang C, Pang C, Yu S. Detection of favorable QTL alleles and candidate genes for lint percentage by GWAS in Chinese Upland cotton. Front Plant Sci. 2016;7:1576. Jiang Y, Guo W, Zhu H, Ruan YL, Zhang T. Overexpression of GhSusA1 increases plant biomass and improves cotton fiber yield and quality. Plant Biotechnol J. 2012;10:301–12. 9. 8. 10. Fang L, Wang Q, Hu Y, Jia Y, Chen J, Liu B, Zhang Z, Guan X, Chen S, Zhou B, Mei G, Sun J, Pan Z, He S, Xiao S, Shi W, Gong W, Liu J, Ma J, Cai C, Zhu X, Guo W, Du X, Zhang T. Genomic analyses in cotton identify signatures of selection and loci associated with fiber quality and yield traits. Nat Genet. 2017;49(7):1089–98. 11. Song C, Li W, Pei X, Liu Y, Ren Z, He K, Zhang F, Sun K, Zhou X, Ma X, Yang D. Dissection of the genetic variation and candidate genes of lint percentage by a genome-wide association study in upland cotton. Theor Appl Genet. 2019;132(7):1991–2002. 12. Su J, Wang C, Ma Q, Zhang A, Shi C, Liu J, Zhang X, Yang D, Ma X. An RTM- GWAS procedure reveals the QTL alleles and candidate genes for three yield related traits in upland cotton. BMC Plant Biol. 2020;20(1):416. 13. Zhang Z, Li J, Jamshed M, Shi Y, Liu A, Gong J, Wang S, Zhang J, Sun F, Jia F, Ge Q, Fan L, Zhang Z, Pan J, Fan S, Wang Y, Lu Q, Liu R, Deng X, Zou X, Jiang X, Liu P, Li P, Iqbal MS, Zhang C, Zou J, Chen H, Tian Q, Jia X, Wang B, Ai N, Feng G, Wang Y, Hong M, Li S, Lian W, Wu B, Hua J, Zhang C, Huang J, Xu A, Shang H, Gong W, Yuan Y. Genome-wide quantitative trait loci reveal the genetic basis of cotton fibre quality and yield-related traits in a Gossypium hirsutum recombinant inbred line population. Plant Biotech- nol J. 2020;18(1):239–53. 14. Wang F, Zhang J, Chen Y, Zhang C, Gong J, Song Z, Zhou J, Wang J, Zhao C, Jiao M, Liu A, Du Z, Yuan Y, Fan S, Zhang J. Identification of candidate genes for key fibre-related QTLs and derivation of favourable alleles in Gossypium hirsutum recombinant inbred lines with G. barbadense introgressions. Plant Biotechnol J. 2020;18:707–20. 15. Zhu G, Hou S, Song X, Wang X, Wang W, Chen Q, Guo W. Genome-wide association analysis reveals quantitative trait loci and candidate genes involved in yield components under multiple field environments in cot- ton (Gossypium hirsutum). BMC Plant Biol. 2021;21:250. 16. Zeng L, Wu J. Germplasm for genetic improvement of lint yield in Upland cotton: genetic analysis of lint yield with yield components. Euphytica. 2012;187:247–61. 17. Du XM, Pan JJ, Wang RH, Zhang TZ, Shi YZ. Genetic analysis of presence and absence of lint and fuzz in cotton. Plant Breeding. 2001;120:519–22. 18. Han Z, Hu Y, Tian Q, Cao Y, Si A, Si Z, Zang Y, Xu C, Shen W, Dai F, Liu X, Fang L, Chen H, Zhang T. Genomic signatures and candidate genes of lint yield and fibre quality improvement in Upland cotton in Xinjiang. Plant Biotechnol J. 2020;18:2002–14. 19. Shi Y, Li W, Li A, Ge R, Zhang B, Li J, Liu G, Li J, Liu A, Shang H, Gong J, Gong W, Yang Z, Tang F, Liu Z, Zhu W, Jiang J, Yu X, Wang T, Wang W, Chen T, Wang K, Zhang Z, Yuan Y. Constructing a high-density linkage map for Gossypium hirsutum percentage. J Integr Plant Biol. 2015;57(5):450–67. Gossypium barbadense and identifying QTLs for lint × 20. Li PT, Rashid MHO, Chen TT, Lu QW, Ge Q, Gong WK, Liu AY, Gong JW, Shang HH, Deng XY, Li JW, Li SQ, Xiao XH, Liu RX, Zhang Q, Duan L, Zou XY, Zhang Z, Jiang X, Zhang Y, Peng RH, Shi YZ, Yuan YL. Transcriptomic and biochemical analysis of upland cotton (Gossypium hirsutum) and a chromosome segment substitution line from G. hirsutum in response to Verticillium dahliae infection. BMC Plant Biol. 2019;19(1):19. G. barbadense × 21. Zhu XF, Wang P, Si ZF, Zhang TZ. QTL mapping for yield components in Gossypium barbadense chromosome segment introgression lines based on Gossypium hirsutum background. Acta Agr Sin (in Chinese). 2017;43(12):1784–90. 22. Wang SP, Xiao SD. Breeding advance of the new cotton line C24 with high lint percentage. Chin Agric Sci Bull (in Chinese). 1996;12(6):27–8. 23. Ma X, Wang Z, Li W, Zhang Y, Zhou X, Liu Y, Ren Z, Pei X, Zhou K, Zhang W, He K, Zhang F, Liu J, Ma W, Xiao G, Yang D. Resequencing core acces- sions of a pedigree identifies derivation of genomic segments and key Niu et al. BMC Plant Biology (2023) 23:179 Page 17 of 18 agronomic trait loci during cotton improvement. Plant Biotechnol J. 2019;17(4):762–75. 24. Li M, Wang ZZ. Characterization, screening and enhancement of cotton germplasm. Zuo Wu Pin Zhong Zi Yuan (in Chinese). 1992;3:11–2. 25. Chen Y, Liu G, Ma H, Song Z, Zhang C, Zhang J, Zhang J, Wang F, Zhang J. Identification of introgressed alleles conferring high fiber quality derived from Gossypium barbadense L. in secondary mapping populations of G. hirsutum L. Front Plant Sci. 2018;9:1023. 26. Zhang Z, Shang H, Shi Y, Huang L, Li J, Ge Q, Gong J, Liu A, Chen T, Wang D, Wang Y, Palanga KK, Muhammad J, Li W, Lu Q, Deng X, Tan Y, Song W, Cai J, Li P, Rashid H, Gong W, Yuan Y. Construction of a high-density genetic map by specific locus amplified fragment sequencing (SLAF-seq) and its application to Quantitative Trait Loci (QTL) analysis for boll weight in upland cotton (Gossypium hirsutum.). BMC Plant Biol. 2005;16:79. 27. Said JI, Song M, Wang H, Lin Z, Zhang X, Fang DD, Zhang J. A comparative meta-analysis of QTL between intraspecific Gossypium hirsutum and inter- specific G. hirsutum × 2015;290(3):1003–25. G. barbadense populations. Mol Genet Genomics. 28. Zhang T, Hu Y, Jiang W, Fang L, Guan X, Chen J, Zhang J, Saski CA, Scheffler BE, Stelly DM, Hulse-Kemp AM, Wan Q, Liu B, Liu C, Wang S, Pan M, Wang Y, Wang D, Ye W, Chang L, Zhang W, Song Q, Kirkbride RC, Chen X, Dennis E, Llewellyn DJ, Peterson DG, Thaxton P, Jones DC, Wang Q, Xu X, Zhang H, Wu H, Zhou L, Mei G, Chen S, Tian Y, Xiang D, Li X, Ding J, Zuo Q, Tao L, Liu Y, Li J, Lin Y, Hui Yu, Cao Z, Cai C, Zhu X, Jiang Z, Zhou B, Guo W, Li R, Chen ZJ. Sequencing of allotetraploid cotton (Gossypium hirsutum L. acc. TM-1) pro- vides a resource for fiber improvement. Nat Biotechnol. 2015;33(5):531–7. 41. Liu R, Wang B, Guo W, Qin Y, Wang L, Zhang Y, Zhang T. Quantitative trait loci mapping for yield and its components by using two immortalized populations of a heterotic hybrid in Gossypium hirsutum L. Mol Breeding. 2012;29:297–311. 42. Wu J, Gutierrez OA, Jenkins JN, McCarty JC, Zhu J. Quantitative analysis and QTL mapping for agronomic and fiber traits in an RI population of upland cotton. Euphytica. 2009;165:231–45. 43. Li C, Dong Y, Zhao T, Li L, Li C, Yu E, Mei L, Daud MK, He Q, Chen J, Zhu S. Genome-wide SNP linkage mapping and QTL analysis for fiber quality and yield traits in the upland cotton recombinant inbred lines popula- tion. Front Plant Sci. 2016;7:1356. 44. Diouf L, Magwanga RO, Gong W, He S, Pan Z, Jia YH, Kirungu JN, Du X. QTL mapping of fiber quality and yield-related traits in an intra-specific upland cotton using genotype by sequencing (GBS). Int J Mol Sci. 2018;19:441. 45. Huang C, Nie X, Shen C, You C, Li W, Zhao W, Zhang X, Lin Z. Population structure and genetic basis of the agronomic traits of upland cotton in China revealed by a genome-wide association study using high-density SNPs. Plant Biotechnol J. 2017;15(11):1374–86. 46. Shen C, Wang N, Huang C, Wang M, Zhang X, Lin Z. Population genomics reveals a fine-scale recombination landscape for genetic improvement of cotton. Plant J. 2019;99(3):494–505. 47. Zhu G, Gao W, Song X, Sun F, Hou S, Liu N, Huang Y, Zhang D, Ni Z, Chen Q, Guo W. Genome-wide association reveals genetic variation of lint yield components under salty field conditions in cotton (Gossypium hirsutum L.). BMC Plant Biol. 2020;20(1):23. 29. Wang M, Tu L, Yuan D, Zhu D, Shen C, Li J, Liu F, Pei L, Wang P, Zhao G, 48. Yu J, Hui Y, Chen J, Yu H, Gao X, Zhang Z, Li Q, Zhu S, Zhao T. Whole- Ye Z, Huang H, Yan F, Ma Y, Zhang L, Liu M, You J, Yang Y, Liu Z, Huang F, Li B, Qiu P, Zhang Q, Zhu L, Jin S, Yang X, Min L, Li G, Chen LL, Zheng H, Lindsey K, Lin Z, Udall JA, Zhang X. Reference genome sequences of two cultivated allotetraploid cottons. Gossypium hirsutum and Gossypium barbadense, Nat Genet. 2019;51:224–9. 30. Huang G, Wu Z, Percy RG, Bai M, Li Y, Frelichowski JE, Hu J, Wang K, Yu JZ, Zhu Y. Genome sequence of Gossypium herbaceum and genome updates of Gossypium arboreum and Gossypium hirsutum provide insights into cot- ton A-genome evolution. Nat Genet. 2020;52:516–24. 31. Ma Z, Zhang Y, Wu L, Zhang G, Sun Z, Li Z, Jiang Y, Ke H, Chen B, Liu Z, Gu Q, Wang Z, Wang G, Yang J, Wu J, Yan Y, Meng C, Li L, Li X, Mo S, Wu N, Ma L, Chen L, Zhang M, Si A, Yang Z, Wang N, Wu L, Zhang D, Cui Y, Cui J, Lv X, Li Y, Shi R, Duan Y, Tian S, Wang X. High-quality genome assembly and resequencing of modern cotton cultivars provide resources for crop improvement. Nat Genet. 2021;53(9):1385–91. 32. Zhang ZS, Xiao YH, Luo M, Li XB, Luo XY, Hou L, Li DM, Pei Y. Construction of a genetic linkage map and QTL analysis of fiber-related traits in upland cotton (Gossypium hirsutum L.). Euphytica. 2005;144:91–9. 33. Liu D, Liu F, Shan X, Zhang J, Tang S, Fang X, Liu X, Wang W, Tan Z, Teng Z, Zhang Z, Liu D. Construction of a high-density genetic map and lint per- centage and cottonseed nutrient trait QTL identification in upland cotton (Gossypium hirsutum L.). Mol Genet Genomics. 2015;290:1683–700. 34. Zhang K, Kuraparthy V, Fang H, Zhu L, Sood S, Jones DC. High-density linkage map construction and QTL analyses for fiber quality, yield and morphological traits using CottonSNP63K array in upland cotton (Gos- sypium hirsutum L.). BMC Genomics. 2019;20:889. 35. Lin ZX, Feng CH, Guo XP, Zhang XL. Genetic analysis of major QTLs and epistasis interaction for yield and fiber quality in upland cotton. Sci Agric Sin (in Chinese). 2009;42(9):3036–47. 36. Lin ZX, Zhang YX, Zhang XL, Guo XP. A high-density integrative linkage map for Gossypium hirsutum. Euphytica. 2009;166:35–45. 37. Wang H, Huang C, Zhao W, Dai B, Shen C, Zhang B, Li D, Lin Z. Identifica- tion of QTL for fiber quality and yield traits using two immortalized back- cross populations in upland cotton. PLoS ONE. 2016;11(12): e0166970. 38. Xia Z, Zhang X, Liu YY, Jia ZF, Zhao HH, Li CQ, Wang QL. Major gene iden- tification and quantitative trait locus mapping for yield-related traits in upland cotton (Gossypium hirsutum L.). J Integr Agr. 2014;13(2):299–309. genome resequencing of 240 Gossypium barbadense accessions reveals genetic variation and genes associated with fiber strength and lint percentage. Theor Appl Genet. 2021;134:3249–61. 49. Feng L, Su Q, Yue H, Wang L, Gao J, Xing L, Xu M, Zhou C, Yang Y, Zhou B. TIP41L, a putative candidate gene conferring both seed size and boll weight, was fine-mapped in an introgression line of Gossypium hirsutum- Gossypium arboretum. Plant Sci. 2022;317: 111197. 50. Niu H, Ge Q, Shang H, Yuan Y. Heredity. QTL mapping and candidate gene isolation of cotton lint percentage, Front Genet. 2022;13: 855574. 51. Hu Y, Chen JD, Fang L, Zhang ZY, Ma W, Niu YC, Ju LZ, Deng JQ, Zhao T, Lian JM, Baruch K, Fang D, Liu X, Ruan YL, Rahman MU, Han JL, Wang K, Wang Q, Wu HT, Mei GF, Zang YH, Han ZG, Xu CY, Shen WJ, Yang DF, Si ZF, Dai F, Zou LF, Huang F, Bai YL, Zhang YG, Brodt A, Ben-Hamo H, Zhu XF, Zhou BL, Guan XY, Zhu SJ, Chen XY, Zhang TZ. Gossypium barbadense and Gossypium hirsutum genomes provide insights into the origin and evolu- tion of allotetraploid cotton. Nat Genet. 2019;51(4):739–48. 52. STRING database: https:// cn. string- db. org. Accessed 21 Sept 2022. 53. CottonGen Database: https:// www. cotto ngen. org. Accessed 21 Sept 2022. 54. Yu J, Jung S, Cheng CH, Lee T, Zheng P, Buble K, Crabb J, Humann J, Hough H, Jones D, Campbell JT, Udall J, Main D. CottonGen: The com- munity database for cotton genomics, genetics, and breeding research. Plants. 2021;10:2805. 55. Chen Y, Gao Y, Chen P, Zhou J, Zhang C, Song Z, Huo X, Du Z, Gong J, Zhao C, Wang S, Zhang J, Wang F, Zhang J. Genome-wide association study reveals novel quantitative trait loci and candidate genes of lint percentage in upland cotton based on the CottonSNP80K array. Theor Appl Genet. 2022;135(7):2279–95. 56. Gowda SA, Katageri IS, Patil RS, Kumar PS, Tiwari GJ, Jena SN, Sawant SV. 63 K and 50 K SNP array based high-density genetic mapping and QTL analysis for productivity and fiber quality traits in cotton. Euphytica. 2022;218:93. 57. Lu Q, Li P, Yang R, Xiao X, Li Z, Wu Q, Gong J, Ge Q, Liu A, Du S, Wang J, Shi Y, Yuan Y. QTL mapping and candidate gene prediction for fiber yield and quality traits in a high-generation cotton chromosome substitu- tion line with Gossypium barbadense segments. Mol Genet Genomics. 2022;297:287–301. 39. Wang M, Li C, Wang Q. Quantitative trait loci mapping and genetic 58. Si Z, Jin S, Chen J, Wang S, Fang L, Zhu X, Zhang T, Hu Y. Construction of dissection for lint percentage in upland cotton (Gossypium hirsutum). J Genet. 2014;93:371–8. 40. Liu R, Wang B, Guo W, Wang L, Zhang T. Differential gene expression and associated QTL mapping for cotton yield based on a cDNA- AFLP transcriptome map in an immortalized F2. Theor Appl Genet. 2011;123:439–54. a high-density genetic map and identification of QTLs related to agro- nomic and physiological traits in an interspecific (Gossypium hirsutum Gossypium barbadense) F2 population. BMC Genomics. 2022;23:307. 59. Zeng L, Meredith WR Jr, Gutiérrez OA, Boykin DL. Identification of asso- ciations between SSR markers and fiber traits in an exotic germplasm × Niu et al. BMC Plant Biology (2023) 23:179 Page 18 of 18 derived from multiple crosses among Gossypium tetraploid species. Theor Appl Genet. 2009;119:93–103. 60. Santos IG, Teodoro PE, Farias FC, Farias FJC, Carvalho LP, Rodrigues JIS, detecting QTNs, and QTN-by-environment and QTN-by-QTN interactions in genome-wide association studies. Mol Plant. 2022;15:630–50. 81. Su Y, Guo A, Huang Y, Wang Y, Hua J. GhCIPK6a increases salt tolerance Cruz CD. Genetic diversity among cotton cultivars in two environments in the State of Mato Grosso. Genet Mol Res. 2017;16(2):16029628. in transgenic upland cotton by involving in ROS scavenging and MAPK signaling pathways. BMC Plant Biol. 2020;20(1):421. 61. Rehman A, Mustafa N, Du X, Azhar MT. Heritability and correlation 82. Gene Ontology (GO; http:// www. geneo ntolo gy. org/ GO. Accessed 21 Sept 2022. 83. Kyoto Encyclopedia of Genes and Genomes: (KEGG; http:// www. genome. jp/ kegg/ KEGG. Accessed 21 Sept 2022. 84. Kanehisa M, Goto S. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 2000;28:27–30 ([PMID:10592173]). 85. Kanehisa M, Furumichi M, Sato Y, Ishiguro-Watanabe M, Tanabe M. KEGG: integrating viruses and cellular organisms. Nucleic Acids Res. 2021;49:D545–51 ([PMID:33125081]). 86. Bolger AM, Lohse M, Usadel B. Trimmomatic: a fexible trimmer for Illu- mina sequence data. Bioinformatics. 2014;30:2114–20. 87. The cotton reference genome: https:// cotto nfgd. org/ about/ downl oad/ assem bly/ genome. Ghir. ZJU. fa. gz. Accessed 21 Sept 2019. 88. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M, Gingeras TR. STAR: ultrafast universal RNA-seq aligner. Bioin- formatics. 2013;29:15–21. 89. Dewey CN, Li B. RSEM: accurate transcript quantifcation from RNA- Seq data with or without a reference genome. BMC Bioinformatics. 2011;12:323. 90. Anders S, Huber W. Differential expression analysis for sequence count data. Genome Biol. 2010;11:R106. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. analysis of morphological and yield traits in genetically modified cotton. J Cotton Res. 2020;3:23. 62. Farias FJC, Carvalho LP, Silva Filho JL, Teodoro PE. Correlations and path analysis among agronomic and technological traits of upland cotton. Genet Mol Res. 2016;15(3):15038239. 63. Zhai H, Gong W, Tan Y, Liu A, Song W, Li J, Deng Z, Kong L, Gong J, Shang H, Chen T, Ge Q, Shi Y, Yuan Y. Identification of chromosome segment substitution lines of Gossypium barbadense introgressed in G. hirsutum and quantitative trait locus mapping for fiber quality and yield traits. PLoS ONE. 2016; 11 (9), e0159101. 64. Chen Q, Zhou SJ, Sun KT, Liu JJ, Yuan BT, Wang YP, Wang W, Wang YW, Wang BH, Zhuang ZM. QTL mapping of lint percentage in Gossypium mustelinum introgression lines. Southwest China J Agric Sci (in Chinese). 2019;32(8):1735–9. 65. Feng L, Chen Y, Xu M, Yang Y, Yue H, Su Q, Zhou C, Feng G, Ai N, Wang N, Zhou B. Genome-wide introgression and quantitative trait locus mapping reveals the potential of Asian cotton (Gossypium arboreum) in improving upland cotton (Gossypium hirsutum). Front Plant Sci. 2021;12: 719371. 66. Zhang J, Huang GQ, Zou D, Yan JQ, Li Y, Hu S, Li XB. The cotton (Gossypium hirsutum) NAC transcription factor (FSN1) as a positive regulator partici- pates in controlling secondary cell wall biosynthesis and modification of fibers. New Phytol. 2018;217:625–40. 67. Ye Z, Qiao L, Luo X, Chen X, Zhang X, Tu L. Genome-wide identification of cotton GRAM family proteins reveals that GRAM31 regulates fiber length. J Exp Bot. 2021;72(7):2477–90. 68. Shangguan X, Yang Q, Wu X, Cao J. Function analysis of a cotton R2R3 MYB transcription factor GhMYB3 in regulating plant trichome develop- ment. Plant Biol. 2021;23:1118–27. 69. Cao JF, Zhao B, Huang CC, Chen ZW, Zhao T, Liu HR, Hu GJ, Shangguan XX, Shan CM, Wang LJ, Zhang TZ, Wendel JF, Guan XY, Chen XY. The miR319-targeted GhTCP4 promotes the transition from cell elongation to wall thickening in cotton fiber. Mol Plant. 2020;13:1063–77. 70. Tang WX, Tu LL, Yang XY, Tan JF, Deng FL, Hao J, Guo K, Lindsey K, Zhang XL. The calcium sensor GhCaM7 promotes cotton fiber elongation by modulating reactive oxygen species (ROS) production. New Phytol. 2014;202(2):509–20. 71. Cheng Y, Lu L, Yang Z, Wu Z, Qin W, Yu D, Ren Z, Li Y, Wang L, Li F, Yang Z. GhCaM7-like, a calcium sensor gene, influences cotton fiber elongation and biomass production. Plant Physiol Biochem. 2016;109:128–36. 72. Oh E, Kang H, Yamaguchi S, Park J, Lee D, Kamiya Y, Choi G. Genome-wide analysis of genes targeted by phytochrome interacting factor 3-LIKE5 during seed germination in Arabidopsis. Plant Cell. 2009;21(2):403–19. 73. Seo E, Lee H, Jeon J, Park H, Kim J, Noh YS, Lee I. Crosstalk between cold response and flowering in Arabidopsis is mediated through the flowering-time gene SOC1 and its upstream negative regulator FLC. Plant Cell. 2009;21(10):3185–97. 74. Emami H, Kempken F. PRECOCIOUS1 (POCO1), a mitochondrial pentatri- copeptide repeat protein affects flowering time in Arabidopsis thaliana. Plant J. 2019;100(2):265–78. 75. Schmalenbach I, Zhang L, Ryngajllo M, Jiménez-Gómez JM. Functional analysis of the Landsberg erecta allele of FRIGIDA. BMC Plant Biol. 2014;14:218. 76. Liu DX, Rajaby R, Wei LL, Zhang L, Yang ZQ, Yang QY, Sung WK. Calling large indels in 1047 Arabidopsis with IndelEnsembler. Nucleic Acids Res. 2021;49(19):10879–94. 77. Zhang L, Jimenez-Gomez JM. Functional analysis of FRIGIDA using naturally occurring variation in Arabidopsis thaliana. Plant J. 2020;103(1):154–65. 78. Allen GC, Flores-Vergara MA, Krasnyanski S, Kumar S, Thompson WF. A modified protocol for rapid DNA isolation from plant tissues using cetyl- trimethylammonium bromide. Nat Protoc. 2006;1(5):2320–5. 79. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. J Mol Biol. 1990;215:403–10. 80. Li M, Zhang YW, Zhang ZC, Xiang Y, Liu MH, Zhou YH, Zuo JF, Zhang HQ, Chen Y, Zhang YM. A compressed variance component mixed model for • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12874-019-0884-8	Tervonen et al. BMC Medical Research Methodology (2019) 19:245 https://doi.org/10.1186/s12874-019-0884-8 R E S E A R C H A R T I C L E Open Access Using data linkage to enhance the reporting of cancer outcomes of Aboriginal and Torres Strait Islander people in NSW, Australia Hanna E. Tervonen, Stuart Purdie and Nicola Creighton* Abstract Background: Aboriginal people are known to be under-recorded in routinely collected datasets in Australia. This study examined methods for enhancing the reporting of cancer incidence among Aboriginal people using linked data methodologies. Methods: Invasive cancers diagnosed in New South Wales (NSW), Australia, in 2010–2014 were identified from the NSW Cancer Registry (NSWCR). The NSWCR data were linked to the NSW Admitted Patient Data Collection, the NSW Emergency Department Data Collection and the Australian Coordinating Register Cause of Death Unit Record File. The following methods for enhancing the identification of Aboriginal people were used: ‘ever-reported’, ‘reported on most recent record’, ‘weight of evidence’ and ‘multi-stage median’. The impact of these methods on the number of cancer cases and age-standardised cancer incidence rates (ASR) among Aboriginal people was explored. Results: Of the 204,948 cases of invasive cancer, 2703 (1.3%) were recorded as Aboriginal on the NSWCR. This increased with enhancement methods to 4184 (2.0%, ‘ever’), 3257 (1.6%, ‘most recent’), 3580 (1.7%, ‘weight of evidence’) and 3583 (1.7%, ‘multi-stage median’). Enhancement was generally greater in relative terms for males, people aged 25–34 years, people with cancers of localised or unknown degree of spread, people living in urban areas and areas with less socio-economic disadvantage. All enhancement methods increased ASRs for Aboriginal people. The weight of evidence method increased the overall ASR by 42% for males (894.1 per 100,000, 95% CI 844.5–945.4) and 27% for females (642.7 per 100,000, 95% CI 607.9–678.7). Greatest relative increases were observed for melanoma and prostate cancer incidence (126 and 63%, respectively). ASRs for prostate and breast cancer increased from below to above the ASRs of non-Aboriginal people with enhancement of Aboriginal status. Conclusions: All data linkage methods increased the number of cancer cases and ASRs for Aboriginal people. Enhancement varied by demographic and cancer characteristics. We considered the weight of evidence method to be most suitable for population-level reporting of cancer incidence among Aboriginal people. The impact of enhancement on disparities in cancer outcomes between Aboriginal and non-Aboriginal people should be further examined. Keywords: Neoplasms, Indigenous, Australia, Data linkage * Correspondence: nicola.creighton@health.nsw.gov.au Cancer Institute NSW, PO Box 41, Alexandria, Sydney, NSW 1435, Australia © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Tervonen et al. BMC Medical Research Methodology (2019) 19:245 Page 2 of 9 Background Aboriginal people are known to be under-recorded in routinely collected datasets [1–3]. Reasons for under- recording are complex and include a lack of awareness and training to ask about Aboriginal status among health staff, and among Aboriginal people concerns about how the question was asked, racism and discrimination, priv- acy, a lack of cultural safety and difficulties in tracing identity [4]. Under-recording of Aboriginal status gener- ally results in under-estimation of absolute measures of health indicators [5, 6]. It is possible to enhance reporting of health outcomes of Aboriginal people by linking data from several sources [7]. For example, Randall and colleagues showed that different enhancement methods using linked data in- creased the number of hospital admissions for Aborigi- nal people with varying impacts on admission and mortality ratios [6]. Several different methods for enhan- cing identification of Aboriginal people have been used, with no consensus on the optimal method. Australian guidelines on data linkage related to Aboriginal people recommend comparing the impact of several methods and choosing the optimal method based on the purpose of the analysis and characteristics of the datasets [7]. Aboriginal people are under-recorded in the New South Wales Cancer Registry (NSWCR) despite increased record- ing of Aboriginal status over time [3]. In the early 1980s, more than 80% of people on the NSWCR had unknown Aboriginal status, which had dropped to approximately 13% by 1999. A previous study examining the feasibility of en- hancement of reporting of Aboriginal people using linked data from several data sources, including NSWCR, found that the number of cancer cases, and hence cancer incidence, for Aboriginal people increased following enhancement [2]. Estimates of health outcomes among Aboriginal people and the size of disparities compared with non-Aboriginal people can change depending on how Aboriginal status is reported and which enhancement method is used [5, 6]. Ac- curate and complete recording of Indigenous status is needed to reliably measure cancer outcomes, identify dispar- ities and produce information about cancer among Indigenous people globally. Cancer registries are a key source of information for reporting cancer outcomes yet there are very few studies examining the impact of under- recording of Indigenous status on cancer incidence [8]. This study examined the impact of linked data enhancement methods on the number of cancer cases and cancer inci- dence rates among Aboriginal people in NSW, Australia, using common algorithms and population-based datasets. Methods Study design and data sources This was a retrospective cohort study using linked invasive routinely-collected health data. All cases of cancer diagnosed and recorded in the NSWCR between 2010 and 2014 were included in the analyses. The NSWCR is a statutory population-based cancer registry which collects information about all invasive cancers di- agnosed in NSW, Australia. Information about Aborigi- nal and Torres Strait Islander status in the NSWCR comes from multiple sources, such as hospital treatment episodes and death registration [3]. Pathology reports do not include information about Aboriginal and Torres Strait Islander status and, therefore, this information is missing if the NSWCR only receives a pathology notifi- cation. The NSWCR uses a progressive positive identifi- cation algorithm with a single notice from any source indicating a person to be Aboriginal or Torres Strait Is- lander taking precedence over any other information. Aboriginal and Torres Strait Islander status is assigned at a person level, rather than individual cancer case level. Torres Strait Islander people are included with Aborigi- nal people throughout this study due to the small num- ber of people from the Torres Strait Islands residing in NSW and in recognition that Aboriginal people are the original inhabitants of NSW [4]. The NSWCR data were linked to the NSW Admitted Pa- tient Data Collection (APDC), the NSW Emergency De- partment Data Collection (EDDC) and the Australian Coordinating Registry Cause of Death Unit Record File (COD URF). The APDC includes records of all hospital ad- missions in NSW public and private hospitals and day pro- information on the EDDC includes cedure centres, presentations to emergency departments of public hospitals in NSW, and the COD URF includes information about deaths occurring in NSW. Data linkage was performed by the Centre for Health Record Linkage (CHeReL). The CHeReL uses Choicemaker software to perform probabilis- tic linkage of personal identifiers using a privacy-preserving protocol (http://www.cherel.org.au). The datasets used in this study are in the CHeReL’s Master Linkage Key. The CHeReL implements quality assurance procedures and per- forms clerical review of a sample of records to keep the es- timated false positive and false negative linkage rate to less than 5 per 1000. The CHeReL provided a unique and arbi- trary “Project Person Number” which enabled the records in each study dataset to be joined for an individual without the researchers accessing personal identifiers. The APDC data covered a period between July 2001 and December 2017, the EDDC between January 2005 and De- cember 2017, and the COD URF between January 1985 and December 2015. Aboriginal status is self-reported in the APDC and EDDC and is provided by the next-of-kin in the COD URF. Population data were based on data from the Australian Bureau of Statistics and obtained through the Se- cure Analytics for Population Health Research and Intelligence (SAPHaRI) data warehouse (Centre for Epi- demiology and Evidence, NSW Ministry of Health). Tervonen et al. BMC Medical Research Methodology (2019) 19:245 Page 3 of 9 This project was approved by the NSW Population and Health Services Research Ethics Committee (HREC/ 15/CIPHS/15) and the Aboriginal Health and Medical Research Council Ethics Committee (HREC Ref. No. 1201/16). Subject matter advice and Aboriginal commu- nity input was sought from the Cancer Institute NSW Aboriginal Advisory Group. Enhancement methods The following methods for enhancing the reporting of ‘ever re- cancer among Aboriginal people were used: ported as Aboriginal’ [7], ‘Aboriginal on most recent rec- ord’ [7], ‘weight of evidence’ [2] and ‘multi-stage median’ [9] (Table 1). These methods were selected because they are among the most commonly used methods, represent a combination of simple and complex enhancement methods and are likely to provide a range of estimates. If a person was recorded as Aboriginal on the NSWCR or on the COD URF, a person was considered to be Abori- ginal in the analyses. Our aim was to correct for under- recording of Aboriginal people in the NSWCR, so we only considered changing the status of those recorded as non-Aboriginal or with unknown status in the NSWCR. We considered the risk of a person being wrongly in the COD URF to be low identified as Aboriginal since the information is provided by the next-of-kin. Otherwise the four enhancement methods were ap- plied to the data according to the descriptions pro- vided in Table 1. Statistical analysis The number, proportion and characteristics of cases re- ported as Aboriginal using the NSWCR information and the four enhancement methods were compared. Character- istics considered in this study were: sex, age at diagnosis, year of diagnosis, cancer site, degree of spread (localised, re- remoteness gional, distant, unknown) (major cities, inner regional, outer regional, remote/very re- mote) [11], and area-based socio-economic disadvantage residential [10], Table 1 The enhancement methods used in the analyses Method Description Ever reported [7] Recorded as being Aboriginal at least once in any of the data sources. Most recent record [7] Weight of evidence [2] Multi-stage median [9] Recorded as being Aboriginal in the most recent record in any of the data sources. Recorded as Aboriginal if 1) there are three or more units of information and at least two indicate that the person is Aboriginal; 2) if there are one or 2 units of information and at least one identifies the person as Aboriginal. The weight of evidence method is applied in a two- step process: firstly to each dataset individually; and then treating the results for each dataset as units of information. (Index of Relative Socio-economic Disadvantage quintiles) [12]. For descriptive analyses, cancer sites were classified using clinical cancer grouping [13]. considered Age-standardised cancer incidence rates (ASR) were calculated for non-Aboriginal and Aboriginal people using the NSWCR Aboriginal status variable before enhancement. Cases with unknown Aboriginal status For Aboriginal non-Aboriginal. were people, cancer incidence was also calculated using the variables created by the four enhancement methods. Direct age-standardisation was calculated using the 2001 Australian standard population and NSW popu- lation data based on data from the Australian Bureau of Statistics [14]. Results were reported as rates per 100,000 with 95% confidence intervals (CIs) for all cancers and for the following sites: (female) breast (International Statistical Classification of Diseases and Related Health Problems, 10th Revision, Australian Modification code C50), colorectal (C18-C20), pros- tate (C61), lung (C34), melanoma (C43), and cervical cancer (C53). The impact of different enhancement methods on the number of cases and on ASRs was examined in relative terms (% increase compared with the NSWCR variable). Analyses were performed using SAS Version 9.4 (SAS Institute, Cary, NC). Results invasive cancer were diag- Overall 204,948 cases of nosed in NSW in 2010–2014. Of these, 2703 (1.3%) were diagnosed among Aboriginal people based on the NSWCR Aboriginal status variable. There were 28,572 cases of cancer with unknown Aboriginal sta- tus (13.9%). After enhancement, the number of cases among Aboriginal people increased to 4184 (2.0%, ‘ever’), 3257 (1.6%, ‘weight ‘multi-stage median’). of evidence’) and 3583 (1.7%, The majority of cancer cases with a status change after enhancement were originally recorded as non- Aboriginal, rather than unknown Aboriginal status. For example, of the 877 cases of cancer with a status enhanced to Aboriginal using the weight of evidence method, 74% (n = 651) were recorded as non- Aboriginal and 26% (n = 226) had unknown Aborigi- nal status on the NSWCR. ‘most recent’), 3580 (1.7%, Relative enhancement (per cent increase) was generally greater for males, people aged 25–34 years, people with cancers of unknown or localised degree of spread, people living in urban areas and areas with less socio- economic disadvantage (Table 2). Overall the ASR among Aboriginal people was 559.9 per 100,000 (95% CI 535.3–585.3) before enhancement. All enhancement methods increased ASRs overall and for both males and females (Table 3, Fig. 1). The greatest Tervonen et al. BMC Medical Research Methodology (2019) 19:245 Page 4 of 9 Table 2 Impact of enhancement on the number of cancer cases and relative increase (%) among Aboriginal people by demographic and cancer characteristics, 2010–2014 NSWCR a Ever reported Most recent record Weight of evidence Multi-stage median n % n % Increase (%) b n % Increase (%) b n % Increase (%) b n % Increase (%) b Sex Female Male Age at diagnosis 0–14 15–24 25–34 35–44 45–54 55–64 65–74 75–84 85+ Year of diagnosis 2010 2011 2012 2013 2014 Clinical cancer group Skin Head and neck Upper gastrointestinal Colorectal Respiratory Bone and connective tissue Breast Gynaecological Urogenital Eye and neurological Thyroid and other endocrine Lymphohaematopoietic Ill-defined and unknown primary sites Degree of spread Localised Regional Distant Unknown Remoteness Major Cities Inner Regional 1329 1.5 1920 2.1 44.5% 1575 1.7 18.5% 1689 1.9 27.1% 1701 1.9 28.0% 1374 1.2 2264 2.0 64.8% 1682 1.5 22.4% 1891 1.7 37.6% 1882 1.6 37.0% 49 47 86 224 521 714 660 332 70 497 532 535 569 570 91 145 340 299 430 25 314 179 416 45 64 258 97 819 668 661 555 4.4 3.0 2.0 2.2 2.2 1.6 1.2 0.8 0.4 1.3 1.3 1.3 1.4 1.4 0.4 2.6 2.1 1.2 2.2 1.7 1.2 2.1 0.9 1.4 1.2 1.2 1.9 1.0 1.5 2.2 1.1 61 68 143 322 727 5.5 4.4 3.4 3.2 3.0 24.5% 44.7% 66.3% 43.8% 39.5% 1076 2.4 50.7% 1058 1.9 60.3% 570 159 766 830 836 851 901 284 205 442 442 553 34 472 249 803 68 109 401 122 1.3 0.8 1.9 2.1 2.0 2.0 2.1 1.3 3.6 2.7 1.8 2.8 2.3 1.9 2.9 1.8 2.2 2.1 1.9 2.3 71.7% 127.1% 54.1% 56.0% 56.3% 49.6% 58.1% 212.1% 41.4% 30.0% 47.8% 28.6% 36.0% 50.3% 39.1% 93.0% 51.1% 70.3% 55.4% 25.8% 54 61 120 276 627 867 785 388 79 594 639 646 683 695 177 171 367 356 464 29 391 207 547 51 81 313 103 4.9 3.9 2.8 2.8 2.6 1.9 1.4 0.9 0.4 1.5 1.6 1.6 1.6 1.7 0.8 3.0 2.3 1.4 2.4 2.0 1.5 2.4 1.2 1.6 1.5 1.5 2.0 10.2% 29.8% 39.5% 23.2% 20.3% 21.4% 18.9% 16.9% 12.9% 19.5% 20.1% 20.7% 20.0% 21.9% 94.5% 17.9% 7.9% 19.1% 7.9% 16.0% 24.5% 15.6% 31.5% 13.3% 26.6% 21.3% 6.2% 57 62 128 289 661 949 878 451 105 664 704 710 745 757 212 186 394 385 508 32 413 222 627 56 93 344 108 5.1 4.0 3.0 2.9 2.8 2.1 1.6 1.0 0.6 1.7 1.7 1.7 1.8 1.8 1.0 3.3 2.4 1.5 2.6 2.2 1.6 2.6 1.4 1.8 1.8 1.6 2.1 16.3% 31.9% 48.8% 29.0% 26.9% 32.9% 33.0% 35.8% 50.0% 33.6% 32.3% 32.7% 30.9% 32.8% 133.0% 28.3% 15.9% 28.8% 18.1% 28.0% 31.5% 24.0% 50.7% 24.4% 45.3% 33.3% 11.3% 56 64 130 293 675 945 876 447 97 661 705 712 747 758 216 184 392 379 502 33 420 223 633 57 94 343 107 5.1 4.1 3.0 2.9 2.8 2.1 1.6 1.0 0.5 1.7 1.7 1.7 1.8 1.8 1.0 3.2 2.4 1.5 2.6 2.2 1.7 2.6 1.4 1.8 1.8 1.6 2.1 14.3% 36.2% 51.2% 30.8% 29.6% 32.4% 32.7% 34.6% 38.6% 33.0% 32.5% 33.1% 31.3% 33.0% 137.4% 26.9% 15.3% 26.8% 16.7% 32.0% 33.8% 24.6% 52.2% 26.7% 46.9% 32.9% 10.3% 1467 1.8 79.1% 1086 1.3 32.6% 1209 1.5 47.6% 1212 1.5 48.0% 964 783 970 2.2 2.6 2.0 44.3% 18.5% 74.8% 767 686 718 1.8 2.3 1.5 14.8% 3.8% 29.4% 857 713 801 2.0 2.3 1.6 28.3% 7.9% 44.3% 850 713 808 2.0 2.3 1.7 27.2% 7.9% 45.6% 1206 0.9 1998 1.4 65.7% 1472 1.1 22.1% 1639 1.2 35.9% 1633 1.2 35.4% 831 1.7 1277 2.6 53.7% 1001 2.0 20.5% 1098 2.2 32.1% 1109 2.2 33.5% Tervonen et al. BMC Medical Research Methodology (2019) 19:245 Page 5 of 9 Table 2 Impact of enhancement on the number of cancer cases and relative increase (%) among Aboriginal people by demographic and cancer characteristics, 2010–2014 (Continued) NSWCR a Ever reported Most recent record Weight of evidence Multi-stage median n % n % Increase (%) b n % Increase (%) b n % Increase (%) b n % Increase (%) b Outer Regional 530 3.4 738 4.7 39.2% Remote/ very remote 136 Socio-economic disadvantage quintilec 12.7 171 15.9 25.7% 620 164 4.0 17.0% 15.3 20.6% 678 165 4.3 27.9% 15.4 21.3% 675 166 4.3 27.4% 15.5 22.1% Q1: Least disadvantaged Q2 Q3 Q4 Q5: Most disadvantaged 131 352 474 843 903 0.3 0.9 1.1 1.8 2.4 280 596 785 0.7 1.6 1.9 113.7% 69.3% 65.6% 157 441 580 0.4 1.1 1.4 19.8% 25.3% 22.4% 194 487 650 0.5 1.3 1.6 48.1% 38.4% 37.1% 195 485 654 0.5 1.3 1.6 48.9% 37.8% 38.0% 1219 2.6 44.6% 1004 2.2 19.1% 1082 2.3 28.4% 1087 2.3 28.9% 1304 3.4 44.4% 1075 2.8 19.0% 1167 3.1 29.2% 1162 3.1 28.7% aNSWCR: Aboriginal status variable in the NSW Cancer Registry bRelative increase compared with the number of cases based on the NSW Cancer Registry Aboriginal status variable cIndex of Relative Socio-economic Disadvantage increases were detected when using the ‘ever reported’ and the smallest increases when using the ‘most recent’ method. Enhancement increased incidence rates more for males than females. For example, the ‘weight of evi- dence’ method increased the ASR by 42% for males (894.1 per 100,000, 95% CI 844.5–945.4) and 27% for fe- males (642.7 per 100,000, 95% CI 607.9–678.7). In site-specific analyses, all enhancement methods in- creased ASRs for all sites compared with rates estimated using the NSWCR Aboriginal status variable (Table 3, Fig. 2). Again, the ‘ever reported’ method demonstrated the greatest increases while the ‘most recent’ method re- sulted in the smallest increases. Greatest relative in- creases were observed for melanoma and prostate cancer incidence, with increases of 126 and 63% respect- ively, using the ‘weight of evidence’ method. Discussion All enhancement methods increased both the number of cancer cases and age-standardised cancer incidence rates among Aboriginal people. The ‘ever reported’ method dem- onstrated the greatest increases and ‘most recent’ method the smallest increases, while the other two methods were very similar to each other and between these two extrem- ities. When using the ‘weight of evidence’ method, the ma- jority (74%) of cases with enhanced Aboriginal status were previously recorded as non-Aboriginal on the NSWCR. This indicates misclassification in the NSWCR Aboriginal status variable and highlights the need to correct this mis- classification and not solely focus on decreasing the num- ber of people with unknown Aboriginal status in the NSWCR and in the information received by the NSWCR from notifiers. Aboriginal and Torres Strait Islander status is self-reported at NSW health facilities and people to identify [4]. There have been may choose not culturally to provide strengthened procedures at a state level to improve the collection of Aboriginal and Torres Strait Islander status in NSW health facilities [15] as well as local initiatives safe health care throughout the study period. These factors are likely to have increased the willingness of people to self- identify as Aboriginal or Torres Strait Islander and improved identification at the point of care in more recent years. Linked data enhances the reporting of Aboriginal status because it brings together informa- tion on Aboriginal status that is not available to the NSWCR through people choosing to identify as Abo- riginal after diagnosis or at facilities that have not provided cancer care. Enhancement was generally greater in relative terms for males, people aged 25–34 years at diagnosis, people living in urban and less disadvantaged areas and for people with a cancer of localised or unknown degree of spread. Several factors are likely to explain these pat- terns, such as sources of cancer notifications and treat- ment patterns (e.g. the likelihood of admission for surgery). People diagnosed with cancers with good prog- nosis are less likely to be hospitalised or die which de- creases the likelihood of recording the Aboriginal status on the NSWCR. If the NSWCR only receives pathology notification, Aboriginal status information will be miss- ing. This is more likely to apply to cancers such as mela- nomas and prostate cancers, both of which showed greater levels of enhancement. A previous NSW study reported that enhancing Abori- ginal status for reporting deaths resulted in greater en- hancements for older people, for people living in urban areas and for those with chronic health conditions [16]. Another NSW study examining the im- pact on enhancement on hospital admissions reported for females, Tervonen et al. BMC Medical Research Methodology (2019) 19:245 Page 6 of 9 ) % ( e s a e r c n I b . 6 3 3 . 9 7 2 . 1 9 3 . 6 1 3 . 4 6 2 . 7 7 3 . 9 7 1 . 4 5 1 . 6 0 2 . 4 0 2 1 . 5 6 3 1 . 7 4 0 1 . 1 5 3 . 0 5 1 . 5 1 6 . – 4 9 1 7 ( . – 4 0 1 6 ( . – 3 7 2 8 ( . 0 8 4 7 . ) 4 7 7 7 . 2 5 4 6 . ) 2 1 8 6 . 6 5 7 8 . ) 7 5 2 9 ) I C % 5 9 ( R S A ) % ( . ) 9 4 9 – 1 4 7 ( . . ) 6 0 9 – 2 4 6 ( . . ) 6 0 1 1 – 6 6 7 ( . . 1 4 8 . 7 6 7 . 7 2 9 . ) 6 1 1 1 – 3 3 8 ( . . 8 6 9 . – 4 1 0 1 ( . 5 2 1 1 . ) 4 4 2 1 . – 5 4 1 1 ( . 3 3 3 1 . ) 0 4 5 1 . ) 7 3 4 – 7 0 3 ( . . ) 4 5 3 – 3 0 2 ( . . ) 6 9 5 – 5 7 3 ( . . 8 6 3 . 2 7 2 . 7 7 4 . – 2 7 2 1 ( . 1 2 4 1 . ) 1 8 5 1 . ) 3 2 2 – 5 2 1 ( . . 9 6 1 . – 6 0 8 1 ( . 6 3 0 2 . ) 3 8 2 2 e s a e r c n I b . 6 4 3 . 4 7 2 . 1 2 4 . 1 5 3 . 5 5 2 . 1 8 4 . 1 0 2 . 2 6 1 . 3 4 2 . 3 6 2 1 . 4 7 1 1 . 6 9 2 1 . 1 3 3 . 6 3 1 . 7 2 6 ) I C % 5 9 ( R S A . – 8 4 2 7 ( . – 9 7 0 6 ( . – 5 4 4 8 ( . 7 3 5 7 . ) 4 3 8 7 . 7 2 4 6 . ) 7 8 7 6 . 1 4 9 8 . ) 4 5 4 9 . ) 4 7 9 – 1 6 7 ( . . ) 0 0 9 – 9 3 6 ( . . ) 0 9 1 1 – 2 2 8 ( . . 3 6 8 . 2 6 7 . 7 9 9 . ) 4 2 1 1 – 0 4 8 ( . . 5 7 9 . – 3 3 0 1 ( . 6 4 1 1 . ) 7 6 2 1 . – 1 8 1 1 ( . 4 7 3 1 . ) 7 8 5 1 . ) 9 4 4 – 4 1 3 ( . . ) 9 2 3 – 4 8 1 ( . . ) 1 7 6 – 7 1 4 ( . . 8 7 3 . 0 5 2 . 5 3 5 . – 2 5 2 1 ( . 0 0 4 1 . ) 0 6 5 1 . ) 1 2 2 – 3 2 1 ( . . 7 6 1 . – 0 2 8 1 ( . 2 5 0 2 . ) 1 0 3 2 ) % ( e s a e r c n I b . 3 9 1 . 6 6 1 . 8 1 2 . 6 1 2 . 1 6 1 . 2 8 2 2 9 . 0 7 . . 6 1 1 . 1 0 7 . 2 2 9 . 9 4 5 . 4 4 2 . 9 0 1 . 0 7 3 ) I C % 5 9 ( R S A . – 1 1 4 6 ( . – 7 5 5 5 ( . – 0 2 2 7 ( . 9 7 6 6 . ) 3 5 9 6 . 5 8 8 5 . ) 6 2 2 6 . 7 6 6 7 . ) 0 3 1 8 . ) 0 8 8 – 1 8 6 ( . . ) 7 3 8 – 7 8 5 ( . . ) 5 3 0 1 – 8 0 7 ( . . 7 7 7 . 5 0 7 . 3 6 8 . ) 7 5 1 1 – 4 3 9 ( . . 2 4 0 1 . ) 1 4 0 1 – 8 6 7 ( . . 8 9 8 . – 1 5 0 1 ( . 3 3 2 1 . ) 4 3 4 1 . ) 2 4 3 – 2 3 2 ( . . ) 0 9 2 – 3 6 1 ( . . ) 7 6 4 – 2 7 2 ( . . 4 8 2 . 1 2 2 . 1 6 3 . – 8 6 1 1 ( . 9 0 3 1 . ) 1 6 4 1 . ) 6 1 2 – 0 2 1 ( . . 3 6 1 . – 8 1 5 1 ( . 7 2 7 1 . ) 3 5 9 1 ) % ( e s a e r c n I b . 0 4 6 . 7 8 4 . 4 0 8 . 5 3 6 . 8 3 4 . 7 9 8 . 0 2 3 . 7 3 2 . 0 1 4 . 1 3 4 2 . 0 3 1 2 . 8 4 6 2 . 1 3 5 . 0 2 3 . 5 3 2 1 ) I C % 5 9 ( R S A ) I C % 5 9 ( R S A ) I C % 5 9 ( R S A . ) 6 1 5 9 – 2 5 8 8 ( . . ) 1 0 9 7 – 2 2 1 7 ( . . 0 8 1 9 . – 3 5 3 5 ( . 5 0 5 7 . – 0 4 7 4 ( . - 0 7 7 0 1 ( . 3 5 3 1 1 . ) 5 5 9 1 1 . – 7 8 8 5 ( . 9 9 5 5 . ) 3 5 8 5 . 6 4 0 5 . ) 5 6 3 5 . 4 9 2 6 . ) 7 1 7 6 . ) 0 3 9 4 – 7 8 8 4 ( . . ) 5 0 2 4 – 0 5 1 4 ( . . ) 0 1 8 5 – 2 4 7 5 ( . . 8 0 9 4 s n o s r e P s r e c n a c l l A . 7 7 1 4 l s e a m e F . 6 7 7 5 l s e a M . ) 0 7 1 1 – 9 2 9 ( . . 5 4 0 1 . ) 2 3 7 – 3 5 5 ( . . ) 3 2 0 1 – 7 3 7 ( . . 3 7 8 . ) 0 3 7 – 8 9 4 ( . . ) 4 0 5 1 – 1 7 0 1 ( . . ) 6 8 3 1 – 9 3 1 1 ( . . ) 2 9 1 1 – 8 9 8 ( . . ) 7 8 7 1 – 7 4 3 1 ( . . 7 7 2 1 . 9 5 2 1 . 8 3 0 1 . 8 5 5 1 . ) 4 2 8 – 9 3 5 ( . . ) 3 6 0 1 – 2 5 8 ( . . ) 8 7 9 – 5 1 7 ( . . 9 3 6 . 7 0 6 . 3 7 6 . 4 5 9 . 9 3 8 . ) 2 9 2 1 – 6 3 9 ( . . 5 0 1 1 . ) 6 8 5 – 1 7 5 ( . . ) 8 9 4 – 9 7 4 ( . . ) 2 9 6 – 9 6 6 ( . . ) 1 3 4 – 8 1 4 ( . . ) 1 4 3 – 5 2 3 ( . . ) 7 4 5 – 6 2 5 ( . . ) 5 6 6 – 9 8 4 ( . . ) 6 5 4 – 7 7 2 ( . . ) 5 3 0 1 – 7 8 6 ( . . 3 7 5 . 0 6 3 . 0 5 8 . ) 6 1 2 – 6 2 1 ( . . ) 2 7 1 – 1 7 ( . . ) 6 2 3 – 8 5 1 ( . . 7 6 1 . 5 1 1 . 3 3 2 . ) 5 1 5 – 1 0 5 ( . . ) 2 1 4 – 4 9 3 ( . . ) 6 4 6 – 3 2 6 ( . . ) 2 8 7 1 – 1 5 4 1 ( . . 1 1 6 1 . ) 3 5 2 – 5 4 1 ( . . 4 9 1 . ) 9 8 1 1 – 6 2 9 ( . . 2 5 0 1 . ) 3 2 2 1 – 2 9 1 1 ( . . 8 0 2 1 . ) 9 9 1 – 5 0 1 ( . . 7 4 1 . ) 2 7 – 4 6 ( . 8 6 . . ) 8 1 1 3 – 5 3 5 2 ( . . 8 1 8 2 . – 9 7 0 1 ( . 1 6 2 1 . ) 1 6 4 1 . ) 5 9 6 1 – 9 5 6 1 ( . . 7 7 6 1 . 8 7 5 . 9 8 4 . 1 8 6 s n o s r e P l s e a m e F l s e a M l a t c e r o o C l g n u L . 5 2 4 s n o s r e P . 3 3 3 . 6 3 5 . 8 0 5 . 3 0 4 . 4 3 6 l s e a m e F l s e a M s n o s r e P l s e a m e F l s e a M a m o n a e M l l ) s e a m e F ( t s a e r B e t a t s o r P i x v r e C i n a d e m e g a t s - i t l u M e c n e d v e i f o t h g e W i d r o c e r t n e c e r t s o M d e t r o p e r r e v E a R C W S N 4 1 0 2 – 0 1 0 2 , l e p o e p l i a n g i r o b A d n a l i a n g i r o b A - n o n g n o m a s e t a r e c n e d c n i i l e p o e p l i a n g i r o b A l i a n g i r o b A - n o N a R C W S N l e p o e p r e c n a c d e s i d r a d n a t s - e g A 3 e l b a T l n o i t a u p o p d r a d n a t s n a i l a r t s u A 1 0 0 2 e h t o t d e s i d r a d n a t s y l t c e r i d ; s l a v r e t n i e c n e d i f n o c % 5 9 h t i w 0 0 0 0 0 1 , r e p e t a r e c n e d i c n i r e c n a c d e s i d r a d n a t s - e g A : ) I C % 5 9 ( R S A l e b a i r a v s u t a t s l i a n g i r o b A y r t s i g e R r e c n a C W S N e h t n o d e s a b e t a r e c n e d i c n i h t i w d e r a p m o c e s a e r c n i y r t s i g e R r e c n a C W S N e h t n i l e b a l i a v a l e b a i r a v s u t a t s l i a n g i r o b A : R C W S N a e v i t a e R b l Tervonen et al. BMC Medical Research Methodology (2019) 19:245 Page 7 of 9 Fig. 1 Age-standardised cancer incidence rates among Aboriginal people using the NSW Cancer Registry (NSWCR) Aboriginal status variable and four enhancement methods, 2010–2014. (see Table 3 for underlying data and 95% confidence intervals) greater enhancement for earlier years of admission, major cities, private hospitals and varying impact by age depending on the enhancement method used [6]. Differ- ent factors impact on enhancement depending on the health outcome of interest and the datasets used in analyses. Lung and cervical cancers saw the smallest increases in incidence rates. Both these cancers have a greater burden in Aboriginal compared with non-Aboriginal people [17]. Due to the poor prognosis, death certificate information is available for most people diagnosed with lung cancer, in- creasing the likelihood of Aboriginal status recording. It is likely that enhancement had a smaller impact on lung cancer incidence rates because the existing NSWCR Abo- riginal status already had relatively good capture. The rela- tively smaller increase in the incidence of cervical cancer may due to relatively good capture on the NSWCR, but may also be due to other factors such the patterns of hos- pitalisation and capture of Aboriginal status at the point of care for what is generally a younger cohort of women. Enhancing the reporting of cancer outcomes of Aborigi- nal people might have a major impact on observed dispar- ities between Aboriginal and non-Aboriginal people. For example, according to national statistics [17] and our Fig. 2 Age-standardised cancer incidence rates by site among Aboriginal people using the NSW Cancer Registry (NSWCR) Aboriginal status variable and four enhancement methods, 2010–2014. (see Table 3 for underlying data and 95% confidence intervals) Tervonen et al. BMC Medical Research Methodology (2019) 19:245 Page 8 of 9 Increased breast cancer analyses using the NSWCR Aboriginal status variable, Abo- riginal people have lower breast and prostate cancer inci- dence rates compared with non-Aboriginal people. This pattern has also been reported among Indigenous peoples in many international jurisdictions and has been proposed as being related to the prevalence of risk factors for these cancers and competing causes of death [18]. After enhance- ment our results indicated higher breast and prostate cancer incidence among Aboriginal people than non- Aboriginal people in NSW. This finding has implications on widely held views on risk of these cancers among Indi- genous peoples. Higher breast cancer rates have been re- ported among Indigenous people (Māori) in New Zealand using the national population-based cancer registry which includes links to a national health database to improve identification [18]. incidence among Indigenous people have been reported in two United States (US) states using data linkage between cancer registries and health service data [19, 20]. Our results also highlight the burden of melanoma among Aboriginal people which warrants further discussion on prevention strategies and actions. After enhancement our results indi- cated substantially higher incidence than when using the NSWCR Aboriginal status variable, but still lower rates compared with non-Aboriginal people (except when using the ‘ever reported’ method). The effect of under-recording of Indigenous status should be investigated in more juris- dictions. Cancer is the second leading cause of death and among the leading causes of burden of disease among Abo- riginal people in Australia [21]. The findings of our study highlight the impact of cancer on Aboriginal people and the need for cancer control to improve health outcomes. Cancer control programs should have a special focus on Aboriginal people considering that their cancer burden may be higher than expected. Australian cancer screening programs are already targeting Aboriginal people due to lower participation rates [17]. Future research should also examine the impact of en- hancement on other cancer outcomes, such as mortality, survival and the likelihood of being diagnosed with ad- vanced stage disease. Studies have shown that Aboriginal people are more likely to be diagnosed with advanced stage cancer than non-Aboriginal people [22, 23]. We found greatest enhancement for people diagnosed with localised or unknown degree of spread, which may impact on the likelihood of Aboriginal people being diagnosed with advanced cancer in comparison with non-Aboriginal people and affect estimates of disparities in survival out- comes since localised cancers have much better prognosis. Based on these results and consultation with the Can- cer Institute NSW Aboriginal Advisory Group, the ‘weight of evidence’ method was considered to be the most suitable for further reporting of cancer outcomes for Aboriginal people. The ‘weight of evidence’ method utilises information from several sources but is still rela- tively straightforward to use and report. It provides a balance between enhancing the identification of Aborigi- nal people and reducing misclassification of non- Aboriginal people as Aboriginal. This method was devel- oped and is also used by the NSW Ministry of Health [6]. Studies have pointed out that ‘ever reported’ may re- sult in misclassification and over-reporting [1, 6]. It should be noted that an enhanced Aboriginal identifier is a statistical construct that enables improved reporting of cancer outcomes using historical data but potentially includes some inaccuracies due to errors in the source datasets and incorrect linkages [2]. Collection of accur- ate information at the point of care remains vital. Limitations include that if a person was recorded as Abo- riginal on the NSWCR or death certificate, this information was accepted. Although there is a possibility for positive misclassification this is likely to be low since the information is provided by the next-of-kin. Numerator-denominator bias is a known issue affecting observed cancer burden in Indi- genous populations internationally because incidence and population data are derived using different data collection methodologies [8]. Population denominators can be unreli- able due to under-participation of Aboriginal people and varying propensity to identify as Aboriginal in censuses. The Australian Bureau of Statistics (ABS) estimates Aboriginal and Torres Strait Islander populations using self-reported information in the Australian Census data with adjustment for undercount using a household survey following the cen- sus [14]. An increase in the number of people self- identifying as Aboriginal or Torres Strait Islander has been observed, with people who did not self-identify in the 2011 Australian Census choosing to identify in the subsequent 2016 Census [24]. In our study, enhancement of the numer- ator is likely to reduce the under-estimation of cancer inci- dence that is common in cancer incidence estimates for Indigenous people [8]. However, without enhancement of the denominator using the same methodologies it may lead to over-estimation of incidence rates. Linkage of the cancer registry, census, hospital and mortality data would enable cancer outcomes for Aboriginal people to be estimated with reduced numerator-denominator bias. Conclusions All data linkage enhancement methods increased the number of cancer cases and cancer incidence rates for Aboriginal people. Enhancement varied by demographic and cancer characteristics. We considered the ‘weight of evidence’ method to be most suitable for future analyses of cancer outcomes of Aboriginal people. Enhancing the reporting of cancer outcomes of Aboriginal people can have major impacts on cancer disparities between Abori- ginal and non-Aboriginal people and this should be fur- ther examined. Tervonen et al. BMC Medical Research Methodology (2019) 19:245 Page 9 of 9 Abbreviations ABS: Australian Bureau of Statistics; APDC: Admitted Patient Data Collection; ASR: Age-standardised cancer incidence rate; CI: Confidence Intervals; COD URF: Cause of Death Unit Record File; EDDC: Emergency Department Data Collection; NSW: New South Wales; NSWCR: New South Wales Cancer Registry; US: United States Acknowledgements The authors would like to thank the Aboriginal Advisory Group of the Cancer Institute NSW for their valuable advice and comments. The Cause of Death Unit Record File (COD URF) is provided by the Australian Coordinating Registry for COD URF on behalf of Australian Registries of Births, Deaths and Marriages, Australian Coroners and the National Coronial Information System. We would also like to thank the Centre for Epidemiology and Evidence, NSW Ministry of Health for providing access to the population data and the Centre for Health Record Linkage for their assistance with this project. Authors’ contributions NC had the original idea for the study. SP and HET analysed the data. HET and NC conducted the literature searches. HET wrote the first draft of the manuscript. All authors contributed to the interpretation of the results, read and approved the final manuscript. Funding Not applicable. Availability of data and materials Restrictions by the data custodians mean that the datasets are not publicly available or able to be provided by the authors. Researchers wanting to access the datasets used in this study should refer to the Centre for Health Record Linkage application process (www.cherel.org.au/apply-for-linked-data). Ethics approval and consent to participate This project was approved by the NSW Population and Health Services Research Ethics Committee (HREC/15/CIPHS/15) and the Aboriginal Health and Medical Research Council Ethics Committee (HREC Ref. No. 1201/16). The data sources were collected under legislation and individual consent was not required for the use of the de-identified data in this project. Subject matter advice and Aboriginal community input was sought from the Cancer Institute NSW Aboriginal Advisory Group. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Received: 5 March 2019 Accepted: 5 December 2019 References 1. Kennedy B, Howell S, Breckell C. Indigenous identification in administrative data collections and the implications for reporting Indigenous health status. Technical Report no. 3. Brisbane: Health Statistics Centre, Queensland Health; 2009. Population and Public Health Division. Improved Reporting of Aboriginal and Torres Strait Islander Peoples on Population Datasets in New South Wales using Record Linkage – a Feasibility Study. Sydney: NSW Ministry of Health; 2012. Cancer Institute NSW. Cancer in NSW Aboriginal peoples: completeness and quality of Aboriginal status data on the NSW Central Cancer registry. Sydney: Cancer Institute NSW; 2012. NSW Aboriginal Affairs. Aboriginal identification: the way forward. An Aboriginal peoples’ perspective. Sydney: NSW Government; 2015. Thompson SC, Woods JA, Katzenellenbogen JM. The quality of Indigenous identification in administrative health data in Australia: insights from studies using data linkage. BMC Med Inform Decis. 2012;12:133. Randall DA, Lujic S, Leyland AH, Jorm LR. Statistical methods to enhance reporting of Aboriginal Australians in routine hospital records using data linkage affect estimates of health disparities. Aust NZ J Publ Health. 2013; 37(5):442–9. 2. 3. 4. 5. 6. 7. 8. 9. 10. Australian Institute of Health and Welfare, Australian Bureau of Statistics. National best practice guidelines for data linkage activities relating to Aboriginal and Torres Strait Islander people. AIHW Cat. No. IHW 74. Canberra: AIHW; 2012. Sarfati D, Robson B. Equitable cancer control: better data needed for indigenous people. Lancet Oncol. 2015;16(15):1442–4. Christensen D, Davis G, Draper G, Mitrou F, McKeown S, Lawrence D, et al. Evidence for the use of an algorithm in resolving inconsistent and missing Indigenous status in administrative data collections. Aust J Soc Issues. 2014; 49:423–49. Esteban D, Whelan S, Laudico A, Parkin DM. Manual for cancer registry personnel. IARC Technical Report No 10. Lyon: International Agency for Research on Cancer; 1995. 11. Australian Bureau of Statistics. 1216.0.15.003 - Australian Standard Geographical Classification (ASGC) Remoteness Area Correspondences. Canberra: ABS; 2011. 12. Australian Bureau of Statistics. 2039.0 - Information Paper: An Introduction to Socio-Economic Indexes for Areas (SEIFA), 2006. Canberra: ABS; 2008. 13. Cancer Institute NSW. Glossary. https://www.cancerinstitute.org.au/ glossary#term-Clinical-cancer-group. Accessed 14 Jun 2019. 14. Australian Bureau of Statistics. 3238.0 - Estimates and Projections, Aboriginal and Torres Strait Islander Australians, 2001 to 2026. Canberra: ABS; 2014. 15. Centre for Aboriginal Health. Aboriginal and Torres Strait Islander Origin - 16. Recording of Information of Patients and Clients. Sydney: NSW Health; 2012. https://www1.health.nsw.gov.au/PDS/pages/doc.aspx?dn=PD2012_042 . Accessed 14 June 2019. Taylor LK, Bentley J, Hunt J, Madden R, McKeown S, Brandt P, et al. Enhanced reporting of deaths among Aboriginal and Torres Strait Islander peoples using linked administrative health datasets. BMC Med Res Methodol. 2012;12(1):91. 17. Australian Institute of Health and Welfare. Cancer in Aboriginal & Torres Strait Islander people of Australia. Web report Cat no. CAN 109: AIHW; 2018. https://www.aihwgovau/reports/cancer/cancer-in-indigenous-australians/ contents/table-of-contents. Accessed 14 Jun 2019 18. Moore SP, Antoni S, Colquhoun A, Healy B, Ellison-Loschmann L, Potter JD, et al. Cancer incidence in indigenous people in Australia, New Zealand, Canada, and the USA: a comparative population-based study. Lancet Oncol. 2015;16(15):1483–92. 19. Partin MR, Rith-Najarian SJ, Slater JS, Korn JE, Cobb N, Soler JT. Improving 20. cancer incidence estimates for American Indians in Minnesota. Am J Public Health. 1999;89(11):1673–7. Foote M, Matloub J, Strickland R, Stephenson L, Vaughan-Batten H. Improving cancer incidence estimates for American Indians in Wisconsin. WMJ. 2007;106(4):196–204. 21. Australian Institute of Health and Welfare. Australian Burden of Disease Study: Impact and causes of illness and death in Aboriginal and Torres Strait Islander people 2011. Australian Burden of Disease Study series no. 6 Cat no. BOD 7. Canberra: AIHW; 2016. 22. Gibberd A, Supramaniam R, Dillon A, Armstrong BK, O'Connell DL. Are 23. Aboriginal people more likely to be diagnosed with more advanced cancer? Med J Australia. 2015;202(4):195–9. Tervonen HE, Walton R, You H, Baker D, Roder D, Currow D, et al. After accounting for competing causes of death and more advanced stage, do Aboriginal and Torres Strait Islander peoples with cancer still have worse survival? A population-based cohort study in New South Wales. BMC Cancer. 2017;17(1):398. 24. Markham D, Biddle N. Indigenous population change in the 2016 census. CAEPR census paper no. 1. Canberra: Centre for Aboriginal Economic Policy Research, Australian National University; 2016. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1177_03010066231175014	Article Effects of cortical distance on the Ebbinghaus and Delboeuf illusions Perception 2023, Vol. 52(7) 459–483 © The Author(s) 2023 Article reuse guidelines: sagepub.com/journals-permissions DOI: 10.1177/03010066231175014 journals.sagepub.com/home/pec Poutasi W. B. Urale and Dietrich Samuel Schwarzkopf School of Optometry & Vision Science, The University of Auckland, New Zealand Abstract The Ebbinghaus and Delboeuf illusions affect the perceived size of a target circle depending on the size and proximity of circular inducers or a ring. Converging evidence suggests that these illusions are driven by interactions between contours mediated by their cortical distance in primary visual cortex. We tested the effect of cortical distance on these illusions using two methods: First, we manipulated retinal distance between target and inducers in a two-interval forced choice design, ﬁnding that targets appeared larger with a closer surround. Next, we predicted that targets pre- sented peripherally should appear larger due to cortical magniﬁcation. Hence, we tested the illu- sion strength when positioning the stimuli at various eccentricities, with results supporting this hypothesis. We calculated estimated cortical distances between illusion elements in each experi- ment and used these estimates to compare the relationship between cortical distance and illusion strength across our experiments. In a ﬁnal experiment, we modiﬁed the Delboeuf illusion to test whether the inﬂuence of the inducers/annuli in this illusion is inﬂuenced by an inhibitory surround. We found evidence that an additional outer ring makes targets appear smaller compared to a sin- gle-ring condition, suggesting that near and distal contours have antagonistic effects on perceived target size. Keywords neural mechanisms, perception, crowding, eccentricity, Ebbinghaus illusion, Delboeuf illusion, size perception Date Received: 7 February 2022; accepted: 12 April 2023 Corresponding author: Dietrich Samuel Schwarzkopf, School of Optometry & Vision Science, The University of Auckland, 85 Park Road, Grafton, Auckland. Email: s.schwarzkopf@auckland.ac.nz 460 Perception 52(7) The Ebbinghaus illusion (see Figure 1) has bamboozled our visual systems for over a century (Ebbinghaus, 1902; Titchener, 1905). Yet, despite a mountain of research on this illusion, the neural mechanisms underlying it remain poorly understood. Filling this lacuna is crucial for under- standing how the brain determines visual object size, which is itself an unresolved question (Schwarzkopf, 2015). Theories of the Ebbinghaus Illusion Several theories have attempted to explain the Ebbinghaus illusion. Most illustrations of this illusion show an apparent “size-contrast” effect, where the center circle (target) surrounded by small inducers appears larger, while large inducers make the target appear smaller (e.g., Massaro & Anderson, 1970, 1971; Obonai, 1954). Many authors have thus described the illusion in terms of this size-contrast mechanism (Aglioti et al., 1995; Haffenden et al., 2001; Yamazaki et al., 2010). However, Todorović and Jovanović (2018) point out that size-contrast is descriptive rather than explanatory and offers an incomplete account for the illusion. They argue that size contrast is nebulously deﬁned, and that there is no explanation for why there is a size-contrast effect instead of an assimilation effect, as found for other visual illusions such as the tilt illusion (Clifford, 2014). Part of this objection stems from the observation that geometrical features other than inducer size also modulate the strength of the Ebbinghaus illusion. One such factor is object-level similarity between targets and inducers, which make the illusion stronger (Coren & Miller, 1974; Massaro & Anderson, 1971; Rose & Bressan, 2002). The illusion also depends on the amount of empty space between inducers, with a more complete ring of inducers around the periphery of the target strengthening the illusion (Girgus et al., 1972; Massaro & Anderson, 1970; Roberts et al., 2005). Roberts et al. (2005) investigated the role of completeness by directly comparing the Ebbinghaus illusion to the Delboeuf illusion (see Figure 1), another size-perception illusion that uses a circular ring that surrounds the target stimulus instead of multiple circular inducers (Delboeuf, 1865; Evans, 1995). They showed that an Ebbinghaus conﬁguration composed of Figure 1. Key stimuli. (A-B) The Ebbinghaus illusion. Most observers will perceive the white ﬁlled-in circle in B (small inducers) as larger than that in A (large inducers). (C) The test targets Experiments 1, 2, 3a, and 3b varies according to a staircase procedure. (D-E) The Delboeuf illusion. Most observers will perceive the white ﬁlled-in circle in E (close ring) as larger than that in D (far ring). (F) Novel two-ring Delboeuf stimulus used in Experiment 3a, featuring near and far annuli. Urale and Schwarzkopf 461 small inducers that formed a complete ring yield about the same illusory effect as a Delboeuf con- ﬁguration, and that both conﬁgurations yielded stronger effects than Ebbinghaus conﬁgurations with less complete inducer annuli. Lastly, Ebbinghaus illusion strength changes with the distance between the target and inducers (target-inducer distance). Massaro and Anderson (1971) found that point of subjective equality (PSE) decreased with greater target-inducer distances, with the farthest distances failing to affect the perceived size of the target at all. Other work (Girgus et al., 1972; Jaeger, 1978; Roberts et al., 2005; Weintraub, 1979) also showed that target-inducer distance mod- ulates the illusion, but with an unexpected reversal of the effect of small inducers in some conﬁg- urations. That is, a sufﬁciently large distance between the target and small inducers causes the target to appear smaller, not larger, compared to control. Similarly, Roberts et al. (2005) found that even when controlling for completeness, increasing the target-inducer distance reversed the effect of small inducers, making the target appear smaller rather than larger. In contrast, large inducers always elicited a perceived shrinkage of the target, and this effect only became stronger with inducer-target distance. These ﬁndings demonstrate that to describe the Ebbinghaus illusion as an example of “size-contrast” is an oversimpliﬁcation. Contour-based accounts (Jaeger, 1978; Jaeger & Long, 2007; Jaeger & Lorden, 1980; Sherman & Chouinard, 2016; Todorović & Jovanović, 2018; Weintraub, 1979; Weintraub & Schneck, 1986) are explanations based on interactions between the low-level contours that make up a stimulus. On the whole, these theories account better for the experimental evidence. Biphasic contour-interaction theory (BCIT) is one such account (Roberts et al., 2005; Sherman & Chouinard, 2016; Weintraub & Schneck, 1986). In contrast to the mid-level size comparison mechanism needed for a size-contrast account, the premise of BCIT explains the illusion in terms of low-level representations of contours: Nearby contours are attracted, whereas distant contours repel each other, hence the effect is “biphasic.” Roberts and colleagues’ (2005) ﬁndings are consistent with this theory, with their data showing a tendency for Ebbinghaus inducers to make the target appear smaller as target-inducer dis- tance is increased. They found a similar effect with the Delboeuf illusion. Relatedly, Sherman and Chouinard (2016) showed a correlation between the Delboeuf illusion and Ebbinghaus illusion. They argue this is incompatible with a size-contrast account because the ring in the Delboeuf illusion is always larger than the target. Todorović and Jovanović (2018) addressed the size-contrast theory more directly by using a novel stimulus. They found that increasing the number of small inducers in an Ebbinghaus conﬁguration can counterintuitively eliminate the illusion if the target is embedded in a grid of inducers (also see Jaeger & Klahs, 2015). Their ﬁnding disputes a size-contrast account that predicts more inducers to amplify the size contrast between the inducers and the target, while supporting a BCIT-based explanation, which posits that attractive and repellent effects of contours located at varying distances from the target cancel out. Nevertheless, BCIT does not offer a complete explanation of the Ebbinghaus illusion. BCIT cannot explain the effect of similarity between inducers and targets when the distribution of near and far contours are controlled for (Coren & Miller, 1974; Deni & Brigner, 1997). As others have noted, there are often multiple contributing factors to visual illusions (Coren & Girgus, 1978), and it is possible that the Ebbinghaus illusion may represent the outcome of several distinct processes along the visual stream. Importantly, Rose and Bressan (2002) replicated the similarity effect and further showed that illusion magnitude was boosted only when both inducers and targets were circles or triangles, but not hexagons or irregular angular shapes. This disputes size contrast is based on the sum of Euclidean distances between contours. Considering this, contour-interaction seems to be a neces- sary but insufﬁcient factor in the Ebbingaus illusion. As Rose and Bressan and others (Coren & Girgus, 1978; Schwarzkopf, 2015) have pointed out, the Ebbinghaus illusion may incorporate non-linear effects arising from top-down feedback, or even multiple contributing mechanisms. theories, as well as any contour-interaction account the complete explanation of that 462 Perception 52(7) Neural Correlates of the Ebbinghaus Illusion Converging evidence suggests that the effect of inducers on perceived size of the target is mediated by processes located in V1. Illusion magnitude was reduced—but not abolished—when inducers and target were shown to separate eyes (Song et al., 2011); indicative of a cortical mechanism in V1 where there are still many monocular neurons, although this cannot rule out a contribution from higher visual areas. Additionally, Schwarzkopf et al. (2011) used functional magnetic reson- ance imaging (fMRI) and retinotopic mapping to show that functional primary visual cortex (V1) surface area can predict Ebbinghaus PSE. V1’s selectivity for local contrast edges makes it a likely candidate site for mediating low-level interactions as posited by the contour-interaction account. They used the classical Ebbinghaus illusion, where observers judged the difference in target size between a large-inducer and small-inducer conﬁguration. In a follow-up study, Schwarzkopf and Rees (2013) also found a correlation between V1 area and the PSEs for large and small inducers tested separately. Both Ebbinghaus conﬁgurations made the target appear relatively larger in indi- viduals with small V1s. The authors surmised this may indicate the effect of local circuits within V1, which are contingent on cortical distance. This could represent an attenuation of the effects of these circuits at greater distances or because of the time taken by those signals to propagate. Moreover, while large inducers reliably made the target appear smaller, small inducers made the target appear larger for some and smaller for other observers depending on their V1 surface area. Relatedly, Moutsiana et al. (2016) found that the expansive effect of the Delboeuf illusion is enhanced when it is encoded by larger population receptive ﬁelds (pRFs) in V1 (Harvey & Dumoulin, 2011). This was true both within observers with variation of pRF size across the visual ﬁeld, and between individuals with different pRF sizes. While not able to explain the repul- sive effect between contours, these results and those of Schwarzkopf et al. (2011) can be concep- tualized as indicative of an antagonistic center-surround ﬁeld of local interactions that deﬁnes a gradient of modulation based on cortical distance (Schwarzkopf, 2015). When target-inducer dis- tance is small, there is an attractive effect, and the target appears larger. Conversely, when the dis- tance is large the repulsive effect dominates, and the target appears smaller. In between is a point of equilibrium where inducers would have neither an attractive nor repulsive effect. The sign change of the illusion with small inducers across observers is consistent with this theory. In the Current Work While Schwarzkopf and Rees (2013) and prior work (Schwarzkopf et al., 2011) has shown evi- dence that the Ebbinghaus illusion depends on between-subject differences in cortical topography, the present work looks at the effect of varying cortical target-inducer distance within individuals. As such, we manipulate cortical distance in two ways: by varying the retinal target-inducer distance in visual space (Experiment 1), and by varying the eccentricity of stimuli when target-inducer distance is constant (Experiment 2). If proximity of contours affects the illusion as described by Schwarzkopf and Rees (2013) then reduced cortical distance will modulate the illusion so that the target appears larger. Furthermore, an account of the Ebbinghaus illusion based on cortical distance should also explain the difference in PSEs between large- and small-inducer Ebbinghaus conﬁgurations. Proponents of contour-based accounts (Sherman & Chouinard, 2016; Todorović & Jovanović , 2018; Weintraub, 1979) claim that large inducers cause a repulsive effect because they possess both near and far contours, while small inducers do not. In Experiments 3a and 3b we investigate this claim by using single- and double-ring conﬁgurations of the Delboeuf illusion. To draw further comparisons with the Ebbinghaus illusion, we also varied the retinal distance between targets and surround as in Experiment 1. If it is true that large inducers in the Ebbinghaus illusion cause Urale and Schwarzkopf 463 Figure 2. Trial procedure for Experiments 1, 2, 3a, and 3b. (A) Edge-to-edge inducer/ring distance(s) from reference target across experiments. (B) Trial sequence for Experiments 1 and 3a/3b. Observers maintained ﬁxation on a cross before being shown reference and test intervals. The target in the test interval changed according to a staircase procedure. The order of these two intervals was counterbalanced across the experiment. Following these observers made a size comparison of targets in the two intervals. (C) Trial sequence for Experiment 2. This was like the other two experiments, except the ﬁrst interval was preceded by an exogenous cue that indicated where the stimuli would appear, and gaze position was monitored using eye-tracking. Stimuli in the ﬁrst and second intervals appeared above and below the horizontal meridian, respectively. repulsion because of the antagonistic effect of near and far contours, then we should observe a similar effect with the addition of a second ring in the Delboeuf illusion (Figure 2). Experiment 1 In Experiment 1 we varied the retinal distance between the target and the inducer. Varying the retinal distance entails changes in distances between representations of visual elements across the visual stream. This is evident in topographically organized areas such as V1 and V2, where we would expect an increase of cortical distance between representations with retinal distance. Our study here is a conceptual replication of the study by Roberts et al. (2005), who varied the dis- tance between inducers/annuli and the target for the Ebbinghaus and Delboeuf illusions, respect- ively. In that study, both inducers and annuli made the target appear smaller at farther distances compared to closer distances. More recently, Knol et al. (2015) also varied the Ebbinghaus illusion along various dimensions, including target size, inducer size, and target-inducer distance, ﬁnding enlargement in cases where small- or medium-sized targets (∼0.5° and ∼1°) were displayed with inducers at short distances. In our study we used a similar manipulation with the addition of some key differences. Firstly, we included shorter target-inducer distances compared to both studies. In Roberts and colleagues’ study, the closest target-inducer distance was 1.9° for small 464 Perception 52(7) inducers, and 2.53° for large inducers. Our study used a minimum distance of 0.14° for both inducer types. Furthermore, Roberts and colleagues limited the target-inducer distance since a closer dis- tance would require overlap between large inducers. In our study, we allowed inducers to overlap to achieve a short target-inducer distance. The inclusion of this smallest distance tests a key hypothesis posited by Schwarzkopf and Rees (2013) who proposed that large inducers could make the target appear larger if sufﬁciently close. Secondly, we showed stimuli close to ﬁxation in two temporal intervals. Most other studies testing the Ebbinghaus illusion typically use a 2-alter- native forced-choice task (like ours, but where two stimuli are presented simultaneously side by side (in opposite hemiﬁelds) and are either ﬂashed brieﬂy (Schwarzkopf & Rees, 2013; Song et al., 2013) or remain on screen until the observer responds (Knol et al., 2015; Roberts et al., 2005; Todorović & Jovanović , 2018). Presenting stimuli in close proximity in separate intervals removes the need to split attention across the two stimulus locations and reduces the possibility of crowding effects of peripherally located stimuli. Materials and Methods Participants. We recruited 12 observers (8 females, age range 21–50), all with normal or corrected-to-normal visual acuity. Observers provided written and informed consent and all proce- dures were approved by the University of Auckland Human Participants Ethics Committee (UAHPEC). Experimental Setup. Stimuli were displayed on a 621 × 341 mm LCD monitor (Expt-1: Dell, S2817Q, USA; Expt-2: Samsung, U28D590D, South Korea), at a resolution of 3840 × 2160 × 8-bit resolution running 60 Hz. Monitors were linearized in software based on measurements made with a photometer (LS100, Konica Minolta, Japan). Stimuli were generated using program- ming environment MATLAB (version 2017B, MathWorks Inc.) and Psychtoolbox 3 (Brainard, 1997; Kleiner et al., 2007; Pelli, 1997) using customized scripts. Observers’ heads were stabilized with a chin rest. Stimuli. A single trial consisted of two stimuli: a reference stimulus, consisting of a target (always 0.56° diameter) and a surround, depending on the condition of the given trial, and a test stimulus, which consisted of only a target which varied in size according to an adaptive staircase procedure (see below). Stimuli were presented on a grey (175 cd/m2) background. Targets were always ﬁlled, white circles (341 cd/m2) inducers were white outlined circles with a ∼0.08° stroke and had dia- meters of 0.84° and 0.2° for large and small inducers, respectively. Example stimuli can be seen in Figure 3. The edge-to-edge distance between the target and inducers (target-inducer distance) could be one of seven possible distances: 0.14°, 0.42°, 0.7°, 1.13°, 1.55°, 2.25°, and 4.5°. In add- ition, there was a control condition without inducers. All Ebbinghaus conﬁgurations in both experi- ments had eight inducers, and large inducers were allowed to overlap in conditions with very short target-inducer distances. The centers of inducers were positioned at evenly spaced radial positions relative to the target ranging from 0° to 315° in steps of 45°. Targets in each interval were shown at a horizontally offset position relative to ﬁxation (see “Procedure” section), so in the 0.42° condition some individual inducers above and below the target overlapped the vertical meridian. Procedure. Observers completed a single session lasting roughly 45 min seated in a darkened room seated at a distance of 82 cm from the screen. Observers were given a brief verbal description of the task prior to commencement of testing. They were told to maintain ﬁxation on a cross (0.05° × 0.05°) in the center of the monitor through- out each block. Blocks consisted of 100 trials and were separated by a rest period of at least 30 s. All Urale and Schwarzkopf 465 Figure 3. Example stimuli at various target-inducer edge-to-edge distances. (A) Large inducer Ebbinghaus conﬁguration at 0.14° and (B) 4.5°. (C) Small inducer conﬁguration 0.14° and (D) 4.5°. (E) Delboeuf conﬁguration with single ring at 0.14° and (F) 4.5°. (G) Two ring Delboeuf conﬁguration, at 0.14°. stimuli were presented on a half-tone background. On each trial, stimuli were displayed sequen- tially with the target centered at 0.42° either left or right of ﬁxation. The ﬁrst interval always appeared just to the left of ﬁxation, followed by the stimulus in the second interval which appeared to the right. The order of the presentation of the reference and test stimuli was decided pseudo- randomly on a per-trial basis. At the beginning of each trial, observers saw a blank (ﬁxation only) screen for 500 milliseconds (ms) before seeing one stimulus for ∼100 ms, followed by a blank screen again for 500 ms before the ﬁnal stimulus for ∼100 ms. They were told that they would be able to respond following pres- entation of all stimuli. Observers pressed the left or right keyboard button to indicate whether the left or right stimulus was larger or smaller. In alternating blocks, observers were instructed to either indicate the target that appeared larger or smaller. They were also told to ignore the inducers. In case of any prior knowledge of the Ebbinghaus illusion, observers were instructed to report on their prima facie experience instead of what they anticipated the correct answer to be. Pressing a button to indicate their response immediately began the next trial. The ratio of the test stimulus diameter relative to the reference diameter was varied using a 1-up-1-down staircase procedure. The procedure was used to determine the PSE for each condition. With two Ebbinghaus conﬁgura- tions, seven target-inducer distances, and a control condition, there were a total of 15 conditions. There were two staircases for each condition, progressing in steps on a binary logarithmic scale. We chose to use a binary logarithm because it linearizes stimulus size increments in line with Weber’s Law. Adjusting sizes in proportions, rather than a binary logarithmic scale as we do here, would be mathematically unsound as the non-linearity of the stimulus size ratios will theor- etically skew statistical and curve ﬁtting analyses. As an example, a stimulus half the size of the reference will have a ratio of 0.5, while a stimulus of the equivalent larger size will have a ratio of 2. The arithmetic mean of these values would be 1.25 above a ratio of 1. However, these two sizes are linearly comparable when represented as binary logarithmic units, that is, −1 (2−1) and 1 (21), respectively. Moreover, in logarithmic units, 0 corresponds to the absence of an illusion (i.e., a size ratio of 1). Nevertheless, some readers might ﬁnd it difﬁcult to interpret logarithmic units; we therefore plot our results in linear units of degrees of visual angle but this is done purely for visualization. On a given trial, the size of the test stimulus in degrees of visual angle was 0.56 × 2g. The stair- case was varied by adjusting g. For each condition, one staircase began with a test diameter 0.2g 466 Perception 52(7) larger than the reference target (i.e., ∼115% of the reference target diameter), and the other 0.2g smaller (i.e., ∼87% of the reference target diameter). The step size of the staircase varied depending on the number of reversals: 0.1g for trials up until the 2nd reversal, then 0.075 until the 4th reversal, followed by 0.05 until the 8th reversal, and then 0.025 for the remaining reversals (25 in total). Trials from each of the 30 staircases were randomly interleaved and discontinued after the requisite number of reversals. The experiment ended when all staircases were complete. We calculated the PSE across conditions for each observer by ﬁtting a cumulative Gaussian psy- chometric function to each condition using the weighted stimulus levels and responses from both staircases (R2 ≥ 0.98 for all ﬁts for the present experiment and ﬁts for psychometric functions in all subsequent experiments in this work). Assigned weighting to each data point was proportionate to the number of trials occurring at that stimulus level. All PSEs were taken as the 50% point of that function. To test the validity of our estimates we compared these values to PSEs calculated by taking the average size of the stimulus level during the last 8 reversals across both staircases for each condition, excluding values beyond twice the median absolute deviation in either direction. Using either method did not meaningfully change the pattern of results or conclusions of this manu- script. We chose a psychometric ﬁt across all experiments as it is a more sensitive and theoretically grounded analysis. Results and Discussion Figure 4 shows the group-level average PSEs for Experiment 1. Prior to analysis, we subtracted the PSE in the control condition from the PSE for both inducer conditions at each distance. These base- lined PSEs were used in all subsequent analyses. For both large and small inducers, we ﬁt a power function of the form axb + c, where a, b, and c are free parameters and x is target-inducer distance. We used a bootstrap technique to calculate the 95% conﬁdence bands for this function by randomly selecting a sample of 12 (with replacement) from the pool of observers and then re-calculating the group means and re-ﬁtting the power function to the new sample. This was repeated for a total of 10,000 times for each inducer type. We calculated goodness-of-ﬁt measures for both small, R2 = .838, and large inducers, R2 = .732, as well as observed model parameters (Supplemental Table 1). A plot containing individual-observer model ﬁts can be viewed in Supplemental Figure 1. In addition to the power function shown here, we also performed the same analysis with a two-term exponential function of the form aebx + cedx. We chose this as an alternative model because of the known exponential relationship between eccentricity and cortical magniﬁca- tion (Duncan & Boynton’s, 2003). This model performed well with small inducers but we chose a power model here because the exponential model performed poorly with large inducers (see Supplemental Table 2). Our results support our hypothesis that shorter target-inducer distances lead to an increase in per- ceived target size (larger, positive PSEs). For targets surrounded by small inducers, there was a clear uptick in PSEs for shorter distances. Moreover, with enough distance the sign of illusion inverted. The pattern for large inducers was more ambiguous. At all target-inducer distances, PSEs were negative, meaning the target was perceived as smaller. Importantly, our results also showed that the basic Ebbinghaus effect occurs with our novel presentation procedure where stimuli are pre- sented near the fovea in separate temporal intervals. Schwarzkopf and Rees (2013) hypothesized that at a short enough distance to the target, large inducers could make the target appear larger. We tested this by allowing large inducers to overlap and display at a distance much closer to the target compared to Roberts et al.’s (2005) study. Our results did not support this hypothesis, with a modest increase in PSE when large inducers were very close to the target. This may be due to the attractive effect of the nearer contours in large indu- cers being counteracted by contours on the far side of the inducers (Todorović & Jovanović , 2018), Urale and Schwarzkopf 467 Figure 4. Group mean PSEs across target-inducer retinal distances in Experiment 1. The horizontal dotted black line indicates the size of the reference stimulus, that is, the absence of any illusion. Solid and dashed lines are the ﬁt to the data for the small- and large-inducer conditions, respectively. Shaded regions show the 95% bootstrapped bands for the power functions for each inducer type. Error bars indicate ±1 standard error of the mean across observers. “dva” = degrees of visual angle. but may also reﬂect an unanticipated effect of allowing large inducers to overlap at short distances from the target. Speciﬁcally, this would also reduce the ﬁgural similarity between the inducers and the target, which has been shown to affect the strength of the illusion (Choplin & Medin, 1999; Coren & Enns, 1993; Deni & Brigner, 1997; Jaeger & Guenzel, 2001; Rose & Bressan, 2002). Generally, our results bear important similarities and differences compared with the results of Roberts et al. (2005). Their study also found a similar pattern when increasing target-inducer dis- tance with large and small inducer conditions. However, they found a more reliable reduction in PSEs at greater target-inducer distances compared to our study. Unlike Roberts et al. we did not vary the numbers of inducers to always form a complete ring around the target, so this discrepancy may reﬂect lower stimulus energy due to the large distances between them. In both studies, the illu- sion for small inducers does invert at greater differences, although the crossover point for Roberts et al.’s study (∼3–3.25°) differs considerably to the crossover seen here (∼1.2°). This may be due to changes in illusion strength related to overall stimulus size, a factor shown to reliably effect illusion strength in other studies (Knol et al., 2015; Massaro & Anderson, 1971). Given that sequential presentation of the elements of the Ebbinghaus illusion can reduce the illu- sion magnitude (Jaeger & Pollack, 1977), a potential concern stems from our choice to present stimuli at nearby locations. Potentially, an afterimage from the target or inducers from the ﬁrst interval could affect perception of the second interval; a persistent image of a target may enhance perceived simi- larity with a second target, and residual images of inducers may introduce an illusory effect on a lone 468 Perception 52(7) target in the second interval. However, we think these concerns are unlikely for the following reasons. Firstly, observers (including the two authors) did not report seeing afterimages. Secondly, we delib- erately offset each interval horizontally (and vertically in Experiment 2, see below), which should reduce the ability for any direct comparisons between stimuli. Thirdly, the temporal order of reference and test stimuli were counterbalanced, meaning any effect of inducers in the ﬁrst interval would be counteracted by trials where the inducer condition was in the second interval. Experiment 1 supports the hypothesized relationship between distance in visual space and PSE in the Ebbinghaus illusion. Speciﬁcally, we predicted that for a given inducer type (i.e., small, large) as cortical distance between target and inducers decreases, perceived size of the target should increase. We observed this effect, albeit more clearly for small inducers. In Experiment 2 we test the relationship between the Ebbinghaus and cortical distance further by taking advantage of the change in cortical magniﬁcation across the visual ﬁeld. Experiment 2 Cortical magniﬁcation in visual cortex falls off with eccentricity (Duncan & Boynton, 2003; Smith et al., 2001). Ebbinghaus stimuli at greater eccentricities therefore reduce cortical distance between representations. Chen et al. (2018) found the Ebbinghaus illusion was stronger when observers ﬁrst viewed a low-spatial frequency prime compared to a high-spatial frequency prime. Sensitivity to low-spatial frequencies increases with eccentricity (Henriksson et al., 2008), and categorization of low-spatial frequency scenes elicits greater activity in brain areas associated with the peripheral visual ﬁeld compared to high-spatial frequency scenes (Musel et al., 2013). We hypothesize that shorter cortical distances between the target and inducers produce an increase in perceived target size, irrespective of the inducer type. Thus, we should observe generally larger PSEs as stimuli are moved further into the periphery. The effect of eccentricity on the Ebbinghaus illusion has been investigated previously by Eymond et al. (2020). In one experiment, observers in their study compared a foveal test circle with a peripheral or foveal reference circle that was either an isolated control circle or an Ebbinghaus conﬁguration with large inducers. They found the PSE for the Ebbinghaus condition did not differ depending on eccentricity while the control condition appeared smaller in the periphery. While this may initially seem inconsistent with our hypotheses, observers in their study compared a foveal test stimulus with a peripheral target, and the authors note there is a general reduction in perceived size when stimuli are placed into the periphery (Baldwin et al., 2016). Therefore, the lack of an effect of eccentricity on PSE in the Ebbinghaus illusion in their experiment may indicate that the effects of the inducers are counteracting a reduc- tion in perceived size in the periphery. Our experiment differs from these studies in two key ways: Firstly, we test perception of the Ebbinghaus illusion at multiple distances from ﬁxation. This will allow us to observe graded effects of eccentricity. Secondly, targets in reference and test stimuli occurred at the same eccentric location. By doing this, account for stimuli varies in size in absolute terms across the visual ﬁeld. Methods Participants. We recruited 12 observers (10 females, age range 22–53) all with normal or corrected-to-normal vision. Observers provided written and informed consent and procedures were approved by UAHPEC. Experimental Setup. We conducted Experiment 2 on the same experimental setup as Experiment 1 with the addition of an Eyelink 1000 Desktop System eye-tracker (operating at 1,000 Hz; SR Research). Urale and Schwarzkopf 469 Stimuli. The retinal dimensions of target and inducer stimuli were identical to Experiment 1. Unlike Experiment 1, in Experiment 2 retinal target-inducer distances were ﬁxed while we manipulated the location of the Ebbinghaus stimuli along the visual ﬁeld’s horizontal meridian. To avoid crowding, we applied Bouma’s law (Bouma, 1970; Pelli & Tillman, 2008), which states that the absence of visual crowding effect can be achieved if the retinal distance between the two visual elements is no less than 50% of the distance between these elements and ﬁxation. Thus, the centers of target stimuli were positioned at a maximum distance of 4.5° from ﬁxation, dictating a suitable target-inducer distance of 2.25°. This distance was used for both large and small-inducer conﬁgura- tions. To avoid inﬂuence from attentional capture on each trial, the presentation of the stimuli was preceded by a primer stimulus to alert the observer to the location of the forthcoming stimuli. Procedure. The procedure for Experiment 2 was mostly the same as Experiment 1, except that the retinal distance between target and inducers was kept constant while the distance between the loca- tion of the target and foveal vision was manipulated. Observers sat in a dimly lit room where they positioned their head on a headrest and chinrest apparatus located in front of a computer monitor where they performed a 9-point calibration routine for the eye-tracker. The experimenter verbally instructed observers to maintain ﬁxation on the ﬁxation cross located in the center of the screen, that the eye-tracker was tracking their eyes, and to try to avoid blinking during stimulus presentation time. Each trial began with presentation of a ﬁxation cross. On a given trial the test and reference target stimuli could either occur at ﬁxation or at an eccentric location close to the horizontal meridian. Eccentric locations could occur either to the left or right hemiﬁeld. Exogenous cueing can affect perceived size in the objects in the periphery (Kirsch et al., 2020), so to avoid any extraneous effects of attentional re-orienting, we ensured that attentional allocation was consistent across con- ditions. We did this with an exogenous cueing stimulus: If on the current trial the target and test targets were to appear at an eccentric location, they were preceded by “×” shaped cue (0.3° × 0.3°) at the location of the forthcoming reference and test stimuli. This cue appeared for 100 ms, followed by a 500 ms interval of only the ﬁxation cross again, followed by the reference and test intervals. The order of test and reference intervals was pseudo-randomly determined on a trial-by-trial basis. Just as in Experiment 1, the reference target stimulus could be either the large- or small-inducer Ebbinghaus conﬁgurations or the control stimulus with no inducers, each with a 0.56° diameter target circle. The test stimulus varied according to the same staircase proced- ure described in Experiment 1. Each interval lasted 100 ms, separated by a 500 ms interval. Unlike Experiment 1, we offset the location of reference and test target in a vertical (rather than horizontal) orientation to avoid extraneous effects of one stimulus occurring at a more central loca- tion than the other. Thus, the center of the target circles in the ﬁrst and second interval always appeared 0.2° above and below the horizontal meridian, respectively. We used the eye-tracker to ensure observers were always looking at ﬁxation during presentation of the reference and test stimuli: A trial would be aborted if, during the reference and stimulus inter- vals, the observer blinked or if their gaze was tracked as deviating more than 1° from ﬁxation. We performed a single-point drift-correction procedure between each 100-trial block. If the reference and test intervals ran to completion, observers were again shown a ﬁxation cross while they responded by pressing a keyboard button to indicate whether they thought the target in the ﬁrst (top) or second (bottom) interval was larger or smaller, depending on the instructions of the current block. This response period was untimed and giving a response would immediately initiate the next trial. Consequently, observers were asked to blink and orient their gaze to the ﬁxation cross before giving their response. If the trial was aborted due to blinking or looking away from ﬁxation, the screen would show the ﬁxation cross for 500 ms before initiating the next trial. 470 Perception 52(7) There were 12 conditions in total with a 3 × 4 design: three types of inducer conditions (large, small, no inducers) and four eccentricities (0°, 1.69°, 2.8°, and 4.5°). There were two staircases for each of these conditions that operated as in Experiment 1. A given staircase ended after 25 reversals and the whole experiment ended when all staircases reached completion. Each block of trials ended after 100 trials or if all staircases were completed, ending the experiment. We calculated PSEs for each condition using the same procedure as Experiment 1. Results and Discussion As before, for a given eccentricity we subtracted the PSE for the control condition from the PSE for both inducer conditions. These baselined PSEs were used in all subsequent analyses. Figure 5 shows the group-level average PSEs for Experiment 2. Our analysis was to investigate whether PSEs increased or decreased with target eccentricity, and for this purpose we determined a linear function of function used in Experiment 1. We ﬁt this to the small, R2 = .732, and large, R2 = .839, inducer conditions. Conﬁdence bounds were generated using the same procedure as Experiment 1 and can be found in Supplemental Table 1. Generally, the target appeared larger as target-ﬁxation distance increased, the form y = a + bx as appropriate, than the power rather Figure 5. Group mean PSEs (illusion magnitude) across target-ﬁxation retinal distances (eccentricity) in Experiment 2 (units as in Figure 4). The horizontal dotted black line indicates the size of the reference stimulus, that is, the absence of any illusion. Shaded regions show the 95% bootstrapped bands for the linear ﬁt to both types of inducers. Solid and dashed lines show ﬁt to small- and large-inducer conditions, respectively. Error bars indicate ±1 standard error of the mean across observers. “dva” = degrees of visual angle. Urale and Schwarzkopf 471 although this effect was not observed to the same extent in large inducers. We see this in the con- ﬁdence interval for the slope parameter for large inducers, b = 0.009 (95% CI [0.025, −0.01]), which overlapped zero. This indicates that there is no clear direction (either positive or negative) of the slope representing the relationship between eccentricity and PSE in the large-inducer condi- tion, and that the slope itself is close to zero. However, the interval for small inducers did not cross zero, b = 0.031 (95% CI [0.0445, 0.0166]), indicating a reliable positive relationship between eccentricity and PSE as determined by our bootstrap procedure. We also observed that for some observers the PSE for small inducers switched sign as target-ﬁxation distance increased, in line with our ﬁndings while increasing target-inducer distance in Experiment 1. Cortical Distance and PSE. We looked at the effect of cortical distance on the Ebbinghaus illusion. Our approach takes inspiration from Mareschal et al.’s (2010) investigation of the effect of cortical distance on the tilt illusion. The tilt illusion (Gibson, 1937) is an illusion where the perceived tilt of a target line is inﬂuenced by the angle of surrounding lines. Mareschal and colleagues estimated cortical distance between target and surround across various retinal distances and concluded that the strength of the tilt illusion increases with cortical proximity. In a similar way, estimates of cor- tical distance allow us to investigate the relationship between cortical distance and PSE in Experiments 1 and 2. To do this, we chose to estimate linear cortical magniﬁcation factor (M ), which is the millimeters of cortex per degree of visual angle (Daniel & Whitteridge, 1961), using Duncan and Boynton’s (2003) formula: M = 9.81 × δ−.083, where δ denotes eccentricity in degrees of visual angle. By subtracting M between two different points (see Mareschal et al., 2010), we can estimate of the cortical distance between inducers and target across inducer condi- tions and experiments. For stimuli presented at ﬁxation, all inducers in each conﬁguration were equidistant to the target both in terms of visual space and cortical distance. However, when stimuli were presented at parafoveal locations in Experiment 2, the distances between individual inducers and the target were asymmetric; for example, cortical distance from the target is greater for the inducers positioned closer to ﬁxation compared to the more peripherally located inducers. To capture these variations, we calculated an index of cortical distance for each condition based on the average edge-to-edge cortical distance between the nearest edge of all eight inducers and the target. We plot the estimates of cortical distance against PSE in Figure 6. We used a bootstrap method to plot conﬁdence bands by taking 10,000 resamples (with replacement) of observers’ PSEs across both experiments. For the observed and each iteration of the bootstrapped data, we ﬁt a linear func- tion of the form y = a + bx, where a is the intercept, and b is the slope coefﬁcient. This was per- formed separately for both target-ﬁxation distance and estimated cortical distance. We chose a linear function as a parsimonious way to characterize a simple relationship between two variables. Cortical distance predicted PSE for both small inducers, R2 = .876, and large inducers, R2 = .304. For the relationship between cortical distance and PSE, slope coefﬁcients (b) for the large and small inducers were −0.01 (95% CI [−0.003, −0.017]) and −0.022 (95% CI [−0.016, −0.028]), respectively. We also ran the same procedure with a Difference-of-Gaussians (DoG) model (see Equation 1), in accordance with our theoretical expectations and to maintain consistency with the analysis in Experiments 3a and 3b (see below). The goodness-of-ﬁt and parameter values (including conﬁdence intervals derived from the bootstrap procedure) can be found in Supplemental Table 3. Upon visual inspection and comparison of goodness-of-ﬁt estimates, we determined that the linear model was a better ﬁt to the data from Experiments 1 and 2. This could be because the data points in the Ebbinghaus experiments fell within the steep portion of this function. Encouragingly, the two separate methods of manipulating cortical distance between target and inducers had comparable effects on the illusion strength. We observed agreement between PSEs 472 Perception 52(7) Figure 6. Group mean PSE as a function of estimated cortical distance. The horizontal dashed black line indicates the size of the reference stimulus, that is, the absence of any illusion. Small and large inducers are shown as circles and triangles, and PSEs from Experiment 1 and 2 are denoted by open and ﬁlled symbols, respectively. The size of the ﬁlled symbols denotes eccentricity in Experiment 2. Error bars indicate standard error (1±) of the mean across observers. Conﬁdence bounds show the 95% upper and lower bounds of the line ﬁt, produced from the bootstrap procedure. “dva” = degrees of visual angle, “mm” = millimeters. across the two experiments in conditions with similar cortical distance estimates, particularly for the small-inducer condition. Speciﬁcally, in Figure 6, markers at a similar position on the x axis, irre- spective of experiment, have similar PSEs. We observed a negative correlation between cortical distance and Ebbinghaus PSE in both large and small inducers, such that smaller cortical distance corresponded to larger perceived target size. These ﬁndings are consistent with the predictions based on previous neuroimaging work showing that smaller cortical extents associate with larger PSEs (Schwarzkopf et al., 2011) and especially perceptually larger stimuli (Schwarzkopf & Rees, 2013). Moreover, the shallower slope seen in the large inducer condition mirrors the results from Experiment 1. This may reﬂect non-linear interactions between the target and inducers due to antagonistic effects of the near and far contours in large inducers (Todorović & Jovanović , 2018). Mareschal et al.’s (2010) study also described opponent processes, which, in the context of the tilt illusion, were antagonistic “repulsive” and “assimilative” forces. These same mechanisms may account for repulsion and attraction in the Ebbinghaus illusion. Experiments 3a and 3b In the next experiment, we investigate why large and small inducers have contrasting effects on perceived target size. We saw support in Experiments 1 and 2 for the link between PSE and Urale and Schwarzkopf 473 estimated cortical distance, but they also replicated the different perceptual effects of large and small inducers. As these results and others (Roberts et al., 2005; Sherman & Chouinard, 2016; Todorović & Jovanović , 2018) have shown, this disparity is unlikely to be caused by a hypothetical size-contrast effect originating in mid or high-level vision. An alternative account is that these dif- ferences are driven by opponent processes which depend on the spatial (or cortical) extent of con- tours around the target. As others have stated (Rose & Bressan, 2002), contour-based accounts offer an incomplete account of the Ebbinghaus illusion, but it may be necessary. Proponents of accounts such as BCIT hold that this difference can be explained in terms of low-level contour interactions (Jaeger, 1978; Jaeger & Grasso, 1993; Sherman & Chouinard, 2016; Todorović & Jovanović , 2018; Weintraub, 1979; Weintraub et al., 1969; Weintraub & Schneck, 1986). The “biphasic” element of BCIT (Sherman & Chouinard, 2016) stipulates that contours nearer to the target have an attractive effect, while contours at more distance locations repel the target. Thus, the reason large inducers make the target look smaller is because large inducers have additional contours at farther distances from the target. An alternative account for these differences sees them as driven by higher level- categorization of inducers as whole objects beyond simple size contrast (Knol et al., 2015; Rose & Bressan, 2002). We test the effect of additional contours with the Delboeuf illusion (Delboeuf, 1865; Evans, 1995), an illusion in which perceived target size is affected by the proximity of a ring surrounding the target. The Delboeuf illusion is suitable for this purpose because it likely shares a common mechanism with the Ebbinghaus illusion. Supporting this, both Pressey (1977) and Sherman and Chouinard (2016) found that the two illusions share around a quarter of their variability. In another study, Roberts et al. (2005) found that a complete ring comprised of small Ebbinghaus inducers had the same illusory effect as a Delboeuf ring at a range of distances from the target. Accordingly, if negative PSEs associated with large inducers in the Ebbinghaus illusion are due to near and far contour placement, and if the Delboeuf and Ebbinghaus share a common mechan- ism, the simple addition of another ring in the Delboeuf illusion (Figure 3) should resemble the effect of large inducers in the Ebbinghaus illusion, such that we observe a downward shift in PSE. We test this hypothesis in Experiment 3a (henceforth “3a”). In Experiment 3a we observed that PSE did not trend towards zero with greater target-ring dis- tances (see “Results and discussion” section). The interaction between the surround and the target in the Ebbinghaus illusion has been conceptualized as sombrero-shaped center-surround of contextual interactions (Schwarzkopf, 2015), and in such a model we would expect the contextual effects to diminish to zero as it approaches the “brim” of the hat (i.e., the boundaries of any suppressive effect). To this end, in Experiment 3b (henceforth “3b”) we increased the ring-target distance further to observe if its effect on the target attenuates at even farther target-ring distances. Methods Participants. We recruited 12 volunteers (seven females, age range 22–49) for 3a and 14 volunteers (9 females, 21–52) for 3b, all with normal or corrected-to-normal vision. Observers provided written and informed consent, and procedures were approved by UAHPEC. Experimental Setup. The experimental setup for 3a was identical to Experiment 1. In order to cover an area of the visual ﬁeld ∼40° in diameter, we reduced the viewing distance 3b from 82 to 42 cm. Stimuli. In both experiments, the reference-interval stimuli consisted of a central circle (0.56° in diameter) with either a single- or double-ring conﬁguration in 3a (see Figure 3) and a single-ring only condition in 3b. Rings in both experiments had a thickness of ∼0.04°. 474 Perception 52(7) For 3a, on a given trial the inner-ring of the double-ring condition could be one of several dis- tances from the edge of the central circle: 0.14°, 0.44°, .7°, 1.13°, 1.55°, 2.25°, and 4.5°. The dis- tances were the same for the single-ring conﬁguration, with the addition of a 5.34° condition (i.e., the distance of the outer ring in the double-ring condition at 4.5°). The distance between the borders of the inner and outer rings was always 0.84°. We chose this distance to match the diameter of the large inducers from Experiment 1, and in doing so emulate the antagonistic effects of near and far contours in those stimuli. Experiment 3b featured the single ring conditions at the following dis- tances: 0.14°, 11°, 15°, and 20°. Procedure. The procedures for both experiments were the same as Experiment 1 (see Figure 2). Observers typically completed the experiment in 45 min for Experiment 3a, and 20 min for 3b. Results and Discussion Prior to analysis, we again subtracted the PSE from the control condition from all other conditions and used those baselined PSEs for all subsequent analyses. Figure 7 shows the group-level average PSEs for 3a and 3b as a function of target-ring distance in degrees of visual angle and estimated cortical distance. We also plotted data from Experiments 3a and 3b separately, as PSE as a function of retinal distance and using the same power function used in Experiment 1 (see Supplemental Figure 5 for plots, and goodness-of-ﬁt and parameter estimates in Supplemental Table 1). In 3a we used the Delboeuf illusion to determine if the difference between large and small indu- cers in the Ebbinghaus illusion are explainable as an interplay of near and far contours (Sherman & Chouinard, 2016; Todorović & Jovanović , 2018). On this basis our results support our hypothesis; compared to a single ring, the two-ring conﬁguration had the effect of numerically reducing the PSE (Figure 7). Looking at Figure 7b, this downward shift is most pronounced for cortical distances between 2 and 8 mm. Biphasic-contour interaction theory explains this as a result of antagonistic effects of contours, with farther contours working to repulse the percept of the target. At the largest cortical distances, the two rings probably fall within the same large peripheral receptive ﬁelds. Thus, the two-ring condition may effectively be a single-ring condition at these distances, only with somewhat increased stimulus energy. Our results also agree with earlier research, albeit with updated psychophysical methods. For example, with a target-inducer distance compar- able to our own, Weintraub and Schneck (1986) observed that the target appeared larger when only the inner-arc of large inducers were visible, but that PSE decreased and eventually changed sign as the outer fragments of those inducers were ﬁlled in with successively more dots. This resembles the addition of the outside contour in 3a, which we observed as shifting the PSE for most target-ring distances. Similarities aside, the effects of the two ring conditions in this experiment and those of small and large inducers in Experiment 1 (Figure 4) are not an exact match. This may be explained in terms of a contour-based account, as these two illusions differ in terms of variations in stimulus energy (i.e., the amount of contour). Unlike Roberts et al. (2005), we did not modify the number of inducers to maintain an uninterrupted surround of inducers in our experiments, meaning that at all distances the Delbouef presented a more complete surround than in Experiment 1. By contrast, Roberts and col- leagues found that an uninterrupted surround of small inducers affected perceived target size almost identically to a Delboeuf ring at various distances from the target. The two conditions in 3a had the same effect at very close distances (0.14°), yet there was a total separation of PSEs between the large- and small-inducer conditions at that same distance in Experiment 1. This, too, may be attrib- utable to a difference in stimulus energy because of intermediary contours between the nearest and farthest edges in large inducers, which are absent in the Delboeuf illusion. Alternatively, the differ- ence we observe between these two illusions may indicate a contribution of a second mechanism, Urale and Schwarzkopf 475 Figure 7. Group mean PSEs across target-ring distances in Experiments 3a and 3b as a function of retinal distance (A) and estimated cortical distance (B). The horizontal dashed black line indicates the size of the reference stimulus, that is, the absence of any illusion. Solid and dashed colored lines show the Difference-of-Gaussians (DoG) function ﬁtted to the data for the two-ring and single-ring conditions, respectively. Shaded regions show the 95% bootstrapped bands for DoG for each inducer type. Error bars indicate ±1 standard error of the mean across observers. “dva” = degrees of visual angle, “mm” = millimeters. possibly located higher in the visual stream. We discuss these possibilities further in the General discussion. DoG Function. We modelled the relationship between PSE and retinal and cortical distance, respect- ively, using a DoG function. DoG has been used to model inhibitory signals in extra-classical recep- tive ﬁelds (Cavanaugh et al., 2002), and it can account for the antagonistic center-surround proposed by Schwarzkopf (2015) and in BCIT (Sherman & Chouinard, 2016; Todorović & Jovanović , 2018; Weintraub & Schneck, 1986). To model a center-surround, we calculated the 476 Perception 52(7) difference of two Gaussian functions with peaks at zero distance. This took for form of Equation 1, where values σa, σb, a, and b, are left as free parameters. The DoG function is also used here to investigate whether the relationship between cortical distance and PSE resembles the hypothesized proﬁle of an antagonistic center-surround mechanism. We ﬁt this function using a least-squares pro- cedure and generated 95% conﬁdence bounds using a bootstrap technique with 10,000 repetitions. A single function was ﬁt to the combined PSEs from the single-ring condition in Experiments 3a and 3b, and another on the two-ring condition from 3a. For PSE as a function of retinal distance we ﬁt DoG functions for small ring, R2 = .955, and large inducer conditions, R2 = .99. For PSE as a function of cortical distance we did the same for the small ring, R2 = .974, and large ring conditions, R2 = .997. f (x) = ae −x2 a − be 2σ2 −x2 2σ2 b (1) Equation 1: DoG function We observe something closely resembling the sagittal cross-section of a sombrero, as described by Schwarzkopf (2015), with an excitatory center and inhibitory surround (Cavanaugh et al., 2002). Due to a theoretical interest relating to the cortical point image (see General discussion), we also calculated zero-crossing for the two functions in Figure 7B (PSE as a function of cortical distance), with 95% conﬁdence bounds generated from the bootstrapping procedure. That crossing was 5.57 mm (95% CI [6.53 4.25]) for the small ring and 3.65 mm (95% CI [4.76 2.98]) for large ring condition. An implication of a putative center-surround zone of interaction (Schwarzkopf, 2015) is a non- monotonic relationship between cortical distance and PSE. Theoretically, the effect should reduce to zero at sufﬁciently large cortical distances as the interaction between contours drops off. Mareschal et al. (2010) found such a “sombrero” pattern when varying cortical distance in the tilt illusion. In our study, even the farthest distances in 3b that did not return to zero. In aggregate, PSEs for distances between 11° and 20° (the three ﬁlled green squares on the right side of Figure 7) averaged slightly closer to zero in log units, that is, no illusion (M = −0.083, SE = 0.029), than the farthest single-ring condition in 3a (5.34°) (M = −0.123, SE = 0.024), although the difference between these was not signiﬁcant, t(24) = 1.045, p = .307. Hence, the PSEs across 3a and 3b trend towards zero, but the cortical distance necessary to see this is evidently beyond the dimen- sions we measured. We note that the results from 3b are not inconsistent with a contour-based account. Speciﬁcally, because of cortical magniﬁcation (Duncan & Boynton, 2003), the large dis- tances used in 3b translate to only minor differences in cortical distance compared to 3a (see Figure 7). General Discussion Numerous studies and many theories have been put forward to explain the Ebbinghaus and Delboeuf illusions (as many as 10; for a review, see Robinson, 1998). Despite that, little is known about what lies behind the illusion, and where in the brain that mechanism occurs. The present research adds to a growing body of research pointing to striate cortex as a promising loca- tion for this substrate. Such work shows that V1 encodes perceived size, like in the cases of the hallway illusion (Murray et al., 2006), and retinal afterimages projected onto near and far surfaces (Sperandio et al., 2012). Additionally, there is partial interocular transfer of the Ebbinghaus illusion (Song et al., 2011), suggesting a cortical process and again implicating V1, as this region is partially monocular, although of course this could also involve monocular and binocular neurons across mul- tiple stages of the visual hierarchy (Dougherty et al., 2019). Finally, there is a correlation between PSE and between-subject variability in cortical magniﬁcation (Schwarzkopf et al., 2011; Urale and Schwarzkopf 477 Schwarzkopf & Rees, 2013). Based on these ﬁndings, Schwarzkopf and Rees (2013) and later Schwarzkopf (2015), raised the possibility that interaction between representations on the topog- raphy of V1 may be the substrate for the Ebbinghaus illusion, and that these interactions depend in part on the cortical distance between those interactions. This proposal offers a plausible neural basis for contour-based accounts of the Ebbinghaus illusions, in which the low-level inter- actions between contours underlies the effect. To query this theory further, in the ﬁrst two experi- ments we tested the effect of cortical magniﬁcation on the Ebbinghaus illusion. Experiment 1 replicated previous work showing that shorter retinal distances between targets and Ebbinghaus inducers increases PSE (Knol et al., 2015; Roberts et al., 2005). Following this, Experiment 2 showed that for small inducers PSE increases when the target is positioned more per- ipherally. The prevailing trend in both experiments is consistent with the predictions that (1) in a general sense, the Ebbinghaus illusion depends in part on cortical magniﬁcation, and the speciﬁc prediction that (2) PSE correlates negatively with cortical distance. We were able to show conver- ging evidence to support this claim by manipulating cortical distances in two ways: ﬁrstly, by adjusting the retinal distance between target and inducer (Experiment 1), and secondly, by taking advantage of a reduction in cortical magniﬁcation across the visual ﬁeld and displaying targets in the periphery (Experiment 2). This culminated in the combined results of both experi- ments and comparing PSE against the estimated cortical distance between targets and inducers (Figure 6). In Experiment 3a, we used the Delboeuf illusion to show that, compared to a single-ring condition, a two-ring condition produced a perceptual shrinkage of the target (PSEs shifted down), lending support to a biphasic-contour account where more distal contours cause a decrease in the perceived size of the target. The attractive or repulsive effect of Ebbinghaus-style inducers may depend on a cortical distance equivalent to the cortical point image. The point image is the cortical representation of a single point stimulus expressed in millimeters (Mcllwain, 1986), calculated as the product of the cortical mag- niﬁcation factor and receptive ﬁeld size. Harvey and Dumoulin (2011) used pRF mapping to show that the cortical point image is near constant in V1, with only small decreases with eccentricity (similar ﬁndings are reported in non-human primates (Palmer et al., 2012). That is, in V1, there is a constancy in the ratio between receptive ﬁeld size and cortical magniﬁcation. Looking at the Ebbinghaus illusion, Schwarzkopf and Rees (2013) found that perceived enlargement of the target in a small-inducer Ebbinghaus conﬁguration occurred in observers with a relatively small V1. Using similar calculations to those used here (Duncan & Boynton, 2003), they estimated the inducer condition) to be cortical distance between target and inducer in their study (small ∼3.3 mm, falling within the 3–4 mm range of given for point images in V1 (Harvey & Dumoulin, 2011). Schwarzkopf and Rees surmised that the stability of the point image across the cortex indicates “constancy in spatial extent of cortical responses” (p. 11), and that the critical cortical distance between target and inducers in order for a perceived enlargement (i.e., attraction between contours) to occur might be equivalent to the cortical point image. Of interest here is whether a shift between attraction and repulsion occurs at a similar cortical distance in the present experiments. We estimated cortical distance at which a sign change (i.e., from perceived smaller to perceived larger) occurs as ∼7.2 mm for the small inducers in Ebbinghaus illusion (Experiment 1 & Experiment 2). These differ considerably to point image size in V1, which may reﬂect inﬂuences of surround completeness. Moreover, other studies have shown (Knol et al., 2015), target size inﬂuences PSE, so these differences may also reﬂect a difference in overall target size, which was 1.03° in Schwarzkopf and Rees’s study, compared to 0.56° in our study. Additionally, the arrangement of small inducers in their study were smaller (small inducer diameter was 26% of target diameter) compared to those used in the current study (35% of target diameter) and formed a more complete ring around the target compared to our study. The relative size of each inducer and completeness of the ring formed are known inﬂuences on PSE 478 Perception 52(7) (Roberts et al., 2005). Supporting this is that when we used the Delboeuf illusion, which consists of an uninterrupted ring, the sign change occurred between ∼5.5 and ∼3.7 mm for the single and double ring conditions, respectively (Experiment 3), which are closer to the estimates of Schwarzkopf and Rees. Thus, if the cortical point image is relevant to the magnitude with these illusions it likely interacts with overall stimulus energy. Despite our evidence for a link between cortical distance and the Ebbinghaus and Delboeuf illu- sions, conversion of retinal distances into cortical distance does not completely account for strength of illusions in other studies. As mentioned above, direct comparisons between the present experi- ments and those of Schwarzkopf and Rees (2013) and other comparable works (Knol et al., 2015; Roberts et al., 2005) are complicated by differences in stimulus dimensions and placement. For instance, Schwarzkopf and Rees (2013), Roberts et al. (2005), and Knol et al. (2015) presented ref- erence and target stimuli at a distance (distances from ﬁxation to target center: 4.65°, ∼15.2°, and 13°, respectively) at either side of ﬁxation, whereas we (with a few exceptions) presented stimuli at the same foveal locations at separate intervals. Compared to the experiments here, PSEs are not affected by cortical distance in the same way in those experiments. For instance, one of the condi- tions in Knol et al.’s (2015) study showed a repulsive effect of inducers (i.e., the target appeared larger) with a 0.48° target, a 1.9° target-inducer distance (measured as distance between target and inducer centers), and inducer radius of 0.09°. Considering the location of the target center in that condition (15.2° horizontal displacement from ﬁxation) and using Duncan and Boynton’s (2003) formula, the estimated cortical distance between targets and the nearest edge of the farthest inducers along the horizontal meridian averages to 0.198 mm. This is well within the range where attraction occurred in our experiments, yet in their experiment this distance coincided with a shrink- age of perceived target size. There are several factors that may account for these differences. Firstly, the stimuli in Roberts et al. (2005), Schwarzkopf and Rees (2013), and Knol et al. (2015) may have been affected by crowding effects; all three studies chose horizontal eccentricities and target-inducer spacing was generally shorter in those experiments than what Bouma’s law (Bouma, 1970) would dictate as necessary to avoid crowding. Indeed, there is evidence that size perception (as opposed to only rec- ognition), is affected by crowding (van den Berg et al., 2007). Crowding thus likely interferes with discriminating peripheral target sizes presented in close proximity to the inducers. This would render measurements of PSEs more variable and potentially obscures other effects. Knol et al. and Schwarzkopf and Rees both failed to ﬁnd substantial repulsive effects on perceived target size, yet Roberts et al., who used large horizontal displacements of their targets, reported an attract- ive effect at short target-inducer distances with small inducers, and effect resembling our ﬁndings from Experiment 1 when stimuli were presented at ﬁxation. It is also a distinct possibility that attractive effects in the illusion and crowding share the same underlying neural mechanism. When cortical distance is sufﬁciently large, this would result in a perceptual overestimation of the target, but when cortical distance is too short it completely disrupts size discrimination. Finally, are potential differences originating in the task: In addition to greater horizontal displace- ment, Roberts et al., Schwarzkopf and Rees, and Knol et al. also presented stimuli simultaneously, while we presented stimuli in separate temporal intervals. We stress that a low-level contour interaction underlying two size illusions of the nature we describe here may fall short of providing a complete account of these illusions. Two decades ago, Rose and Bressan (2002) observed that research with inducers the same size or larger than the target does not have effects on perceived size consistent with a static zone of repulsion and attraction (Girgus et al., 1972; Jaeger & Grasso, 1993; Massaro & Anderson, 1971; Weintraub, 1979). This is reﬂected in the current study, with the large inducer condition being less responsive than small inducers to manipulations of retinal size and eccentricity in Experiments 1 and 2, respect- ively. A constant gradient of interaction between contours would not explain this discrepancy Urale and Schwarzkopf 479 between conditions, even when considering potential antagonistic effects between relatively near and far contours. This is even more apparent when comparing Experiment 1 (Ebbinghaus illusion) with Experiment 3 (Delboeuf illusion). We have suggested that these inconsistencies may be attrib- utable to less consistent spacing (and thus differences in stimulus energy) of Ebbinghaus inducers versus the continuous ring in the Delboeuf illusion, but mid- and high-level cognitive factors may also ﬁll this explanatory gap. Last but not least, several studies suggest that cognitive factors modulate the strength of size illu- sions, including the Ebbinghaus illusion. For example, Gestalt grouping of a surround reduces PSE in the Ebbinghaus illusion (Rashal et al., 2020), and several studies have observed that size is affected by ﬁgural similarity between inducers and targets, even while controlling for the proximity and distribution of contours (Choplin & Medin, 1999; Coren & Enns, 1993; Deni & Brigner, 1997; Jaeger & Guenzel, 2001; Rose & Bressan, 2002). These ﬁgural effects have commonly been explained in terms of object-level categorization and attention. Indeed, attention in other contexts has been known to affect perceived size; Kirsch et al. (2020) found that a peripheral target appears small while attending to a central target. Fang et al. (2008) used fMRI to show that spatial distri- bution of activity in V1 reﬂected the perceived size of two attended targets embedded in the hallway illusion, and that this activity was signiﬁcantly diminished when attention was narrowed by a demanding central task. There are also reports that semantic knowledge affects the Ebbinghaus illu- sion, with objects of a known size biasing their perceived size when surrounded by Ebbinghaus-style inducers (Hughes & Fernandez-Duque, 2010). This relates to ﬁndings that ventral temporal cortex (Konkle & Oliva, 2012) show selectivity for real-world size of objects, independent of image transformations. Consistent with known models of predictive processing across various stages of visual processing (Ballard & Jehee, 2012; Rao & Ballard, 1999), these size-encoded representations could drive feedback signals that boost predicted signals in earlier visual areas in V1 consistent with these predictions. Thus, while a mechanism that involves spa- tially contingent interactions between low-level representations shows promise in explaining some of the known characteristics of the Ebbinghaus and Delboeuf illusions, more work is needed to determine whether contour-interaction alone can explain these illusions or if there is a need for the addition of other (potentially cognitive) mechanisms. Conclusion In addition to showing a relationship between cortical distance and the Ebbinghaus illusion (Schwarzkopf & Rees, 2013), our results broadly support biphasic contour-based accounts of the Ebbinghaus and Delboeuf illusions. Speciﬁcally, shorter cortical distances between inducers and target in the Ebbinghaus illusion, whether due to retinal distance (Experiment 1) or cortical mag- niﬁcation (Experiment 2), associate with perceptual enlargement of the target. We did not conﬁrm the prediction that large Ebbinghaus inducers produce perceptual enlargement at short dis- tances, a ﬁnding potentially due to repulsive effects of distal contours (on the far side of the inducer relative to the target) counteracting attractive effects of nearer contours. We noted that predictions based on estimated cortical distance did not align with select ﬁndings from other studies (Knol et al., 2015; Roberts et al., 2005; Schwarzkopf & Rees, 2013), possibly due to differences in stimu- lus dimensions, stimulus location, and task design. In the Delboeuf illusion, we showed that the addition of a second, more distant contour reliably decreases perceived target size compared to a single-ring, a ﬁnding aligned with an antagonistic surround described in contour-based accounts, such as BCIT (Roberts et al., 2005; Sherman & Chouinard, 2016; Todorović & Jovanović , 2018; Weintraub & Schneck, 1986). Lastly, we found that at large retinal distances (>11°), a single-ring Delboeuf ring still decreased perceived target size, a ﬁnding that may reﬂect low cortical magniﬁcation (the relatively minor changes in cortical distance) at points ranging into peripheral 480 Perception 52(7) vision. Future studies should continue to characterize effects of the surround in these illusions, including interactions between surround elements, potential inﬂuences of task design, and contri- butions of mid- and high-level vision. Author contribution(s) Poutasi W. B. Urale: Conceptualization; Formal analysis; Investigation; Methodology; Project administra- tion; Software; Visualization; Writing – original draft; Writing – review & editing. D. Samuel Schwarzkopf: Conceptualization; Formal analysis; Investigation; Methodology; Resources; Software; Supervision; Writing – review & editing. Declaration of Conﬂicting Interests The authors declared no potential conﬂicts of interest with respect to the research, authorship, and/or publica- tion of this article. Funding The authors received no ﬁnancial support for the research, authorship, and/or publication of this article. Data Availability Data and scripts used to conduct this analysis can be viewed at Open Science Framework: Data and analysis for Effects of cortical distance on the Ebbinghaus and Delboeuf illusions. https://doi.org/10.17605/OSF.IO/ GUHSF. ORCID iDs Poutasi W. B. Urale Dietrich Samuel Schwarzkopf https://orcid.org/0000-0001-6106-2297 https://orcid.org/0000-0003-3686-1622 Supplemental Material Supplemental material for this article is available online. References Aglioti, S., DeSouza, J. F. X., & Goodale, M. A. (1995). Size-contrast illusions deceive the eye but not the hand. Current Biology, 5, 679–685. https://doi.org/10.1016/S0960-9822(95)00133-3 Baldwin, J., Burleigh, A., Pepperell, R., & Ruta, N. (2016). The perceived size and shape of objects in periph- eral vision. I-Perception, 7, 2041669516661900. https://doi.org/10.1177/2041669516661900 Ballard, D. H., & Jehee, J. (2012). Dynamic coding of signed quantities in cortical feedback circuits. Frontiers in Psychology, 3, 254. https://doi.org/10.3389/FPSYG.2012.00254/BIBTEX Bouma, H. (1970). Interaction effects in parafoveal letter recognition. Nature, 226, 177–178. https://doi.org/10. 1038/226177a0 Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vision, 10, 433–436. http://color.psych.ucsb.edu/ psychtoolbox Cavanaugh, J. R., Bair, W., & Movshon, A. J. (2002). Nature and interaction of signals from the receptive ﬁeld center and surround in macaque V1 neurons. Journal of Neurophysiology, 88, 2530–2546. https://doi.org/ 10.1152/jn.00692.2001 Chen, L., Qiao, C., & Jiang, Y. (2018). Low-spatial-frequency bias in context-dependent visual size percep- tion. Journal of Vision, 18, 2–9. https://doi.org/10.1167/18.8.2 Choplin, J. M., & Medin, D. L. (1999). Similarity of the perimeters in the Ebbinghaus illusion. Perception & Psychophysics, 61, 3–12. https://doi.org/10.3758/BF03211944 Clifford, C. W. G. (2014). The tilt illusion: Phenomenology and functional implications. Vision Research, 104, 3–11. https://doi.org/10.1016/J.VISRES.2014.06.009 Urale and Schwarzkopf 481 Coren, S., & Enns, J. T. (1993). Size contrast as a function of conceptual similarity between test and inducers. Perception & Psychophysics, 54, 579–588. https://doi.org/10.3758/BF03211782 Coren, S., & Girgus, J. S. (1978). Seeing is deceiving: The psychology of visual illusions. Routledge. Coren, S., & Miller, J. (1974). Size contrast as a function of ﬁgural similarity. Perception & Psychophysics, 16, 355–357. https://doi.org/10.3758/BF03203955 Daniel, P. M., & Whitteridge, D. (1961). The representation of the visual ﬁeld on the cerebral cortex in monkeys. The Journal of Physiology, 159, 203–221. https://doi.org/10.1113/JPHYSIOL.1961.SP006803 Delboeuf, F. J. (1865). Note sur certaines illusions d’optique: Essai d’une théorie psychophysique de la maniere dont l’oeil apprécie les distances et les angles. Bulletins de l’Académie Royale Des Sciences, Lettres et Beaux-Arts de Belgique, 19, 195–216. Deni, J. R., & Brigner, W. L. (1997). Ebbinghaus illusion: Effect of ﬁgural similarity upon magnitude of illu- sion when context elements are equal in perceived size. Perceptual and Motor Skills, 84, 1171–1175. https://doi.org/10.2466/pms.1997.84.3c.1171 Dougherty, K., Cox, M. A., Westerberg, J. A., & Maier, A. (2019). Binocular modulation of monocular V1 neurons. Current Biology, 29, 381–391.e4. https://doi.org/10.1016/J.CUB.2018.12.004 Duncan, R. O., & Boynton, G. M. (2003). Cortical magniﬁcation within human primary visual cortex corre- lates with acuity thresholds. Neuron, 38, 659–671. https://doi.org/10.1016/S0896-6273(03)00265-4 Ebbinghaus, H. (1902). The principles of psychology. Veit, Leipzig, 437. Evans, R. B. (1995). Joseph Delboeuf on visual illusions: A historical sketch. The American Journal of Psychology, 108, 563–574. https://doi.org/10.2307/1423073 Eymond, C., Seidel Malkinson, T., & Naccache, L. (2020). Learning to see the Ebbinghaus illusion in the per- iphery reveals a top-down stabilization of size perception across the visual ﬁeld. Scientiﬁc Reports, 10, 1– 10. https://doi.org/10.1038/s41598-020-69329-9 Fang, F., Boyaci, H., Kersten, D., & Murray, S. O. (2008). Attention-dependent representation of a size illusion in human V1. Current Biology, 18, 1707–1712. https://doi.org/10.1016/j.cub.2008.09.025 Gibson, J. J. (1937). Adaptation, after-effect, and contrast in the perception of tilted lines. II. Simultaneous contrast and the areal restriction of the after-effect. Journal of Experimental Psychology, 20, 553–569. https://doi.org/10.1037/H0057585 Girgus, J. S., Coren, S., & Agdern, M. (1972). The interrelationship between the Ebbinghaus and Delboeuf illusions. Journal of Experimental Psychology, 95, 453–455. https://doi.org/10.1037/H0033606 Haffenden, A. M., Schiff, K. C., & Goodale, M. A. (2001). The dissociation between perception and action in the Ebbinghaus illusion: Nonillusory effects of pictorial cues on grasp. Current Biology, 11, 177–181. https://doi.org/10.1016/S0960-9822(01)00023-9 Harvey, B. M., & Dumoulin, S. O. (2011). The relationship between cortical magniﬁcation factor and popu- lation receptive ﬁeld size in human visual Cortex: Constancies in cortical architecture. Journal of Neuroscience, 31, 13604–13612. https://doi.org/10.1523/JNEUROSCI.2572-11.2011 Henriksson, L., Nurminen, L., Hyvärinen, A., & Vanni, S. (2008). Spatial frequency tuning in human retino- topic visual areas. Journal of Vision, 8, 5–5. https://doi.org/10.1167/8.10.5 Hughes, M., & Fernandez-Duque, D. (2010). Knowledge inﬂuences perception: Evidence from the Ebbinghaus illusion. Journal of Vision, 10, 954–954. https://doi.org/10.1167/10.7.954 Jaeger, T. (1978). Ebbinghaus illusions: Size contrast or contour interaction phenomena? Perception & Psychophysics, 24, 337–342. https://doi.org/10.3758/BF03204250 Jaeger, T., & Grasso, K. (1993). Contour lightness and separation effects in the Ebbinghaus illusion. Perceptual and Motor Skills, 76, 255–258. https://doi.org/10.2466/pms.1993.76.1.255 Jaeger, T., & Guenzel, N. (2001). Similarity and lightness effects in Ebbinghaus illusion created by keyboard characters. Perceptual and Motor Skills, 92, 151–156. https://doi.org/10.2466/pms.2001.92.1.151 Jaeger, T., & Klahs, K. (2015). The Ebbinghaus illusion: New contextual effects and theoretical considera- tions. Perceptual and Motor Skills, 120, 177–182. https://doi.org/10.2466/24.27.PMS.120v13x4 Jaeger, T., & Long, S. (2007). Effects of contour proximity and lightness on Delboeuf illusions created by cir- cumscribed letters. Perceptual and Motor Skills, 105, 253–260. https://doi.org/10.2466/PMS.105.1.253- 260 Jaeger, T., & Lorden, R. (1980). Delboeuf illusions: Contour or size detector interactions? Perceptual and Motor Skills, 50, 376–378. https://doi.org/10.1177/003151258005000205 482 Perception 52(7) Jaeger, T., & Pollack, R. H. (1977). Effect of contrast level and temporal order on the Ebbinghaus circles illu- sion. Perception & Psychophysics, 21, 83–87. https://doi.org/10.3758/BF03199473 Kirsch, W., Pﬁster, R., & Kunde, W. (2020). On why objects appear smaller in the visual periphery. Psychological Science, 31, 88–96. https://doi.org/10.1177/0956797619892624 Kleiner, M., Brainard, D., & Pelli, D. G. (2007). What’s new in Psychtoolbox-3? Perception 36 ECVP Abstract Supplement. http://www.psychtoolbox.org. Knol, H., Huys, R., Sarrazin, J.-C., & Jirsa, V. K. (2015). Quantifying the Ebbinghaus ﬁgure effect: Target size, context size, and target-context distance determine the presence and direction of the illusion. Frontiers in Psychology, 6, 1679. https://doi.org/10.3389/fpsyg.2015.01679 Konkle, T., & Oliva, A. (2012). A real-world size organization of object responses in occipitotemporal cortex. Neuron, 74, 1114–1124. https://doi.org/10.1016/J.NEURON.2012.04.036 Mareschal, I., Morgan, M. J., & Solomon, J. A. (2010). Cortical distance determines whether ﬂankers cause crowding or the tilt illusion. Journal of Vision, 10, 13–13. https://doi.org/10.1167/10.8.13 Massaro, D. W., & Anderson, N. H. (1970). A test of a perspective theory of geometrical illusions. Source: The American Journal of Psychology, 83, 567–575. Massaro, D. W., & Anderson, N. H. (1971). Judgmental model of the Ebbinghaus illusion. Journal of Experimental Psychology, 89, 147–151. Mcllwain, J. T. (1986). Point images in the visual system: New interest in an old idea. Trends in Neurosciences, 9, 354–358. https://doi.org/10.1016/0166-2236(86)90113-X Moutsiana, C., de Haas, B., Papageorgiou, A., van Dijk, J. A., Balraj, A., Greenwood, J. A., & Schwarzkopf, D. S. (2016). Cortical idiosyncrasies predict the perception of object size. Nature Communications, 7, 1–12. https://doi.org/10.1038/ncomms12110 Murray, S. O., Boyaci, H., & Kersten, D. (2006). The representation of perceived angular size in human primary visual cortex. Nature Neuroscience, 9, 429–434. https://doi.org/10.1038/nn1641 Musel, B., Bordier, C., Dojat, M., Pichat, C., Chokron, S., le Bas, J.-F., & Peyrin, C. (2013). Retinotopic and lateralized processing of spatial frequencies in human visual cortex during scene categorization. Journal of Cognitive Neuroscience, 25, 1315–1331. https://doi.org/10.1162/JOCN_A_00397 Obonai, T. (1954). Induction effects in estimates of extent. Journal of Experimental Psychology, 47, 57–60. https://doi.org/10.1037/H0057223 Palmer, C. R., Chen, Y., & Seidemann, E. (2012). Uniform spatial spread of population activity in primate parafoveal V1. Journal of Neurophysiology, 107, 1857–1867. https://doi.org/https://doi.org/https:// doi.org/ 10.1152/jn.00117.2011 Pelli, D. G. (1997). The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10, 437–442. Pelli, D. G., & Tillman, K. A. (2008). The uncrowded window of object recognition. Nature Neuroscience, 11, 1129–1135. https://doi.org/10.1038/nn.2187 Pressey, A. W. (1977). Measuring the Titchener circles and Delboeuf illusions with the method of adjustment. Bulletin of the Psychonomic Society, 10, 118–120. https://doi.org/10.3758/BF03329298 Rao, R. P. N., & Ballard, D. H. (1999). Predictive coding in the visual cortex: A functional interpretation of some extra-classical receptive-ﬁeld effects. Nature Neuroscience, 2, 79–87. https://doi.org/10.1038/4580 Rashal, E., Cretenoud, A. F., & Herzog, M. H. (2020). Perceptual grouping leads to objecthood effects in the Ebbinghaus illusion. Journal of Vision, 20, 1–15. https://doi.org/10.1167/JOV.20.8.11 Roberts, B., Harris, M. G., & Yates, T. A. (2005). The roles of inducer size and distance in the Ebbinghaus illusion (Titchener circles). Perception, 34, 847–856. https://doi.org/10.1068/p5273 Robinson, J. O. (1998). The Psychology of Visual Illusion. Hutchinson and Co. Lt. Rose, D., & Bressan, P. (2002). Going round in circles: Shape effects in the Ebbinghaus illusion. Spatial Vision, 15, 191–203. Schwarzkopf, D. S. (2015). Where is size in the brain of the beholder? Multisensory Research, 28, 285–296. https://doi.org/10.1163/22134808-00002474 Schwarzkopf, D. S., & Rees, G. (2013). Subjective size perception depends on central visual cortical magni- ﬁcation in human V1. PLoS ONE, 8, e60550. https://doi.org/10.1371/journal.pone.0060550 Schwarzkopf, D. S., Song, C., & Rees, G. (2011). The surface area of human V1 predicts the subjective experi- ence of object size. Nature Neuroscience, 14, 28–30. https://doi.org/10.1038/nn.2706 Urale and Schwarzkopf 483 Sherman, J. A., & Chouinard, P. A. (2016). Attractive contours of the Ebbinghaus illusion. Perceptual and Motor Skills, 122, 88–95. https://doi.org/10.1177/0031512515626632 Smith, A. T., Singh, K. D., Williams, A. L., & Greenlee, M. W. (2001). Estimating receptive ﬁeld size from fMRI data in human striate and extrastriate visual cortex. Cerebral Cortex, 11, 1182–1190. https://doi.org/ 10.1093/cercor/11.12.1182 Song, C., Samuel Schwarzkopf, D., Lutti, A., Li, B., Kanai, R., & Rees, G. (2013). Effective connectivity within human primary visual cortex predicts interindividual diversity in illusory perception. Journal of Neuroscience, 33, 18781–18791. https://doi.org/10.1523/JNEUROSCI.4201-12.2013 Song, C., Schwarzkopf, D. S., & Rees, G. (2011). Interocular induction of illusory size perception. BMC Neuroscience, 12, 27. https://doi.org/10.1186/1471-2202-12-27 Sperandio, I., Lak, A., & Goodale, M. A. (2012). Afterimage size is modulated by size-contrast illusions. Journal of Vision, 12, 18–18. https://doi.org/10.1167/12.2.18 Titchener, E. B. (1905). Experimental psychology: A manual of laboratory practice (Vol. 2). Johnson Reprint Company. Todorović , D., & Jovanović , L. (2018). Is the Ebbinghaus illusion a size contrast illusion? Acta Psychologica, 185, 180–187. https://doi.org/10.1016/J.ACTPSY.2018.02.011 van den Berg, R., Roerdink, J. B. T. M., & Cornelissen, F. W. (2007). On the generality of crowding: Visual crowding in size, saturation, and hue compared to orientation. Journal of Vision, 7, 14–14. https://doi.org/ 10.1167/7.2.14 Weintraub, D. J. (1979). Ebbinghaus illusion: Context, contour, and age inﬂuence the judged size of a circle amidst circles. Journal of Experimental Psychology: Human Perception and Performance, 5, 353–364. https://doi.org/10.1037/0096-1523.5.2.353 Weintraub, D. J., & Schneck, M. K. (1986). Fragments of Delboeuf and Ebbinghaus illusions: Contour/context explorations of misjudged circle size. Perception & Psychophysics, 40, 147–158. https://doi.org/10.3758/ BF03203010 Weintraub, D. J., Wilson, B. A., Greene, R. D., & Palmquist, M. J. (1969). Delboeuf illusion: Displacement versus diameter, arc deletions, and brightness contrast. Journal of Experimental Psychology, 80, 505–511. https://doi.org/10.1037/h0027424 Yamazaki, Y., Otsuka, Y., Kanazawa, S., Yamaguchi, M. K., & Kanazawa, S. O. (2010). Perception of the Ebbinghaus illusion in 5- to 8-month-old infants. Japanese Psychological Research, 52, 33–40. https:// doi.org/10.1111/j.1468-5884.2009.00420.x
10.1186_s12871-021-01333-6	Kalu et al. BMC Anesthesiology (2021) 21:114 https://doi.org/10.1186/s12871-021-01333-6 R E S E A R C H A R T I C L E Open Access Effect of preoperative versus postoperative use of transversus abdominis plane block with plain 0.25 % bupivacaine on postoperative opioid use: a retrospective study Richard Kalu1, Peter Boateng2, Lauren Carrier2, Jaime Garzon2, Amy Tang3, Craig Reickert4 and Amalia Stefanou4* Abstract Background: Enhanced recovery protocols optimize pain control via multimodal approaches that include transversus abdominis plane (TAP) block. The aim of this study was to evaluate the effect of preoperative vs. postoperative plain 0.25 % bupivacaine TAP block on postoperative opioid use after colorectal surgery. Methods: A retrospective cohort study comparing postoperative opioid use in patients who received preoperative (n = 240) vs. postoperative (n = 22) plain 0.25 % bupivacaine TAP blocks. The study was conducted in a single tertiary care institution and included patients who underwent colorectal resections between August 2018 and January 2020. The primary outcome of the study was postoperative opioid use. Secondary outcomes included operative details, length of stay, reoperation, and readmission rates. Results: Patients who received postoperative plain 0.25 % bupivacaine TAP blocks were less likely to require postoperative patient-controlled analgesia (PCA) (59.1 % vs. 83.3 %; p = 0.012) and opioid medications on discharge (6.4 % vs. 16.9 %; p = 0.004) relative to patients who received preoperative TAP. When needed, a significantly smaller amount of opioid was prescribed to the postoperative group (84.5 vs. 32.0 mg, p = 0.047). No significant differences were noted in the duration of postoperative PCA use, amount of oral opioid use, and length of stay. Conclusions: Plain 0.25 % bupivacaine TAP block administered postoperatively was associated with significantly lower need for postoperative PCA and discharge opioid medications. The overall hospital length of stay was not affected by the timing of TAP block. Because of the limited sample size in this study, conclusions cannot be generalized, and more research will be required. Keywords: Transversus abdominis plane block, Preemptive analgesia, TAP block * Correspondence: Astefan2@hfhs.org 4Division of Colon and Rectal Surgery, Henry Ford Hospital, 2799 West Grand Blvd Detroit, 48202 Detroit, MI, USA Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Kalu et al. BMC Anesthesiology (2021) 21:114 Page 2 of 6 Background Enhanced recovery protocols (ERP) after surgery have the aim of reducing morbidity and the surgical stress re- sponse while advancing early return of patients to their baseline functioning [1]. Components of ERP include optimal pain control and surgical stress reduction with regional anesthesia, early mobilization, and early enteral nutrition [2, 3]. Multiple studies including randomized controlled trials have shown a reduction in hospital length of stay, duration of postoperative ileus, reduced morbidity, and an earlier return of normal function after ERP implementation [1, 4–6]. Optimal pain management is an integral part of an ERP, particularly after colon surgery. Poor pain control may lead to longer length of stay, cost, and patient dis- satisfaction [6]. Many ERPs use a multimodal approach to achieve an optimal pain control, employing neuraxial and regional anesthesia techniques and lower utilization of opioids as the primary analgesic [7–9]. Transversus abdominis plane (TAP) block is an example of a regional anesthetic technique that has been used extensively in abdominal surgery [10, 11]. The TAP block involves injecting plain 0.25 % bupivacaine into the fascial plane between the internal oblique and transversus abdominis muscles. The duration of action for plain bupivacaine ranges from 2 to 10 h with peak effect noted around 30 to 45 min {Beiranvand, 2018 #348} [12]. A TAP block may be administered at any time during the immediate perioperative period. However, whether plain bupivacaine TAP block is effective in colorectal surgery remains to be elucidated. There have been no studies assessing the timing of TAP block administration for optimal postoperative pain control. In the present study, we assessed the effect of preoperative vs. postop- erative administration of TAP block using plain 0.25 % bupivacaine on postoperative opioid use in patients undergoing colorectal surgery. Furthermore, we assessed for any effects on length of stay, rates of reoperation, and readmission. Methods Patient selection All patients who received TAP blocks during a transab- dominal colorectal resection between August 2018 and January 2020 were identified through hospital chart re- view. The TAP procedure was performed by the regional anesthesiologist on duty that day. The majority of TAP blocks are performed preoperatively in our institution. At times, the TAP block was done at the conclusion of the operation often due to timing issues. All included patients underwent colorectal resection done by colon and rectal surgeons at our tertiary care hospital. We ab- stracted patient demographics, medical comorbidities, preoperative diagnosis, past medical and surgical history, procedure-related details, and postoperative opioid use after the index operation until discharge. The study was approved by the Henry Ford Health System Institutional Review Board. This manuscript was conducted in ac- cordance with the ethical standards laid down in the amended 1964 Declaration of Helsinki. Patients were divided in 2 groups based on timing of TAP block procedure: before (preoperative) or immedi- ately after (postoperative) their surgical procedure. in acute pain management TAP block technique Under ultrasound guidance, TAP blocks were performed by, or under supervision of, the attending anesthesiologist who specializes and regional anesthesia. The blocks were performed either 1 h preopera- tively or at the conclusion of the surgery procedure within 30 min of reaching the post-anesthesia care unit. The TAP block consisted of bilateral injection of 20 mL of 0.25 % plain bupivacaine in the fascial plane between the internal oblique and transversus abdominis muscles in the midaxillary line. None of the patients in the 2 groups received additional re- gional anesthesia, including epidural or spinal anesthesia. Additional multimodal adjuncts Our postoperative pain management protocol consisted of scheduled acetaminophen, muscle relaxant (methocarbamol or cyclobenzaprine), gabapentin, and ketorolac in the ab- sence of any contraindication. Patient-controlled analgesia (PCAs) were initiated for patients who had inadequate pain control, had pain-related hypertension or tachycardia, or were unable to ambulate or participate in pulmonary toilet. All patients were routinely assessed for their pain levels. However, the use of numerical pain scoring was not consist- ent amongst the patients and hence, was not included in the study. While there may be minor differences in pain manage- ment based on specific surgeon-preference, most of the pa- tients were treated using this protocol. Finally, the decision to prescribe discharge opioid was based on several factors such as surgeon’s discretion, the level patient’s pain control, and history of chronic opioid use. Outcomes The primary outcome of the study was postoperative opioid use. The incidence, type, and total amount of opi- oid in milligram morphine equivalent (MME) and non- the opioid analgesic medications administered after index procedure were recorded. Secondary outcomes in- cluded procedure-related operative details, length of hospital stay, rates of readmission, and reoperation. Statistical analysis Statistical analyses were conducted to compare the baseline characteristics of the patients who received preoperative TAP blocks with those who received postoperative TAP Kalu et al. BMC Anesthesiology (2021) 21:114 Page 3 of 6 blocks. Continuous variables were described using the mean and standard deviation (SD), and categorical variables were described with the frequency and percentage. Analysis of variance or Kruskal–Wallis tests were used for continuous variables and chi-square tests or Fisher’s exact tests were used for categorical variables as appropriate. P < 0.05 was considered statistically significant. All analyses were done in SAS 9.4 (SAS Institute, Cary, NC). Results Descriptive analysis A total of 262 patients were identified through chart re- view. A total of 240 patients received preoperative TAP blocks and 22 received postoperative TAP blocks. The mean (SD) patient age was 57.8 years (16 years), 45 % were men, and the mean (SD) body mass index was 28.4 kg/m2 (7.32 kg/m2). There were no significant dif- ferences in the 2 groups with regard to age, sex, body mass index, American Society of Anesthesiology classifi- cation, history of cancer or inflammatory bowel disease, and opioid use at the time of the index procedure (Table 1). The 2 groups were similar in terms of comor- bidities, including history of hypertension, diabetes mel- litus, hyperlipidemia, congestive heart failure, chronic pulmonary obstructive disease, smoking, and alcohol use. The surgical indications and surgical approaches were similar between the 2 groups (Table 1). Analgesic Requirements Table 2 shows the postoperative analgesics used by the two groups. The patients who received plain bupivacaine Table 1 Patient characteristics based on timing of plain transversus abdominis plane (TAP) block Variables Age, years, mean (SD) Male, no. (%) Body mass index, kg/m2, mean (SD) Overall (N = 262) 57.76 (16.10) 118 (45.0) 28.38 (7.32) Diabetes mellitus, no. (%) Hyperlipidemia, no. (%) Cigarette smoking, no. (%) History of COPD, no. (%) Alcohol use, no. (%) Hypertension, no. (%) History of CHF, no. (%) Current opioid use, no. (%) Steroid use, no. (%) Previous abdominal surgery, no. (%) Blood thinner use, no. (%) Aspirin Warfarin Plavix Others ASA class, no. (%) ASA class 1 ASA class 2 ASA class 3 ASA class 4 ASA > 2 Surgical indication, no. (%) Malignancy IBD Benign disease 67 (25.6) 156 (59.5) 85 (32.4) 21 (8.0) 126 (48.1) 132 (50.4) 22 (8.4) 43 (16.5) 37 (14.1) 170 (64.9) 52 (19.8) 7 (2.7) 11 (4.2) 25 (9.5) 1 (0.4) 85 (32.4) 169 (64.5) 7 (2.7) 176 (67.2) 129 (49.2) 52 (19.8) 128 (48.9) Preoperative TAP (n = 240) Postoperative TAP (n = 22) P-value 57.83 (15.87) 107 (44.6) 28.51 (7.48) 56.95 (18.90) 11 (50.0) 27.03 (5.27) 64 (26.7) 144 (60.0) 80 (33.3) 19 (7.9) 113 (47.1) 121 (50.4) 21 (8.8) 41 (17.2) 31 (12.9) 157 (65.4) 47 (19.6) 6 (2.5) 11 (4.6) 22 (9.2) 1 (0.4) 78 (32.5) 156 (65.0) 5 (2.1) 161 (67.1) 119 (49.6) 46 (19.2) 116 (48.3) 3 (13.6) 12 (54.5) 5 (22.7) 2 (9.1) 13 (59.1) 11 (50.0) 1 (4.5) 2 (9.1) 6 (27.3) 13 (59.1) 5 (22.7) 1 (4.5) 0 (0.0) 3 (13.6) 0 (0.0) 7 (31.8) 13 (59.1) 2 (9.1) 15 (68.2) 10 (45.5) 6 (27.3) 12 (54.5) 0.807 0.791 0.366 0.278 0.786 0.436 0.692 0.392 0.99 0.705 0.499 0.126 0.718 0.941 0.463 0.607 0.451 0.241 0.99 0.882 0.401 0.738 ASA American Society of Anesthesiology, COPD chronic obstructive pulmonary disease, CHF congestive heart failure, IBD inflammatory bowel disease, SD standard deviation Kalu et al. BMC Anesthesiology (2021) 21:114 Page 4 of 6 Table 2 Postoperative analgesics use based on when the plain TAP block was administered Variables PCA use, no. (%) PCA type Morphine Hydromorphone Morphine + hydromorphone PCA amount, MME, mean (SD) Morphine PCA duration, hours, mean (SD) IV opioid use, no. (%) IV opioid amount, mg, mean (SD) Overall (N = 262) 213 (81.3) 172 (65.6) 31 (11.8) 10 (3.8) 27.40 (57.37) 33.97 (64.30) 54 (20.6) 1.80 (10.39) Preoperative TAP (n = 240) Postoperative TAP (n = 22) 200 (83.3) 13 (59.1) 165 (68.8) 26 (10.8) 9 (3.8) 27.09 (51.87) 34.66 (66.22) 46 (19.2) 1.38 (9.07) 7 (31.8) 5 (22.7) 1 (4.5) 30.77 (101.32) 26.36 (37.48) 8 (36.4) 6.42 (19.50) Oral opioid amount, MME, mean (SD) 93.94 (153.61) 87.71 (117.82) 161.93 (360.53) Discharge opioid, MME, mean (SD) 123.52 (123.01) 128.09 (126.78) 73.64 (48.03) Prescription muscle relaxants on discharge, no. (%) Opioid use on first postoperative follow-up, no. (%) Opioid refill request at first postoperative follow up, no. (%) 108 (41.4) 19 (7.3) 3 (1.1) 97 (40.6) 18 (7.5) 2 (0.8) 11 (50.0) 1 (4.5) 1 (4.5) IV intravenous, PCA patient-controlled analgesia, SD standard deviation, TAP transversus abdominis plane, MME morphine milligram equivalent P-value 0.012 0.003 0.019 0.263 0.102 0.051 0.743 0.005 0.528 0.99 0.605 TAP blocks postoperatively experienced a statistically significant reduction in the overall use of PCA compared with those who received preoperative TAP blocks (59.1 % vs. 83.3 %; p = 0.012). However, when given a PCA, the postoperative TAP group used a significantly higher amount of morphine compared to their counter- parts (30.77 MME vs. 27.09 MME; p = 0.019). The post- operative TAP group was less likely to be prescribed opioid medication at the time of discharge (6.4 % vs. 16.9 %; p = 0.004). For patients who received prescription opioid at the time of discharge, the patients who had postoperative TAP received a smaller amount of opioid (128.09 MME vs. 73.64 MME; p = 0.047). There were no differences between the groups with regard to duration of PCA or intravenous and oral opioid use. significantly Procedure-related details, length of stay, reoperation, and readmission Table 3 presents the procedure-related details, length of stay, and reoperation and readmission rates for the 2 groups. Surgical approach did not differ based on timing of the regional anesthesia. There was no statistically sig- nificant difference in procedure length, estimated blood loss, length of hospital stays, reoperation, or readmission rates between the 2 groups. Discussion Postoperative pain management is an integral part of achieving the goals of ERPs after colorectal surgery. TAP blocks are an attractive approach for minimizing the use of opioids, especially given their low risk of adverse ef- fects. Although usually done preoperatively, our study showed that postoperative TAP block with plain bupiva- caine appeared to be at least as efficacious as preopera- tive TAP block in reducing postoperative intravenous opioid use, both PCA and administered intravenous in- jections. Furthermore, postoperative plain TAP block was associated with a reduced total amount of prescrip- tion opioids needed to the time of discharge from the hospital. Other variables such as length of stay, esti- mated operative blood loss, procedure length, reopera- tion and readmission rates did not differ between preoperative or postoperative administration of the TAP. The effectiveness and feasibility of TAP blocks in colo- rectal surgery has been shown in multiple studies [7, 13–17]. In a randomized, placebo-controlled clinical trial, Tikuisis et al. showed that patients who received ropivacaine TAP blocks had significantly lower pain scores at 2, 4, and 12 h at rest, and at 2- and 4-hours during movement. The TAP group also used signifi- cantly less fentanyl and ketorolac following a hand- assisted laparoscopic left hemicolectomy for colon can- cer compared to those who received placebo [13]. Pir- rera et al. compared the use of preoperative ropivacaine TAP block vs. thoracic epidural analgesia in patients be- fore elective laparoscopic colon resection. Both patient groups were a part of a standard enhanced recovery after surgery pathway. Albeit a case-control study, pain con- trol was comparable between the 2 groups. Additionally, the TAP group had significantly lower rates of postoper- ative nausea, vomiting, ileus, and paresthesia. There was Kalu et al. BMC Anesthesiology (2021) 21:114 Page 5 of 6 Table 3 Procedure-related details, length of stay, reoperation, and readmission Variables Surgical procedure, no. (%) Surg_type1a Surg_type2b Surg_type3c Surg_type4d Surgical approach, no. (%) Open Laparoscopic Robotic Procedure details Overall (N = 262) 195 (74.4) 38 (14.5) 11 (4.2) 18 (6.9) 69 (26.3) 79 (30.2) 114 (43.5) Emergent procedure, no. (%) 2 (0.8) Procedure length, minutes (SD) EBL in mL, mean (SD) LOS in days, mean (SD) Readmission rate, no. (%) Reoperation rate, no. (%) 208.38 (96.27) 100.05 (185.54) 4.92 (5.95) 29 (11.2) 5 (1.9) Preoperative TAP (n = 240) Postoperative TAP (n = 22) 178 (74.2) 36 (15.0) 9 (3.8) 17 (7.1) 62 (25.8) 75 (31.2) 103 (42.9) 1 (0.4) 207.79 (97.16) 100.43 (190.96) 4.82 (5.62) 28 (11.8) 4 (1.7) 17 (77.3) 2 (9.1) 2 (9.1) 1 (4.5) 7 (31.8) 4 (18.2) 11 (50.0) 1 (4.5) 214.82 (87.76) 95.91 (113.14) 6.00 (8.87) 1 (4.5) 1 (4.5) P-value 0.554 0.439 0.161 0.655 0.855 0.836 0.485 0.901 EBL estimated blood loss, LOS length of day, SD standard deviation, TAP transversus abdominis plane asurg_type1 includes hemicolectomy, sigmoidectomy, low anterior resection, total abdominal colectomy and abdominoperineal resection bsurg_type2 includes ostomy reversal csurg_type3 includes ostomy creation dsurg_type4 includes appendectomy, exploratory laparotomy, lysis of adhesion, and rectopexy no significant difference in hospital length of stay or 30- day readmission rate. In a prospective, randomized, double-blind study, Keller et al. assessed the effect of TAP blocks on postoperative pain in patients following laparoscopic colorectal resections. Compared to their counterparts, the TAP group had significantly lower pain scores and used fewer opioids. However, there was no difference in hospital length of stay and readmission rate between the groups [17]. These research study findings are consistent with our current results. Considering the short duration of action of plain bupi- vacaine, the timing of its administration for TAP blocks can be planned to provide optimal analgesia. Preoperative administration is easily performed in the preoperative area and does not affect surgical planning such as ostomy pro- cedures. It is also in line with preemptive analgesia. Oli- veira et al., through a meta-analysis that included a variety of abdominopelvic surgeries, reported a greater postopera- tive pain control, reduced pain at rest, and decreased opi- oid used with preoperative TAP compared to placebo or no treatment [18]. The majority of the TAP blocks per- formed in our study were performed preoperatively, but we found more benefit with postoperative TAP blocks. While the numbers are small in the postoperative TAP group, the postoperative TAP group did not show inferior pain control compared to preoperative administrated block. In fact, our results showed that plain bupivacaine TAP administered postoperatively led to a reduced postoperative intravenous opioid use and lesser amount of discharge prescription opioid. The study was limited by its retrospective design, the small number of patients who received postoperative plain bupivacaine TAP block procedures, and by having been performed in a single center. Although the TAP blocks were performed by, or under supervision of, the attending anesthesiologists who specialize in acute pain management and regional anesthesia, we acknowledge the possibility that performer-related differences could lead to differences in pain relief. However, the regional block was administered by well-trained and experienced anesthesiologists who routinely perform the procedure, hence minimizing the effect of performer-related differ- ences. Finally, our study did not directly measure the pa- Instead, we used the amount of tient pain scores. postoperative opioid used as a surrogate for the ad- equacy of pain control. Per our ERP, additional analgesia is prescribed on an as-needed basis and coincides with the patient’s assessment of pain on a numeric scale. Conclusions TAP blocks provide an effective and feasible means of optimizing pain control in the era of enhanced recovery after surgery. When administered postoperatively, plain bupivacaine TAP may be as effective as preoperative TAP blocks, offering the same analgesic effect and min- imizing intravenous opioid use. Kalu et al. BMC Anesthesiology (2021) 21:114 Page 6 of 6 Abbreviations ERP: Enhanced recovery protocols; PCA: Patient-controlled analgesia; TAP: Transversus abdominis plane Acknowledgements The authors thank Stephanie Stebens, MLIS, at Sladen Library, Henry Ford Hospital, for her additional input in reviewing this manuscript and Karla D Passalacqua, PhD, at Henry Ford Hospital for editorial assistance. Authors’ contributions AT analyzed the patient data obtained and assisted in interpreting the statistical results. PB and LC ensured compliance of the transversus abdominis plane (TAP) block procedure and kept record of patient participants. RK contributed to the literature review; performed chart reviews of the participants; ensured completeness of data; assisted in interpreting the statistical results; and was a major contributor in writing the manuscript. JG ensured compliance and consistency of the TAP procedure. R reviewed the manuscript for completeness before submission. AS contributed to the literature review, interpreted that patient data and was a major contributor in writing the manuscript. All authors read and approved the final manuscript. Funding None. Availability of data and materials The datasets used and/or analyzed during the current study are not publicly available due to patient privacy and institutional policy but are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This retrospective chart review research study was categorized as exempt and approved by the Henry Ford Hospital System Institutional Review board (Research Administration, Henry Ford Health System, Detroit, Chairperson Dr. Jonathan Ehrman, IRB #14286).In addition, the requirement for written informed consent was waived by the Henry Ford Health System Institutional Review board. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1Department of Surgery, Henry Ford Hospital, 2799 W. Grand Blvd, MI 48202 Detroit, USA. 2Department of Anesthesiology, Henry Ford Hospital, 2799 W. Grand Blvd, 48202 Detroit, MI, USA. 3Department of Public Health Sciences, Henry Ford Health System, One Ford Place, 48202 Detroit, MI, USA. 4Division of Colon and Rectal Surgery, Henry Ford Hospital, 2799 West Grand Blvd Detroit, 48202 Detroit, MI, USA. Received: 18 December 2020 Accepted: 23 March 2021 8. 7. 6. Muller S, Zalunardo MP, Hubner M, Clavien PA, Demartines N. A fast-track program reduces complications and length of hospital stay after open colonic surgery. Gastroenterology. 2009;136(3):842–7. Stokes AL, Adhikary SD, Quintili A, Puleo FJ, Choi CS, Hollenbeak CS, et al. Liposomal Bupivacaine Use in Transversus Abdominis Plane Blocks Reduces Pain and Postoperative Intravenous Opioid Requirement After Colorectal Surgery. Dis Colon Rectum. 2017;60(2):170–7. Park JS, Choi GS, Kwak KH, Jung H, Jeon Y, Park S, et al. Effect of local wound infiltration and transversus abdominis plane block on morphine use after laparoscopic colectomy: a nonrandomized, single-blind prospective study. J Surg Res. 2015;195(1):61–6. Gustafsson UO, Scott MJ, Hubner M, Nygren J, Demartines N, Francis N, et al. Guidelines for Perioperative Care in Elective Colorectal Surgery: Enhanced Recovery After Surgery (ERAS(®)) Society Recommendations: 2018. World J Surg. 2019;43(3):659–95. 9. 11. 10. Charlton S, Cyna AM, Middleton P, Griffiths JD. Perioperative transversus abdominis plane (TAP) blocks for analgesia after abdominal surgery. Cochrane Database Syst Rev. 2010;8(12):Cd007705. Felling DR, Jackson MW, Ferraro J, Battaglia MA, Albright JJ, Wu J, et al. Liposomal Bupivacaine Transversus Abdominis Plane Block Versus Epidural Analgesia in a Colon and Rectal Surgery Enhanced Recovery Pathway: A Randomized Clinical Trial. Dis Colon Rectum. 2018;61(10):1196–204. 12. Beiranvand S, Moradkhani MR. Bupivacaine Versus Liposomal Bupivacaine 13. For Pain Control. Drug Res (Stuttg). 2018;68(7):365–9. Tikuisis R, Miliauskas P, Lukoseviciene V, Samalavicius N, Dulskas A, Zabuliene L, et al. Transversus abdominis plane block for postoperative pain relief after hand-assisted laparoscopic colon surgery: a randomized, placebo- controlled clinical trial. Tech Coloproctol. 2016;20(12):835–44. 14. Pirrera B, Alagna V, Lucchi A, Berti P, Gabbianelli C, Martorelli G, et al. Transversus abdominis plane (TAP) block versus thoracic epidural analgesia (TEA) in laparoscopic colon surgery in the ERAS program. Surg Endoscopy. 2018;32(1):376–82. 16. 15. Conaghan P, Maxwell-Armstrong C, Bedforth N, Gornall C, Baxendale B, Hong LL, et al. Efficacy of transversus abdominis plane blocks in laparoscopic colorectal resections. Surg Endoscopy. 2010;24(10):2480–4. Smith SR, Draganic B, Pockney P, Holz P, Holmes R, McManus B, et al. Transversus abdominis plane blockade in laparoscopic colorectal surgery: a double-blind randomized clinical trial. Int J Colorectal Dis. 2015;30(9):1237–45. Keller DS, Ermlich BO, Schiltz N, Champagne BJ, Reynolds HL Jr., Stein SL, et al. The effect of transversus abdominis plane blocks on postoperative pain in laparoscopic colorectal surgery: a prospective, randomized, double- blind trial. Dis Colon Rectum. 2014;57(11):1290–7. 17. 18. De Oliveira GS, Jr., Castro-Alves LJ, Nader A, Kendall MC, McCarthy RJ. Transversus abdominis plane block to ameliorate postoperative pain outcomes after laparoscopic surgery: a meta-analysis of randomized controlled trials. Anesth Analg. 2014;118(2):454–63. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References 1. 2. 3. 4. 5. Jakobsen DH, Sonne E, Andreasen J, Kehlet H. Convalescence after colonic surgery with fast-track vs conventional care. Colorectal Dis. 2006;8(8):683–7. Basse L, Hjort Jakobsen D, Billesbølle P, Werner M, Kehlet H. A clinical pathway to accelerate recovery after colonic resection. Ann Surg. 2000; 232(1):51–7. Kehlet H, Mogensen T. Hospital stay of 2 days after open sigmoidectomy with a multimodal rehabilitation programme. Br J Surg. 1999;86(2):227–30. Kehlet H, Dahl JB. Anaesthesia, surgery, and challenges in postoperative recovery. Lancet (London, England). 2003;362(9399):1921–8. Khoo CK, Vickery CJ, Forsyth N, Vinall NS, Eyre-Brook IA. A prospective randomized controlled trial of multimodal perioperative management protocol in patients undergoing elective colorectal resection for cancer. Ann Surg. 2007;245(6):867–72.
10.1186_s12872-022-02488-x	Son et al. BMC Cardiovascular Disorders (2022) 22:44 https://doi.org/10.1186/s12872-022-02488-x ORIGINAL RESEARCH Open Access Risk of aortic aneurysm and aortic dissection with the use of fluoroquinolones in Korea: a nested case–control study Nayeong Son†, Eunmi Choi†, Soo Youn Chung, Soon Young Han and Bonggi Kim* Abstract Background: Recent studies have raised concern about the association of fluoroquinolones with an increased risk of aortic aneurysm and aortic dissection. We aimed to evaluate such risk in a Korean population. Methods: We conducted a nested case–control study using data from the National Health Insurance Service col- lected from 2013 to 2017 in Korea. The study cohort included patients older than 40 years and excluded patients who had used fluoroquinolones or been diagnosed with aortic aneurysm, aortic dissection, or related diseases 1 year prior to the cohort entry date. We randomly matched four controls in the risk set with each case of aortic aneurysm and aortic dissection (same sex, age, and cohort entry date). We assessed the risk of aortic aneurysm and aortic dissection from fluoroquinolones and adjusted for potential confounders using a conditional logistic regression model. Results: A total of 29,638 aortic aneurysm and aortic dissection patients were identified between 2014 and 2017. The use of fluoroquinolones within a year was associated with a 10% increased risk of aortic aneurysm and aortic dissec- tion (adjusted odds ratio: 1.10, 95% CI 1.07–1.14, p < 0.05) compared with nonusers. The risk was higher in patients who had used fluoroquinolones within 60 days (adjusted odds ratio: 1.53, 95% CI 1.46–1.62, p < 0.05). The risk of aortic aneurysm and aortic dissection positively correlated with the cumulative dose and duration of fluoroquinolone therapy (p < 0.001). Conclusions: Our study provides real-world evidence of the risk of aortic aneurysm and aortic dissection from fluo- roquinolones in Korea. Patients and medical professionals should be aware that fluoroquinolones can increase the risk of aortic aneurysm and aortic dissection, which may be acerbated by high dosage and duration of use. Keywords: Fluoroquinolone, Aortic aneurysm, Aortic dissection, Drug safety, Pharmacovigilance, Adverse effect Background Fluoroquinolones (FQs) are among the most widely used antibiotics in Korea, and their use has consistently increased to account for 9 to 11% of all antibiotic use [1]. Although FQs are powerful antibiotics with a wide antibacterial spectrum [2], they induce degradation of collagen and other structural components of the extra- cellular matrix by stimulating matrix metalloproteinases [3]. The possibility of excessive tissue breakdown by this mechanism has raised concern about the risk of adverse Correspondence: bgkim@drugsafe.or.kr †Nayeong Son and Eunmi Choi are co-first authors and contributed equally. All the authors take responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation Korea Institute of Drug Safety and Risk Management, 6th FL, 30, Burim-ro 169beon-gil, Dongan-gu, Anyang-si, Gyeonggi-do, Republic of Korea © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Son et al. BMC Cardiovascular Disorders (2022) 22:44 Page 2 of 9 reactions, such as aortic aneurysm (AA) and aortic dis- section (AD). In December 2018, the U.S. Food and Drug Adminis- tration warned that FQs can increase the occurrence of rare but serious ruptures or tears in the aorta. The warn- ing included special caution for patients with a history of aneurysms, blockages, or hardening of the arteries, high blood pressure, or genetic conditions such as Mar- fan or Ehlers–Danlos syndrome and instructed patients to inform their health-care professional before starting a fluoroquinolone prescription [4]. Following that warning, the Ministry of Food and Drug Safety in Korea issued a safety letter warning about the potential association of fluoroquinolone use and the risk of AA/AD [5]. Many observational studies have suggested that fluo- roquinolone use could be significantly associated with an increased risk of AA/AD [3, 6–9]. Recently, a systemic review and meta-analysis showed that fluoroquinolone use incurs a risk of developing three collagen-associated diseases, including AA/AD [10]. However, it has not yet been established whether fluoroquinolone use increases the risk of AA/AD in the Korean population. This study aims to evaluate the association between fluoroquinolone use and the risk of AA/AD in the Korean population. Methods Data source Insurance Service We conducted a nested case-control study using National (NHIS)-customized data Health (NHIS-2019-1-024). The NHIS database covers almost 98% of the total population in Korea. It contains patient demographic information such as sex, date of birth, date of death, and medical treatment records, including details of disease and prescriptions [11]. The authors declare no conflicts of interest with NHIS. Study population The study population comprised all patients aged 40 to 99 years 2014–2017 in the NHIS database. The date of 1 January 2014 was defined as the cohort entry date for patients aged 40 years or older in 2014. For patients aged less than 40 years in 2014, we established the cohort entry date as the first day of the year that the patient became 40 years old. We excluded 510,805 patients who: • Had taken FQs more than once during the year prior to the cohort entry date • Were diagnosed with AA/AD during the year prior to the cohort entry date • Were diagnosed with underlying related diseases (atherosclerosis of the aorta, arteritis, aortitis, Lerche’s syndrome, coarctation of the aorta, Marfan’s syndrome, valve diseases, endocarditis, congenital malformations of valves, heart failure) (Additional file 1: Table S1) during the year prior to the cohort entry date. Case selection From the cohort, we identified 29,638 patients aged 40 years or older who had experienced AA/AD from 2014 to 2017 according to the definition of health outcomes of interest “Statistical analysis” section . Patients in the case group were observed from the cohort entry date to the index date, which was defined as the first date of diagno- sis of AA/AD. Control selection After we stratified the case group based on age and sex, we created a risk set for each case using patients who were of the same sex and age as those with AA/AD and did not have a history of an AA/AD diagnosis. The size of the risk set was 20 times the sample size of each stra- tum. We randomly matched four controls in the risk set. Patients in the control group were observed from the cohort entry date to the index date of matched cases. Health outcomes of interest The outcome of the main analysis was defined as a diag- nosis of AA/AD after entry to the cohort. Incident cases were defined as those who had received an ICD 10 code I71 (ICD 10 I71.0–I71.9) for all kinds of AA/AD. The outcome for the sensitivity analysis was redefined as a diagnosis of AA/AD in addition to having received a laboratory test specific for AA/AD (abdominal/thoracic aortography, computed tomography (CT), magnetic res- onance imaging, ultrasonography, Doppler echocardiog- raphy, transesophageal/transthoracic echocardiography, abdominal vascular ultrasonography, or aorta Doppler ultrasonography) within 28 days prior to the diagnosis. The diagnosis and treatment of AA/AD were based on the ACCF/AHA/AATS/ACR/ASA/SCA/SCAI/SIR/STS/ SVM guideline in the general Korean hospitals [12]. The first date of diagnosis was defined as the index date for cases and matched controls. Exposure The exposure of interest was the use of a fluoroquinolone (balofloxacin, ciprofloxacin, enoxacin, gatifloxacin, gemi- floxacin, levofloxacin, lomefloxacin, moxifloxacin, nor- floxacin, ofloxacin, tosufloxacin, and zabofloxacin) in the year prior to the index date. We categorized fluoroquinolone users as current, recent, or past users according to the time from the end of supply of the fluoroquinolone prescription to the index date. In this definition, ‘termination of fluoroquinolone Son et al. BMC Cardiovascular Disorders (2022) 22:44 Page 3 of 9 exposure’ means the end of supply of the fluoroquinolone prescription. Current users were defined as patients who had terminated fluoroquinolone exposure within the 60 days prior to the index date. Recent users were defined as patients who had terminated fluoroquinolone expo- sure 61–120 days prior to the index date. Past users were defined as patients who had terminated fluoroquinolone exposure 121–365 days prior to the index date. To investigate the effects of cumulative dose and dura- tion of FQ exposure on the prevalence of AA/AD, we categorized fluoroquinolone users into three groups according to the quantiles of duration and into four groups according to the quantiles of cumulative dose. The duration was calculated as the sum of the total days of supply for each prescription in the year prior to the index date. The first and third quantiles of the cumulative days of supply were found to be 2 and 14 days, respectively. We represented the cumulative dose in the year prior to the index date in terms of the defined daily dose (DDD), as defined by the anatomical therapeutic chemical clas- sification system. The first, second, and third quantiles of cumulative dose were found to be 4 DDD, 7.5 DDD, and 15 DDD, respectively. The NHIS dataset included the Korean ingredient code of the drug, the date the prescription was written, the number of days of supply, and the quantity. We used these data to identify prescriptions for FQs and any con- comitantly used drugs. Statistical analysis Pearson Chi-square tests and Fisher’s exact test were used for the analysis of categorical variables. The odds ratios of the association between FQ use and AA/AD were calculated using multivariate conditional logistic regression analysis. We considered covariates known to be related to AA/AD or fluoroquinolone use from pre- vious studies and included them as confounders in the model [3, 6–9]. The covariates are listed in Table 1. We also tested the tendency of AA/AD to occur with changes in timing, cumulative dose, and duration of FQ use using the Cochran-Armitage trend test. All data processing and statistical analyses were performed using SAS 9.4 and R 5.3.1 using two-sided tests, and a p value of < 0.05 was considered significant. Results Demographic and clinical characteristics The final study population was composed of 148,190 patients, including 29,638 cases and 118,552 controls. Table 1 shows the baseline characteristics of the study population. This cohort comprised 92,645 male patients (62.5%) and 55,545 female patients (37.5%). More than half of the study population was 60–69 years old (23.7%) or 70–79 years old (32.0%). Patients in the AA/AD case group had a higher prevalence of cerebrovascular dis- ease and cardiovascular disorders such as arterial disease and ischemic heart disease. In the year prior to the index date. Patients in the AA/AD case group were more often users of angiotensin-converting enzyme inhibitors, anti- arrhythmics, anticonvulsants, etc. from the cohort entry date to the index date and experienced more cardiac or aortic procedures and surgeries in the previous year. Association between AA/AD and FQ use During the observation period (1 year before the index date), 8562 cases (28.9%) and 25,387 controls (21.4%) received at least one prescription for FQs. Table 2 and Figure 1 show the results of the conditional logistic regression analysis. The adjusted odds ratio was 1.10 (95% CI 1.07–1.14, p < 0.05) during the 1-year observa- tion period. However, the risk was substantially higher in current users (adjusted OR 1.53, 95% CI 1.46–1.62, p < 0.05). FQ use did not have a significant association with AA/AD in recent users (adjusted OR 1.00, 95% CI 0.93–1.07, p < 0.05). The risk was even lower in past users (adjusted OR 0.92, 95% CI 0.87–0.96, p < 0.05). The risk of AA/AD was studied according to the dura- tion of exposure and the cumulative dose of FQs. In this study, 25% of FQ users were exposed to FQs for 2 days or less. On the other hand, 25% of the FQ users were exposed to FQs for more than 14 days. Among them, 50% of the FQ users were exposed to FQs for between 3 days and 13 days. We used the same covariates as those adopted for the primary analysis. Patients who used FQs for less than three days had a lower risk of AA/AD than nonusers (adjusted OR 0.87, 95% CI 0.82–0.92, p < 0.05). However, the risk was significantly higher in patients who had used FQs for between three days and 13 days (adjusted OR 1.14, 95% CI 1.09–1.19, p < 0.05) and was highest in patients who used FQs for more than 14 days (adjusted OR 1.33, 95% CI 1.26–1.40, p < 0.05). FQ users were categorized into four groups with regard to dose (low, mid-low, mid-high, or high) according to the quantiles of cumulative dose. Patients in the low- dose group had used FQs less than 4 DDDs during the observation period. Patients in the mid-low dose group and mid-high dose group had used 4 DDDs to 7.5 DDDs and 7.5 DDDs to 15 DDDs, respectively. Patients in the high-dose group had used more than 15 DDDs during the observation period. Compared with nonusers, the risk of AA/AD in the low-dose group (<4 DDDs) was not significantly higher (adjusted OR 0.97, 95% CI 0.89–1.04, p > 0.05). However, the risk was significantly higher in patients who had used FQs at more than 4 DDDs. Spe- cifically, the adjusted odds ratio of AA/AD was 1.25 (95% CI 1.16–1.34, p < 0.05) in the mid-low dose group, 1.29 Son et al. BMC Cardiovascular Disorders (2022) 22:44 Page 4 of 9 Table 1 Demographics and clinical characteristics of the study population Case (29,638) Control (118,552) p value Sex Male Female Age (year) 40–49 50–59 60–69 70–79 80–89 90 + Underlying disease Cerebrovascular disease Arterial disease Ischemic heart disease Cardiac valve disease Conduction disorder Heart failure or cardiomyopathy Chronic obstructive pulmonary disease Pneumonia Cancer Liver disease Renal disease Rheumatism Psychiatric disorder Diabetes Hypertension Lipid disorder Trauma Obstructive sleep apnea Asthma Obesity Seizure disorder Decubitus ulcer Infectious disease Hypothyrodism Inflammatory bowel disease Urinary tract infection Ehlers–Danlos syndrome Charlson comorbidity Index Mean(SD) 0 1 2 3 + Myocardial infarction Congestive heart failure Peripheral vascular disease Cerebrovascular disease N 18,529 11,109 2080 4422 7031 9479 5748 878 3124 8227 7845 884 115 2723 10,626 3579 3422 10,520 1773 2005 13,498 9518 19,716 16,673 15,191 73 6747 41 1225 9 13,949 1729 3876 2287 1 (%) 62.5 37.5 7 14.9 23.7 32 19.4 3 10.5 27.8 26.5 3 0.4 9.2 35.9 12.1 11.5 35.5 6 6.8 45.5 32.1 66.5 56.3 51.3 0.2 22.8 0.1 4.1 0 47.1 5.8 13.1 7.7 0 2.67(2.41) 1.86(2.07) 5371 5781 5425 13,061 1377 3731 7376 6913 18.1 19.5 18.3 44.1 4.6 12.6 24.9 23.3 N 74,116 44,436 8320 17,688 28,124 37,916 22,992 3512 6418 22,080 14,715 444 272 4108 31,726 9047 8462 32,012 2902 5720 41,650 33,700 60,520 50,268 53,280 190 19,825 116 3015 26 46,881 5087 11,773 5442 0 37,377 27,124 19,785 34,266 1924 6315 20,782 16,487 (%) 62.5 37.5 7 14.9 23.7 32 19.4 3 5.4 18.6 12.4 0.4 0.2 3.5 26.8 7.6 7.1 27 2.4 4.8 35.1 28.4 51 42.4 44.9 0.2 16.7 0.1 2.5 0 39.5 4.3 9.9 4.6 0 31.5 22.9 16.7 28.9 1.6 5.3 17.5 13.9 1 1 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 0.002 < 0.001 0.069 < 0.001 0.526 < 0.001 < 0.001 < 0.001 < 0.001 0.2 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 Son et al. BMC Cardiovascular Disorders (2022) 22:44 Page 5 of 9 Table 1 (continued) Dementia Chronic pulmonary diseases Connective tissue disease Peptic ulcer Mild liver diseases Uncomplicated diabetes Diabetes complicated with retinopathy, neuropathy, renal disease Hemiplegia Moderate or severe renal diseases Nonmetastatic solid cancer, leukemia, lymphoma, multiple myeloma Moderate or severe liver diseases Metastatic solid cancer AIDS/HIV Medication use** Angiotensin-converting enzyme inhibitors Antiarrhythmic Anticonvulsant Antidepressant Immunodepressant Anticoagulant β-blocker Oral hypoglycemic agent Benzodiazepine Calcium Channel Blockers corticosteroid Disease-modifying antirheumatic drugs Insulin Loop diuretics Nonsteroidal anti-inflammatory drugs Antipsychotic Peripheral vasodilators Lipid-lowering agent Parkinson medication Hydroxyzine Cardiac or aortic procedure/surgery The p values are results from Chi-square or Fisher’s exact tests Case (29,638) Control (118,552) p value N 3995 13,066 1827 10,817 9697 106 2868 936 1773 3288 180 337 7 1119 13,410 3375 5542 8901 20,898 7664 3935 10,934 13,194 16,632 618 1136 3483 23,354 7588 3425 11,079 1990 2616 287 (%) 13.5 44.1 6.2 36.5 32.7 0.4 9.7 3.2 6 11.1 0.6 1.1 0 3.8 45.2 11.4 18.7 30 70.5 25.9 13.3 36.9 44.5 56.1 2.1 3.8 11.8 78.8 25.6 11.6 37.4 6.7 8.8 1 N 11,955 40,316 5172 33,712 29,509 427 10,911 2016 2902 8082 545 765 25 2174 34,164 8889 14,066 27,220 68,727 15,373 18,862 30,683 38,216 56,174 1559 2876 5350 82,504 23,261 4115 31,002 6220 8049 251 (%) 10.1 34 4.4 28.4 24.9 0.4 9.2 1.7 2.4 6.8 0.5 0.6 0 1.8 28.8 7.5 11.9 23 58 13 15.9 25.9 32.2 47.4 1.3 2.4 4.5 69.6 19.6 3.5 26.2 5.2 6.8 0.2 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 0.991 0.012 < 0.001 < 0.001 < 0.001 0.001 < 0.001 0.965 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 *Information on underlying disease were derived from data recorded prior to the index date and after cohort entry. Information on medication use were derived from data recorded in 1 year prior to the index date Table 2 Results of conditional logistic regression analysis of the association between AA/AD and FQ use Case N 21,076 8562 Main analysis Nonusers Users Control Crude OR Adjusted OR % N % OR 95% CI OR 95% CI 71.1 28.9 93,165 25,387 78.6 21.4 1 1.51 – 1.47–1.56 1 1.10 – 1.07–1.14 Adjusted for covariates presented in Table 1 (sex, age, underlying disease, Charlson comorbidity index, medication use, history of procedure/surgery) p < 0.05 Son et al. BMC Cardiovascular Disorders (2022) 22:44 Page 6 of 9 Fig. 1 Results of conditional logistic regression analysis of the association between AA/AD and FQ use (95% CI 1.20–1.38, p < 0.05) in the mid-high dose group, and 1.36 (95% CI 1.26–1.45) in the high dose group. Subgroup analysis and sensitivity analysis From the subgroup analysis by sex (Fig. 1), we found that the association between AA/AD and FQ use remained statistically significant in both the male and female sub- groups. In particular, the risk was high in female patients (adjusted OR 1.15, 95% CI 1.09–1.21, p < 0.05) compared with female nonusers. When ages were grouped into 10-year bands, the association between AA/AD and FQ use remained statistically significant in every age group. To verify the consistency of the results, we performed sensitivity analysis (Table 3) by changing the definition of AA/AD occurrence. The AA/AD occurrence in the primary analysis was identified using the ICD 10 code for AA/AD. For the sensitivity analysis, we changed the definition of an AA/AD case to a diagnosis of AA/AD in addition to having received a laboratory test specific for AA/AD within the 28 days prior to the initial diag- nosis of AA/AD. Among 29,648 AA/AD cases, 21,528 (72.6%) received the laboratory test specific for AA/ AD within the 28 days prior to the initial diagnosis of AA/AD. Among those 21,528 patients, 17,875 (83.0%) were diagnosed with AA/AD the day they took the tests. Abdominal/thoracic CT, aortography, and tran- sthoracic echocardiography were found to have been the commonly performed procedures. The results remained consistent with the primary results under the new definition. The risk of AA/AD by FQs was substantially higher in current users. The risk increased as the duration of exposure and cumulative dose increased. The associa- tion remained statistically significant in every subgroup Son et al. BMC Cardiovascular Disorders (2022) 22:44 Page 7 of 9 Table 3 Results of sensitivity analysis of the association between AA/AD and FQ use Case N 15,294 6234 Sensitivity analysis Nonusers Users Controls Crude OR Adjusted OR % N % OR 95% CI OR 95% CI 71.0 29.0 67,570 18,542 78.5 21.5 1 1.51 – 1.46–1.56 1 1.10 – 1.06–1.14 Cases that received laboratory tests specific for AA/AD within 28 days prior to the initial diagnosis of AA/AD and their matched controls were included Adjusted for covariates presented in Table 1 (sex, age, underlying disease, Charlson comorbidity index, medication use, history of procedure/surgery) *p < 0.05 by sex and age. See Additional file 1: Table S2 for numeric results. Discussion In this study, FQ use showed a trend to be associated with an increased risk of AA/AD during the 1-year observation period, but the effect size was not remark- able. However, the risk of AA/AD in current users of FQs was relatively considerable. This result is in line with preceding research in many ways. In an in vitro study that assessed the effect of FQs on MMP activi- ties in human aortic smooth muscle cells, 48 hours of treatment with ciprofloxacin significantly increased total MMP activity. Observational studies using Tai- wanese and Swedish databases also showed that the risk of AA/AD within 60 days after FQ use was signifi- cantly higher than that of nonusers [3, 7, 8]. In addition, a cohort study in Ontario, Canada and a signal analysis using U.S. FAERS data also indicated significant asso- ciations between FQ use and AA/AD [6, 9]. This trend is consistent with the results of a systematic literature review and meta-analysis conducted in 2019 [10]. In particular, Pasternak et al. [3] showed that the cumula- tive incidence of AA/AD increased significantly during the first 10 days after FQ use. Given these findings, fur- ther studies are needed to evaluate the risk in the early period of FQ use. Studies that utilized the Taiwanese database [7, 8] reported that the risk of AA/AD increased as the dura- tion of drug use increased. In this study, the adjusted odds ratio of AA/AD also increased as the cumula- tive duration of FQ use increased. In addition, while no prior study has determined the effect of the cumu- lative dose of FQs on the risk for AA/AD, this study showed that an increased cumulative dose of FQs could increase the risk for AA/AD. The dose–response rela- tionship and duration-response relationship can be interpreted as considerable evidence of the causal rela- tionship between FQ use and the occurrence of AA/ AD. Therefore, the patient’s condition should be care- fully monitored, keeping in mind that the risk of AA/ AD may increase as the cumulative dose or duration of FQ use increases. Our study suggests some different results from the general understanding of AA/AD. In general, AA/AD progresses slowly over several years, and men and old age are known as risk factors. However, we found that the risk of AA/AD from FQ use was significant (1) in the early period of FQ use, (2) in female patients and male patients, and (3) in all age groups. In this research, the risk of AA/AD was 8% higher in male FQ users and 15% higher in female FQ users than in nonusers of each sex. Although the risk difference between female patients and male patients was not statistically significant, it gives us a reasonable inference that female patients may have a higher risk of FQ-induced AA/AD, contrary to general knowledge that the incidence of AA/AD is higher in male patients. A previous study also showed that the risk was higher in female patients [7]. For age, the risk was signifi- cant in all age groups, but the differences between sub- groups were not statistically significant. Given that the risk was higher in patients aged 70 or older in a previous study [7], we recommend that further research be under- taken to understand the risk factors for FQ-induced AA/ AD. The sensitivity analysis supported the robustness of the results, as they were very similar to the results before the definition of the study population was changed. Strengths and limitations As an indication of the strength of this research, it was conducted using the national health insurance claim data of all adults aged 40 or older in Korea during the five years from 2013 to 2017. The NHIS-customized data are well accumulated in the form of detailed medi- cal activities and drugs, making it easy to generalize the research results, as nearly all domestic AA/AD patients were included in the study population. Additionally, we comprehensively considered various confounding fac- tors, such as underlying diseases, medication use, and Son et al. BMC Cardiovascular Disorders (2022) 22:44 Page 8 of 9 procedures and surgeries related to AA/AD. Moreover, we performed a sensitivity analysis by changing the def- inition of health outcomes of interest to minimize the effect of classification errors on the results. The preced- ing research results showed a 92% positive prediction for the identification of AA/AD cases when defining a group of cases considering both examination and diag- nosis [7]. Our work clearly has some limitations. First, the results may have been affected by confounding indica- tions. To reduce the bias from confounding indications, we excluded patients who had taken FQs during the year prior to the cohort entry date and included major indications of FQs as covariates in the adjusted model. However, the results may still have been affected by unmeasured underlying indications or the severity of the indication. The result must be carefully interpreted considering that patients who take FQs are possibly at higher risk of AA/AD due to unmeasured underlying conditions, indications for the drug, and important risk factors such as smoking. We emphasize that this result should not be interpreted as explicit evidence for causal effects. It is clear that more studies would be necessary to determine whether there is a causal relationship. Second, due to the nature of the claim data, it is dif- ficult to pinpoint the exact timing of treatment and drug use, and it is not possible to analyze drug use, procedures, or surgeries that are not covered by NHIS. Third, socioeconomic and clinical confounding fac- tors that could not be measured or predicted may have affected the results. For example, the difference in base- line characteristics of cases and controls can affect the results. To minimize the effect of known risk factors for AA/AD, we excluded patients with a history of AA/AD or related diseases during the year prior to the cohort entry date. However, some risk factors generally known to affect AA/AD, such as blood pressure, smoking sta- tus, and family history, were not considered in this study. In this sense, further studies are needed to evalu- ate the risk by patients’ baseline health status and par- ticular medical conditions, such as known risk factors for AA/AD. Finally, the various clinical types and characteristics of AA/AD were not analyzed because clinical information such as severity and detailed disease symptoms could not be fully determined by the diagnosis code alone. Thus, the results of this study are not appropriate for direct application to individuals, as patients may present with a variety of clinical characteristics. To overcome the potential bias introduced by confounding factors and the definition of exposure and health outcome of interest, we performed subgroup analysis, sensitivity analysis, and examined the dose–response relationship. Conclusion In this nested case–control study, we found that the use of FQs within a year was associated with a 10% increased risk of AA/AD in the Korean population. AA/ AD is a life-threatening disease accompanied by severe complications such as low blood pressure, shock, myo- cardial infarction, stroke, lower limb paralysis, and acute renal failure, which can lead to sudden death. In particular, early diagnosis and prompt treatment of abdominal AA/AD are critical, as 65% of patients die from cases of rupture [13]. Therefore, if patients feel symptoms such as chest pain in the early period of FQ use, even if the patient is not in the previously known risk group, medical professionals should suspect acute FQ-induced AA/AD, make a close diagnosis and con- sider changing or stopping the prescription. Moreover, if FQs are used in patients with already identified AA/ AD, medical professionals should review the patient’s history and carefully monitor them after drug admin- istration, keeping in mind that FQs could increase the risk of AA/AD and that the cumulative dose or dura- tion of FQ use may affect the risk. Abbreviations AA: Aortic aneurysm; AD: Aortic dissection; DDD: Defined daily dose; FQs: Fluoroquinolones; NHIS: National Health Insurance Service. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12872- 022- 02488-x. Additional file 1: Table S1. ICD 10 code of AA/AD–related dis- ease. Table S2. Results of conditional logistic regression analysis of the association between AA/AD and FQ use. Table S3. Frequency of underlying disease and mediation use in exposed and unexposed con- trols. Table S4. Association between AA/AD and FQ use in patients with cardiovascular diseases or indications of FQs. Acknowledgements We would like to thank the Benefits Strategy Department of the National Health Insurance Service for support. Authors’ contributions NYS and EMC are co-first authors and contributed equally. NYS contributed to the conceptualization, methodology, software, statistical analysis, and writing of the original draft. EMC contributed to conceptualization, methodology, and writing—review & editing. SYH and SYC contributed to supervision. BGK contributed to the supervision, project administration, writing, reviewing, and editing of the manuscript. All authors read and approved the final manuscript. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Availability of data and materials The data that support the findings of this study are available from the National Health Insurance Service in the Republic of Korea, but restrictions apply to the availability of these data, which were used under license for the current study and so are not publicly available. Data are, however, available from the authors Son et al. BMC Cardiovascular Disorders (2022) 22:44 Page 9 of 9 upon reasonable request and with permission from the National Health Insur- ance Service. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. Declarations Ethics approval and consent to participate Ethics approval for this study was obtained from the institutional review board of Korea Institute of Drug Safety and Risk Management, which waived informed consent (IRB approval number 2019-4). The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Received: 4 March 2021 Accepted: 31 January 2022 References 1. Kim Y, Park Y, Youk T, Lee S, Son Y. A study on the use of antibiotics in Korea and the resistance of major pathogens to antibiotics. NHIS Ilsan hospital Report. 2016:20-001. 2. Redgrave LS, Sutton SB, Webber MA, Piddock LJ. Fluoroquinolone resist- ance: mechanisms, impact on bacteria, and role in evolutionary success. Trends Microbiol. 2014;22(8):438–45. 3. Pasternak B, Inghammar M, Svanström H. Fluoroquinolone use and 4. risk of aortic aneurysm and dissection: nationwide cohort study. BMJ. 2018;360:k678. Food and Drug Administration (FDA). Drug Safety Communication: FDA warns about increased risk of ruptures or tears in the aorta blood vessel with fluoroquinolone antibiotics in certain patients. 2018. https:// www. fda. gov/ Drugs/ DrugS afety/ ucm62 8753. htm. Accessed 14 Jan 2022. 5. Ministry of Food and Drug Safety (MFDS). A notification on the distribu- tion of safety letters on fluoroquinolone antibiotics. https:// www. mfds. go. kr/ brd/m_ 545/ view. do? seq % ED% 94% 8C% EB% A3% A8% EC% 98% A4% EB% A1% 9C% ED% 80% B4% EB% 86% 80% EB% A1% A0& srchTp 0& itm_ seq_1 = itm_ seq & compa ny_ nm & page 0& itm_ seq_2 & Data_ stts_ gubun = 1. Accessed 14 Jan 2022. 0& compa ny_ cd 286& srchFr & srchW ord 0& multi_ & srchTo = = = = = = = = = C9999 6. Daneman N, Lu H, Redelmeier DA. Fluoroquinolones and collagen = 7. 8. associated severe adverse events: a longitudinal cohort study. BMJ Open. 2015;5(11):e010077. Lee C-C, Lee MG, Chen Y-S, Lee S-H, Chen Y-S, Chen S-C, et al. Risk of aortic dissection and aortic aneurysm in patients taking oral fluoroqui- nolone. JAMA Intern Med. 2015;175(11):1839–47. Lee C-C, Lee MG, Hsieh R, Porta L, Lee W-C, Lee S-H, et al. Oral fluo- roquinolone and the risk of aortic dissection. J Am Coll Cardiol. 2018;72(12):1369–78. 9. Meng L, Huang J, Jia Y, Huang H, Qiu F, Sun S. Assessing fluoroquinolone- associated aortic aneurysm and dissection: data mining of the public version of the FDA adverse event reporting system. Int J Clin Pract. 2019;73(5):e13331. 10. Singh S, Nautiyal A. Aortic dissection and aortic aneurysms associated with fluoroquinolones: a systematic review and meta-analysis. Am J Med. 2017;130(12):1449-57.e9. 11. Lee J, Lee JS, Park S-H, Shin SA, Kim K. Cohort Profile: The National Health Insurance Service-National Sample Cohort (NHIS-NSC), South Korea. Int J Epidemiol. 2016;46(2):e15. 12. Foundation ACoC, Guidelines AHATFoP, Surgery AAfT, Radiology ACo, Association AS, Anesthesiologists SoC, et al. 2010 ACCF/AHA/AATS/ ACR/ASA/SCA/SCAI/SIR/STS/SVM guidelines for the diagnosis and management of patients with thoracic aortic disease. J Am Coll Cardiol. 2010;55(14):e27–129. 13. Sakalihasan N, Limet R, Defawe OD. Abdominal aortic aneurysm. Lancet. 2005;365(9470):1577–89. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12884-021-03852-z	Bruckner et al. BMC Pregnancy and Childbirth (2021) 21:478 https://doi.org/10.1186/s12884-021-03852-z R E S E A R C H A R T I C L E Open Access Stillbirth as left truncation for early neonatal death in California, 1989–2015: a time-series study Tim A. Bruckner1, Samantha Gailey2* Gary M. Shaw7 and Jennifer Zeitlin8 , Abhery Das3, Alison Gemmill4, Joan A. Casey5, Ralph Catalano6, Abstract Background: Some scholars posit that attempts to avert stillbirth among extremely preterm gestations may result in a live birth but an early neonatal death. The literature, however, reports no empirical test of this potential form of left truncation. We examine whether annual cohorts delivered at extremely preterm gestational ages show an inverse correlation between their incidence of stillbirth and early neonatal death. Methods: We retrieved live birth and infant death information from the California Linked Birth and Infant Death Cohort Files for years 1989 to 2015. We defined the extremely preterm period as delivery from 22 to < 28 weeks of gestation and early neonatal death as infant death at less than 7 days of life. We calculated proportions of stillbirth and early neonatal death separately by cohort year, race/ethnicity, and sex. Our correlational analysis controlled for well-documented declines in neonatal mortality over time. Results: California reported 89,276 extremely preterm deliveries (live births and stillbirths) to Hispanic, non-Hispanic (NH) Black, and NH white mothers from 1989 to 2015. Findings indicate an inverse correlation between stillbirth and early neonatal death in the same cohort year (coefficient: -0.27, 95% CI of − 0.11; − 0.42). Results remain robust to alternative specifications and falsification tests. Conclusions: Findings support the notion that cohorts with an elevated risk of stillbirth also show a reduced risk of early neonatal death among extremely preterm deliveries. Results add to the evidence base that selection in utero may influence the survival characteristics of live-born cohorts. Keywords: Stillbirth, Neonatal death, Live birth, Left truncation Bias Background Infants born alive at extremely early gestational ages face substantial risk of imminent death. Extremely preterm births (i.e., delivery at less than 28 weeks' gestational age [GA]), for instance, account for less than 1% of all live births but over 40% of neonatal mortality [1, 2]. The ma- jority of these infant deaths occur in the early neonatal * Correspondence: sgailey@uci.edu 2School of Social Ecology, University of California Irvine, 209 Social Ecology I, Irvine, CA 92697, USA Full list of author information is available at the end of the article period which extends to less than 7 days after birth. Sub- stantial advancements in obstetric monitoring, neonatal care, and medical technology since the 1980s in high- income countries correspond with reductions over time in neonatal mortality [2, 3]. However, the incidence of early neonatal mortality among extremely preterm live births in the US remains between 15 and 20% [4, 5]. Epidemiologists continue to debate how to best esti- mate the population at risk in the perinatal period [6– 11]. Some argue that all fetuses which pass through a particular GA “starting point” (e.g., > 22 weeks) represent © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Bruckner et al. BMC Pregnancy and Childbirth (2021) 21:478 Page 2 of 9 a risk set, or denominator, of gestations at risk of ending with a neonatal death. According to this reasoning, those that die in utero and receive a classification of stillbirth would also appear in this risk set. This logic appears reflected in the rationale for the use of a composite out- come of perinatal death in randomized trials of obstet- rical interventions in which both stillbirths and early neonatal deaths represent “cases” [6, 12]. This composite perinatal death outcome coheres with the argument that stillbirths near week 22, at the threshold of viability, would have been at elevated risk for early neonatal death had they been born live. This “left truncation” argument [13, 14], if distilled to the realm of clinical decision- making, assumes that attempts to avert imminent still- birth among threatened gestations may “convert” a sub- set of them to a live-born delivery but result in an early neonatal death. The literature, however, also includes reports in which stillbirth and early neonatal death may be considered as distinct entities [7, 15]. The argument arises from two strands of evidence. First, in high-income countries that use consistent definitions and classification schemes, risk factors differ for stillbirth and early neonatal death. For instance, in the 1990s in Canada, congenital anomalies reportedly accounted for over 45% of early neonatal deaths but only ~ 9% of stillbirths reaching 25 weeks' GA [15]. Second, the risk of early neonatal death in the US has fallen, but the risk of stillbirth at GAs in the ex- tremely preterm period (< 28 weeks' GA) has remained unchanged [16]. These divergent population-level pat- terns indicate distinct antecedents of early neonatal death and stillbirth among extremely preterm gestations. Taken together, the field continues to debate various cir- cumstances under which researchers should regard still- birth and early neonatal death as joint or distinct outcomes [11, 17]. A recent report using data from California finds that an abrupt downward shift in stillbirths over time coin- cides with an upward shift in live births delivered in the extremely preterm period [18]. This report, while sug- gestive of left truncation, has no information on infant death and therefore cannot address whether early neo- natal death in the extremely preterm period falls in preg- nancy cohorts in which the risk of stillbirth rises. Understanding this potential relation would inform the extent to which intensity of fetal selection shapes the survival characteristics of In this paper, we contribute to the literature by testing in Cali- fornia over a 26-year period whether annual cohorts de- livered at extremely preterm GAs show an inverse correlation between their proportions of stillbirth and early neonatal death. live-born cohorts. We apply methods that adjust for the secular decline since 1990 in the risk of early neonatal death before test. conducting our correlational In addition, given well-documented differences in risks of stillbirth and early neonatal death by race/ethnicity and fetal sex, we stratify pregnancy cohorts by maternal race/ethnicity and sex of gestation [15, 16, 19–23]. Our analysis fo- cuses on California because, in addition to accounting for ~ 15% of all US births, the state uses consistent defi- nitions and data collection practices for recording still- births over a long time period (i.e., 1989–2015) [24]. The US file, by contrast, makes available fewer years of cohort mortality data and reflects a mix of states with substantial differences in quality and reporting practices of stillbirth. Methods We retrieved live birth and infant death information from the California Linked Birth and Infant Death Co- hort File (BCF) from 1989 to 2015. The cohort nature of the BCF allows for the estimation of incident early neo- natal death. Our study period began in 1989 and ended in 2015. This time period uses consistent definitions for live births and early neonatal deaths. The methodology of reporting births and infant deaths in California has not changed over the time period and remains nearly 100% complete [22, 24]. For administrative reasons, the California Department of Health Services did not create a BCF for 1998. As a result, we did not include the 1998 birth cohort in our analysis. The institutional review boards at the California Department of Public Health (# 2018–065) and the University of California, Irvine (# 2013–9716) approved the use of these data for our study. We retrieved stillbirth information from the California Fetal Death file. The State of California defines a still- birth as a “death prior to the complete expulsion or ex- traction from its mother of a product of human conception. The death is indicated by the fact that the fetus does not breathe or show any other evidence of life such as beating of the heart, pulsation of the umbilical cord, or definite movement of voluntary muscles [25].” California’s Health and Safety Code requires reporting of all stillbirths after the 20th week of gestation except for induced abortions [25–28]. The California Department of Health Services uses a standard protocol to perform quality control checks and data processing. We, as with previous research [26], calculated the proportion of still- births for each year by dividing the count of stillbirths by the sum of live births and stillbirths among extremely preterm deliveries. Consistent with the definition from the World Health Organization [29], we specified the extremely preterm period as delivery from 22 weeks 0/7 days to 27 weeks 6/ 7 days of gestation. Previous literature uses this span of gestational ages to define extreme preterm delivery for Bruckner et al. BMC Pregnancy and Childbirth (2021) 21:478 Page 3 of 9 two reasons [2, 30, 31]. First, many clinicians argue that 22 0/7 weeks represents the lowest gestational age cutoff for a viable delivery [32, 33]. Second, the risk of infant death among live births beginning at 28 0/7 weeks falls below 5% [34]. We therefore restricted our analysis to stillbirths and live births from 22 0/7 to 27 6/7 weeks of gestation. In addition, we restricted the sample to single- ton gestations owing to the shorter mean gestational age of multiple births and the greater risk of perinatal death among them regardless of gestational age [34]. We used early neonatal death as a key indicator of perinatal health whose causes likely originate during pregnancy [35]. Early neonatal death is defined as an in- fant death at less than 7 days of life [36, 37]. Consistent with definitions used for surveillance in California and elsewhere, only live births represent the risk set of pos- sible early neonatal deaths [35, 38]. We therefore calcu- lated the proportion of early neonatal deaths by dividing the number of infant deaths in the first 7 days of life by the total number of live births. The risk of stillbirth and early neonatal death vary sub- stantially by race/ethnicity and by sex [16, 39, 40]. For this reason, we arrayed all extremely preterm deliveries by race/ethnicity and sex before conducting the statis- tical analyses. Given that the notion of left truncation represents a cohort concept, we did not include individual-level controls in our aggregate-level test. We excluded records with missing or unknown race/ethni- city (0.87%) or sex (<.0001%) as well as live birth records with implausible birthweight for gestational age informa- tion (1.2%) [41] and stillbirths with missing values for GA (10.2%). The welcomed rarity of extremely preterm deliveries creates an analytic challenge in providing stable estimates, by race/ethnicity and sex, of the annual incidence of stillbirth and early neonatal death. To minimize the role of stochastic variation in our analysis, we focused only on race/ethnicities with a minimum of 100 extremely preterm deliveries per sex in each study year. This restriction yielded three race/ethnicities: non- Hispanic (NH) Black, NH white, and Hispanic. Statistical analysis We first plotted, by race/ethnicity and sex, the annual incidence (1989–2015) of stillbirth and early neonatal death among extremely preterm deliveries. Second, given the well-documented declines over time in perinatal mortality, we removed trend from these series (if de- tected by a Dickey-Fuller test) by employing ordin- ary least squares linear regression analysis to fit a year from 1 to 26, where 1989 = 1, variable (continuous, 1990 = 2, etc.) [42]. We removed trend to minimize con- founding due to secular improvements in perinatal care over time that could induce a positive correlation be- tween the risk of fetal and early neonatal death. Third, we tested our hypothesized inverse association between yearly risks of fetal and early neonatal death by calculat- ing the Pearson correlation coefficient between the de- trended annual values of the two series. Given the three race/ethnicities, two sexes, and 26 years studied, the op- erational sample size for the correlational analysis is 156 (i.e., 3 × 2 × 26 = 156). We conducted additional sensitivity checks including autoregressive, integrated, moving average (ARIMA) time-series analyses if we discovered an inverse correl- ation. We applied a transfer function within the ARIMA context [43], which identifies and removes patterns from the early neonatal death series before inserting the inde- pendent variable (i.e., residual values of stillbirth) into the test equation. ARIMA transfer functions provide more efficient estimation of standard errors than do sim- ple ordinary least squares correlational analyses since they remove autocorrelation. Next, we repeated all ana- lyses but examined the correlation coefficient between stillbirth and neonatal death (i.e., death within first 28 days after birth), rather than early neonatal death, given that a small but non-negligible fraction of frail and ex- tremely preterm infants die between 7 and 28 days of life [4]. Results Over the test period, California recorded 89,276 ex- tremely preterm deliveries (live births and stillbirths) to Hispanic, NH Black, and NH white mothers. Table 1 de- scribes the annual mean and range of live births, still- births, and early neonatal deaths by race/ethnicity. The crude incidence of early neonatal death is 20.2 per 100 Table 1 Annual mean and range of live births, stillbirths, and early neonatal deaths delivered extremely preterm (22 to 27 weeks of gestational age), by race/ethnicity, in California, 1989 to 2015 N Annual mean (SD)a Annual rangeb Non-Hispanic Black Live births Fetal deaths 12,604 485 (89) 2911 112 (18) Early neonatal deaths 2525 97 (29) Non-Hispanic white Live births Fetal deaths 21,725 836 (148) 6602 254 (67) Early neonatal deaths 4540 175 (60) Hispanic Live births Fetal deaths 36,058 1387 (178) 9376 361 (36) Early neonatal deaths 7133 274 (35) Abbreviation: SD standard deviation aData for 1998 not available bValues rounded to nearest integer 377–667 82–147 65–158 632–1172 155–409 102–328 972–1760 253–433 224–341 Bruckner et al. BMC Pregnancy and Childbirth (2021) 21:478 Page 4 of 9 extremely preterm live births. NH whites show the greatest incidence of neonatal death (20.9 per 100 ex- tremely preterm live births). Table 2 describes the socio- demographic characteristics of the study population. Figure 1 a through c plot, by race/ethnicity and sex, the annual stillbirths among extremely preterm deliver- ies. We caution against comparing mean levels of still- birth across the panels given that missing GA occurs disproportionately among racial/ethnic minorities and therefore underestimates stillbirth especially among NH Blacks. The plots, rather, are useful in highlighting (within each race/ethnicity) the substantial variation over time in stillbirth. For NH whites, stillbirths show a downward trend over time. For Hispanics, stillbirths also decline over time, but this decline begins with a down- ward shift in 2007. By contrast, the mean level of still- births among NH Blacks is not lower in 2010–2015 relative to 1989–1994. For NH whites and Hispanics (but not NH Blacks), the proportion of stillbirths among females is greater than that of males. Table 2 Maternal and pregnancy characteristics among extremely preterm deliveries (22 to 27 weeks of gestational age) in California, 1989 to 2015 Maternal age 18 or younger 18 to 24 25 to 29 30 to 34 35 or older Maternal education Less than high school High school graduate Some college College graduate Maternal race/ethnicity Non-Hispanic Black Non-Hispanic white Hispanic Expected source of payment Medicaid Private insurance Other Fetal sex Male Female Na 4736 27,082 21,444 19,641 16,114 30,637 27,222 23,065 4760 16,251 25,472 47,553 43,054 34,638 11,575 47,090 42,186 aValues from 1998 not available bColumn percentages may not sum to 100 due to missing values for that variable %b 5.3 30.3 24.0 22.0 18.0 34.3 30.5 25.8 5.3 18.2 28.5 53.3 48.2 38.8 13.0 52.7 47.3 Fig. 1 Incidence of stillbirth among extremely preterm deliveries for females (red) and males (blue), by race/ethnicity, in California, 1989 to 2015. a Non-Hispanic Black; b Non-Hispanic white; c Hispanic The risk of early neonatal death declines over time for all race/ethnicities (Fig. 2a through c). Most of this re- duction occurs before 2000. After 2000, NH Blacks show a leveling off of early neonatal death, but male risk con- sistently falls below female risk (i.e., for 13 of the 15 years 2001–2015). This sex-specific pattern, after 2000, in early neonatal death also occurs in Hispanics (i.e., male incidence is less than female incidence for 11 of the 15 years 2001–2015). The correlation coefficient between the stillbirth and early neonatal death series, after removal of trend, sup- ports left truncation in that it is negative and shows a Bruckner et al. BMC Pregnancy and Childbirth (2021) 21:478 Page 5 of 9 Fig. 3 Scatter plot and best fitting line of detrended incidence of stillbirth and early neonatal death among extremely preterm deliveries across 156 race/ethnicity-sex-year cohorts, 1989–2015 binary indicator variables for each race/ethnicity-gender group to remove mean differences in early neonatal death. Second, we used autocorrelation and partial auto- correlation function routines (as outlined by Box and Jenkins) to identify and remove patterns from the early neonatal death series (i.e., dependent variable) [44]. Pat- terns detected by these routines include secular trends, cycles, oscillations, and the tendency for high or low values to be “remembered” in subsequent time periods [44]. The residuals of each of the time series (after ARIMA routines) show no patterns, have a mean of 0, and have values statistically independent of one another (Additional file 1: Tables S1 and S2). Third, we inserted the unpatterned values of the stillbirth series (i.e., inde- pendent variable) into the test equation and estimated its relation with early neonatal death. ARIMA results show that a 1-unit change in stillbirth varies inversely with a 0.40-unit change in early neonatal death (coeffi- cient: -0.40, 95% CI of − 0.16; − 0.63). Note that the scale of this ARIMA coefficient differs from that of the ori- ginal test, thus precluding direct comparisons of their magnitude. In addition, we remind the reader that al- though ARIMA time-series routines increase the effi- ciency of estimates and rule out confounding by autocorrelation, they are conservative in that they re- move patterns from both series without consideration of whether one series (e.g., stillbirth) may have induced a pattern in the other series (e.g., early neonatal death). As a falsification check we inspected whether the lead and lag cross-correlation coefficients (i.e., stillbirths in year t-1 and early neonatal deaths in year t, and still- births in year t + 1 and early neonatal deaths in year t) differ from 0 [45]. These lead and lag tests show no de- tectable difference from 0 (Table 3). Findings indicate that the discovered inverse association appears specific to pregnancy cohorts which share the same year of delivery. Fig. 2 Incidence of early neonatal death among extremely preterm live births for females (red) and males (blue), by race/ethnicity, in California, 1989 to 2015. a Non-Hispanic Black; b Non-Hispanic white; c Hispanic confidence interval (CI) that does not contain 0 (coeffi- cient.: -0.27, 95% CI of − 0.11; − 0.42). This inverse cor- extremely preterm relation indicates deliveries, incidence of stillbirth above trend in a par- ticular year corresponds with fewer early neonatal deaths among live births in that year. Figure 3 displays the scat- ter plot and best fitting line of this inverse correlation. among that, We conducted an additional sensitivity analysis to as- sess robustness of findings. We used a transfer function approach within an ARIMA time-series context which proceeded with the following steps. First, we inserted Bruckner et al. BMC Pregnancy and Childbirth (2021) 21:478 Page 6 of 9 Table 3 Cross-correlation coefficients (standard errors in parentheses) of the detrended incidence of stillbirth and early neonatal death among extremely preterm deliveries (22 to 27 weeks of gestational age) in California, 1989 to 2015 Stillbirth precedes early neonatal death by 1 year −0.07 (0.08) Both series in same year −0.27 (0.08) Stillbirth follows early neonatal death by 1 year −0.00 (0.08) Given that some frail extremely preterm live births die between the 7th and 28th day after birth, we then assessed whether results appear similar when using the de-trended incidence of neonatal death (i.e., death within first 28 days after birth) instead of early neonatal death. As with the main findings, we observe an inverse correl- ation between detrended annual values of fetal and neo- natal death. The result (coefficient: -0.33, 95% CI of −0.17; − 0.49) is farther from the null than that of the original test using early neonatal death. Discussion The literature reports no tests of whether, at the popula- tion level, stillbirth represents a form of left truncation for the risk of early neonatal death among live births. We used the longest annual time series available to us— in California, from 1989 to 2015—to test among ex- tremely preterm births whether the incidence of early neonatal death varies inversely with the incidence of stillbirth. Results, which control for well-documented secular trend, support the hypothesis. Annual cohorts which experience relatively lower stillbirth in the ex- treme preterm period also show elevated risk of neonatal death among live births. This finding builds on recent work documenting an inverse relation between stillbirth and live births in the periviable period [18] and further supports that selection in utero may affect the infant health profile of live-born cohorts [23]. Strengths of the analysis include the use of a long an- nual time series in a populous state with a consistent definition, classification, and reporting protocol for still- births. Methods also adjust for well-documented secular declines in early neonatal death, which minimizes the risk of confounding due to medico-technological im- provements in perinatal care. In addition, our test rules out confounding due to the changing racial/ethnic com- position of cohorts over time since we stratified the series by race/ethnicity. Lastly, the fact that we observed an inverse correlation at the synchronous pregnancy co- hort—but not between asynchronous cohorts—further minimizes results arising due to chance. the possibility of stillbirths Limitations include that remain largely under-reported, especially earlier in the series and dur- ing the extreme preterm period when the fetus is smaller [25–28, 40]. Whereas the extent of this under-reporting is unknown, this circumstance likely improved substan- tially over time [46, 47]. Lack of GA reporting for stillbirth also appears more common among NH Blacks, which precludes direct comparison of GA-specific inci- dence of stillbirth across race/ethnicity [1]. For example, the incidence of all recorded NH Black stillbirths (in- cluding missing and non-missing GA) in California is greater than that of NH whites, but exclusion of cases with missing GA reverses this difference. We, however, know of no evidence that vigilance of reporting stillbirth falls in particular race/ethnicities and years when report- ing of early neonatal death increases. Reductions over time in missing/unknown GA in the California Fetal Death File make it challenging to inter- pret whether any observed reductions in the risk of still- birth represent true perinatal health improvements. We note, however, that this circumstance—or other clinical or cultural shifts in reporting—are unlikely to drive our results. Findings remain robust to ARIMA time-series methods which removed such patterns in the series be- fore testing the synchronous correlation. We, neverthe- less, note substantial shifts over time in obstetrical practice. Deliveries by cesarean section among periviable births, for instance, have increased substantially over our test period (e.g., from 32% [1989–1997] to over 50% [2007 to 2015] of deliveries between 24 to 27 weeks oc- curring by cesarean section; see Additional File, Fig. S1). The potential influence of these shifts in clinical practice on stillbirths warrant further investigation. A recent workshop panel from European countries en- courages vital statistics agencies to routinely collect clin- ical data that could classify stillbirths as occurring either before the initiation of labor (i.e., antepartum) or during labor (i.e., intrapartum) [48]. Such information may as- sist with identifying a subset of stillbirths whose selec- tion in utero affects the risk of early neonatal death [15]. California and other US states do not routinely collect In addition, unlike other countries this information. (e.g., France) [49], termination of pregnancies after 22 weeks (due to, for instance, structural anomaly) is rare and not routinely reported in vital statistics [50, 51]. We encourage collection of this and other information on cause of the stillbirth and reason for induction of labor (taken from medical records) so that researchers can better understand the components of these losses as well as their potential influence on neonatal death. We focused on deliveries in the extremely preterm period. Excess stillbirths may also induce left truncation for neonatal mortality among live births greater than 28 weeks’ GA. Although the risk of neonatal mortality Bruckner et al. BMC Pregnancy and Childbirth (2021) 21:478 Page 7 of 9 declines substantially with each advancing week of GA [4], we encourage replication and extension of our work beyond the extreme preterm period. We suspect, how- ever, that any discovered “signal” would appear attenu- ated relative to the inverse correlation we report for the extreme preterm period. A recent analysis in California of over 11 million births finds a lower-than-expected frequency of spontaneous preterm live births among NH Black males [52]. The Authors speculate that elevated selection in utero of NH Black males in particular may contribute to the “miss- ing” number of NH Black males born preterm. Intri- guingly, the discovered outlier in Fig. 3 (bottom right corner)—exceptionally high stillbirth but low early neo- natal death—occurs among NH Black males in 2000. In addition, exploration among NH Black males indicates that the correlation coefficient for the stillbirth and early neonatal death series (1989–2015) at the synchronous lag is −0.40 (vs. -0.27 for the overall coefficient across all race/ethnicities and sexes). This evidence, albeit post hoc and exploratory, would appear to warrant further inquiry on the role of late selection in utero on the risk of neonatal death especially among NH Black males born preterm. We acknowledge the descriptive nature of our investi- gation in that we do not identify underlying causes of stillbirth or early neonatal death. We recommend add- itional work to identify individual-level risk factors pre- sumed to cause either outcome. In addition, given the population-based nature of our investigation, we caution against using findings to infer individual frailty of spe- cific live births who may have been delivered early in ef- forts to avoid imminent stillbirth. The annual resolution of our cohorts, moreover, indicates that we cannot align pregnancies by estimated month of conception to estab- lish clear temporal order between stillbirth and early neonatal death. Our work, rather, complements other re- search examining the potential role of selection in utero in shaping the survival characteristics of live-born co- horts [2, 11, 23, 53]. We encourage subsequent analyses of cohorts using larger datasets with sufficient counts of fetal and early neonatal death per month to establish such temporal order between loss in utero and the risk of death among extremely preterm live births. Abbreviations GA: Gestational age; NH: Non-Hispanic; CI: Confidence interval Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12884-021-03852-z . Additional file 1: Table S1. Coefficients (standard errors in parentheses) for the lagged values of the autocorrelation (ACF) and partial autocorrelation (PACF) functions of the residualized value of the incidence of stillbirths (n = 126). Table S2. Coefficients (standard errors in parentheses) for the lagged values of the autocorrelation (ACF) and partial autocorrelation (PACF) functions of the residualized value of the incidence of early neonatal deaths (n = 126). Fig. S1. Proportion of C- sections by perviable gestational age (22–27 weeks) in California by three time periods between 1989 and 2015. Acknowledgments This manuscript benefitted from discussions of earlier drafts with the following groups: Obstetrical, Perinatal and Pediatric Epidemiology Research Team at the Institut national de la santé et de la recherche médicale (INSE RM), and colleagues attending the Public Health Seminar at the Portland State University / Oregon Health Sciences University School of Public Health. Authors’ contributions TB, JZ, and RC developed the research question, TB, SG and AD compiled the data, and TB, AG, JC, RC, GS, and JZ analyzed the data. All Authors interpreted the results TB drafted the manuscript, and all authors wrote sections of the manuscript. All authors read and approved the final manuscript. Funding None. Availability of data and materials The datasets generated and/or analyzed during the current study are not publicly available to guarantee the anonymity of individuals. Please contact Tim A. Bruckner (Tim.bruckner@uci.edu) to request access to the datasets used in this study. Declarations Ethics approval and consent to participate The Authors have adhered to ethical standards in this work and obtained Institutional Review Board approvals from the California Department of Public Health (CDPH) (#2018–065) and University of California, Irvine (UCI) (#2013–9716) to access and use the California linked birth and infant death cohort file (BCF) 1989 to 2015. Since we used administrative data, the research ethics review committees of the CDPH and UCI waived the need for informed consent. Data were collected and analyzed using anonymous identifiers instead of maternal names to ensure confidentiality. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Conclusions Annual pregnancy cohorts which experience relatively greater stillbirth in the extremely preterm period also show lower risk of early neonatal death among live births. Results, which remain robust to alternative speci- fications and falsification tests, add to growing evidence that elevated selection in utero contributes to improved survival in live-born cohorts. Author details 1Program in Public Health & Center for Population, Inequality, and Policy, University of California Irvine, 653 E. Peltason Dr., Irvine, CA 92697, USA. 2School of Social Ecology, University of California Irvine, 209 Social Ecology I, Irvine, CA 92697, USA. 3Program in Public Health, University of California Irvine, 653 E. Peltason Dr., Irvine, CA 92697, USA. 4Population, Family and Reproductive Health, Johns Hopkins Bloomberg School of Public Health, 615 N. Wolfe St., Baltimore, MD 21205, USA. 5Mailman School of Public Health, Columbia University, 722 W. 168th St., New York, NY 10032, USA. 6School of Public Health, University of California Berkeley, Berkeley, CA 94720, USA. 7School of Medicine, Stanford University, Stanford, CA 94305, USA. Bruckner et al. BMC Pregnancy and Childbirth (2021) 21:478 Page 8 of 9 8Université de Paris, CRESS, Obstetrical, Perinatal and Pediatric Epidemiology Research Team, EPOPé, INSERM, INRA, F-75004 Paris, France. Received: 5 September 2020 Accepted: 5 May 2021 References 1. 2. 3. 4. 5. 6. Glass HC, Costarino AT, Stayer SA, Brett C, Cladis F, Davis PJ. Outcomes for extremely premature infants. Anesth Analg. 2015;120(6):1337–51. https://doi. org/10.1213/ANE.0000000000000705. Stoll BJ, Hansen NI, Bell EF, Walsh MC, Carlo WA, Shankaran S, et al. Trends in care practices, morbidity, and mortality of extremely preterm neonates, 1993-2012. JAMA. 2015;314(10):1039–51. https://doi.org/10.1001/jama.2015.1 0244. Iams JD, Romero R, Culhane JF, Goldenberg RL. Primary, secondary, and tertiary interventions to reduce the morbidity and mortality of preterm birth. Lancet. 2008;371(9607):164–75. https://doi.org/10.1016/S0140-673 6(08)60108-7. Stoll BJ, Hansen NI, Bell EF, Shankaran S, Laptook AR, Walsh MC, et al. Neonatal outcomes of extremely preterm infants from the NICHD neonatal research network. Pediatrics. 2010;126(3):443–56. https://doi.org/10.1542/ peds.2009-2959. Patel RM, Kandefer S, Walsh MC, Bell EF, Carlo WA, Laptook AR, et al. Causes and timing of death in extremely premature infants from 2000 through 2011. N Engl J Med. 2015;372(4):331–40. https://doi.org/10.1056/NEJMoa14 03489. Platt RW, Joseph KS, Ananth CV, Grondines J, Abrahamowicz M, Kramer MS. A proportional hazards model with time-dependent covariates and time- varying effects for analysis of fetal and infant death. Am J Epidemiol. 2004; 160(3):199–206. https://doi.org/10.1093/aje/kwh201. 8. 7. Wilcox AJ, Weinberg CR. Invited commentary: analysis of gestational-age- specific mortality—on what biologic foundations? Am J Epidemiol. 2004; 160(3):213–4. https://doi.org/10.1093/aje/kwh204. Joseph KS, Kramer MS. The fetuses-at-risk approach: survival analysis from a fetal perspective. Acta Obstet Gynecol Scand. 2018;97(4):454–65. https://doi. org/10.1111/aogs.13194. Auger N, Gilbert NL, Naimi AI, Kaufman JS. Fetuses-at-risk, to avoid paradoxical associations at early gestational ages: extension to preterm infant mortality. Int J Epidemiol. 2014;43(4):1154–62. https://doi.org/10.1093/ ije/dyu011. 9. 13. 11. 12. 10. Basso O. Implications of using a fetuses-at-risk approach when fetuses are not at risk. Paediatr Perinat Epidemiol. 2016;30(1):3–10. https://doi.org/1 0.1111/ppe.12254. Snowden JM, Bovbjerg ML, Dissanayake M, Basso O. The curse of the perinatal epidemiologist: inferring causation amidst selection. Curr Epidemiol Rep. 2018;5(4):379–87. https://doi.org/10.1007/s40471-018-0172-x. Kramer MS, Zhang X, Platt RW. Analyzing risks of adverse pregnancy outcomes. Am J Epidemiol. 2014;179(3):361–7. https://doi.org/10.1093/aje/kwt285. Lisonkova S, Joseph KS. Left truncation bias as a potential explanation for the protective effect of smoking on preeclampsia. Epidemiology. 2015;26(3): 436–40. https://doi.org/10.1097/EDE.0000000000000268. Schisterman EF, Cole SR, Ye A, Platt RW. Accuracy loss due to selection bias in cohort studies with left truncation. Paediatr Perinat Epidemiol. 2013;27(5): 491–502. https://doi.org/10.1111/ppe.12073. Kramer MS, Liu S, Luo Z, Yuan H, Platt RW, Joseph KS. Analysis of perinatal mortality and its components: time for a change? Am J Epidemiol. 2002; 156(6):493–7. https://doi.org/10.1093/aje/kwf077. 14. 15. 16. MacDorman MF, Gregory ECW. Fetal and perinatal mortality: United States, 17. 18. 2013. Natl Vital Stat Rep. 2015;64(8):1–24. Joseph KS. Towards a unified perinatal theory: reconciling the births-based and fetus-at-risk models of perinatal mortality. Paediatr Perinat Epidemiol. 2019;33(2):101–12. https://doi.org/10.1111/ppe.12537. Elser H, Gemmill A, Casey JA, Karasek D, Bruckner T, Mayo JA, et al. Stillbirths and live births in the periviable period. Ann Epidemiol. 2020;49:8–12. https://doi.org/10.1016/j.annepidem.2020.07.002. 19. Gregory ECW, MacDorman MF, Martin JA. Trends in fetal and perinatal mortality in the United States, 2006-2012. NCHS Data Brief. 2014;(169):1–8. 20. MacDorman MF, Kirmeyer SE, Wilson EC. Fetal and perinatal mortality, United States, 2006. Natl Vital Stat Rep. 2012;60(8):1–22. 21. MacDorman MF, Kirmeyer S. The challenge of fetal mortality. NCHS Data Brief. 2009;:1–8. 22. Catalano R, Bruckner T, Anderson E, Gould JB. Fetal death sex ratios: a test of the economic stress hypothesis. Int J Epidemiol. 2005;34(4):944–8. https:// doi.org/10.1093/ije/dyi081. 23. Bruckner TA, Catalano R. Selection in utero and population health: theory and typology of research. SSM Popul Health. 2018;5:101–13. https://doi. org/10.1016/j.ssmph.2018.05.010. 24. California Department of Health Services Center for Health Statistics. Linked birth and infant death cohort file, 1989-2001. Sacramento; 2004. 25. California Department of Health Services, Office of Vital Records. Handbook for Local Registrars of Births and Deaths. Sacramento, 1996. 26. Martin JA, Hoyert DL. The national fetal death file. Semin Perinatol. 2002; 26(1):3–11. https://doi.org/10.1053/sper:2002.29834. 27. Gaudino JA, Blackmore-Prince C, Yip R, Rochat RW. Quality assessment of fetal death records in Georgia: a method for improvement. Am J Public Health. 1997;87(8):1323–7. https://doi.org/10.2105/AJPH.87.8.1323. 28. Gould JB. Vital records for quality improvement. Pediatrics. 1999;103(1 Suppl E):278–90. 29. Preterm birth. https://www.who.int/news-room/fact-sheets/detail/preterm- 30. 31. birth. Accessed 20 Jan 2020. Johanzon M, Ödesjö H, Jacobsson B, Sandberg K, Wennerholm U-B. Extreme preterm birth: onset of delivery and its effect on infant survival and morbidity. Obstet Gynecol. 2008;111(1):42–50. https://doi.org/10.1097/01.A OG.0000295866.97499.35. Fellman V, Hellström-Westas L, Norman M, Westgren M, Källén K, Lagercrantz H, et al. One-year survival of extremely preterm infants after active perinatal care in Sweden. Obstet Anesth Dig. 2010;30(1):22–3. https:// doi.org/10.1097/01.aoa.0000367003.25266.35. 32. Morgan MA, Goldenberg RL, Schulkin J. Obstetrician-gynecologists’ practices regarding preterm birth at the limit of viability. J Matern Fetal Neonatal Med. 2008;21(2):115–21. https://doi.org/10.1080/14767050701866971. 33. Markestad T, Kaaresen PI, Rønnestad A, Reigstad H, Lossius K, Medbø S, et al. Early death, morbidity, and need of treatment among extremely premature infants. Pediatrics. 2005;115(5):1289–98. https://doi.org/10.1542/peds.2004-1482. 34. Chen H-Y, Chauhan SP. Risk of neonatal and infant mortality in twins and singletons by gestational age. Am J Perinatol. 2019;36(08):798–805. https:// doi.org/10.1055/s-0038-1675218. Lehtonen L, Gimeno A, Parra-Llorca A, Vento M. Early neonatal death: a challenge worldwide. Semin Fetal Neonatal Med. 2017;22(3):153–60. https:// doi.org/10.1016/j.siny.2017.02.006. 35. 36. United Nations Interagency Group for Child Mortality Estimation. Levels and trends in child mortality. Report 2014. New York: UNICEF. p. 2014. http:// www.unicef.org/media/files/Levels_and_Trends_in_Child_Mortality_2014.pdf 37. Barfield WD, Newborn C on FA. Standard terminology for fetal, infant, and perinatal deaths. Pediatrics. 2016;137(5):e20160551. https://doi.org/10.1542/ peds.2016-0551. 38. Bruckner TA, Cheng YW, Caughey AB. Increased neonatal mortality among normal-weight births beyond 41 weeks of gestation in California. Am J Obstet Gynecol. 2008;199:421.e1–421.e7. https://doi.org/10.1016/j.ajog.2008.05.015. 39. MacDorman MF, Hoyert DL, Mathews TJ. Recent declines in infant mortality in the United States, 2005-2011. NCHS Data Brief. 2013:1–8. 40. Mondal D, Galloway TS, Bailey TC, Mathews F. Elevated risk of stillbirth in males: systematic review and meta-analysis of more than 30 million births. BMC Med. 2014;12(1):220. https://doi.org/10.1186/s12916-014-0220-4. 41. Alexander GR, Himes JH, Kaufman RB, Mor J, Kogan M. A United States national reference for fetal growth. Obstet Gynecol. 1996;87(2):163–8. https://doi.org/10.1016/0029-7844(95)00386-X. 42. Dickey D, Fuller W. Distribution of the estimators for autoregressive time series with a unit root. J Am Stat Assoc. 1979;74(366a):427–31. https://doi. org/10.1080/01621459.1979.10482531. Scientific Computing Associates Corp. [computer program]. Chicago, IL; 2009. 43. 44. Box GE, Jenkins GM, Reinsel GC, Ljung GM. Time series analysis: forecasting and control. Hoboken: Wiley; 2015. 45. Haugh LD. Checking the independence of two covariance-stationary time series: a univariate residual cross-correlation approach. J Am Stat Assoc. 1976;71(354):378–85. https://doi.org/10.1080/01621459.1976.10480353. 46. Alexander GR. The accurate measurement of gestational age--a critical step toward improving stillbirth reporting and perinatal health. Am J Public Health. 1997;87(8):1278–9. https://doi.org/10.2105/AJPH.87.8.1278. Bruckner et al. BMC Pregnancy and Childbirth (2021) 21:478 Page 9 of 9 47. 48. Schoendorf KC, Branum AM. The use of United States vital statistics in perinatal and obstetric research. Am J Obstet Gynecol. 2006;194(4):911–5. https://doi.org/10.1016/j.ajog.2005.11.020. Smith LK, Blondel B, Zeitlin J. Producing valid statistics when legislation, culture and medical practices differ for births at or before the threshold of survival: report of a European workshop. BJOG. 2020;127(3):314–8. https:// doi.org/10.1111/1471-0528.15971. 49. Monier I, Lelong N, Ancel PY, Benachi A, Khoshnood B, Zeitlin J, et al. Indications leading to termination of pregnancy between 22+0 and 31+6 weeks of gestational age in France: a population-based cohort study. Eur J Obstet Gynecol Reprod Biol. 2019;233:12–8. https://doi.org/10.1016/j. ejogrb.2018.11.021. 50. Grimes DA. The continuing need for late abortions. JAMA. 1998;280(8):747– 51. 52. 50. https://doi.org/10.1001/jama.280.8.747. Jatlaoui TC, Eckhaus L, Mandel MG, Nguyen A, Oduyebo T, Petersen E, et al. Abortion surveillance - United States, 2016. MMWR Surveill Summ. 2019; 68(11):1–41. https://doi.org/10.15585/mmwr.ss6811a1. Shaw GM, Mayo JA, Eisenberg ML, Catalano R, Stevenson DK. Male-to- female ratios, race/ethnicity, and spontaneous preterm birth among 11 million California infants. Am J Perinatol. 2019. https://doi.org/10.1055/s-003 9-3400449. 53. Catalano R, Bruckner TA, Karasek D, Yang W, Shaw GM. Reproductive suppression, birth defects, and periviable birth. Evol Appl. 2018;11(5):762–7. https://doi.org/10.1111/eva.12585. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12872-019-1064-9	Li et al. BMC Cardiovascular Disorders (2019) 19:90 https://doi.org/10.1186/s12872-019-1064-9 R E S E A R C H A R T I C L E Open Access Unique electrocardiographic pattern “w” wave in lead I of idiopathic ventricular arrhythmias arising from the distal great cardiac vein Teng Li1,2, Qiong Xu1, Xian-zhang Zhan2, Yu-mei Xue2, Hong-tao Liao2, Yi-fu Li1, Konstantinos P. Letsas3 and Shu-lin Wu2* Abstract Background: The ECG characteristics of the distal coronary venous system ventricular arrhythmias (VAs) share common features with VAs arising from the aortic cusps or the endocardial left ventricular outflow tract (LVOT) beneath the cusps. The purpose of this study was to identify specific electrocardiographic and electrophysiological characteristics of VAs originating from the distal great cardiac vein (GCV). Methods: Based on the successful ablation site, patients with idiopathic VAs from the distal GCV, left coronary cusp (LCC) or the subvalvular left ventricular outflow tract (LVOT) area were included in the present study. Results: The final population consisted of 39 patients (35 males, mean age 51 ± 23 years). All VAs displayed a right bundle branch block (RBBB) morphology with inferior axis. Among these patients, 15 were successfully ablated at the GCV, 15 at the LCC and 9 at the subvalvular region. A “w” pattern in lead I was present in 12 out of 15 (80%) VAs originating from the distal GCV compared to none of VAs arising from the other two sites (p < 0.01). VAs with a GCV origin exhibited more commonly increased intrinsicoid deflection time, higher maximum deflection index and wider QRS duration compared to LCC and subvalvular sites (p < 0.05). Acceptable pace mapping at the successful ablation site was achieved in 10 patients. After an average of 36 ± 24 months follow up, 14 (93.3%) patients were free from VAs recurrence. Conclusion: A “w” pattern in lead I may distinguish distal GCV VAs from VAs arising from the LCC or the subvalvular region. Keywords: Idiopathic, Ventricular arrhythmias, Great cardiac vein, Catheter ablation Background Idiopathic ventricular arrhythmias (VAs) can arise from the left ventricular (LV) endocardium and epicardium. The incidence of an epicardial origin may be as high as 9% [1]. The coronary venous system provides an alternative route to target epicardial VAs [2–7]. The ECG characteris- tics of the distal coronary venous system VAs share common features with VAs arising from the aortic cusps or the endocardial left ventricular outflow tract (LVOT) beneath the cusps [2, 8–11]. Pre-interventional ECG screening of idiopathic VAs for possible epicardial origin is crucial for a targeted ablation. In this study, the ECG and electrophysiological characteristics of VAs originating from the distal great cardiac vein (GCV) were compared with VAs successfully ablated from the left coronary cusp (LCC) or the subvalvular LVOT region. * Correspondence: doctorwushulin@163.com 2Cardiovascular Department, Guangdong Cardiovascular Institute, Guangdong General Hospital & Guangdong Academy of Medical Sciences, Guangzhou 510010, Guangdong, China Full list of author information is available at the end of the article Methods Study population In this observational study, a total of 980 patients underwent radiofrequency catheter ablation (RFCA) of symptomatic © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Li et al. BMC Cardiovascular Disorders (2019) 19:90 Page 2 of 8 idiopathic VAs at the Guang-dong General Hospital be- tween January 2012 and August 2014. All patients were re- fractory to one or two antiarrhythmic drugs. Structural heart disease was ruled out in all patients by means of trans- thoracic echocardiography, exercise stress test or coronary angiography (CAG) and cardiac magnetic resonance im- aging in selected cases. Based on the successful ablation site, patients with VAs from the distal GCV, LCC or the subvalv- ular LVOT area were included in the present study. All pa- tients provided written informed consent before the procedure, and the study was approved by the institutional review board of Guang-dong General Hospital. ECG analysis The 12-lead ECGs were analyzed at a paper speed of 100 mm/s, and signals were amplified at 10 mm/mv. ECG ana- i. QRS lysis was focused on the following parameters: morphology including bundle-branch block pattern and axis deviation; ii. QRS duration; iii. R wave amplitude in leads II and III; iv. R wave amplitude ratio in leads III to II (III/II); v. QS wave amplitude in leads aVL and aVR; vi. QS wave amplitude ratio in leads aVL to aVR (aVL/aVR); vii. The presence of pseudo-delta wave [12]; viii. Intrinsicoid deflection time (IDT) defined as the interval measured from the earliest ventricular activation to the peak of the R wave in V2; ix. Maximum deflection index (MDI) de- fined as the interval measured from the beginning of the QRS complex to the maximum deflection in the precor- dial leads divided by the QRS duration [1]; and ix. The presence of a notch at downslope of an initial q wave re- sembling a “w” pattern in lead I. Electrophysiological study and ablation The electrophysiological study (EPS) was performed with the patients in a fasting non-sedated state. Antiarrhythmic drugs were discontinued for at least five half-lives prior to the study. A 3.5-mm tip irrigated ablation catheter (Navi-s- tar, Biosense Webster, Diamond Bar, CA, USA, or Cool Path, IBI, St. Jude Medical, Irvine, CA, USA) compatible with a three-dimensional mapping system (CARTO3, Bio- sense Webster, Inc. or Ensite NavX Velocity, St. Jude Med- ical) was advanced via percutaneous access through a femoral artery to the aortic root and LVOT region for mapping and ablation. An additional long sheath was placed in the right femoral vein to access the coronary sinus (CS) for epicardial mapping and ablation, if neces- sary. A diagnostic catheter with 10 electrodes (Biosense Webster, Inc. or St. Jude Medical) was introduced into the CS as deeply as possible to map the GCV. Activation times were measured from the onset of the earliest local bipolar electrogram to the earliest onset of the QRS complex in any of the 12 ECG leads. Pace mapping was additionally performed at the same site. If spontaneously, clinical arrhythmias failed to occur programmed ventricular stimulation was performed. If VAs were not inducible at baseline, intravenous iso- proterenol infusion (2 to 5 μg/min) was administered to provoke clinical arrhythmias. Radiofrequency (RF) en- ergy applications (30 W, temperature limit of 43°Cand ir- rigation rate of 17 ml/min) were stopped if the VAs did not terminate within 20 s. Ablation was not recom- mended to perform first in the GCV to avoid the risk of injury to coronary artery even the activation of CS cath- eter preceded the QRS onset by more than 20 ms. If ab- lation within the aortic cusps or endocardial at the LVOT failed to eliminate the VAs, then additional map- ping was performed within the coronary venous system. If the earliest activation within the GCV preceded the QRS onset by more than 20 ms and pacing from that site produced an acceptable QRS match (> 11/12 leads), ab- lation within the GCV was attempted (Fig. 1). Prior to ablation within the GCV, CAG was performed to access the potential risk of coronary artery damage. RF energy applications were never delivered when the distance be- tween the earliest activation site and an epicardial coronary artery was less than 5 mm. RF energy was ap- plied at the earliest activation site in the GCV using an irrigated ablation catheter with a power of 20 W, temperature limit of 43 °C and an irrigation rate of 30 ml/min. The RF power was titrated to a maximal of 30 W. The end-point of the catheter ablation was the elim- ination and non-inducibility of VAs during an iso- infusion and burst pacing from the right proterenol ventricle (to a cycle length as short as 300 ms). Following GCV ablation, CAG was repeated to ensure that there was no evidence of injury to the coronary artery. The Fig. 1 Flowchart displaying the ablation strategy. VAs, ventricular arrhythmias; GCV, great cardiac vein; CS, coronary sinus; LCC, left coronary cusp; AMC, aortomitral continuity; CAG, Coronary angiography Li et al. BMC Cardiovascular Disorders (2019) 19:90 Page 3 of 8 site of origin of VAs was determined based on successful elimination with RF energy application. Follow-up After the procedure, continuous ECG monitoring was performed for 24 h. Antiarrhythmic drugs were discon- tinued in the absence of arrhythmia recurrence. Patients were evaluated every 3 months in an outpatient clinic. 24-h Holter monitoring was carried out in every visit. If patients report any chest discomfort symptoms must undergo treadmill exercise testing or event CAG again to eliminate the possibility of coronary artery damage. Statistical analysis Continuous variables (expressed as mean ± SD) were com- pared with Student t test (or Wilcoxon when necessary). Categorical variables were compared with chi-squared or Fisher exact test, as appropriate. A probability value < 0.05 was considered statistically significant. Results Patient characteristics The final population consisted of 39 patients (35 males, mean age 51 ± 23 years). Among these patients, 15 were successfully ablated at the distal GCV, 15 at the LCC and 9 at the subvalvular region (below the LCC). The baseline characteristics of the study cohort are depicted in Table 1. None of the patients displayed a family his- tory of sudden cardiac death or a history of previous cardiovascular disease and decreased left ventricular ejection fraction. None of the patients had failed previ- ous ablation procedures. The dominant arrhythmia was symptomatic VAs in all GCV group patients with high burden (23 ± 6%). ECG characteristics The ECG characteristics of VAs with to respect to suc- cessful site of ablation are shown in Table 1. All GCV VAs displayed a right bundle branch block (RBBB) pat- tern with inferior axis (Fig. 2). In particular, GCV VAs exhibited positive QRS forces in all precordial leads with no S waves in leads V5 and V6. An Rs pattern in leads V2-V4 was evident in all but two patients. VAs with a GCV origin exhibited more commonly a positive QRS complex concordance in all precordial leads, dominant negative waves in leads I and aVL, increased IDT, higher MDI and wider QRS duration (172 ± 19 ms) compared to LCC and subvalvular sites (p < 0.05). There were no significant differences of the (III/II) R wave amplitude ratio and the (aVL/aVR) QS wave amplitude ratio between GCV group and LCC or subvalvular groups (p > 0.05). A trend towards a higher (aVL/aVR) QS wave amplitude ratio was seen in VAs originating from the GCV compared to VAs arising from the LCC (1.72 ± 0.39 vs. 1.51 ± 0.2, p < 0.05). A “w” pattern in lead I (Fig. 2) was present in 12 out of 15 (80%) VAs originating from the GCV compared to VAs arising from the other two sites (P < 0.01). An rS Table 1 Differences in baseline patient characteristics and VAs electrocardiographic morphologies GCV group (n = 15) LCC group (n = 15) Subvalvular region group (n = 9) P value* P value# Characteristic Male, n (%) Age (years) VAs burden (%) LVEF (%) Structural heart disease (%) Limb leads 14(93%) 51 ± 23 23 ± 6 64.8 ± 4.3 0 172 ± 19 110 ± 16 52 ± 11 R-wave amplitude in lead II (mV) 1.45 ± 0.46 R-wave amplitude in lead III(mV) 1.74 ± 0.53 III/II R-wave amplitude ratio 1.21 ± 0.12 Q-wave amplitude in lead aVR(mV) 0.65 ± 0.22 Q-wave amplitude in lead aVL(mV) 1.05 ± 0.31 aVL/aVR Q-wave amplitude ratio 1.72 ± 0.39 Duration (ms) QRS duration IDT Delta wave (ms) MDI 13(87%) 48 ± 21 20 ± 6 63.2 ± 4.8 0 2.04 ± 0.36 2.22 ± 0.29 1.11 ± 0.09 1.26 ± 0.23 1.32 ± 0.12 1.51 ± 0.2 148 ± 5 96 ± 9 46 ± 7 8(89%) 50 ± 14 23 ± 5 62.0 ± 5.1 0 2.28 ± 0.20 2.72 ± 0.30 1.19 ± 0.07 0.97 ± 0.10 1.58 ± 0.19 1.67 ± 0.43 152 ± 6 93 ± 8 48 ± 8 0.71 0.13 0.34 < 0.05 < 0.05 0.06 < 0.05 < 0.05 0.07 < 0.05 < 0.05 0.86 < 0.05 0.81 0.85 0.17 < 0.05 < 0.05 0.65 < 0.05 < 0.05 0.77 < 0.05 < 0.05 0.16 < 0.05 VAs ventricular arrhythmias, GCV great cardiac vein, LCC left coronary cusp, IDT intrinsicoid deflection time, MDI maximum deflection index. *Comparison between GCV and LCC group, #Comparison between GCV and Subvalvular region group 0.64 ± 0.05 0.60 ± 0.04 0.61 ± 0.02 Li et al. BMC Cardiovascular Disorders (2019) 19:90 Page 4 of 8 Fig. 2 Twelve-lead ECG of the ventricular arrhythmias arising from the distal great cardiac vein are shown. The presence of a notch at downslope of an initial q wave resembling a “w” pattern in lead I is shown (arrow) pattern in lead I was evident in the remaining 3 GCV patients. A QS or rS pattern in lead I was present in pa- tients with LCC and endocardial LVOT origins. Mapping and ablation Access to the left ventricle for mapping and ablation was performed retrogradely. Among the 15 patients with GCV VAs (Fig. 3), mapping and ablation of proceeded via right femoral vein in 13 patients at GCV and via the left subclavian vein in two patients. The targeted poten- tial at the final ablation site preceded QRS complex on- set of clinical VAs by 25 ± 5 ms (Figs. 3 and 4). Of these 15 patients, Pace mapping at the successful ablation site in GCV was obtained in only 10 patients. During sinus rhythm, a clear atrial potential was al- ways observed before the ventricular activation (Fig. 4). During VAs, a pre-potential was observed at the initial ventricular activation in 5 of 15 patients (33%). 12-lead ECG and local electrogram of successful ablation site in three representative cases of the distal GCV, LCC, and subvalvular LVOT were compared (Additional file 1: Fig- ure S1). Mean time to clinical VAs disappearance was 6 ± 3 s. Among 15 patients with GCV VAs, impedance in the epicardium was much higher than LV endocardium (187 ± 9 Ω vs. 117 ± 9 Ω, P < 0.05). The VAs was success- fully abolished with a mean of 2.3 ± 1.4 RF applications. No ST segment or T wave changes were observed during ablation in any of the 15 patients. Procedure time was 127 ± 57 min with fluoroscopy time of 12.3 ± 6.4 min. The distance between the successful ablation site in the GCV and the best (but failed) site in the endocardium was 14 ± 4 mm. In the initial study population, 24 patients exhibited the earliest activation site at the distal GCV. Nine of these patients were excluded from the current analysis due to failure to advance the ablation catheter at the dis- tal GCV (n = 4), failure to eliminate the VAs with RF ap- plications (n = 3) and abandoned ablation because of the close proximity to a major coronary artery (n = 2). A schematic diagram shows the sites of origin of 15 GCV-VAs determined by the CAG of the left coronary artery and three-dimensional mapping (Additional file 2: Figure S2). Follow-up After an average of 36 ± 24 months follow up, 14 (93.3%) patients were free from VAs recurrence. No complica- tions occurred immediately after the procedure or dur- ing follow up. None of the patients reported chest pain or discomfort symptoms at follow-up and therefore sub- sequent CAG was not repeated. Discussion Major findings The main findings of the present study are the following: i. a “w” pattern in lead I distinguish a GCV origin from LCC or endocardial origins; ii. GCV VAs display an increased IDT and MDI compared to LCC or endocardial sites. ECG morphology criteria The ECG characteristics of VAs originating from the distal GCV varies. In the proximal segment, VAs display a RBBB morphology and in the distal segment as well in the anter- ior interventricular vein a LBBB morphology [2, 8, 11]. In our series, all cases displayed a RBBB, and therefore the proximal segment of distal GCV appears to be the accur- ate site of origin. Previous studies addressing the QRS morphology in lead I have demonstrated more commonly an rS and less commonly a QS pattern [2, 8, 11]. In this study, the presence of a notch at downslope of the initial q wave resembling a “w” pattern in lead I was present in 80% of VAs originating from the distal GCV compared to none of VAs arising from the LCC and the endocardial LVOT areas. GCV VAs are mostly located at epicardium of superior-lateral LV. The presence of initial negative forces (q wave) in lead I is indicative of an initial Li et al. BMC Cardiovascular Disorders (2019) 19:90 Page 5 of 8 Fig. 3 Example of a successful ablation of premature ventricular contractions originating from the distal GCV. a Activation time at the GCV (30 ms before QRS onset). b Pace-map at the GCV demonstrating an excellent match. c Corresponding fluoroscopic view (RAO 30°and LAO 45°) of the ablation catheter (ABL) positions during coronary angiography of the left coronary artery (Additional files 3 and 4: Video S1 and S2). The ablation catheter tip is ≥5 mm distant from any coronary artery. d The anatomical distance from the earliest activation site at the distal GCV to the closest anatomical point at the LCC and the LV endocardium were 24 mm and 9 mm, respectively. ABL d (p), the distal and proximal electrode pairs of the ablation catheter; GCV, great cardiac vein; CS, coronary sinus; RAO, right anterior oblique projection; LAO, left anterior oblique projection; LCC, left coronary cusp, LV, left ventricle rightward activation of the LV base from an epicardial ori- gin. VAs originating from the epicardium showed a q wave in lead I as initial activation which means that the net vec- tor of the global activation pattern from left to right and superior to inferior [13]. The presence of a q wave in lead I and absence of q waves in the inferior leads appear to be a very sensitive criterion for identifying an epicardial site of origin [13]. Therefore, a “w” pattern in lead I raises a high degree of suspicion for a distal GCV origin. A trend towards a higher (aVL/aVR) QS wave amplitude ratio was seen in VAs originating from the GCV compared to VAs arising from the LCC, which is consistent with previous findings [8]. Finally, we showed that distal GCV VAs display an increased MDI and IDT compared to LCC or endocardial sites. A MDI > 0.59 has been previously shown to identify VAs of an epicardial origin [1, 13]. In our study, all patients had a MDI > 0.60 (100% sensitivity). Previous reports with VAs arising from the distal GCV have consistently demonstrated a higher MDI in relation to endocardial sites [2, 14]. Li et al. have demonstrated that pseudo wave time and IDT were significantly longer for VAs of distal GCV origin compared to VAs from the cusps or the endocardium [14]. Mapping and ablation In our study, catheter ablation of VAs arising from the dis- tal GCV was highly effective, in the cases where RF appli- cations could be successfully delivered. However, catheter Li et al. BMC Cardiovascular Disorders (2019) 19:90 Page 6 of 8 Fig. 4 Fluoroscopy, activation, and 3-dimensional mapping in a 55-year-old man with successful ablation of premature ventricular contractions at the distal of great coronary vein (GCV). a (A1 and A2): Corresponding fluoroscopic views (RAO 30°and LAO 45°) of the ablation catheter (ABL) at the successful ablation site at the distal of GCV during left coronary angiography. b surface ECG and intracardiac recordings from the mapping catheter (ABL) at the site of earliest ventricular activation at the distal GCV. Note that the local potential precedes the QRS by 20 ms during ventricular extrasystoles. c (C1 and C2): Electroanatomic maps of the same patient demonstrating the successful ablation point (red tag) at the distal GCV. d Termination of PVCs within 4 s of radiofrequency energy application at the distal GCV. Blue point indicates left main coronary artery. L, left coronary cusp. CS, coronary sinus catheter; LAO, left anterior oblique; LAD, left anterior descending artery; LAO, left anterior oblique; LCX, left circumflex artery ablation of VAs arising from the distal GCV is challenging. We showed that in a significant number of patients (9/24) certain limitations do not allow catheter ablation. First, it is not always possible to advance the ablation catheter to the distal GCV. Second, the high impedance and/or the limited cooling from the blood flow may lead to ineffective RF applications. Third, the close proximity to the coronary arteries does not permit a safe ablation procedure. Nagashima et al. reported occlusion requiring stenting of a marginal branch of the circumflex artery in 2 who re- ceived RF energy at sites 2 mm and 5–7 mm distant from the vessel, respectively [11]. A distance of 5–12 mm between the ablation catheter and the coronary artery may reduce the risk of injury[15]. VAs displayed a RBBB morphology with inferior axis could safely be eliminated in the GCV, but sometimes Li et al. BMC Cardiovascular Disorders (2019) 19:90 Page 7 of 8 ablation can be challenging because of the close proxim- ity to the coronary arteries and failure to advance the ablation catheter at the distal GCV. RFCA within GCV may be limited by a high impedance. Potential risks (Cor- onary vasospasm, coronary artery damage or perforation) of ablation within the GCV are much more than in the endocardium. Repeated CAG and careful titration of irri- gated radiofrequency energy for successful ablation with- out acute complications. Even no complications were reported in this study, we recommend catheter ablation performed within the distal of GCV with strong clinical indications including symptomatic high burden VAs, arrhythmia induced cardiomyopathy and special profes- sion such as pilot and sports athletes. Besides, catheter ablation was performed in the GCV only after attempting to eliminate the VAs in the endocardium has failed. A few cases of GCV VAs might be eliminated through LVOT endocardium ablation. Because the breakout site of GCV VAs maybe located adjacent LV epicardium, but the true origin maybe located adjacent LV endocardium. So these VAs originating from the GCV could be ablated from the LCC/adjacent LV endocardium. We could not determine the true origin site of the VAs even after they were eliminated. As a result, the site of origin of VAs was determined based on successful elimination in our study. Study limitations This is a single-center observational and retrospectively study including a small number of patients with rela- tively uncommon VAs. Although none of the patients reported symptoms during follow-up, coronary artery damage in terms of stenosis cannot be excluded. Due to the signal interference of our catheter labs, the unipolar electrogram was very difficult to identify during mapping and ablation, so the data about unipolar electrogram was not available in this study. Conclusions VAs originating from the GCV represents a rare sub- group of idiopathic VAs. Among patients with outflow tract VAs exhibiting a RBBB morphology, inferior axis, and QRS complex predominantly positive in all precor- dial leads, a “w” pattern in lead I, an increased IDT, a higher MDI and a wider QRS duration predicted a GCV origin compared to LCC or endocardial sites. Additional files Additional file 1: Figure S1. Example of three successful ablations of premature ventricular contractions originating from the distal GCV, LCC, and subvalvular LVOT, respectively. A: Activation time (30 ms before QRS onset) at the GCV of case 1 with VAs originated from the distal GCV. B: Activation time (24 ms before QRS onset) at the LCC of case 2 with VAs originated from LCC. C: Activation time (28 ms before QRS onset) beneath LCC of case 3 with VAs originated from the distal GCV. B: Activation time (24 ms before QRS onset) at the LCC of case 2 with VAs originated from subvalvular LVOT. ABL d (p), the distal and proximal electrode pairs of the ablation catheter; GCV, great cardiac vein; LCC, left coronary cusp; LVOT, left ventricular outflow tract. (JPG 162 kb) Additional file 2: Figure S2. A schematic representation of the anatomic distribution of successful ablation sites in the 15 GCV-VAs. Left lateral view of the heart. Black circles represented the successful ablation sites. GCV, great cardiac vein; LAD, left anterior descending; CS, coronary sinus; LV, left ventricle; LAA, left atrial appendage; LSPV, left superior pulmonary vein; LIPV, left inferior pulmonary vein; RV, right ventricle. (JPG 177 kb) Additional file 3: Video S1. Fluoroscopic view (LAO 45°) of the ablation catheter (ABL) positions during coronary angiography of the left coronary artery. (AVI 4868 kb) Additional file 4: Video S2. Fluoroscopic view (RAO 30°) of the ablation catheter (ABL) positions during coronary angiography of the left coronary artery. (AVI 4100 kb) Abbreviations CAG: Coronary angiography; CS: Coronary sinus; EPS: Electrophysiologic study; GCV: Great cardiac vein; IDT: Intrinsicoid deflection time; LCC: Left coronary cusp; LVEF: Left ventricular ejection fraction; LVOT: Left ventricular outflow tract; MDI: Maximum deflection index; PVC: Premature ventricular contraction; RBBB: Right bundle branch block; RF: Radiofrequency; RFCA: Radiofrequency catheter ablation; RV: Right ventricular; VAs: Ventricular arrhythmias Acknowledgements Not applicable. Funding No funding was obtained for this study. Availability of data and materials The datasets used and analysed during the current study are available from the corresponding author on reasonable request. Authors’ contributions TL and QX have done the patient’s follow-up and drafted the manuscript. TL, XZZ, YMX and HTL have done the ablation and provided the photographs of the ablation. YFL, KPL and SLW have helped on the manuscript drafting and revision. All authors read and approved the final manuscript. Ethics approval and consent to participate The retrospective and observational study was approved by the Guangdong Cardiovascular Institute, Guangdong General Hospital, and performed in accordance with the Declaration of Helsinki. All patients provided written informed consent. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Author details 1Arrhythmia Department, Fuwai Hospital Chinese Academy of Medical Sciences, Shenzhen 518057, Guangdong, China. 2Cardiovascular Department, Guangdong Cardiovascular Institute, Guangdong General Hospital & Guangdong Academy of Medical Sciences, Guangzhou 510010, Guangdong, China. 3Second Department of Cardiology, Laboratory of Cardiac Electrophysiology, Evangelismos General Hospital of Athens, Athens, Greece. Li et al. BMC Cardiovascular Disorders (2019) 19:90 Page 8 of 8 Received: 7 October 2018 Accepted: 24 March 2019 References 1. Daniels DV, Lu YY, Morton JB, Santucci PA, Akar JG, Green A. Idiopathic epicardial left ventricular tachycardia originating remote from the sinus of Valsalva: electrophysiological characteristics, catheter ablation, and identification from the 12-lead electrocardiogram. Circulation. 2006;113:1659–66. 2. Mountantonakis SE, Frankel DS, Tschabrunn CM, Hutchinson MD, Riley MP, 3. 4. 5. 6. 7. 8. 9. 10. Lin D. Ventricular arrhythmias from the coronary venous system: prevalence, mapping, and ablation. Heart Rhythm. 2015;12:1145–53. Houmsse M, Daoud EG. Techniques to ablate premature ventricular ectopy arising from the coronary sinus system. Pacing Clin Electrophysiol. 2011;34:e74–7. Baman TS, Ilg KJ, Gupta SK, Good E, Chugh A, Jongnarangsin K. Mapping and ablation of epicardial idiopathic ventricular arrhythmias from within the coronary venous system. Circ Arrhythm Electrophysiol. 2010;3:274–9. Kaseno K, Tada H, Tanaka S, Goto K, Yokokawa M, Hiramatsu S. Successful catheter ablation of left ventricular epicardial tachycardia originating from the great cardiac vein: a case report and review of the literature. Circ J. 2007;71:1983–8. Obel OA, D'Avila A, Neuzil P, Saad EB, Ruskin JN, Reddy VY. Ablation of left ventricular epicardial outflow tract tachycardia from the distal great cardiac vein. J Am Coll Cardiol. 2006;48:1813–7. Hirasawa Y, Miyauchi Y, Iwasaki YK, Kobayashi Y. Successful radiofrequency catheter ablation of epicardial left ventricular outflow tract tachycardia from the anterior interventricular coronary vein. J Cardiovasc Electrophysiol. 2005; 16:1378–80. Jauregui AM, Campos B, Park KM, Tschabrunn CM, Frankel DS, Park RE. Ablation of ventricular arrhythmias arising near the anterior epicardial veins from the left sinus of Valsalva region: ECG features, anatomic distance, and outcome. Heart Rhythm. 2012;9:865–73. Li YC, Lin JF, Li J, Ji KT, Lin JX. Catheter ablation of idiopathic ventricular arrhythmias originating from left ventricular epicardium adjacent to the transitional area from the great cardiac vein to the anterior interventricular vein. Int J Cardiol. 2013;167:2673–81. Kimura T, Takatsuki S, Fukumoto K, Nishiyama N, Aizawa Y, Miyoshi S. Idiopathic ventricular tachycardia cured by radiofrequency application from the distal great cardiac vein and the left coronary cusp. Heart Lung Circ. 2014;23:193–6. 11. Nagashima K, Choi EK, Lin KY, Kumar S, Tedrow UB, Koplan BA. Ventricular arrhythmias near the distal great cardiac vein: a challenging arrhythmia for ablation. Circ Arrhythm Electrophysiol. 2014;7(5):906–12. 12. Berruezo A, Mont L, Nava S, Chueca E, Bartholomay E, Brugada J. Electrocardiographic recognition of the epicardial origin of ventricular tachycardias. CIRCULATION. 2004;109:1842–7. 13. Valles E, Bazan V, Marchlinski FE. ECG criteria to identify epicardial ventricular tachycardia in nonischemic cardiomyopathy. Circ Arrhythm Electrophysiol. 2010;3:63–71. Li JW, Chen XL, Li YC, Chen XX, Chen XS, Lin JF. Distinct ECG characteristics of idiopathic ventricular arrhythmias originating from four regions of left coronary veins. Int J Cardiol. 2014;175:181–2. Sosa E, Scanavacca M, D'Avila A. Transthoracic epicardial catheter ablation to treat recurrent ventricular tachycardia. Curr Cardiol Rep. 2001;3:451–8. 14. 15.
10.1186_s12872-020-01654-3	Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 https://doi.org/10.1186/s12872-020-01654-3 R E S E A R C H A R T I C L E Open Access Impact of plasma potassium normalization on short-term mortality in patients with hypertension and hypokalemia or low normal potassium Maria Lukács Krogager1* Anders Bonde5, Jesper Q. Thomassen6, Gunnar Gislason7,8,9, Manan Pareek2,10,11 and Kristian Kragholm1,4,12 , Peter Søgaard1, Christian Torp-Pedersen2, Henrik Bøggild3,4, Christina Ji-Young Lee2,5, Abstract Background: Hypokalemia is common in patients treated with antihypertensive drugs, but the impact of correcting hypokalemia is insufficiently studied. We examined the consequences of hypokalemia and borderline hypokalemia correction in patients with hypertension. Methods: We identified 8976 patients with hypertension and plasma potassium concentrations ≤3.7 mmol/L within 100 days from combination antihypertensive therapy initiation. The first measurement between 6 and 100 days after the episode with potassium ≤3.7 mmol/L was retained. We investigated all-cause and cardiovascular mortality within 60-days from the second potassium measurement using Cox regression. Mortality was examined for seven predefined potassium intervals derived from the second measurement: 1.5–2.9 mmol/L (n = 271), 3.0–3.4 mmol/L (n = 1341), 3.5–3.7 (n = 1982) mmol/L, 3.8–4.0 mmol/L (n = 2398, reference), 4.1–4.6 mmol/L (n = 2498), 4.7–5.0 mmol/ L (n = 352) and 5.1–7.1 mmol/L (n = 134). Results: Multivariable analysis showed that potassium concentrations 1.5–2.9 mmol/L, 3.0–3.4 mmol/L, 4.7–5.0 mmol/L and 5.1–7.1 mmol/L were associated with increased all-cause mortality (HR 2.39, 95% CI 1.66–3.43; HR 1.36, 95% CI 1.04–1.78; HR 2.36, 95% CI 1.68–3.30 and HR 2.62, 95% CI 1.73–3.98, respectively). Potassium levels <3.0 and > 4.6 mmol/L were associated with increased cardiovascular mortality. The adjusted standardized 60-day mortality risks in the seven strata were: 11.7% (95% CI 8.3–15.0%), 7.1% (95% CI 5.8–8.5%), 6.4% (95% CI 5.3–7.5%), 5.4% (4.5–6.3%), 6.3% (5.4–7.2%), 11.6% (95% CI 8.7–14.6%) and 12.6% (95% CI 8.2–16.9%), respectively. Conclusions: Persistent hypokalemia was frequent and associated with increased all-cause and cardiovascular mortality. Increase in potassium to levels > 4.6 mmol/L in patients with initial hypokalemia or low normal potassium was associated with increased all-cause and cardiovascular mortality. Keywords: Hypokalemia, Borderline hypokalemia, Hypokalemia correction, Mortality, Low potassium. * Correspondence: lkcsmaria@yahoo.com; maria.krogager@rn.dk 1Department of Cardiology, Aalborg University Hospital, Aalborg, Denmark Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 2 of 11 Novelty and Significance What is new? (cid:1) Correcting plasma potassium concentrations ≤3.7 mmol/L to levels between 3.5–4.6 mmol/L was associated with improved short-term prognosis What is relevant? (cid:1) Increased mortality risk was observed in patients who initially had borderline hypokalemia, partly because they developed hypokalemia. This emphasizes that potassium supplementation might be relevant in patients with low normal potassium concentrations (cid:1) Correcting hypokalemia and borderline hypokalemia shortly was associated with good prognosis (cid:1) Low potassium concentrations have previously been associated with arrhythmogenesis and increased mortality risk in patients with hypertension. Summary In this register based study we investigated the impact of correcting hypokalemia and borderline hypokalemia on 60-day mortality among 8976 patients treated with combination antihypertensive therapy. We observed that: (1) persistent hypokalemia was common and asso- ciated with increased all-cause and cardiovascular mor- tality; (2) Increase in potassium to levels > 4.6 mmol/L in patients with initial hypokalemia or low normal potas- sium was associated with increased all-cause and cardio- vascular mortality; (3) Among patients with borderline hypokalemia initially, development of hypokalemia or hyperkalemia was associated with increased mortality risk; (4) Correcting hypokalemia associated with in- creased survival. Background Several common clinical conditions and drugs are known to cause or precipitate hypokalemia [1]. Among patients with hypertension, thiazides are the antihyper- tensive drugs most frequently associated with hypokal- emia [2–4]. We recently demonstrated a U-shaped relationship be- tween potassium levels and mortality among patients with hypertension. We observed an increased mortality risk even in patients with low and high normal serum potassium concentrations, suggesting a narrower than previously thought normal interval for potassium of 4.1– 4.7 mmol/L. [5] However, at present there is no evidence regarding the consequences of potassium normalization in patients with hypertension and hypokalemia. There- fore, it is essential to examine how correction and even overcorrection of hypokalemia affect prognosis in pa- tients with hypertension. Using Danish national registers, we investigated the 60-day mortality among patients with hypertension and hypokalemia or low normal potassium concentrations, according to their subsequent plasma potassium concen- trations measured within 6–100 days following the initial episode with low potassium levels. Methods Data sources In Denmark, a unique and personal identification num- ber is allocated to all individuals at the time of birth or immigration. This unique identifier allows linkage of health and administrative data at the individual level [6] and ensures nearly complete follow-up. We used anon- ymized data from five different registers made available by Statistics Denmark after central encryption of the unique identifiers [7]. An overview of the registers used in this study is available in Supplementary Table S1. In Denmark, register-based studies using anonymized data provided by Statistics Denmark are not warranted ap- proval from the ethics committee. Study population We defined hypertension as redemption of minimum two antihypertensive agents in two consecutive quarters. This definition has previously been validated [8]. Pa- tients entered the present study in the second quarter, referred to as the date of hypertension. An overview of the Anatomical Therapeutic Chemical Classification System (ATC) codes used to identify patients with hypertension is available in Supplementary Table S2. We required a plasma potassium measurement ≤3.7 mmol/L within 100 days from the date of hypertension for inclu- sion. The first measurement within this time interval was retained and referred to as the first potassium meas- urement (K1). The second potassium measurement (K2) was identified in the interval 6–100 days from K1 and the first draw within this timeframe was retained. We did not include potassium concentrations within 1–5 days from K1 as, potassium disarrays are usually cor- rected within a few days, regardless of the strategies ap- plied. Patients below 18 years of age were excluded from the study. Supplementary Figure S1 illustrates the popu- lation flowchart. Comorbidities and medication We identified comorbidities and medications regarded as confounders when studying the association between changes in potassium levels and short-term mortality. The following comorbidities dated up to 5 years before the index date (K2 date) were identified: hospitalization for heart failure, ischemic heart disease, stroke, chronic Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 3 of 11 obstructive pulmonary disease, chronic liver disease, dia- betes mellitus, inflammatory bowel disease and malig- nancy. Furthermore, patients with a past history of primary adrenal insufficiency, primary hyperaldosteron- ism, and diabetes insipidus were excluded. The Inter- national Classification of Disease (ICD) codes used to identify above-mentioned comorbidities are shown in Supplementary Table S3. We used the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [9] to calculate renal function, and an estimated glom- erular filtration rate (eGFR) < 30 mL/min/1.73 m2 de- insufficiency. Patients were scribed significant excluded if no creatinine concentrations were available the same day as or within a week from the index date. Patients with missing serum sodium measurements on the index date were also excluded. renal Prescriptions redeemed up to 90-days before the index date were identified for the following drugs: po- tassium supplements, non-steroidal anti-inflammatory drugs, corticosteroids, laxatives, xanthines, and anti- microbials. See Supplementary Table S3 for relevant ATC codes. Exposure variable Serum and plasma measurements yield similar results, but for serum samples there is a risk of contamination with potassium from burst platelets during coagulation in the range of 0.1–0.5 mmol/L due to non-standard sample handling [10]. Therefore, we only used plasma potassium measurements. There is not a consensus on the normal plasma potas- sium interval, as it can vary from population to popula- tion. Supplementary Table S4 gives an overview on the three most used reference intervals in serum and plasma originating from different populations. We defined hypo- kalemia as plasma potassium concentrations below 3.5 mmol/L and borderline hypokalemia as potassium levels within the interval 3.5–3.7 mmol/L. Hyperkalemia was defined as potassium levels above 4.6 mmol/L. [11] For K2, potassium intervals were constructed: 1.5–2.9 mmol/L, 3.0–3.4 mmol/L, 3.5–3.7 4.7–5.0 mmol/L, mmol/L and 5.1–7.1 mmol/L. Plasma potassium interval K: 3.8–4.0 mmol/L was used as the reference for statis- tical analyses. We chose this interval as the reference group because it had one of the largest number of pa- tients and lowest mortality rate. 4.1–4.6 mmol/L, 3.8–4.0 mmol/L, predefined seven Outcome The primary outcome was all-cause mortality within 60 days from K2. The secondary outcome was presumed cardiovascular death within 60 days from K2. Statistical analyses Categorical variables were presented as counts and per- centages, and continuous variables as median with corre- sponding 25th and 75th percentiles. Differences between variables were compared using chi-squared and Kruskal- Wallis tests, as appropriate. To illustrate survival probability, Kaplan Meier curves were plotted for the seven potassium intervals. A re- stricted cubic spline curve was constructed to investigate the relationship between potassium as a continuous vari- able and absolute mortality risk in an age, sex, comor- bidity and drug standardized population. Cox proportional hazard modeling was used to analyze the association between the seven predefined potassium intervals and 60-day all-cause and presumed cardiovas- cular mortality. Based on the Cox regression principle, we modelled an average effect to estimate the 60-day ab- solute risk of all-cause mortality, with potassium interval 3.8–4.0 mmol/L as reference. The multivariable model was adjusted for: age, sex, serum sodium, renal insufficiency, malignancy, heart fail- ure, chronic liver disease, chronic obstructive pulmonary disease, diabetes mellitus, stroke, atrial flutter/fibrilla- tion, ischemic heart disease, inflammatory bowel disease, antihypertensive therapy, corticosteroids, antimicrobials, non-steroidal anti-inflammatory drugs, xanthines, laxa- tives, and potassium supplements. The proportional hazard assumption was tested by plotting Schoenfeld re- siduals and was not violated. Interactions on mortality were tested by comparing the likelihood ratio of the Cox regression model with and without the interaction term. The following variables were tested for interaction with plasma potassium on mortality: age, sex, and renal insuf- ficiency. A two-sided p-value < 0.01 was considered statistically significant for interactions. We found no sta- tistically significant interactions. For other analyses, a two-sided p-value < 0.05 was considered statistically sig- nificant. Linearity of age on mortality was also assessed through a likelihood ratio test comparing a linear description with a categorical one. Age was found to violate linearity and was included as a categorical variable with five levels, using cut-off values from every 20th percentiles (55, 64, 72, 79 and 101 years). Hazard ratios (HR) and absolute risks (AR) were estimated with 95% confidence intervals (95% CI). All data management and analyses were performed using SAS, version 9.4 and R, version 3.5.0 [12]. Results Demographics We identified 8976 patients treated with combination antihypertensive therapy who had potassium concentra- tions ≤3.7 mmol/L within the first 100 days from com- bination therapy initiation. Baseline characteristics for Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 4 of 11 the cohort stratified on the seven predefined K2 intervals are presented in Table 1. Females accounted for 53% of the total population and median age was 68.3 years (range 18.2–100.8 years). Of the patients with borderline hypokalemia at K1, 13% developed hypokalemia and 5.7% hyperkalemia at K2. As for patients with hypokal- emia at K1, we observed that 28.5% remained hypokal- emic at the second blood draw and 4.8% developed hyperkalemia. Approximately half of the population was hospitalized at K1 and four fifths at K2. See supplemen- tary Figure S2 displaying the distribution of K1, average of potassium measurements drawn within 1–5 days from K1, and K2. A low number of patients (n = 572) had renal insufficiency at the time of second potassium draw. Median time from K1 to K2 was 22 days (range: 6–100 days). As for diuretic treatment, thiazides were common in patients with potassium concentrations ≤3.7 mmol/L, whereas loop diuretics were more common among pa- tients with high potassium levels. Thiazide-like diuretics accounted for 4.4% of total prescriptions of the thiazides. insufficiency, Demographics stratified on survival status showed that age, renal lower sodium concentrations, hospitalization at the time K1, prior history of malig- nancy, chronic liver disease, chronic obstructive pul- monary disease, atrial flutter, heart failure, and stroke were predominant among the de- ceased (Supplementary Table S5). fibrillation/atrial 60-day survival after the second potassium measurement During 60-day follow-up after K2, 627 (7.0%) patients died, 331 from a cardiovascular cause. Mortality in the seven strata was: 14.4, 7.0, 6.3, 5.2, 6.7, 13.6 and 21.6%, respectively. The restricted cubic spline curve revealed a U-shaped relationship between potassium and mortality (Fig. 1). The results of the multivariable Cox regression, with plasma potassium 3.8–4.0 mmol/L as the reference group are shown in Fig. 2. All-cause mortality was sig- nificantly increased in patients with hypokalemia (1.5– 2.9 mmol/L HR 2.39, 95% CI 1.66–3.43 and 3.0–3.4 mmol/L HR 1.36, 95% CI 1.04–1.78) when compared with the reference. We observed a trend towards increased mortality in patients with borderline hypokal- emia and with potassium levels within the interval 4.1– 4.6 mmol/L (HR 1.24, 95% CI 0.97–1.59 and HR 1.20, 95% CI 0.95–1.51, respectively). All-cause mortality was also elevated in patients with hyperkalemia (4.7–5.0 mmol/L HR 2.36, 95% CI 1.68–3.30; 5.1–5.7 mmol/L HR 2.62, 95% CI 1.73–3.98). The univariable analysis showed similar results. We observed no interaction between K1 and K2 on 60-day mortality. Cardiovascular mortality accounted for nearly 53% of all deaths. We observed increased risk of cardiovascular death in patients with initial hypokalemia or low normal potassium levels who had potassium concentrations < 3.0 mmol/L second measurement. and > 4.6 mmol/L the at The standardized 60-day absolute risk of all-cause mortality was lowest in patients with potassium concen- trations between 3.8–4.0 mmol/L (AR 5.4, 95% CI 4.5– 6.3%, Table 2). Significant differences in risks (reported against the reference) were observed for the following potassium intervals: 1.5–2.9 mmol/L risk difference 6.3% (95% CI 2.9–9.7%); 4.7–5.0 mmol/L risk difference 6.2% (95% CI 3.2–9.3%); 5.1–7.1 mmol/L risk difference 7.2% (95% CI 2.8–11.6%). Subgroup and sensitivity analyses We performed eleven additional analyses to test the ac- curacy and robustness of the main results (Table S6). First, multivariable analysis performed on a subgroup of patients without kidney insufficiency showed that po- tassium levels within the intervals 1.5–2.9 mmol/L and 3.0–3.4 mmol/L were associated with increased mortality risk compared with the reference (3.8–4.0 mmol/L) (HR 2.33, 95% CI 1.56–3.46 and HR 1.35, 95% CI 1.02–1.79, respectively). Second, in a subpopulation without history of malig- nancy, adjusted analyses showed that potassium concen- interval 3.0–4.6 mmol/L were trations outside associated with increased risk of death compared with the reference. the Third, subgroup analysis on patients without history of heart failure and no loop diuretic prescription showed that patients with hypokalemia and hyperkalemia had an increased mortality risk compared with patients with po- tassium levels in the interval 3.8–4.0 mmol/L. Fourth, analysis performed on a subgroup of patients without ischemic heart disease showed that patients with severe hypokalemia, and hyperkalemia had increased risk short-term mortality risk when compared with the reference. Fifth, looking at patients with borderline hypokalemia at the first potassium measurement, we observed that patients who developed hypokalemia (1.5–2.9 mmol/L: HR 2.16, 95% CI 1.25–3.73; 3.0–3.4 mmol/L: HR 1.70, 95% CI 1.22–2.37), or hyperkalemia (4.7–5.0 mmol/L: HR 1.84, 95% CI 1.18–2.86; 5.1–7.1 mmol/L: HR 2.81, 95% CI 1.68–4.71) had an increased risk of death within 60-days when compared with the reference. Sixth, among patients with hypokalemia at K1, analyses showed that potassium concentrations within the inter- vals 1.5–2.9 mmol/L, 4.1–4.6 mmol/L and 4.7–5.0 mmol/ L were associated with increased short-term mortality risk. Seventh, by performing the analyses on the last avail- able potassium measurement within 6–100 days from K1 Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 5 of 11 l e u a v - p L / l o m m 1 7 – 1 5 . . L / l o m m 0 5 – 7 4 . . L / l o m m 6 4 – 1 4 . . L / l o m m 0 4 – 8 3 . . ) 4 3 1 = n n ( ) 2 5 3 = ( ) 8 9 4 2 = ( ) 8 9 3 2 = n n ( L / l o m m 7 3 – 5 3 . . ) 2 8 9 1 = n ( L / l o m m 4 3 – 0 3 . . ) 1 4 3 1 = n ( L / l o m m 9 2 – 5 1 . . ) 1 7 2 = n ( y p a r e h t e v i s n e t r e p y h i t n a n o i t a n b m o c i h t i w d e t a e r t s t n e i t a p 6 7 9 8 f o t r o h o c a n i s l a v r e t n i m u i s s a t o p a m s a p l d e n i f e d e r p t h g e i e h t o t i g n d r o c c a d e i f i t a r t s i s c h p a r g o m e D 1 e l b a T . 1 0 0 < . ) 8 7 9 , . 7 7 2 ( 9 1 7 . . ) 6 8 9 , . 2 0 2 ( 9 9 6 . . ) 9 9 9 , . 2 8 1 ( 7 9 6 . . ) 5 7 9 , . 2 9 1 ( 7 7 6 . . ) 7 7 9 , . 1 2 2 ( 7 6 . ) 8 0 0 1 , . 2 9 1 ( 7 6 . ) 9 4 9 , . 4 1 2 ( 6 0 7 . . 1 0 0 < . ) 5 2 4 ( 7 5 . ) 4 8 2 ( 8 3 . ) 9 8 4 ( 2 7 1 . ) 0 0 5 ( 0 5 2 1 . ) 2 3 5 ( 6 7 2 1 . ) 8 4 5 ( 7 8 0 1 . ) 5 6 5 ( 7 5 7 . ) 9 7 5 ( 7 5 1 . ) 2 2 1 ( 3 4 ) 7 5 ( . 2 4 1 . ) 3 5 ( 7 2 1 ) 0 6 ( . 8 1 1 ) 8 5 ( . 8 7 ) 6 9 ( . 6 2 . 1 0 0 < ) 9 4 1 , 2 1 1 ( 6 3 1 ) 6 6 1 , 4 1 1 ( 8 3 1 ) 9 5 1 , 7 0 1 ( 9 3 1 ) 1 6 1 , 1 0 1 ( 0 4 1 ) 5 5 1 , 5 0 1 ( 0 4 1 ) 9 7 1 , 7 1 1 ( 9 3 1 ) 7 5 1 , 1 1 1 ( 8 3 1 . ) 4 6 6 ( 9 8 . ) 3 3 7 ( 8 5 2 . ) 1 5 7 ( 7 7 8 1 . ) 7 3 7 ( 8 6 7 1 . ) 7 6 6 ( 2 2 3 1 . ) 2 2 5 ( 0 0 7 . ) 8 5 3 ( 7 9 i n a d e m ) e g n a r ( l e a m e F i n a d e m ) e g n a r ( L / l o m m 7 3 – 5 3 . . . 1 0 0 < . 1 0 0 < . ) 6 3 3 ( 5 4 . ) 8 4 2 ( 2 3 . ) 7 6 2 ( 4 9 . ) 7 3 1 ( 7 4 . ) 9 4 2 ( 1 2 6 ) 9 6 ( . 6 6 1 . ) 3 6 2 ( 0 3 6 . ) 5 5 ( 7 2 1 . ) 3 3 3 ( 0 6 6 ) 2 6 ( . 7 1 1 ) 9 6 ( . 7 8 ) 8 9 ( . 5 2 . ) 8 7 4 ( 1 4 6 . ) 2 4 6 ( 4 7 1 L / l o m m 5 3 < . . 1 0 0 < . ) 1 8 8 ( 8 1 1 . ) 1 6 8 ( 3 0 3 . ) 1 2 8 ( 0 5 0 2 . ) 6 7 7 ( 0 6 8 1 . ) 7 7 7 ( 8 3 5 1 . ) 4 8 7 ( 0 5 0 1 . ) 6 5 8 ( 2 3 2 5 8 7 7 2 8 2 9 7 7 7 1 i e n n i t a e r c g n i s s i m ) t n e m e r u s a e m d n o c e s ( i m u d o s m u r e S ) t m e r u s a e m d n o c e s ( ) t n e m e r u s a e m t s r i f ( m u i s s a t o p a m s a P l m u i s s a t o p d n o c e s f o e m i t e h t t a n o i t a z i l a t i p s o H t n e m e r u s a e m ) t m e r u s a e m t s r i f ( i y c n e c i f f u s n i l a n e R y c n e c i f f i u s n i e g A l a n e R x e S . 1 0 0 < ) 7 9 , 6 ( 4 1 ) 0 0 1 , 6 ( 5 3 1 . ) 0 0 1 , 6 ( 1 2 ) 0 0 1 , 6 ( 6 2 ) 0 0 1 , 6 ( 5 2 ) 0 0 1 , 6 ( 1 2 ) 7 9 , 6 ( 4 1 . 1 0 0 < . 1 0 0 < . 1 0 0 < 6 6 0 . . 1 0 0 < 1 0 0 . 1 0 0 . . 1 0 0 < . 1 0 0 < . 1 0 0 < 4 0 0 . 7 6 0 . 2 1 0 . . ) 6 1 2 ( 9 2 . ) 4 0 1 ( 4 1 . ) 9 0 2 ( 8 2 . ) 9 0 2 ( 8 2 . ) 9 6 2 ( 6 3 . ) 4 9 1 ( 6 2 ) 0 9 ( . 2 1 . ) 1 2 3 ( 3 4 . ) 1 6 2 ( 5 3 . ) 0 4 4 ( 9 5 . ) 6 9 3 ( 3 5 ) 7 9 ( . 3 1 ) 5 4 ( . 6 . ) 6 3 1 ( 8 4 . ) 0 8 ( 8 2 . ) 2 0 2 ( 1 7 . ) 9 7 1 ( 3 6 . ) 9 1 2 ( 7 7 . ) 8 0 1 ( 8 3 . ) 0 6 ( 1 2 . ) 1 4 2 ( 5 8 . ) 4 4 3 ( 1 2 1 . ) 0 9 2 ( 2 0 1 . ) 8 5 3 ( 6 2 1 . ) 4 1 1 ( 0 4 . ) 4 1 ( 5 ) 7 6 ( . 8 6 1 . ) 6 3 ( 1 9 ) 2 5 ( . 4 2 1 ) 5 2 ( . 9 5 ) 3 6 ( . 5 2 1 ) 4 3 ( . 8 6 ) 0 7 ( . 4 9 ) 7 3 ( . 0 5 . ) 1 9 1 ( 8 7 4 . ) 8 5 1 ( 5 9 3 . ) 9 7 1 ( 9 2 4 . ) 1 2 1 ( 0 9 2 . ) 6 9 1 ( 9 8 3 . ) 0 2 1 ( 8 3 2 . ) 8 8 1 ( 2 5 2 . ) 4 3 1 ( 0 8 1 . ) 1 8 1 ( 3 5 4 . ) 0 8 1 ( 2 3 4 . ) 3 6 1 ( 4 2 3 . ) 5 6 1 ( 1 2 2 . ) 8 7 ( 4 9 1 ) 0 5 ( . 4 2 1 . ) 4 1 2 ( 5 3 5 . ) 7 7 3 ( 1 4 9 . ) 0 5 2 ( 4 2 6 . ) 2 9 2 ( 9 2 7 . ) 8 2 1 ( 0 2 3 . ) 5 1 ( 8 3 ) 1 7 ( . 0 7 1 ) 5 4 ( . 9 0 1 . ) 6 8 1 ( 6 4 4 . ) 3 7 3 ( 5 9 8 . ) 9 8 1 ( 3 5 4 . ) 8 4 2 ( 4 9 5 . ) 2 2 1 ( 2 9 2 ) 6 1 ( . 9 3 ) 9 8 ( . 6 7 1 ) 5 4 ( . 0 9 . ) 0 7 1 ( 7 3 3 . ) 1 8 3 ( 5 5 7 . ) 6 7 1 ( 8 4 3 . ) 2 1 2 ( 0 2 4 . ) 2 2 1 ( 1 4 2 ) 2 2 ( . 3 4 ) 7 8 ( . 6 1 1 ) 1 4 ( . 5 5 . ) 6 5 1 ( 9 0 2 . ) 4 5 3 ( 5 7 4 . ) 8 5 1 ( 2 1 2 . ) 2 9 1 ( 7 5 2 . ) 0 1 1 ( 7 4 1 ) 4 1 ( . 9 1 . ) 4 4 1 ( 9 3 . ) 7 7 ( 1 2 . ) 4 1 2 ( 8 5 . ) 1 5 1 ( 1 4 . ) 9 5 1 ( 3 4 ) 9 5 ( . 6 1 ) 5 8 ( . 3 2 . ) 7 3 1 ( 7 3 . ) 3 1 4 ( 2 1 1 . ) 3 7 1 ( 7 4 . ) 9 9 1 ( 4 5 . ) 3 3 1 ( 6 3 . ) 5 1 ( 4 i n a d e m ) e g n a r ( ) s y a d ( t n e m e r u s a e m m u i s s a t o p d n o c e s o t t s r i f m o r f e m T i s y a d 0 6 - h t a e D y t i l a t r o m l r a u c s a v o d r a c i y a d - 0 6 i s e i t i d b r o m o C e v i t c u r t s b o i c n o r h C e s a e s i d y r a n o m u p l s e t e b a D i e s a e s i d y e n d i k i c n o r h C e s a e s i d r e v i l i c n o r h C y c n a n g i l a m y n A r e t t u l f l a i r t A / n o i t a l l i r b i f l a i r t A ) 0 1 - D C I ( n o i s n e t r e p y H e r u l i a f t r a e H e s a e s i d t r a e h i c m e h c s I e k o r t S l e w o b y r o t a m m a l f n I e s a e s i d Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 6 of 11 l e u a v - p . 1 0 0 < . 1 0 0 < 9 5 0 . 0 3 0 . 0 9 0 . . 1 0 0 < 6 2 0 . 7 6 0 . 4 2 0 . 4 7 0 . . 1 0 0 < . 1 0 0 < . 1 0 0 < . 1 0 0 < . 1 0 0 < 8 3 0 . . 1 0 0 < . 1 0 0 < L / l o m m 1 7 – 1 5 . . L / l o m m 0 5 – 7 4 . . L / l o m m 6 4 – 1 4 . . L / l o m m 0 4 – 8 3 . . ) 4 3 1 = n n ( ) 2 5 3 = ( ) 8 9 4 2 = ( ) 8 9 3 2 = n n ( L / l o m m 7 3 – 5 3 . . ) 2 8 9 1 = n ( L / l o m m 4 3 – 0 3 . . ) 1 4 3 1 = n ( L / l o m m 9 2 – 5 1 . . ) 1 7 2 = n ( . ) 7 9 5 ( 0 8 ) 7 9 ( . 3 1 . ) 2 8 5 ( 8 7 . ) 4 8 2 ( 8 3 . ) 6 4 2 ( 3 3 ) 0 3 ( . 4 ) 7 3 ( . 5 . ) 5 4 5 ( 2 9 1 . ) 8 8 4 ( 9 1 2 1 . ) 2 5 4 ( 3 8 0 1 . ) 0 9 1 ( 7 6 . ) 1 3 2 ( 7 7 5 . ) 1 8 2 ( 4 7 6 . ) 2 5 4 ( 5 9 8 . ) 1 8 2 ( 7 5 5 . ) 0 8 5 ( 4 0 2 . ) 4 7 5 ( 4 3 4 1 . ) 2 6 5 ( 7 4 3 1 . ) 6 4 5 ( 2 8 0 1 . ) 4 6 2 ( 3 9 . ) 0 3 2 ( 1 8 . ) 5 6 ( 3 2 . ) 3 4 ( 5 1 . ) 0 5 2 ( 5 2 6 . ) 4 1 2 ( 4 3 5 . ) 0 3 ( 4 7 . ) 2 3 ( 1 8 . ) 0 4 2 ( 6 7 5 . ) 8 0 2 ( 9 9 4 ) 4 2 ( . 8 5 ) 3 2 ( . 6 5 . ) 8 2 2 ( 1 5 4 . ) 0 1 2 ( 7 1 4 ) 6 2 ( . 1 5 ) 9 2 ( . 7 5 . ) 1 2 5 ( 9 9 6 . ) 9 0 3 ( 5 1 4 . ) 3 5 5 ( 2 4 7 . ) 1 5 2 ( 7 3 3 . ) 2 1 2 ( 4 8 2 ) 1 3 ( . 1 4 ) 2 3 ( . 3 4 . ) 1 0 6 ( 3 6 1 . ) 7 4 2 ( 7 6 . ) 1 6 5 ( 2 5 1 . ) 4 1 2 ( 8 5 . ) 9 9 1 ( 4 5 . ) 6 2 ( 7 . ) 1 4 ( 1 1 . ) 7 6 5 ( 6 7 . ) 0 6 5 ( 7 9 1 . ) 6 7 5 ( 8 3 4 1 . ) 8 6 5 ( 2 6 3 1 . ) 0 5 5 ( 0 9 0 1 . ) 1 5 5 ( 9 3 7 . ) 5 6 5 ( 3 5 1 B 2 1 A : C T A 3 0 C : C T A ) 0 3 ( . 4 ) 0 0 ( . 0 . ) 0 3 5 ( 1 7 . ) 4 8 2 ( 8 3 . ) 4 0 6 ( 1 8 . ) 9 4 6 ( 7 8 . ) 9 4 1 ( 0 2 3 ≤ 3 ≤ . ) 4 9 4 ( 4 7 1 . ) 6 8 4 ( 3 1 2 1 ) 0 0 ( . 0 3 ≤ . ) 0 1 ( 6 2 ) 0 0 ( . 0 . ) 2 1 3 ( 0 8 7 ) 1 1 ( . 6 2 ) 0 0 ( . 0 . ) 5 0 4 ( 2 7 9 . ) 2 5 3 ( 3 4 8 ) 9 0 ( . 8 1 3 ≤ . ) 1 9 3 ( 5 7 7 . ) 7 4 3 ( 7 8 6 . ) 4 6 2 ( 3 9 . ) 4 9 5 ( 9 0 2 . ) 9 0 5 ( 9 7 1 . ) 0 7 2 ( 5 9 ) 1 1 ( . 4 ) 0 2 ( . 7 . ) 3 0 6 ( 6 0 5 1 . ) 0 2 6 ( 7 8 4 1 . ) 4 8 5 ( 8 5 1 1 . ) 5 4 4 ( 2 1 1 1 . ) 1 5 3 ( 1 4 8 . ) 4 6 3 ( 9 0 9 . ) 3 6 4 ( 0 1 1 1 . ) 5 1 ( 7 3 . ) 9 1 ( 8 4 ) 0 2 ( . 7 4 ) 3 2 ( . 4 5 . ) 3 6 3 ( 0 2 7 . ) 1 7 4 ( 4 3 9 ) 3 2 ( . 6 4 ) 5 2 ( . 0 5 ) 6 1 ( . 1 2 ) 0 0 ( . 0 . ) 1 7 3 ( 7 9 4 . ) 3 8 3 ( 4 1 5 . ) 4 3 5 ( 6 1 7 . ) 7 9 3 ( 3 3 5 . ) 9 0 5 ( 2 8 6 ) 1 2 ( . 8 2 . ) 0 3 ( 0 4 . ) 1 6 2 ( 5 3 . ) 6 3 2 ( 3 8 . ) 9 4 1 ( 2 7 3 . ) 7 1 1 ( 1 8 2 . ) 6 2 1 ( 0 5 2 . ) 5 3 1 ( 1 8 1 . ) 8 5 3 ( 7 9 . ) 2 3 4 ( 7 1 1 . ) 4 9 4 ( 4 3 1 . ) 4 2 4 ( 5 1 1 . ) 9 6 4 ( 7 2 1 . ) 6 2 ( 7 ) 5 5 ( . 5 1 . ) 8 8 1 ( 1 5 . ) 5 1 ( 4 . ) 0 0 ( 0 t n e m e p p u s l m u i s s a t o P y p a r e h t o c a m r a h P ) d e u n i t n o C ( s t s i n o g a 2 - a t e B i s d o c i t r o C s e v i t a x a L s e n i t n a X I s D A S N s g u r d i c g r e n e r d a i t n A i s l a b o r c m i i t n A s r o t a l i d o s a V s r o t i b h n i i m e t s y s n i s n e t o g n a i i n n e R s c i t e r u d i e d i z a h T i s c i t e r u d i p o o L s c i t e r u d i g n i r a p s m u i s s a t o P s t s i n o g a t n a r o t p e c e r l a r e n M i s c i t e r u d i e k i l - e d l i z a h T i s r e k c o b l l e n n a h c m u c a C i l s r e k c o b l a t e B y p a r e h t e v i s n e t r e p y h i t n a n o i t a n b m o c i h t i w d e t a e r t s t n e i t a p 6 7 9 8 f o t r o h o c a n i s l a v r e t n i m u i s s a t o p a m s a p l d e n i f e d e r p t h g e i e h t o t i g n d r o c c a d e i f i t a r t s i s c h p a r g o m e D 1 e l b a T s g u r d y r o t a m m a l f n i - i t n a l i a d o r e t s - n o N s D A S N I , m e t s y S n o i t a c i f i s s a C l l a c i m e h C c i t u e p a r e h T l a c i m o t a n A C T A , n o i s r e v h t 0 1 e s a e s i D f o n o i t a c i f i s s a C l l a n o i t a n r e t n I 0 1 - D C I a t a d e h t f o n o i t a z i m y n o n a e r u s n e o t r e d r o n i 3 r o 2 , 1 s i y c n e u q e r f e h t e r e h w s l l e c i d e n a t r e c s a s I - 3 ≤ Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 7 of 11 Fig. 1 Age, sex, comorbidity and drug standardized 60-day risk of all-cause death in relation to plasma potassium as a continuous variable. Model adjusted for age, gender, plasma sodium, renal insufficiency, malignancy, heart failure, chronic liver disease, chronic obstructive pulmonary disease, diabetes mellitus, atrial flutter/fibrillation, stroke and ischemic heart disease, antihypertensive therapy, corticosteroids, antimicrobials, non- steroidal anti-inflammatory drugs, potassium supplement, xanthines, laxatives instead of the first measurement, we noted that potas- sium levels below 3.8 mmol/L were associated with in- creased short-term mortality. Eighth, analyses on patients with available K2 measure- ments within 6–45 days from K1, showed that severe hypokalemia and hyperkalemia were associated with 60- day all-cause mortality. Ninth, analyses on patients with available K2 measure- ments above 45 days from K1, showed that potassium interval 3.0–3.4 mmol/L was associated with 60-day all- cause mortality. Tenth, we stratified K2 in three intervals: 1.5–3.4 mmol/L (hypokalemia), 3.5–4.6 mmol/L (normokalemia) and 4.7–7.1 mmol/L (hyperkalemia). Mortality within 60-days was increased both in patients with hypokalemia (HR 1.36, 95% CI 1.12–1.66) and in patients with hyperkalemia (HR 2.13, 95% CI 1.66–2.74) at K2 meas- urement compared with patients with normal potassium concentrations. Eleventh, multivariable analysis on patients with avail- able magnesium measurements at the time of plasma potassium draws, showed significant association of po- tassium levels below 3.0 mmol/L and mortality (HR 2.46, 95% CI 1.05–5.74). In addition, we also observed a trend towards increased risk of death in patients with potas- sium between 3.0–3.4 mmol/L. Discussion This Danish register-based cohort study investigated 60- day mortality among 8976 patients with hypertension and hypokalemia or low normal potassium in relation to a subsequent potassium measurement. The major find- ings were: (1) Persistent hypokalemia following low potassium was more than twice as frequent as develop- ment of hyperkalemia. (2) Persistent hypokalemia was common and associated with increased all-cause and presumed cardiovascular mortality; (3) Increase in potas- sium to levels > 4.6 mmol/L in patients with initial hypo- kalemia or low normal potassium was associated with increased all-cause and cardiovascular mortality; (4) Among patients with borderline hypokalemia initially, development of hypokalemia or hyperkalemia was asso- ciated with increased mortality risk; (5) Correcting hypo- kalemia associated with increased survival. In the current study, we observed significantly higher risk in patients with potassium 60-day mortality Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 8 of 11 Fig. 2 All-cause and cardiovascular mortality after hypokalemia or borderline hypokalemia according to subsequent potassium measurements in patients treated with combination antihypertensive therapy (60-days follow-up, n = 8976). Potassium interval K: 3.8–4.0 mmol/L represented the reference range. Adjusted for age, gender, serum sodium, renal insufficiency, malignancy, heart failure, chronic liver disease, chronic obstructive pulmonary disease, diabetes mellitus, atrial flutter/fibrillation, stroke and ischemic heart disease, antihypertensive therapy, corticosteroids, antimicrobials, non-steroidal anti-inflammatory drugs, potassium supplement, xanthines, laxatives concentrations < 3.5 or > 4.6 mmol/L after an episode with hypokalemia or low normal potassium. This finding was not surprising as we previously observed an appar- ent optimal potassium range within 4.1–4.7 mmol/L in a similar population [5]. Of 8976 patients with initial plasma potassium ≤3.7 mmol/L, 18% had potassium con- centrations ≤3.7 mmol/L at the second measurement and 5.4% > 4.6 mmol/L, suggesting that potassium deficit is frequently underestimated than overestimated by phy- sicians. Notably, 13% of the patients with borderline hypokalemia (K: 3.5 and 3.7 mmol/L) at the first measurement experienced a further decrease in potas- sium (< 3.5 mmol/L) at the second measurement. This suggests that the association of low normal potassium Table 2 60-day standardized absolute risk for all-cause death after hypokalemia or borderline hypokalemia according to subsequent potassium measurements in patients treated with combination antihypertensive therapy (n = 8976). Potassium interval K: 3.8–4.0 mmol/L represented the reference range. Adjusted for age, gender, serum sodium, renal insufficiency, malignancy, heart failure, chronic liver disease, chronic obstructive pulmonary disease, diabetes mellitus, atrial flutter/fibrillation, stroke and ischemic heart disease, antihypertensive therapy, corticosteroids, antimicrobials, non-steroidal anti-inflammatory drugs, potassium supplement, xanthines, laxatives P(K) 1.5–2.9 mmol/L P(K) 3.0–3.4 mmol/L P(K) 3.5–3.7 mmol/L P(K) 3.8–4.0 mmol/L P(K) 4.1–4.6 mmol/L P(K) 4.7–5.0 mmol/L P(K) 5.1–7.1 mmol/L Absolute risk %, (95% CI) 11.7% (8.3–15.0) 7.1% (5.8–8.5) 6.4% (5.3–7.5) 5.4% (4.5–6.3) 6.3% (5.4–7.2) 11.6% (8.7–14.6) 12.6% (8.2–16.9) 60-d Risk difference %, (95%CI) 6.3 (2.9–9.7) 1.7 (0.1–3.4) 1.0 (− 0.3–2.4) REF. 0.9 (−0.3–2.2) 6.2 (3.2–9.3) 7.2 (2.8–11.6) p-value < 0.001 0.03 0.14 0.13 < 0.001 0.001 Average risk ratio %, (95%CI) 2.17 (1.46–2.88) 1.32 (0.99–1.66) 1.19 (0.91–1.47) REF. 1.18 (0.92–1.44) 2.17 (1.51–2.82) 2.34 (1.45–3.22) p-value 0.001 0.06 0.17 0.17 < 0.001 0.003 Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 9 of 11 concentrations with mortality that we previously ob- served [5] can partly be explained by further declines in potassium levels, and that low normal potassium con- centrations might be a marker for an ongoing decrease in potassium. Our results also suggest that correction of hypokal- emia is important in relation to short-term mortality, as patients in the middle of the normal reference interval had good prognosis. Guidelines recommend supplemen- tation with potassium when plasma potassium levels are below 3.5 mmol/L. [13] However, in this study we can- not elucidate because of the low follow-up time the mechanism through which patients increased or de- creased in potassium concentrations. It is also difficult to state whether potassium is a risk factor or a risk marker regarding mortality. Our population is relatively old, patients are treated with at least two antihyperten- sive drugs, and about 20% of the patients have history of heart failure, ischemic heart disease, atrial fibrillation/ flutter, chronic obstructive pulmonary disease and dia- at non- betes. Possibly, potassium concentrations cardiotoxic levels more likely are a risk marker of great disease burden, which is very important to recognize and identify. Potassium supplementation of asymptomatic patients with low normal concentrations is controversial. Guide- lines in the US recommend a stricter standard for potas- sium replacement therapy (< 4.0 mmol/L) especially in patients with cardiovascular disease who are at high risk of ventricular tachyarrhythmias [13]. Our study suggests that potassium concentrations in the middle of the refer- ence interval are beneficial even in patients with potas- sium levels ≤3.7 mmol/L. Various studies have previously demonstrated that hypokalemia among patients with cardiovascular disease is associated with an increased mortality risk [14–18]. However, no prior studies have investigated the impact of potassium normalization on short-term survival. Though, one study examined the impact of correcting hypokalemia within 24 h on the risk of cardiac arrhyth- mias in hospitalized patients without coronary syn- dromes or history of arrhythmias [19]. The authors did not find increased odds of arrhythmia in patients with hypokalemia whose potassium levels were not corrected ≥3.5 mmol/L. Although, the study does not describe or account for the cause of admission, comorbidities or pharmacotherapy. The investigators excluded patients with history of ischemic heart disease and arrhythmia, but included patients with heart failure who have a high arrhythmia risk. Overall, both the study population and the outcome measure differed in this paper compared with our study. Another study performed on 5916 individuals from the general population found no significant associations between borderline hypokalemia (3.4–3.6 mmol/L) and risk of all-cause mortality, risk of stroke or risk of acute myocardial infarction [20]. Comparing the results of our study with this study is difficult due to major differences in study population, methodology and aim. First, our population was characterized by redemption of at least two antihypertensive drugs. Mattsson et al. [20] enrolled participants from the general population, where 49.6% had high blood pressure at baseline, 13.9% were pre- scribed heart medication and 10.9% were treated with diuretics. In our population, we observed higher burden of cardiovascular disease and use of diuretics. Second, our aim was to investigate the impact of correcting hypokalemia or borderline hypokalemia on short-term all-cause and cardiovascular mortality. In terms of mor- tality, Mattsson et al. [20] followed participants from their fourth examination in 2001–2003 until November 2014 or death, having a median follow-up of 11.9 years (Q1-Q3: 11.4–12.5 years). As potassium is varying over time especially in patients with cardiovascular disease or treated with antihypertensive drugs, use of one potas- sium measurement to assess mortality over more than 10 years can provide results that are difficult to interpret. Shorter follow-up time or time varying analysis where the authors accounted for both multiple measurements over time and change in relevant medication would have provided better methodology. Although, it is important to acknowledge that correcting hypokalemia and low normal potassium might not have the same impact in general population compared to a population with heart disease. Another study investigated the influence of dyskalemia at admission and early dyskalemia correcting on short- term survival and cardiac events among intensive care unit (ICU) patients [21]. The authors concluded that pa- tients with persisting hypokalemia or hyperkalemia within the first 2 days in ICU had increased risk of death. The two populations are not comparable, however both studies emphasize the importance of rapid correc- tion of hypokalemia to improve short-term mortality. Limitations The limitations are related to the observational nature of register-based studies, which imply non-causal interpret- ation of the results. We did not have information about comorbidities and risk factors from the primary sector. Therefore, patients who did not redeem any medication of interest or were not registered an ICD-code from the secondary sector could have been misclassified as “healthy”. Patients with complications related to hypertension have a larger like- lihood for being referred to the secondary sector and therefore also a higher probability for being diagnosed with other conditions (compared with patients with Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 10 of 11 uncomplicated hypertension), leading to an ascertain- ment/surveillance bias and non-differential misclassifica- tion bias. To reduce this bias, we defined hypertension as use of at least two antihypertensive drugs in two con- comitant quarters. Whether hypertension was resistant, controlled or uncontrolled was unkown, and data about ejection fraction and type of heart failure was not available. We cannot exclude that the blood draws may contain hemolysis. However, in case of significant hemolysis the samples submitted are rejected and no potassium value is available. the patients were hospitalized at We could not investigate the effect of any potential treatment or drug dosage adjustment in the time be- tween the first and second potassium measurement. The Danish National Prescription Registry records filled pre- scriptions; thus, changes in dosage cannot be identified, unless a new drug is prescribed. In addition, the majority of the time of potassium measurement and any treatment during hospitalization is not registered in the Danish National Prescription Registry. Moreover, it was also difficult to identify the cause of hypokalemia using the registers. Hypokalemia might have occurred due to administra- tion of diuretics, alkalosis, derangements in the renin angiotensin aldosterone gastroenteritis or other pathologies. However, the purpose of this study was neither to investigate the cause of hypokalemia, nor to assess the strategies used to correct low potas- sium concentrations. The purpose of this study was to find a clue, whether normalization of potassium had an effect on short-term mortality, whether we should increase potassium concentrations in patients with borderline hypokalemia and whether potassium actually increased. system, It is also important to acknowledge that plasma potas- sium is not always a good predictor of the whole body potassium. Yet, it is the most commonly used method to assess potassium and only in patients with persistent hypokalemia over a longer period of time total body po- tassium is calculated. Conclusion Persistent hypokalemia was frequent and associated with increased all-cause and cardiovascular mortality. In- crease in potassium to levels > 4.6 mmol/L in patients with initial hypokalemia or low normal potassium was associated with increased all-cause and cardiovascular mortality. Perspectives We were not able to report the initiatives medical doc- tors undertook after observing potassium levels below the first measurement. However, our 3.8 mmol/L at of the emphasize importance results potassium normalization after an episode with hypokalemia and low normal potassium and that overcorrection is associ- ated with an increased risk of death. Potassium concen- trations in the middle of the normal reference interval are associated with good prognosis. Possibly, potassium supplementation, use of mineral receptor antagonists or thiazide-like diuretics instead of thiazide-type in patients with potassium concentrations ≤3.7 mmol/L could be of clinical importance, but requires further study, prefera- bly through a randomized controlled trial. Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s12872-020-01654-3. Additional file 1. Abbreviations ATC: Anatomical therapeutic chemical classification system; K1: First potassium measurement within 100 days from combination antihypertensive therapy initiation; K2: First potassium measurement within 6–100 days from K1; ICD: International classification of disease; CKD-EPI: Chronic kidney disease epidemiology collaboration; eGFR: Estimated glomerular filtration rate; HR: Hazard ratio; AR: Absolute risk; 95% CI: 95% confidence interval; NSAI Ds: Non-steroidal anti-inflammatory drugs Acknowledgements None. Authors’ contributions Conception or design of the work: MLK, PS, CTP, KK, JQT, AB, HB. Acquisition of data: MLK, PS, CTP, KK, AB, CJYL. Analysis and interpretation of data: MLK, PS, CTP, AB, KK, CJYL, HB, JQT. Draftet the manuscript: MLK. Critically revised the manuscript: Peter Søgaard, Christian Torp-Pedersen, Henrik Bøggild, Christina Ji-Young Lee, Anders Bonde, Jesper Q. Thomassen, Gunnar Gislason, Manan Pareek, Kristian Kragholm, All authors gave final approval and agree to be accountable for all aspects of work ensuring integrity and accuracy. Funding This study was funded using departmental funding sources only. The funding covered labor costs. Availability of data and materials Due to restrictions related to Danish law and protecting patient privacy, the combined set of data used in this study can only be made available through a trusted third party, Statistics Denmark. This state organisation holds the data used for this study. University-based Danish scientific organisations can be authorized to work with data within Statistics Denmark and such organ- isation can provide access to individual scientists inside and outside of Denmark. Data are available upon request to authorized scientists by con- tacting Statistics Denmark: http://www.dst.dk/en/OmDS/organisation/Tele- fonbogOrg.aspx?kontor=13&tlfbogsort=sektion or the Danish Data Protection Agency: https://www.datatilsynet.dk/english/the-danish-data-protection- agency/contact/. More information regarding data access is available at https://www.dst.dk/en/TilSalg/Forskningsservice. Ethics approval and consent to participate Retrospective studies do not require ethics approval in Denmark and all data were deidentified and only available through Statistics Denmark. Approval from the Danish Data Protection Agency was secured, and the need for patient informed consent was not needed. Consent for publication Not applicable. Krogager et al. BMC Cardiovascular Disorders (2020) 20:386 Page 11 of 11 trial. Hypertension. 2012;59:926–33 Department of Epidemiology and Social Medicine, Albert Einstein College of Medicine, Bronx, NY, USA. 15. Aldahl M, Jensen A-SC, Davidsen L, Eriksen MA, Moller Hansen S, Nielsen BJ, 16. Krogager ML, Kober L, Torp-Pedersen C, Sogaard P. Associations of serum potassium levels with mortality in chronic heart failure patients. Eur Heart J England. 2017;38:2890–6. Krogager ML, Eggers-Kaas L, Aasbjerg K, Mortensen RN, Køber L, Gislason G, Torp-Pedersen C, Søgaard P. Short-term mortality risk of serum potassium levels in acute heart failure following myocardial infarction. Eur Hear J - Cardiovasc Pharmacother. 2015;1:245–51. 17. Hagengaard L, Søgaard P, Espersen M, et al. Association between serum potassium levels and short-term mortality in patients with atrial fibrillationor flutter co-treated with diuretics and rate- or rhythm-controlling drugs. Eur Heart J Cardiovasc Pharmacother. 2020;6(3):137-44. Tishler M, Armon S. Nifedipine-induced hypokalemia. Drug Intell Clin Pharm United States. 1986;20:370–1. 18. 19. Harkness W, Watts P, Kopstein M, Dziadkowiec O, Hicks G, Scherbak D. Correcting hypokalemia in hospitalized patients does not decrease risk of cardiac arrhythmias. Adv Med. 2019;2019:1–4. 20. Mattsson N, Nielsen OW, Johnson L, Prescott E, Schnohr P, Jensen GB, Kober L, Sajadieh A. Prognostic impact of mild hypokalemia in terms of death and stroke in the general population-a prospective population study. Am J Med United States. 2018;131:318.e9–318.e19. 21. Bouadma L, Mankikian S, Darmon M, Argaud L, Vinclair C, Siami S, Garrouste-Orgeas M, Papazian L, Cohen Y, Marcotte G, Styfalova L, Reignier J, Lautrette A, Schwebel C, Timsit JF, Timsit JF, Azoulay E, Garrouste-Orgeas M, Zahar JR, Adrie C, Darmon M, Clec’h C, Alberti C, Francąis A, Vesin A, Ruckly S, Bailly S, Lecorre F, Nakache D, Vannieuwenhuyze A, et al. Influence of dyskalemia at admission and early dyskalemia correction on survival and cardiac events of critically ill patients. Crit Care. 2019. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Competing interests None to declare. Author details 1Department of Cardiology, Aalborg University Hospital, Aalborg, Denmark. 2Department of Cardiology and Clinical Research, Nordsjællands Hospital, Hillerød, Denmark. 3Public Health and Epidemiology Group, Department of Health Science and Technology, Aalborg University, Aalborg, Denmark. 4Unit of Epidemiology and Biostatistics, Aalborg University Hospital, Aalborg, Denmark. 5Department of Cardiology, Copenhagen University Hospital, Herlev and Gentofte, Hellerup, Denmark. 6Department of Clinical Biochemistry, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark. 7Department of Cardiology, Herlev and Gentofte University Hospital, Hellerup, Denmark. 8The Danish Heart Foundation, Copenhagen, Denmark. 9The National Institute of Public Health, University of Southern Denmark, Copenhagen, Denmark. 10Department of Internal Medicine, Yale New Haven Hospital, Yale University School of Medicine, New Haven, USA. 11Brigham and Women’s Hospital, Heart & Vascular Center, Harvard Medical School, Boston, USA. 12Department of Cardiology, Regionshospital Nordjylland, Hjørring, Denmark. Received: 16 April 2020 Accepted: 4 August 2020 References 1. 2. 3. 4. 5. 6. 7. 8. 9. Veltri KT, Mason C. Medication-induced hypokalemia. P T United States. 2015;40:185–90. Sica DA, Carter B, Cushman W, Hamm L. Thiazide and loop diuretics. J Clin Hypertens (Greenwich) United States. 2011;13:639–43. Tamargo J, Segura J, Ruilope LM. Diuretics in the treatment of hypertension. Part 2: loop diuretics and potassium-sparing agents. Expert Opin Pharmacother. 2014;15(5):605–21. Rodenburg EM, Visser LE, Hoorn EJ, Ruiter R, Lous JJ, Hofman A, Uitterlinden AG, Stricker BH. Thiazides and the risk of hypokalemia in the general population. J Hypertens. 2014;32:2092–7. Krogager ML, Torp-Pedersen C, Mortensen RN, Køber L, Gislason G, Søgaard P, Aasbjerg K. Short-term mortality risk of serum potassium levels in hypertension: a retrospective analysis of nationwide registry data. Eur Heart J. 2017;38:104–12. Thygesen LC, Daasnes C, Thaulow I, Bronnum-Hansen H. Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving. Scand J Public Health Sweden. 2011;39:12–6. Pedersen CB. The Danish civil registration system. Scand J Public Health. 2011;39:22–5. Olesen JB, Lip GY, Hansen ML, Hansen PR, Tolstrup JS, Lindhardsen J, Selmer C, Ahlehoff O, Olsen AM, Gislason GH, Torp-Pedersen C. Validation of risk stratification schemes for predicting stroke and thromboembolism in patients with atrial fibrillation: nationwide cohort study. BMJ. 2011;342:d124 Department of Cardiology, Copenhagen University Hospital Gentofte, 2900 Hellerup, Denmark. jo@heart.dk. Levey AS, Stevens LA, Schmid CH, Zhang Y, Castro AF, Feldman HI, Kusek JW, Eggers P, Van Lente F, Greene T, Coresh J. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150:604–12. 10. Burtis CA, Ashwood ER, Bruns DE. Tietz textbook of clinical chemistry and molecular diagnostics. 5th ed. St. Louis, Missouri, USA: Elsevier Saunders; 2012. 11. Drogies T, Ittermann T, Lüdemann J, Klinke D, Kohlmann T, Lubenow N, Greinacher A, Völzke H, Nauck M. Potassium - reference intervals for lithium- heparin plasma and serum from a population-based cohort. LaboratoriumsMedizin. 2010;34:39–44. 12. Article citationsMore>> R Core Team. R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2018. https://www.R-project.org. 13. Cohn JN, Kowey PR, Whelton PK, Prisant LM. New guidelines for potassium replacement in clinical practice: a contemporary review by the National Council on potassium in clinical practice. Arch Intern Med United States. 2000;160:2429–36. 14. Alderman MH, Piller LB, Ford CE, Probstfield JL, Oparil S, Cushman WC, Einhorn PT, Franklin SS, Papademetriou V, Ong ST, Eckfeldt JH, Furberg CD, Calhoun DA, Davis BR, Group A and L-LT to PHATCR. Clinical significance of incident hypokalemia and hyperkalemia in treated hypertensive patients in the antihypertensive and lipid-lowering treatment to prevent heart attack
10.1186_s12879-022-07247-z	Gelanew et al. BMC Infectious Diseases (2022) 22:261 https://doi.org/10.1186/s12879-022-07247-z RESEARCH Open Access High seroprevalence of anti-SARS-CoV-2 antibodies among Ethiopian healthcare workers Tesfaye Gelanew1, Berhanu Seyoum1, Andargachew Mulu1, Adane Mihret1, Markos Abebe1, Liya Wassie1, Baye Gelaw2, Abebe Sorsa3, Yared Merid4, Yilkal Muchie1, Zelalem Teklemariam5, Bezalem Tesfaye1, Mahlet Osman1, Gutema Jebessa1, Abay Atinafu1, Tsegaye Hailu1, Antenehe Habte1, Dagaga Kenea3, Anteneh Gadisa4, Desalegn Admasu5, Emnet Tesfaye4, Timothy A. Bates6, Jote Tafese Bulcha7, Rea Tschopp1,8, Dareskedar Tsehay1, Kim Mullholand9, Rawleigh Howe1, Abebe Genetu1, Fikadu G. Tafesse6† and Alemseged Abdissa1† Abstract Background: COVID-19 pandemic has a devastating impact on the economies and health care system of sub- Saharan Africa. Healthcare workers (HWs), the main actors of the health system, are at higher risk because of their occupation. Serology-based estimates of SARS-CoV-2 infection among HWs represent a measure of HWs’ exposure to the virus and could be used as a guide to the prevalence of SARS-CoV-2 in the community and valuable in combating COVID-19. This information is currently lacking in Ethiopia and other African countries. This study aimed to develop an in-house antibody testing assay, assess the prevalence of SARS-CoV-2 antibodies among Ethiopian high-risk frontline HWs. Methods: We developed and validated an in-house Enzyme-Linked Immunosorbent Assay (ELISA) for specific detec- tion of anti-SARS-CoV-2 receptor binding domain immunoglobin G (IgG) antibodies. We then used this assay to assess the seroprevalence among HWs in five public hospitals located in different geographic regions of Ethiopia. From consenting HWs, blood samples were collected between December 2020 and February 2021, the period between the two peaks of COVID-19 in Ethiopia. Socio-demographic and clinical data were collected using questionnaire-based interviews. Descriptive statistics and bivariate and multivariate logistic regression were used to determine the overall and post-stratified seroprevalence and the association between seropositivity and potential risk factors. Results: Our successfully developed in-house assay sensitivity was 100% in serum samples collected 2- weeks after the first onset of symptoms whereas its specificity in pre-COVID-19 pandemic sera was 97.7%. Using this assay, we analyzed a total of 1997 sera collected from HWs. Of 1997 HWs who provided a blood sample, and demographic and clinical data, 51.7% were females, 74.0% had no symptoms compatible with COVID-19, and 29.0% had a history of contact with suspected or confirmed patients with SARS-CoV-2 infection. The overall seroprevalence was 39.6%. The lowest (24.5%) and the highest (48.0%) seroprevalence rates were found in Hiwot Fana Specialized Hospital in Harar and ALERT Hospital in Addis Ababa, respectively. Of the 821 seropositive HWs, 224(27.3%) of them had a history of Correspondence: tesfaye.gelanew@ahri.gov.et; tafesse@ohsu.edu; alemseged.abdissa@ahri.gov.et †Fikadu G. Tafesse and Alemseged Abdissa contributed equally 1 Armauer Hansen Research Institute, Addis Ababa, Ethiopia 6 Department of Molecular Microbiology and Immunology, Oregon Health & Sciences University, Portland, OR, USA Full list of author information is available at the end of the article © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Gelanew et al. BMC Infectious Diseases (2022) 22:261 Page 2 of 9 symptoms consistent with COVID-19 while 436 (> 53%) of them had no contact with COVID-19 cases as well as no history of COVID-19 like symptoms. A history of close contact with suspected/confirmed COVID-19 cases is associated 1.4, 95% CI 1.1–1.8; p with seropositivity (Adjusted Odds Ratio (AOR) 0.015). = = Conclusion: High SARS-CoV-2 seroprevalence levels were observed in the five Ethiopian hospitals. These findings highlight the significant burden of asymptomatic infection in Ethiopia and may reflect the scale of transmission in the general population. Keywords: SARS-CoV-2, COVID-19, RBD, ELISA, Seroprevalence, Antibodies, Ethiopia Background Despite the total population of 1.3 billion, Africa stands out as the region least affected by the severe acute respir- atory syndrome-Corona-Virus-2 (SARS-CoV-2) and cor- onavirus disease-2019 (COVID-19) pandemic. As of May 23rd, 2021[1], the total reported case number had risen to 4,748,581 with 128,213 reported deaths, representing 2.9% and 3.7% of global cases and deaths, respectively. The low number of reported cases and deaths in Africa has been attributed to low testing capacity, younger population, warmer environments, and the successful implementation of control measures [2]. Also, pre-exist- ing cross-protective immunity due to the four other less pathogenic human coronaviruses (HCoVs) [3], Bacil- lus Calmette-Guérin (BCG)-vaccination [4], or recent history of malaria infection may offer some protection against infection or severe forms of COVID-19[5]. As of May 21, 2021, Ethiopia has performed over 2,682,758 real-time reverse transcription-polymerase chain reactions (RT-PCR) tests for SARS-CoV-2 and reported 268,901 cases and 4068 deaths since the first case was detected in the country on March 13, 2020. Almost all testing has been done to confirm SARS-CoV-2 infection in suspected cases and contacts, as well as both outbound and inbound travelers. Given the difficulty and cost of RT-PCR-based testing in resource-limited coun- tries like Ethiopia, mildly affected or asymptomatic indi- viduals are not usually screened, and so the number of confirmed SARS-CoV-2 infections is likely vastly under- estimated [6]. In this context, seroprevalence surveys are of the utmost importance to assess the proportion of the population that have already developed antibodies against the virus. Evidence has shown that healthcare workers (HWs) are at higher risk of acquiring the infection than the general population. This is because their work is likely to require close contact with SARS-CoV-2 infected patients at COVID-19 treatment centers, in emergency rooms and wards, and via virus-contaminated surfaces. If infected, they can pose a significant risk to vulnerable patients and co-workers [7]. Thus, assessing the seroprevalence of SARS-CoV-2 antibodies among HWs in Ethiopia will help us understand COVID-19 spread among health care facilities and to measure the success of public health interventions. It will also provide an opportunity to compare the disease trajectory in a low-income set- ting. A report from London, UK suggested that the rate of asymptomatic SARS-CoV-2 infection among HWs reflects general community transmission rather than in- hospital exposure [8]. Currently, numerous commercial COVID-19 antibody tests, including Enzyme Linked Immunosorbent Assay (ELISA) are available. However, most of them lack vali- dation and evaluation of their sensitivity and specific- ity in large-scale studies, particularly in Africa, where interpretation of results from these tests is challenging due to pre-existing cross-reactive antibodies induced by other pathogens [9, 10]. However, in resource-limited countries, including Ethiopia, data generated from vali- dated antibody tests using local clinical samples could provide information valuable in combating COVID-19. Unfortunately, the price and supply of antibody tests are unpredictable. To overcome these challenges, we have developed a relatively cheaper in-house ELISA to detect anti–SARS-CoV-2 Receptor Binding Domain (RBD) immunoglobulin G (IgG) antibodies in serum samples. We then conducted a serosurvey amongst HWs in five public hospitals to estimate the seroprevalence of SARS- CoV-2 in urban Ethiopia using our validated assay. We discussed the implications of our SARS-CoV-2 serosur- veillance findings for frontline HWs and the Ethiopian population at large. Methods Participant recruitment This cross-sectional study represents a joint effort between the Armauer Hansen Research Institute (AHRI) and five public hospitals in Ethiopia, namely All Africa Leprosy and Tuberculosis Rehabilitation and Training Center (ALERT Center); Hawassa, Gondar, Asella, and Hiwot Fana (located in Harar) hospitals. These partici- pating hospitals were selected because they are among the 11 hospitals located in different regional states of the country and are linked to the AHRI’s Clinical Research Network. Similar serosurvey studies for the remain- ing hospitals liked to the AHRI’s CRN are ongoing. All Gelanew et al. BMC Infectious Diseases (2022) 22:261 Page 3 of 9 hospital staff (n = 7898) from all five public hospitals were invited to take part in the study through office memos and notice board announcements. However, only 1997 (24.4%) out of 7898 invited hospital staff were volun- teered to provide 5 ml blood and demographic and clini- cal data. Demographic and clinical data were obtained using a structured questionnaire based on WHO SARS- CoV-2 seroprevalence studies (World Health Organiza- tion (WHO, the “Solidarity II” global serologic study for COVID-19) [11]. Sample collection, storage, transportation, and inactivation Five milliliters of blood were collected in a serum collec- tion tube from each participant using standard proce- dures. Sera were separated by centrifugation and stored at − 20 °C until transferred to AHRI laboratory in Addis Ababa, Ethiopia, in a cold box. Inactivation of infectious viruses in serum was performed by incubation with Tri- ton X-100 to a final concentration of 1% for 1 h [12] and stored at − 80 °C until testing for the presence of SARS- CoV-2 specific antibodies. Serum samples were collected from December 2020 to February 2021 between the two peaks of SARS-CoV-2 transmissions in Ethiopia (https:// covid 19. who. int/ region/ afro/ count ry/ et). Enzyme‑Linked Immunosorbent Assay (ELISA) The SARS-CoV-2 spike protein RBD-containing plas- mid construct was cloned as described previously [13]. The RBD protein was then expressed in EXPi293 cells using previous methods [13]. Then, the purified RBD protein was used as a target antigen to develop our in- house anti-SARS-CoV-2 RBD IgG detection ELISA. We used 1 µg/ml of RBD to coat the microwell plate over- night at 4 °C. The assay is an indirect ELISA, measuring serum IgG against RBD of spike protein SARS-CoV-2, using a horseradish peroxidase-linked anti-human IgG secondary antibody (Invitrogen, USA). Supplementary method shows the detail procedure description of our assay (Supplementary Method). We validated this ELISA using pre-COVID-19 pandemic sera/plasma samples (n = 365), WHO “Solidarity II” plasma panels (n = 5), and sera/plasma samples (n = 401) collected from a cohort of mild (majority) and severe COVID-19 patients confirmed by RT-PCR. Detection of RBD-specific IgG antibodies in each serum sample was done in duplicate microwells of ELISA plate. In each ELISA run, we included posi- tive and negative controls. Positive and negative control samples were selected by matching their optical density (OD) readouts with WHO solidarity II plasma panels developed by the United Kingdom’s National Institute for Biological Standards and Control (NIBSC;20/130, single donor, high-titer antibody), 20/120 (single donor, relatively high-titer antibody), 20/122 (pool of five donor samples, mid-titer antibody), 20/124 (low S1, high-nucle- ocapsid protein antibody titer), 20/126 (low-titer anti- body, 20/128, negative control). Optimization and validation of in‑house anti‑RBD IgG detection ELISA We noted background signal from the negative controls at a 1:100 dilution of serum. Of 365 pre-COVID-19 sera, 30 showed optical density (OD) values comparable to the low reactive convalescent WHO plasma samples. We further optimized the assay by increasing the concentra- tion of skimmed milk powder and Tween-20 in blocking buffer from 3 to 4% and from 0.05% to 0.1%, respectively, and serum dilution at 1:200. Except for fourteen pre- COVID-19 samples, the background was significantly reduced when re-tested false positives, which in turn increased the specificity of our assay without compro- mising its sensitivity in WHO positive control samples and serum samples obtained from a cohort of COVID-19 patients. Using our optimized ELISA protocol, we calculated the cut-off value for positivity using pre-COVID-19 pandemic sera collected between 2012 and 2018, and plasma/serum samples collected from cohort of con- firmed COVID-19 patients at different time points of post-onset of symptoms (dps). The definition of sero- positivity represents a greater than 2.5 ratio of sample OD value to the mean OD value of the negative controls (Fig. 1). This definition provides specificity of 97.7% (95% CI, 95.6–99.0) (Additional file 1: Table S1). Our anti-RBD IgG detection ELISA showed a sensitivity of 67.3% (95% CI 62.3.0–72.3), 75. 8% (95% CI 61.0–86.0), 100% (95% CI 84.0–100) in serum/plasma samples collected at 1–7 dps, 8–14 dps and ≥ 15dps, respectively from mostly (> 90%) mild and moderate COVID-19 cases confirmed by RT- PCR (Additional file 1: Table S2). This performance is in line with those published for both in-house and commer- cial assays approved for emergency use by the FDA [14] and (https:// covid- 19- diagn ostics. jrc. ec. europa. eu/). In‑house IgG ELISA comparison with commercial anti‑SARS‑CoV‑2 serologic assays We further compared the relative sensitivity and specific- ity of our assay with commercially available SARS-CoV-2 antibody tests: one lateral flow assay (LFA) (Hangzhou Realy Tech Co., LTD) and one ELISA (Beijing Wantai Biological Pharmacy Enterprise Co., Ltd) following the manufacturers’ instructions using randomly selected small panels (pre-pandemic; n = 40, and COVID-19; n = 40) from the large size panels that were used for our assay validation. We found a comparable sensitivity and specificity to those commercially available COVID-19 Gelanew et al. BMC Infectious Diseases (2022) 22:261 Page 4 of 9 Multivariate regression analysis was applied for those variables with a p-value < 0.25 in bivariate analysis to evaluate the strength of association between independ- ent variables and seropositivity, the outcome variable. A p-value of < 0.05 was considered statistically significant. Results Characteristics of study participants The total number of HWs in the five participating hospi- tals was 7898. Of these, we enrolled 1997 (24.4%) HWs [from ALERT (n = 308); Hawassa (n = 414); Gondar (n = 453); Assela (n = 484); and Haromaya (n = 338)] in the study. Almost half (51.7%) of the study participants were females. The majority (85.7%) of the participants belonged to the age groups 25–34 and 35–49 years with the mean age 34 years (range 20–60 years). Of the par- ticipants, 559 (28.3%) were nurses, 368 (18.7%) were doctors, 223 (11.3%) were medical laboratory personnel, 345 (17.4%) were administrative staff, and the remain- ing 24.2% (n = 478) did not specify their occupation. In the cohort, 1490 (74.0%) participants were asymp- tomatic, 507 (26.0%) had reported one or more symp- toms compatible with COVID-19 during the preceding 4 weeks, and 557 (29.0%) had a history of close contact with suspected or confirmed COVID-19 cases. Overall, 133 (6.7%) of the participants reported having a history of comorbid medical conditions, with obesity (1.9%), asthma (1.7%), hypertension (1.5%), and Human Immu- nodeficiency Virus (HIV) (1.3%) being the most common. These and other demographic and clinical characteristics of study participants are summarized in Table 1. Seroprevalence of SARS‑CoV‑2 antibodies by geographic locations of participating hospitals, age, sex, healthcare cadre and clinical factors The overall seroprevalence of SARS-CoV-2 antibod- ies among HWs from all five studied public hospitals was 821 of 1997 (39.8% [95% CI 37.4–41.7]). Of the 821 seropositive HWs, 224 (27.3%) of them had a history of symptoms consistent with COVID-19 while 436 (> 53%) them had no contact with COVID-19 cases as well as no a history of COVID-19 like symptoms. The estimated seroprevalence with 95% CI for each of the participat- ing hospitals was shown in Fig. 2 and Table 1, ranging from 24.5% to 48.0%. There was no statistically significant seroprevalence difference among females (42.4% [95% CI 39.4–45.55]) and males (39.6% [95% CI 36.6–42.7]). We did not find association between seropositivity, and participants’ demographic and clinical features given in Table 1, except history of contact with suspected or con- firmed COVID-19 case. We noted higher [48.5% [95% CI 44.3–52.6)] seroprevalence in HWs who had close contact with COVID-19 case than in HWs who reported Fig. 1 Validation of the SARS-CoV-2 RBD specific IgG antibody detection ELISA. The value on the y-axis represents the ratio of OD450 nm to the average mean OD450 nm of the negative controls. The broken black line represents the cut-off value (2.5). We tested a total of 405 serum/plasms samples collected from cohort of mild and moderate (93.6%) and severe Ethiopian COVID-19 patients confirmed by RT-PCR (represented in red color). Of these 325 samples were collected during 0–7 days post-onset of symptoms (dps); 52 were collected during 8–14 dps, and 17 were collected within 15–28 dps (Additional file 1: Table S2). We also tested serum/plasma samples collected from 365 Ethiopian individuals before the global COVID-19 pandemic, represented in blue color (Additional file 1: Table S1) antibody detection kits depending on the sample col- lection date (Additional file 1: Table S3, S4). We then utilized this assay to estimate the seroprevalence of anti- SARS-CoV-2 spike protein RBD IgG antibodies among HWs. Data analysis The data were double entered into REDCap Database Version 8.11. Following data verification and validation, analysis was done using STATA Version 15.0. Descrip- tive statistics and the actual number of cases were used to describe frequency outputs for categorical variables. Figures were generated using GraphPad Prism Version 9.1. Cross-tabulations were performed to explore and display relationships between two categorical variables. The overall seroprevalence with 95% CI for anti-SARS- CoV-2 RBD IgG was calculated by dividing the num- ber of seropositive cases divided by the total number of study participants from all five hospitals. Apparent SARS-CoV-2 prevalence was stratified by the geographic location of hospitals, age, sex, self-reported previous his- tory exposure, symptoms, comorbidities, and further by occupation/department where HWs are working. Bivari- ate logistic regression was done between seroprevalence with independent variables such as sex, age, occupation, comorbidity, history of close contact, and symptoms. Gelanew et al. BMC Infectious Diseases (2022) 22:261 Page 5 of 9 Table 1 Seroprevalence of anti-SARS-CoV-2 RBD IgG antibodies by participant (n 1997) characteristics and hospitals, Ethiopia, 2021 = Characteristics N % Seroprevalence (%), 95% CI Gender Male Female Age (in years) 19–24 25–34 35–49 50 ≥ Morbidity Yes No COVID-19 Symptomatic Asymptomatic Contact Yes No Hospitals ALRET Hawassa Gondar Asella Hiwot Fana Occupation Doctor Nurse Lab Technician Administrator Others 980 169 918 792 115 133 1864 507 1490 557 1362 308 414 453 484 338 368 559 223 345 478 49.3 51.7 8.8 46.0 39.7 5.8 6.7 93.3 26.0 74.0 29.0 71.0 15.4 20.7 22.6 24.2 17.0 18.7 28.3 11.3 17.4 24.2 39.6 (36.6–42.7) 42.37 (39.4–45.5) 44.0 (36.7–51.7) 41.6 (38.4–44.8) 39.7 (36.4–43.0) 41.7 (33.00–51.0) 44.4 (36.1–52.9) 40.9 (38.7–43.3) 39.9 (37.4–42.4) 45.2 (40.9–49.5) 48.5 (44.3–52.6) 38.1 (35.6–40.7) 48.1 (40.3–53.6) 44.8 (40.05–49.) 44.7 (40.12–49.3) 40.7 (36.4–45.1) 24.6 (20.3–29.4) 40.5 (35.6–45.6) 41.9 (37.7–45.8) 46.2 (39.7–52.8) 39.1 (34.1–44.4) 43.5 (38.7–48.4) N is the total number of participants included in each category % indicates proportion of participants that fell within each category no contact (38.1% [95% CI 35.6–40.7]). Seroprevalence was similar amongst different cadres of the health sys- tem, and amongst different age groups of HWs (Table 1). Although not statistically significant, a relatively higher (44.4% [95% CI 36.1–52.9]) seropositivity against SARS- CoV-2 was found in comorbid HWs than in HWs who had no comorbidity (40.9% [95% CI 38.7–43.3]). Factors associated with anti‑SARS‑CoV‑2 RBD IgG antibodies positivity [(Adjusted odds ratio HWs working at ALERT (AOR) = 2.7, 95% CI 1.6–3.1; p = 0.001]; Hawassa (AOR = 2.1, 95% CI 11.5–3.2; p = 0.001); Gondar (AOR = 2.8, 95% CI 1.99–3.87; p = 0.001), and Assela (AOR = 2.1, 95% CI 1.6–3.1; p = 0.001) were at higher odds of seropositivity compared to HWs working at Hiwot Fana Specialized University Hospital (Table 2). Association with seropositivity was further tested for correlation with gender, age, contact, morbidity, previous COVID-19 symptoms, and occupation using both bivari- ate and multivariate analyses. However, only previous history of contact with confirmed or suspected COVID- 19 case [Crude odds ratio (COR) 1.5 95% (1.3–1.9; p = 0.0001) and AOR 1.4 (1.1–1.8; p = 0.015)] and hav- ing symptoms compatible with COVID-19 in preceding 4 weeks [COR 1.3 (1.0–1.5] were found to be associated with seropositivity (Table 2). Discussion Interpretation of SARS-CoV-2 serologic test results, except pan Igs Wanti ELISA, has been reported to be very challenging in Africa due to pre-existing cross-reac- tive antibodies induced by other pathogens such as non- SARS-CoV-2 human coronaviruses and malaria parasites [9]. Given the rapid decline of anti-SARS-CoV-2 nucle- ocapsid antibodies as compared to the anti-RBD IgG antibody [13], we developed and optimized an in-house ELISA that detects anti-SARS-CoV-2 IgG antibodies. Our assay, unlike other commercially available serologic assays, is affordable and has been validated with a large number of Ethiopian sera from both pre-COVID-19 and COVID-19 patients from the same regions. Its sensitivity on convalescent sera from COVID-19 patients confirmed by RT-PCR was found to be as sensitive as the Wantai pan Ig ELISA (100%), and superior to Realy Tech’s IgM/ IgG LFA (90%). Also, our in-house assay displayed 97.7% specificity in randomly selected pre-COVID-19 Ethio- pian origin sera, which is superior to Realy Tech (92.5%). Seroprevalence studies provide information about the extent of individuals who had exposure to the virus and help to understand the future course of the pandemic and are key to providing target prevention and control measures in reducing transmission and severe outcomes [15]. In this study, the overall seroprevalence of SARS- CoV-2 spike RBD IgG antibodies among HWs was 39.6%, ranging from 24.5% in the Hiwot Fana Specialized Hos- pital, Harar to 48·0% in ALERT Hospital located in the capital city, Addis Ababa. This is not a surprise given Addis Ababa is the epicenter of SARS-CoV-2 transmis- sion in Ethiopia, and SARS-CoV-2 has been introduced 4 months later in Harar. As a result, it is expected that a higher proportion of HWs in hospitals located in Addis Ababa, including ALERT are more frequently exposed to COVID-19 cases than that HWs working in hospitals located in Harar, where fewer number cases and deaths had been reported. Gelanew et al. BMC Infectious Diseases (2022) 22:261 Page 6 of 9 Fig. 2 A map of Ethiopia showing the location of the study hospitals with corresponding SARS-CoV-2 seroprevalence. a Shows the location of five hospitals from which a total of 1997 healthcare workers enrolled between December 2020 and February 2021. b Shows the corresponding seroprevalence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The y-axis of Fig. 2b represents the study hospitals. The x-axis of Fig. 2b shows crude seroprevalence rates (%) with 95% confidence intervals estimated by dividing the number of participants tested seropositive for immunoglobin G (IgG) antibodies elicited against the receptor binding domain (RBD) of the spike protein of SARS-CoV-2 to the total number of participants who provided sera and were tested According to our findings, at least 4 in 10 urban Ethio- pian HWs had already been exposed to SARS-CoV-2 by February 2021 in Ethiopia. This result contrasts with a serosurvey in asymptomatic individuals from the gen- eral population conducted in March 2020 in Addis Ababa (8.8%) [16] and from the household serosurveys in Jimma (2%) and Addis Ababa (5%) that were conducted during the first wave of the pandemic-i.e., four months after the first COVID-19 case in Ethiopia [17]. Although this stark seroprevalence difference between our study and these two previous studies might be explained by differences in the types of assays employed, lack of personal protective equipment (PPE) and/or cohort types. The most plausible explanation is that the sera for the present serosurveil- lance study had been collected after the first wave of the pandemic in Ethiopia, between March 2020 and February 2021. While the high seroprevalence rates observed among the different geographically located hospitals are approaching those of high-incidence countries like Bra- zil [18], they are in agreement with several other SARS- CoV-2 seroprevalence studies from sub-Saharan Africa that, like Ethiopia, have reported much lower rates of RT-PCR confirmed cases and deaths. For example, higher anti-SARS-CoV-2 antibody seroprevalence has been reported in South Sudan (30–60.6%) [19], Democratic Republic of Congo (8–36%) [20] and Nigeria (25–45%) [21] depending on the population sampled and the sero- logical test used. Taken together these studies indicate that SARS-CoV-2 has spread widely in sub-Saharan Africa [22]. However, the majority (74.0%) of our study participants never had any symptoms compatible with COVID-19, suggesting the occurrence of significant bur- den of asymptomatic infections and its transmissions in the country, which is now, being reflected in the trend of increasing RT-PCR positivity since January 2021 (https:// covid 19. who. int/ region/ afro/ count ry/ et). The higher pro- portion of younger HWs (mean age of 34 years), and the fewer participants with comorbidities (6.7%) may have contributed to the observed high burden of asympto- matic infection among the studied HWs. Malaria, BCG- vaccination, warmer environment, and high prevalence of pre-existing cross-reactivate against HCoVs may have also contributed[3]. A report from Spain showed a higher (38.3%) sero- prevalence of SARS-CoV-2 among HWs [23]. This is comparable with the present report from Ethiopia, where there were a relatively fewer severe cases and deaths Gelanew et al. BMC Infectious Diseases (2022) 22:261 Page 7 of 9 Table 2 Adjusted odds ratios (AOR) and 95% confidence intervals for anti-SARS-CoV-2 RBD IgG antibodies positivity and associated factors, Ethiopia, 2021 Variable Hospital ALERT Hawassa Gondar Assela Hiwot Fana Sex Male Female Age (in years) 19–24 25–34 35–49 50 ≥ Contact No Yes COVID-19 Asymptomatic Symptomatic Occupation Doctor Nurse Lab Technician Administration Others AOR (95% CI) p‑value 2.7 (1.6–3.1) 2.2 (1.5–3.2) 2.8(2.0–3.9) 2.2 (1.6–3.1) 1 1 1.1 (0.92–1.4) 1.27 (0.9–1.9) 1.1 (0.9–1.5) 1 1.2 (0.8–1.8) 1 1.4 (1.1–1.8) 1 0.0001 0.0001 0.0001 0.0001 0.222 0.226 0.254 0.479 0.015 0.97 (0.8–1.2) 0.785 1 1.0 (0.8–1.4) 1.3 (0.9–1.9) 1.01 (0.8–1.5) 1.3 (0.9–1.7) 0.809 0.131 0.766 0.150 p > 0.05 shows absence of statistical significance. On the other hand, p-values in bold show presence of statistical significance: p < 0.015 and p < 0.0001 indicate statistically moderate and highly significances, respectively (https:// covid 19. who. int/ region/ afro/ count ry/ et). Simi- larly, higher seroprevalence among frontline HWs has been reported in other sub-Saharan African countries such as in Malawi [24]. These findings and ours high- light the importance of asymptomatic infections in the African countries. Interestingly, we found no seropreva- lence differences between healthcare occupations includ- ing administrative staff. The lack of a dramatic difference among frontline HWs and administrators may be a reflec- tion of the frontline administrative staff are also at high risk and are poorly protected, or may suggest the level of virus transmission in the general population at large as previously observed in UK [8]. Nevertheless, further well-designed investigations are required to implement occupation-specific public health strategies in healthcare facilities. In the present study, a history of previous close contact with a suspected or confirmed COVID-19 case was found to be strongly associated with seropositivity; however, this finding contradicts the observed similar seroposi- tivity among front line HWs and administrators. Similar odds of seropositivity among males and females were also found although several studies elsewhere reported higher odds of seropositivity in males [25]. A similar contradic- tory finding with no seroprevalence differences by sex was reported in the Spanish general population [23]. Our study has several strengths. These include its use of an in-house developed assay which we optimized to significantly minimize false positive responses by validat- ing it with both pre-pandemic and pandemic samples of Ethiopian origin. Most importantly, the study involved a relatively large sample size from five hospitals located in different geographical locations, providing much needed information about the COVID-19 pandemic in sub-Saha- ran Africa. Despite its strengths, this study has several limita- tions. First, we did not employ a random sampling tech- nique. Instead, we invited all hospital staff to take part in the study on a voluntary basis. Hence selection bias might have affected our results. For example, although we invited all staff to reach the desired sample size per cadre, some staff who had exposure to SARS-CoV-2 might not had been enrolled in our study in fear/mis- conception that being seropositive would lead them to undergo quarantine. If this were the case, it would have led to underreporting of seropositive cases. Second, recall bias might have affected the responses to the his- tory of symptoms compatible with COVID-19, and close contact with a confirmed COVID-19 case, and thereby contributed to the absence of a strong correlation between seropositivity and these covariates, albeit having close contact with COVID-19 case. Third, although the sensitivity of our assay was 100% in convalescent serum samples from COVID-19 patients, its specificity is not 100%. Our assay’s specificity, determined by using pre- COVID-19 pandemic sera/plasmata those were collected before December 2019 was found to be 97.7% and hence could lead to overestimating of HWs tested positive to anti- SARS-CoV-2 RBD IgG antibodies. For example, for every 1000 people tested by our assay, 23 HWs who never had SARS-CoV-2 infection might have been incor- rectly identified as they had antibodies specific to SARS- CoV-2. However, even this slight overestimation of the apparent seroprevalence associated with the assay speci- ficity is likely to be matched by the proportion of study participants who might be infected and yet not produce humoral immune responses at the time of blood sample collection. In conclusion, we developed an IgG ELISA that meets the WHO requirements to be uti- lized for SARS-CoV-2 serosurveillance studies. This in-house Gelanew et al. BMC Infectious Diseases (2022) 22:261 Page 8 of 9 seroprevalence study revealed a remarkably high sero- prevalence (40–48%) of SARS-CoV-2 among HWs in the five public hospitals; with slight differences amongst hos- pitals, except Hiwot Fana Specialized Hospital in which relatively lowest (24.5%) seroprevalence was found. We found no seroprevalence rate differences between front line HWs and administrative staff, indicating the observed high seroprevalence of SARS-CoV-2 might also be a reflection of the community transmission. Taken together these findings suggest extensive cryptic circu- lation (asymptomatic transmission) of SARS-CoV-2 in Ethiopia. Whether the detected anti-SARS-CoV-2 anti- bodies can persist adequately and confer protection from subsequent infections to those HWs who had or had not received COVID-19 vaccine will require further immu- nological investigation. had no role on the study design, execution, interpretation, or where these data were published. Availability of data and materials All data available for this study are presented in the manuscript. Declarations Ethics approval and consent to participate We obtained ethical approvals from Armauer Hansen Research Institute/All Africa Leprosy and Tuberculosis Rehabilitation and Training Center (ALERT Center) Hospital Ethical Review Committee (PO/32/20), Gondar University Institutional Review Board (V/P/RCS/05/00/2020) and Arsi University College of Health Sciences Institutional Review Board (A/U/H/S/C/20/6155/2C while written permissions were obtained from, Hawassa University and Haromaya University. HWs participation in the study was on the voluntary basis, and writ- ten informed consent was received from each participant in their native lan- guages. All methods were performed in accordance with Helsinki Declaration, and relevant health regulations and guidelines. Positive and negative standard samples were obtained from WHO solidarity II plasma panels developed by the United Kingdom’s National Institute for Biological Standards and Control. Abbreviations COVID-19: Coronavirus disease 2019; SARS-CoV-2: Severe acute respiratory syndrome-coronavirus-2; RBD: Receptor binding domain; RT-PCR: Reverse transcription polymerase chain reaction; BCG: Bacille Calmette Guerin; IgG: Immunoglobulin G; ELISA: Enzyme-Linked Immunosorbent Assay; WHO: World health organization; OD: Optical density; LFA: Lateral follow assay; HWs: Healthcare workers. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12879- 022- 07247-z. Additional file 1: Table S1. The Specificity of RBD IgG ELISA among pre-COVID-19 pandemic sera (n ELISA among cohort COVID-19 patients (n Table S3. Percentage of positive specimens (n tested positive for SARS-CoV-2 by DAAn RT–PCR. Table S4. Specificity of RBD IgG ELISA in pre-covid plasma/serum specimens (n 40) collected before COVID-19 pandemic. 365). Table S2. Sensitivity of RBD IgG 405) confirmed by RT-PCR. 40) from patients who = = = = Acknowledgements We thank study participants and AHRI’s SARS-CoV-2 diagnostic and research team members. We also thank Mr. Mekonnen Ashagarie, Director of American Health and Home Care, MA, USA for the non-technical support. Authors’ contributions TG, AM, AMi MA, AA and FGT conceived the study. TG, BS, AM, and AA wrote the first draft of the protocol, and revised by all authors. TG, AM, MA, AA and FGT developed the serologic assay, TAB purified the antigen. TG, BS, AM, AMi and AA coordinated the sample collection. BG, AS, YM, YM, ZT, DK, AG, DA and ET collected the blood samples and data. BT, MO, GJ, AA, AH and DT con- ducted the sample testing under the supervision of TG. TG and TH analyzed the data, TG accessed and verified the data underlying the study and take responsibility for the data. TG and AA drafted the manuscript. TG, AA and FGT were responsible for decision where to submit for publication. TG was respon- sible for submission, editing and reviewing. All authors read and approved the final manuscript. Funding This seroprevalence study was funded by Noard, and Sida core fund, and the Ethiopian Ministry of Health. Other support was obtained from the Oregon Health & Science University Innovative IDEA grant 1018784 (to FGT) and National Institutes of Health training grant T32AI747225 (to TAB). The funders Consent for publication Not applicable. Competing interests The authors have no competing interests as defined by BMC, or other interests that might be perceived to influence the results and/or discussion reported in this paper. Author details 1 Armauer Hansen Research Institute, Addis Ababa, Ethiopia. 2 Department of Medical Microbiology, School of Biomedical and Laboratory Sciences, Col- lege of Medicine and Health Sciences, University of Gondar, Gondar, Ethiopia. 3 Arsi University, Asella College of Health Sciences, Asella, Ethiopia. 4 College of Medicine and Health Sciences, Department of Medical Microbiology, Hawassa University, Hawassa, Ethiopia. 5 Department of Medical Laboratory Sciences College of Health and Medical Sciences, Haramaya University, Harar, Ethiopia. 6 Department of Molecular Microbiology and Immunology, Oregon Health & Sciences University, Portland, OR, USA. 7 Horae Gene Therapy Center, University of Massachusetts Medical School, Worcester, MA, USA. 8 Swiss Tropi- cal and Public Health Institute, Basel, Switzerland. 9 London School of Hygiene and Tropical Medicine, London, UK. Received: 1 July 2021 Accepted: 7 March 2022 References 1. Coronavirus Disease 2019 (COVID-19) – Africa CDC. https:// afric acdc. org/ covid- 19/. Accessed 18 Mar 2021. 2. Gaye B, Khoury S, Cene CW, Kingue S, N’Guetta R, Lassale C, et al. 3. Socio-demographic and epidemiological consideration of Africa’s COVID-19 response: what is the possible pandemic course? Nat Med. 2020;26:996–9. Tso FY, Lidenge SJ, Peña PB, Clegg AA, Ngowi JR, Mwaiselage J, et al. High prevalence of pre-existing serological cross-reactivity against severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) in sub-Saharan Africa. Int J Infect Dis. 2021;102:577–83. 4. Yitbarek K, Abraham G, Girma T, Tilahun T, Woldie M. The effect of Bacillus Calmette-Guérin (BCG) vaccination in preventing severe infectious res- piratory diseases other than TB: implications for the COVID-19 pandemic. Vaccine. 2020;38:6374–80. 5. Kalungi A, Kinyanda E, Akena DH, Kaleebu P, Bisangwa IM. Less Severe Cases of COVID-19 in Sub-Saharan Africa: could co-infection or a recent history of Plasmodium falciparum infection be protective? Front Immunol. 2021;12:565625. Gelanew et al. BMC Infectious Diseases (2022) 22:261 Page 9 of 9 Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. 6. Verity R, Okell LC, Dorigatti I, Winskill P, Whittaker C, Imai N, et al. Estimates of the severity of coronavirus disease 2019: a model-based analysis. Lancet Infect Dis. 2020;20:669–77. 7. Garcia-Basteiro AL, Moncunill G, Tortajada M, Vidal M, Guinovart C, Jimé- 8. 9. nez A, et al. Seroprevalence of antibodies against SARS-CoV-2 among health care workers in a large Spanish reference hospital. Nat Commun. 2020;11:3500. Treibel TA, Manisty C, Burton M, McKnight Á, Lambourne J, Augusto JB, et al. COVID-19: PCR screening of asymptomatic health-care workers at London hospital. Lancet Lond Engl. 2020;395:1608–10. Emmerich P, Murawski C, Ehmen C, von Possel R, Pekarek N, Oestereich L, et al. Limited specificity of commercially available SARS-CoV-2 IgG ELISAs in serum samples of African origin. Trop Med Int Health TM IH. 2021. https:// doi. org/ 10. 1111/ tmi. 13569. 10. Kontou PI, Braliou GG, Dimou NL, Nikolopoulos G, Bagos PG. Antibody tests in detecting SARS-CoV-2 infection: a meta-analysis. Diagnostics. 2020;10:319. 11. “Solidarity 2” global serologic study for COVID-19. https:// www. who. int/ emerg encies/ disea ses/ novel- coron avirus- 2019/ global- resea rch- on- novel- coron avirus- 2019- ncov/ solid arity-2- global- serol ogic- study- for- covid- 19. Accessed 18 Mar 2021. Isho B, Abe KT, Zuo M, Jamal AJ, Rathod B, Wang JH, et al. Persistence of serum and saliva antibody responses to SARS-CoV-2 spike antigens in COVID-19 patients. Sci Immunol. 2020;5. 12. 13. Bates TA, Weinstein JB, Farley S, Leier HC, Messer WB, Tafesse FG. Cross- reactivity of SARS-CoV structural protein antibodies against SARS-CoV-2. Cell Rep. 2021;34:108737. 14. OpenFDA. https:// open. fda. gov/ apis/ device/ covid 19ser ology/. Accessed 15 Apr 2021. 15. Clapham H, Hay J, Routledge I, Takahashi S, Choisy M, Cummings D, et al. Seroepidemiologic study designs for determining SARS-COV-2 transmis- sion and immunity. Emerg Infect Dis. 2020;26:1978–86. 16. Alemu BN, Addissie A, Mamo G, Deyessa N, Abebe T, Abagero A, et al. Sero-prevalence of anti-SARS-CoV-2 Antibodies in Addis Ababa, Ethiopia. BioRxiv. 2020;41:1821. 17. Abdella S, Riou S, Tessema M, Assefa A, Seifu A, Blachman A, et al. Preva- lence of SARS-CoV-2 in urban and rural Ethiopia: randomized household serosurveys reveal level of spread during the first wave of the pandemic. EClinicalMedicine. 2021;35:100880. 18. Buss LF, Prete CA, Abrahim CM, Mendrone A, Salomon T, de Almeida- Neto C, et al. COVID-19 herd immunity in the Brazilian Amazon. medRxiv. 2020;4:856. 19. Wiens KE, Mawien PN, Rumunu J, Slater D, Jones FK, Moheed S, et al. Seroprevalence of anti-SARS-CoV-2 IgG antibodies in Juba, South Sudan: a population-based study. MedRxiv Prepr Serv Health Sci. 2021. https:// doi. org/ 10. 1101/ 2021. 03. 08. 21253 009. 20. Mukwege D, Byabene AK, Akonkwa EM, Dahma H, Dauby N, Buhendwa J-PC, et al. High SARS-CoV-2 seroprevalence in healthcare Workers in Bukavu, Eastern Democratic Republic of Congo. Am J Trop Med Hyg. 2021;104:1526–30. 21. Olayanju O, Bamidele O, Edem F, Bola Eseile, Amoo A, Nwaokenye J, et al. SARS-CoV-2 seropositivity in asymptomatic frontline health workers in Ibadan, Nigeria. Am J Trop Med Hyg. 2020;104(1). https:// www. ajtmh. org/ view/ journ als/ tpmd/ 104/1/ artic le- p91. xml. Accessed 13 Apr 2021. 22. Usuf E, Roca A. Seroprevalence surveys in sub-Saharan Africa: what do they tell us? Lancet Glob Health. 2021;9:e724–5. 23. Pollán M, Pérez-Gómez B, Pastor-Barriuso R, Oteo J, Hernán MA, Pérez- Olmeda M, et al. Prevalence of SARS-CoV-2 in Spain (ENE-COVID): a nationwide, population-based seroepidemiological study. Lancet. 2020;396(10250):535–44. https:// doi. org/ 10. 1016/ S0140- 6736(20) 31483-5. 24. Chibwana MG, Jere KC, Kamng’ona R, Mandolo J, Katunga-Phiri V, Tembo D, et al. High SARS-CoV-2 seroprevalence in health care workers but relatively low numbers of deaths in urban Malawi. Wellcome Open Res. 2020;5:199https:// doi. org/ 10. 12688/ wellc omeop enres. 16188.2. 25. Galanis P, Vraka I, Fragkou D, Bilali A, Kaitelidou D. Seroprevalence of SARS-CoV-2 antibodies and associated factors in healthcare workers: a systematic review and meta-analysis. J Hosp Infect. 2021;108:120–34. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12913-023-09326-6	Ninsiima et al. BMC Health Services Research (2023) 23:333 https://doi.org/10.1186/s12913-023-09326-6 BMC Health Services Research RESEARCH Open Access Acceptability of integration of cervical cancer screening into routine HIV care, associated factors and perceptions among HIV-infected women: a mixed methods study at Mbarara Regional Referral Hospital, Uganda Mackline Ninsiima1, Agnes Nyabigambo2 and Joseph Kagaayi1,3 Abstract Background Integrating cervical cancer screening into routine Human Immunodeficiency Virus (HIV) care has been endorsed as an effective strategy for increasing uptake of cervical cancer screening, facilitating early detection and treatment of pre-cancerous lesions among HIV-infected women. In Uganda, this strategy has not been implemented yet in most HIV clinics. Assessing acceptability of this intervention among HIV-infected women is of great relevance to inform implementation. We assessed acceptability of integration of cervical cancer screening into routine HIV care, associated factors and perceptions among HIV-infected women enrolled in the HIV clinic at Mbarara Regional Referral Hospital. Methodology A mixed methods study utilizing explanatory sequential approach was conducted among 327 eligible HIV-infected women. Acceptability of integration of cervical cancer screening into routine HIV care was measured based on Theoretical Framework of Acceptability. Quantitative data was collected using a pre-tested question- naire. We conducted focus group discussions to explore perceptions regarding the intervention among purposively selected HIV-infected women. Modified Poisson regression with robust variance analysis was utilized to determine factors associated with acceptability of the intervention. Statistical significance was determined at p-value <0.05. Thematic analysis utilizing inductive coding was applied to analyse qualitative data. Results The majority of HIV-infected women (64.5%) accepted integration of cervical cancer screening into routine HIV care. Religion, perceived risk of developing cervical cancer and ever screened for cervical cancer were statistically significantly associated with acceptability of integration of cervical cancer screening into routine HIV care. Perceived benefits of the proposed intervention were: convenience to seek for cervical cancer screening, motivation to undergo cervical cancer screening, improved archiving of cervical cancer screening results, confidentiality of HIV patient infor- mation, and preference to interact with HIV clinic health workers. Shame to expose their privacy to HIV clinic health workers and increased waiting time were the only perceived challenges of the integrated strategy. Correspondence: Mackline Ninsiima nmackline@musph.ac.ug Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 2 of 16 Conclusion Study findings highlight the need to take advantage of this acceptability to prioritize implementation of integration of cervical cancer screening into routine HIV care. HIV-infected women should be reassured of confiden- tiality and reduced waiting time to increase uptake of integrated cervical cancer screening and HIV services among HIV-infected women along the continuum of HIV care and treatment services. Keywords Acceptability, Integration, Cervical Cancer Screening, HIV Background Globally, cervical cancer is the fourth most frequently diagnosed cancer and fourth leading cause of cancer death in women with an estimated 570,000 incident cases and 311,000 deaths [1]. World Health Organisa- tion (WHO) reported a high mortality rate from cervical cancer globally at an age-standardized rate of 6.9/100,000 [2]. In low and medium Human Development Index regions, cervical cancer ranks second to breast cancer among females at world age-standardized incidence and mortality rates of 18.2/100,000 and 12.0/100,000 respec- tively [1]. In Africa, cervical cancer is the most common cause of cancer accounting for 22% of all female cancers [3]. Of note, 34 out of every 100,000 women are diag- nosed with cervical cancer, and 23 out of every 100,000 women die from cervical cancer every year [3]. The cer- vical cancer incidence and mortality rates have remained high in Sub-Saharan Africa [1, 4]. In Eastern Africa, age- standardized incidence and mortality rates of cervical cancer are 40.1/100,000 and 30.0/100,000 respectively [1]. In Uganda, prevalence of human papillomavirus among women is 33.6% with one of the highest cervical cancer incidence rates in the world of 47.5 per 100,000 per year [5]. Human Immunodeficiency Virus (HIV) has been asso- ciated with high vulnerability of developing cervical can- cer, a great public health challenge. Since 1989, research studies have reported an increased incidence of cervical cancer among HIV-infected women [6]. HIV-infected women are at increased risk of new and persistent human papillomavirus infections hence an accelerated advance- ment and incidence of cervical cancer compared to HIV uninfected women [2, 7–12]. HIV-infected women develop cervical cancer 5 to 10 years earlier compared to HIV uninfected women [2]. Susceptibility to human pap- illomavirus infection and progression to cervical cancer is highly attributed to weakened cellular immunity among HIV-infected women [2, 6, 8]. Due to the association between HIV and cervical cancer, WHO recommended that all sexually active girls and women should undergo cervical cancer screening as soon as they are diagnosed HIV positive and further emphasized regular screening and treatment of pre-cancer lesions among HIV infected women [13]. With alignment to WHO guidelines, cervi- cal cancer screening using Visual Inspection with Acetic acid and treatment of pre-cancerous cervical lesions by cryotherapy was recommended for all HIV-infected sexually active girls and women at enrolment into HIV care and repeated annually in Uganda [14]. Neverthe- less, reports have indicated persistent low uptake of cervical cancer screening among this highly susceptible population [15, 16]. A nationally representative popula- tion-based survey carried out in Uganda reported that only 30.3% HIV-infected women had received cervical cancer screening [17]. Integration of cervical cancer screening into routine HIV care has been endorsed as an effective strategy to increase uptake of cervical cancer screening, achieve early detection and treatment of pre-cancerous lesions among HIV-infected women in Sub Saharan Africa where cervical cancer burden parallels that of HIV [4, 15–22]. In Uganda, this strategy has not been imple- mented yet in most HIV clinics. It has been noted that published studies from Uganda explored perceptions and preferences of health care providers, policy makers, and community members including women, men, and village health teams, regarding integration of HIV and cervical cancer screening services in a single-visit approach with- out focussing on HIV-infected women [23, 24]. Utilizing these findings might be difficult to understand percep- tions of HIV-infected women enrolled in HIV care which would be instrumental in guiding implementation phase. Furthermore, acceptability of this intervention has not been assessed among HIV-infected women using Theo- retical Framework of Acceptability (TFA); the approved framework for assessing acceptability of healthcare interventions. This indicates the need to generate scien- tific evidence regarding acceptability of the intervention among targeted recipients. Assessing acceptability of this proposed intervention is of great relevance to inform implementation into comprehensive HIV programming [25]. HIV-infected women enrolled in the HIV clinic are either referred or voluntarily undergo cervical cancer screening conducted by health workers in the cervi- cal cancer screening department at Mbarara Regional Referral Hospital. Despite establishing referral or cli- ent-initiated cervical cancer screening system at Mba- rara Regional Referral Hospital for the last 10 years, only 11.3% of HIV-infected women enrolled in the Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 3 of 16 HIV clinic at Mbarara Regional Referral Hospital had undergone cervical cancer screening by September 30, 2019. HIV-infected women miss cervical cancer screening opportunities for early detection of pre-can- cerous cervical lesions at enrolment and during subse- quent annual intervals despite their frequent visits to HIV clinics for medical reviews, viral load monitoring, and monthly Anti–Retroviral Treatment (ART) refills. Such missed opportunities increase risk of presenting late with advanced cervical cancer and poor progno- sis among this vulnerable population. Implementation of integration of cervical cancer screening into routine HIV care is expected to increase uptake of cervical can- cer screening services at enrolment and during sub- sequent annual HIV clinic visits among HIV-infected women enrolled in the HIV clinic at Mbarara Regional Referral Hospital compared to currently practiced referral or client-initiated cervical cancer screening system. We assessed acceptability of integration of cervical cancer screening into routine HIV care, asso- ciated factors and perceptions among HIV-infected women enrolled in the HIV clinic at Mbarara Regional Referral Hospital. Methods Study site The study was conducted in the HIV clinic at Mbarara Regional Referral Hospital in Mbarara District, South Western Uganda. Cervical cancer is the leading gynae- cological cancer at Mbarara Regional Referral Hospi- tal [26]. With a proportion of 25.2%, cervical cancer is the single leading cancer; contributing to 10.1% of all diseases on the gynaecological ward and 73.9% of all gynaecological cancers at Mbarara Regional Refer- ral Hospital [26]. The HIV clinic in Mbarara Regional Referral Hospital has provided HIV care since Novem- ber 1998. The clinic has two sections: adult care section, under the Department of Internal Medicine and paediat- ric and adolescents care section under the Department of Paediatrics. A total of 7,212 HIV-infected women were reported to be active in HIV care by September 30, 2019. On average 168 HIV-infected women receive HIV care and treatment services on a typical clinic day. HIV-infected women enrolled in the HIV clinic are either referred or voluntarily undergo cervical cancer screening conducted by health workers in the cervical cancer screening department at Mbarara Regional Refer- ral Hospital. It is on these grounds that the HIV clinic at Mbarara Regional Referral Hospital was selected to study acceptability of integration of cervical cancer screening into routine HIV care, associated factors and perceptions among HIV-infected women. Study design Mixed methods study design utilizing explanatory sequential approach was used to assess acceptability of integration of cervical cancer screening into routine HIV care, associated factors and perceptions among HIV-infected women. Using explanatory sequential approach, the quantitative phase of the study was first conducted followed by analysis of data. HIV-infected women were then selected based on generated quanti- tative results of acceptability of integration of cervical cancer screening into routine HIV care; to participate in qualitative investigations. Study population This study was conducted among HIV-infected women receiving HIV care and treatment services from the HIV clinic at Mbarara Regional Referral Hospital. Only HIV-infected women aged 18 and above who turned up for ART refills at the HIV clinic on interview dates were included in the study. HIV-infected women aged 18 and above who turned up for ART refills on unscheduled dates were excluded based on electronic appointment lists generated for the respective appointment dates. Rationale HIV-infected women who turned up on unsched- uled dates were excluded because they were not on the appointment list from which study respondents were sys- tematically selected to participate in the study. Quantitative phase Measurement of study variables Dependent variable The dependent variable for this study was acceptability of integration of cervical cancer screening into routine HIV care. The Theoretical Frame- work of Acceptability was adapted to measure acceptabil- ity of integration of cervical cancer screening into routine HIV care. The Theoretical Framework of Acceptability was proposed as a multi-component framework and systematic approach to assess intervention acceptability across prospective, concurrent, and retrospective tem- poral perspectives [25]. Questions based on Theoretical Framework of Acceptability constructs namely: affective attitude, burden, perceived effectiveness, ethicality, inter- vention coherence, opportunity costs, and self-efficacy were administered to eligible participants. Responses were based on a 5-point rating ordinal scale per con- struct. The summated score of constructs of acceptability per study participant was computed by summing weights Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 4 of 16 assigned to construct responses; henceforth a continuous dependent variable. Information on sociodemo- Independent variables graphic characteristics, awareness of cervical cancer, knowledge of risk factors of cervical cancer, knowledge of symptoms of cervical cancer, cervical cancer screening awareness, uptake of cervical cancer screening, and per- ceived risk of cervical cancer was obtained. Sociodemo- graphic characteristics included age, marital status, high- est education level, religion, area of residence, number of living biological children, and duration of HIV disease. To assess knowledge of risk factors and symptoms of cer- vical cancer, participants were requested to select from response options “1. Yes” “2. No” for each item. A correct answer was awarded “1” whereas a wrong answer was awarded “0”. According to the African Women Awareness of CANcer tool, codes to response options were assigned without any meaning. So, out of two options, if the par- ticipant selected an appropriate answer, one received an award of "1" for the item because it was the correct answer. If the participant selected an inappropriate answer, one received an award of "0" for the item because it was the wrong answer. A summated score was obtained for variables: knowledge of risk factors and knowledge of signs and symptoms of cervical cancer. Knowledge of risk factors of cervical cancer was catego- rised into two categories: “1. Poor Knowledge” and “2. Good Knowledge”; based on the mean score as evidenced in recent published studies [27, 28]. “1. Poor Knowledge” was awarded to HIV-infected women whose summated score of weights of items of knowledge of risk factors of cervical cancer was less than the mean score value of 8.34. “2. Good Knowledge” was awarded to HIV-infected women whose summated score of weights of items of knowledge of risk factors of cervical cancer was greater than or equal to the mean score value of 8.34. Knowledge of signs and symptoms of cervical cancer was also categorised into two categories: “1. Poor Knowl- edge” and “2. Good Knowledge”; based on the mean score as demonstrated in recent published studies [27, 28]. “1. Poor Knowledge” was awarded to HIV-infected women whose summated score of weights of items of knowledge of signs and symptoms of cervical cancer was less than the mean score value of 7.99. “2. Good Knowledge” was awarded to HIV-infected women whose summated score of weights of items of knowledge of signs and symptoms of cervical cancer was greater than or equal to the mean score value of 7.99. Sample size and sampling For the quantitative component, sample size was calcu- lated using sample size estimation formula [29, 30]. The sample size was determined by assuming a 2-sided type 1 error rate of 5%. The calculated standard deviation in cervical cancer screening was 8.75 based on sample size estimation for epidemiological studies [30]. Assuming a marginal error of less than one, the minimum number of participants required was 294. After considering a 10% non-response rate, the resulting sample size was 327 par- ticipants. Systematic sampling method was used to select participants among HIV-infected women who had been scheduled to turn up at the HIV clinic for ART refills based on respective appointment dates. Data collection Quantitative data were collected using administered questionnaires. The pre-coded questionnaire was pre- tested at Kawaala Health Centre IV to check for suit- ability of various aspects such as translations, skip procedures, filtering questions and modifications were made thereafter. The questionnaire was translated into Runyankole/Rukiiga. Selection and recruitment of experienced research assistants was based on compe- tence, quantitative data collection skills and qualifica- tions. Research assistants received training on how to administer study questions before data collection. After meeting selected participants to take part in the study, research assistants introduced themselves and explained the purpose of the study. Eligible participants provided written informed consent before the interview was con- ducted. Each questionnaire was completed within 15 - 20 minutes. During data collection, questionnaires were reviewed on completion of each interview so that correc- tions are made in addition to checking for completeness before departure of participants. Supervision was con- ducted to ensure compliance throughout the study. Data management Data capture screens with in-built checks for consistency, logical flow, range and accuracy of data were designed in EpiData version 3.0 for data entry. Data were cleaned and stored daily. Double-entry of data was done to check for any errors. Stored data was backed up on different flash discs. Questionnaires were kept securely throughout the study and thereafter. Data analysis Linear regression analysis was the recommended method of modelling since the dependent variable, acceptability of integration of cervical cancer screening into routine HIV care, was a continuous variable. Summated scores of Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 5 of 16 acceptability of integration of cervical cancer screening into routine HIV care were converted into percentages. The regression modelling predicting acceptability of inte- gration of cervical cancer screening into routine HIV care from all independent variables was conducted. Residuals were generated followed by testing assumptions of linear regression: normality of residuals and homogeneity of variance. Linear regression analysis was not utilised for this study because both assumptions of normality and homogeneity of variance of residuals were violated. Acceptability of integration of cervical cancer screen- ing into routine HIV care was dichotomized into two categories: “0. Not Accepted” and “1. Accepted” based on 26.25, the 75th percentile of the highest possible sum- mated score of 35 from seven constructs of Theoreti- cal Framework of Acceptability. The 75th percentile was recommended as the cut-off value based on a published study which utilised likert-type scale responses assess- ing constructs to measure composite dependent variable [31]. “0. Not Accepted” was awarded to HIV-infected women whose summated score of weights of constructs of acceptability of integration of cervical cancer screening into routine HIV care was less than 26.25. “1. Accepted” was awarded to HIV-infected women whose summated score of weights of constructs of acceptability of integra- tion of cervical cancer screening into routine HIV care was greater than or equal to 26.25. Frequencies and per- centages of HIV-infected women who had accepted and those who had not accepted integration of cervical can- cer screening into routine HIV care were computed. Univariable analysis All variables were analysed to describe study respond- ents. For independent categorical variables, data were presented as frequencies and percentages. Bivariable analysis Cross tabulations of outcome variable and independ- ent variables were done to obtain frequencies and cor- responding percentages of HIV-infected women who accepted and those who did not accept the integration of cervical cancer screening into routine HIV care per independent variable. Selection of regression model to use at bivariable analysis was conducted. Using occupa- tion as an example, logistic regression, log binomial and Modified Poisson regression models were applied. Logis- tic regression and log binomial models overestimated measures of association compared to Modified Poisson regression model. Given that the outcome had a higher percentage of 64.5%, Modified Poisson regression with robust variance was selected for bivariable analysis to obtain crude prevalence ratios, 95% confidence interval, and corresponding p-values. Multivariable analysis Multicollinearity among independent variables was checked for using variance inflation factor. Regres- sion modelling predicting acceptability of integration of cervical cancer screening into routine HIV care from all independent variables was conducted and then vif. command applied to check for multicollinearity. Vari- ance inflation factors for all independent variables were less than 2.5. Ultimately, all independent variables were included in multivariable analysis. Modified Poisson regression with robust variance using stepwise logi- cal model building technique was conducted to obtain adjusted prevalence ratios, 95% confidence interval and corresponding p-values. Checking for goodness of fit of the model was done based on Akaike Information Crite- ria (AIC). Significance of independent variables was set at p-value <0.05. Statistical analysis was conducted using Stata/SE Version 14.0. Qualitative phase Sample size and sampling With reference to the 5–point likert-type scale, accept- ability of integration of cervical cancer screening into routine HIV care was categorised into five categories: “1. Very Unacceptable”, “2. Unacceptable”, “3. Neutral”, “4. Acceptable” and “5. Very Acceptable” based on sum- mated score of weights of constructs of acceptability. “1. Very Unacceptable” was awarded to respondents whose summated score of weights of constructs of acceptabil- ity was 7 and below. “2. Unacceptable” was awarded to respondents whose overall score of weights of constructs of acceptability was 8-14. “3. Neutral” was awarded to respondents whose summated score of weights of con- structs of acceptability was 15-21. “4. Acceptable” was awarded to respondents whose overall score of weights of constructs of acceptability was 22-28. “5. Very Accepta- ble” was awarded to respondents whose summated score of weights of constructs of acceptability was 29 and above out of the 35 criteria. Based on the category of acceptability of integration of cervical cancer screening into routine HIV care, 6 focus group discussions were conducted among HIV- infected women. The number of focus group discussions per level of acceptability of integration of cervical cancer screening into routine HIV care was based on number of HIV-infected women within respective categories. Sample size of 6 focus group discussions was further determined by data saturation point; a point at which further sampling did not generate any new concepts or ideas about the phenomenon under investigation. Pur- posive sampling was used to select HIV-infected women from the quantitative phase to participate in focus group discussions. Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 6 of 16 Data collection The focus group discussion guide was specifically devel- oped for this study with questions aimed at eliciting HIV- infected womens’ perceptions of integration of cervical cancer screening into routine HIV care. The focus group discussion guide was pre-tested and recommended changes were made to ensure that the guide captured relevant and appropriate information before use. Focus group discussion guide was translated into Runyankole/ Rukiiga, the commonly used local language among study respondents. Key questions in the focus group discussion guide explored participants’ impressions, anticipated benefits, and challenges of integration of cervical cancer screening into routine HIV care in a single visit approach in HIV clinics in Uganda. An experienced research assis- tant at collecting qualitative data through conducting focus group discussion was recruited. Data management Focus group discussions were audio-recorded. The col- lected audio-recorded data were adequately and appro- priately backed up. All audio recorded local language interviews were translated into English and transcribed verbatim simultaneously. The transcripts were proof-read before importing them into Atlas.ti Version 6.0, a qualita- tive data management software. Data analysis Thematic analysis method using inductive coding was used to analyse qualitative data. Exploration of data and synthesis of codes, subthemes and themes was done in Atlas.ti Version 6.0. Relevant verbatim quotations were selected as evidence to support generated themes. Results Socio‑demographic characteristics of study respondents Table 1 shows characteristics of study respondents. The majority of respondents (33.0%) were aged 40 – 49. Based on self-reports, 59.0% of respondents had lived with HIV for 10 years and above since they were diagnosed HIV positive. Of note, 75.8% reported that they were aware about cervical cancer screening programme whereas 65.1% had undergone cervical cancer screening. Acceptability of integration of cervical cancer screening into routine HIV care The majority of HIV-infected women (64.5%) accepted integration of cervical cancer screening into routine HIV care. Of note, 35.5% did not accept integration of cervical cancer screening into routine HIV care. Factors associated with acceptability of cervical cancer screening into routine HIV care Table 2 shows adjusted prevalence ratios with corre- sponding confidence intervals. At multivariable analysis, based on 95% confidence intervals, religion, perceived risk of developing cervical cancer and ever screened for cervical cancer were statistically significantly associ- ated with acceptability of integration of cervical cancer screening into routine HIV care among HIV-infected women at Mbarara Regional Referral Hospital. Mus- lims were 47% more likely to accept integration of cer- vical cancer screening into routine HIV care compared to Protestants. HIV-infected women with “Much Above Average” perceived risk of developing cervical cancer were 43% more likely to accept integration of cervi- cal cancer screening into routine HIV care compared to HIV-infected women with “Much Below Average” per- ceived risk of developing cervical cancer. HIV-infected women who had not undergone cervical cancer screen- ing were 29% less likely to accept integration of cervi- cal cancer screening into routine HIV care compared to HIV-infected women who had undergone cervical cancer screening. Perceptions of HIV infected women regarding integration of cervical cancer screening into routine HIV care Characteristics of focus group discussion participants Table 3 shows characteristics of Focus Group Discus- sion (FGD) participants. FGD participants were grouped into 6 groups based on level of acceptability to ensure homogeneity of groups. FGD 1 comprised of 2 and 4 par- ticipants from “Very Unacceptable” and “Unacceptable” levels of acceptability respectively. FGD 2 comprised of 6 participants from “Neutral” level of acceptability. FGD 3 comprised of 6 participants from “Acceptable” level of acceptability. FGD 4 comprised of 6 participants from “Acceptable” level of acceptability. FGD 5 comprised of 6 participants from “Very Acceptable” level of acceptability. FGD 6 comprised of 6 participants from “Very Accept- able” level of acceptability. Table 4 shows codes, sub-themes and themes of per- ceptions of HIV-infected women regarding integration of cervical cancer screening into routine HIV care. Percep- tions of HIV-infected women were presented in two the- matic areas: i) perceived benefits of integration of cervical cancer screening into routine HIV care and ii) perceived challenges of integration of cervical cancer screening into routine HIV care. Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 7 of 16 Table 1 Characteristics of study respondents Study variables Completed age 18-29 30-39 40-49 50 and above Residence Rural Urban Highest education level None Primary Education Secondary or Higher Education Marital Status Not Married Married Religion Anglican Catholic Muslim Others (Pentecostal & SDA) Occupation Not working Employed (Paid) Self Employed (Businesswoman) Self Employed (Agriculture) Number of children None 1 – 3 4 and above HIV Duration (Number of years since diagnosed HIV positive) 1 – 4 5 – 9 10 and above Awareness of cervical cancer Yes No Knowledge of risk factors of cervical cancer Poor Knowledge Good Knowledge Knowledge of signs and symptoms of cervical cancer Poor Knowledge Good Knowledge Perceived risk of developing cervical cancer Much Below Average Below Average Average Above Average Much Above Average Frequencies (n 327) = Percentages (%) 43 91 108 85 197 130 70 147 110 205 122 153 107 34 33 28 29 115 155 26 183 118 50 84 193 311 16 149 178 119 208 67 28 31 31 170 13.2 27.8 33.0 26.0 60.2 39.8 21.4 45.0 33.6 62.7 37.3 46.8 32.7 10.4 10.1 8.6 8.9 35.1 47.4 8.0 56.0 36.0 15.3 25.7 59.0 95.1 4.9 45.6 54.4 36.4 63.6 20.4 8.6 9.5 9.5 52.0 Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 8 of 16 Table 1 (continued) Study variables Awareness of cervical cancer screening Yes No Ever screened for cervical cancer Yes No Frequencies (n 327) = Percentages (%) 248 79 213 114 75.8 24.2 65.1 34.9 Perceived benefits of integration of cervical cancer screening into routine HIV care Convenience to seek cervical cancer screening services Convenience to seek cervical cancer screening services was the most predicted advantage of integration of cer- vical cancer screening into routine HIV care. In all focus group discussions, HIV-infected women acknowledged that integration of cervical cancer screening into routine HIV care would grant them an opportunity to receive both cervical cancer screening in addition to HIV care and treatment services from the HIV clinic compared to currently practiced referral or client – initiated cervical cancer screening conducted in the cervical cancer screen- ing unit at Mbarara Regional Referral Hospital. HIV- infected women in the majority of focus group discussions revealed that referral or client – initiated cervical cancer screening services have been quite hectic and inconven- iencing to an extent that some of them have missed sev- eral opportunities to undergo cervical cancer screening; hence supporting integrated mode of delivery of cervical cancer screening and HIV related services. Furthermore, HIV-infected women claimed that the proposed strategy of delivering cervical cancer screening would reduce dis- ruption and movement from one clinic to another since all cervical cancer screening and HIV related services would be available and received under one roof in a single HIV clinic visit comprehensively. “Sending you down there wouldn’t have a problem but it can somehow disturb you …, but if they are here, you can know that I am going to pick drugs and then test so I get all services from one place...” (Respondent 1, FGD 6 _ Very Acceptable) HIV-infected women in most focus group discussions affirmed that convenience attributed to integration of cervical cancer screening into routine HIV care would eventually save time spent at the health facility compared to currently delivered stand-alone cervical cancer screen- ing and HIV services. “… because when she comes here, she has to pick drugs and then slope down there or first slope there and hurry back to get drugs and in this she may be caught by time, or even by the time she gets down, she may find that they have already closed but if those services are here, she can get drugs and later enter the room and they test for cancer of the cervix.” (Respondent 1, FGD 3 _ Acceptable) HIV-infected women in most focus group discussions believe that scenarios of missing HIV care and treat- ment services due to disruption of seeking cervical cancer screening services from the cervical cancer screening unit at Mbarara Regional Referral Hospital would be minimised by implementation of integrated cervical cancer screen- ing and HIV related services. It was further revealed that integrated services would save time since a client would easily receive both HIV and cervical cancer screening ser- vices in a single HIV clinic compared to spending a lot of time while travelling to the health facility twice; to seek for cervical cancer screening and HIV related services from two separate clinics on different days which would in due course reduce travel costs. “Sending you down there wouldn’t have a problem but it can somehow disturb you because sometimes you can be home without money waiting to use the one you have the time you are coming to pick drugs, ... It can disturb you a bit if you go down there to get tested and then by the time you come back, you find the health workers here have closed. And you have to come back the next day to pick drugs… but if they are here, you can know that I am going to pick drugs and then test so I get all services from one place, I test for cancer of the cervix and at the same time, pick my drugs on the same day.” (Respondent 1, FGD 5 _ Very Acceptable) Motivation to undergo cervical cancer screening HIV-infected women in all focus group discussions reported that integration of cervical cancer screen- ing into routine HIV care would motivate HIV-infected women to undergo cervical cancer screening services. Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 9 of 16 Table 2 Factors associated with acceptability of integration of cervical cancer screening into routine HIV care among HIV-infected women Independent variables Acceptability of integration of cervical cancer screening into routine HIV care Not Accepted n (%) Accepted n (%) Adjusted prevalence ratios 95% confidence intervals Completed age 18-29 30-39 40-49 50 and above Residence Rural Urban Highest education level None Primary Education Secondary or Higher Education Marital status Not Married Married Religion Protestant Catholic Muslim Others (Pentecostal & SDA) Occupation None Employed (Paid) Self Employed (Business) Self Employed (Agriculture) Number of children None 1 – 3 4 and above 19 (44.2) 30 (33.0) 37 (34.3) 30 (35.3) 77 (39.1) 39 (30.0) 28 (40.0) 58 (39.5) 30 (27.3) 75 (36.6) 41 (33.6) 66 (43.1) 34 (31.8) 5 (14.7) 11 (33.3) 7 (25.0) 7 (24.1) 38 (33.0) 64 (41.3) 13 (50.0) 50 (27.3) 53 (44.9) HIV Duration (Number of years since diagnosed HIV positive) 1 – 4 5 – 9 10 and above Awareness of cervical cancer Yes No 24 (48.0) 26 (31.0) 66 (34.2) 104 (33.4) 12 (75.0) Knowledge of risk factors of cervical cancer Poor Knowledge Good Knowledge 61 (40.9) 55 (30.9) Knowledge of signs and symptoms of cervical cancer Poor Knowledge Good Knowledge 52 (43.7) 64 (30.8) Perceived risk of developing cervical cancer Much Below Average Below Average Average 33 (49.3) 10 (35.7) 15 (48.4) 24 (55.8) 61 (67.0) 71 (65.7) 55 (64.7) 120 (60.9) 91 (70.0) 42 (60.0) 89 (60.5) 80 (72.7) 130 (63.4) 81 (66.4) 87 (56.9) 73 (68.2) 29 (85.3) 22 (66.7) 21 (75.0) 22 (75.9) 77 (67.0) 91 (58.7) 13 (50.0) 133 (72.7) 65 (55.1) 26 (52.0) 58 (69.0) 127 (65.8) 207 (66.6) 4 (25.0) 88 (59.1) 123 (69.1) 67 (56.3) 144 (69.2) 34 (50.7) 18 (64.3) 16 (51.6) Ref. 1.11 1.16 1.15 Ref. 1.13 Ref. 0.88 0.99 Ref. 0.97 Ref. 1.16 1.47 1.14 Ref. 1.03 0.91 0.81 Ref. 1.25 0.95 Ref. 1.18 1.08 Ref. 0.46 Ref. 1.01 Ref. 1.14 Ref. 1.27 1.13 [0.83 – 1.49] [0.86 – 1.56] [0.86 – 1.55] [0.97 – 1.31] [0.70 – 1.11] [0.78 – 1.26] [0.82 – 1.15] [0.96 – 1.39] [1.21 – 1.78] *** [0.88 – 1.48] [0.80 – 1.33] [0.73 – 1.14] [0.64 – 1.03] [0.86 – 1.83] [0.63 – 1.43] [0.89 – 1.58] [0.81 – 1.44] [0.20 – 1.02] [0.85 – 1.20] [0.96 – 1.35] [0.91 – 1.78] [0.75 – 1.70] Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 10 of 16 Table 2 (continued) Independent variables Above Average Much Above Average Awareness of cervical cancer screening Yes No Ever screened for cervical cancer Yes No * p<0.05 p<0.01 * p<0.001 Acceptability of integration of cervical cancer screening into routine HIV care Adjusted prevalence ratios 95% confidence intervals Not Accepted n (%) 11 (35.5) 47 (27.6) 76 (30.7) 40 (50.6) 59 (27.7) 57 (50.0) Accepted n (%) 20 (64.5) 123 (72.4) 172 (69.3) 39 (49.4) 154 (72.3) 57 (50.0) 1.23 1.43 Ref. 1.03 Ref. 0.71 [0.87 – 1.73] [1.11 – 1.85] [0.74 – 1.45] [0.58 – 0.86] Table 3 Characteristics of focus group discussion participants Study variables Frequencies (n 36) = Percentages (%) Completed Age Age [Mean Age (SD)] 20-29 30-39 40-49 50 and above Residence Rural Urban 40.0 (9.8) 7 10 12 7 16 20 Highest Education Level None Primary Education Secondary or Higher Education Occupation Not working Employed (Paid) Self Employed (Businesswoman) Self Employed (Agriculture) 6 17 13 6 6 11 13 19.4 27.9 33.3 19.4 44.4 55.6 16.7 47.2 36.1 16.7 16.7 30.5 36.1 Level of acceptability of integration of cervical cancer screening into routine HIV care Very Unacceptable Unacceptable Neutral Acceptable Very Acceptable 2 4 6 12 12 5.6 11.1 16.7 33.3 33.3 This would ultimately increase uptake of cervical cancer screening among HIV-infected women; hence achieving the primary goal of implementing integration of cervical cancer screening into routine HIV care. “When they bring the cervical cancer screening clinic here, it will be really so good for it will give us morale to access and undergo cervical cancer screen- ing.” (Respondent 2, FGD 2 _ Neutral) Furthermore, HIV-infected women in the major- ity of the focus group discussions alleged that integra- tion of cervical cancer screening into routine HIV care would inspire HIV-infected women, targeted recipients to comply and adhere to recommended annual cervi- cal cancer screening services. Since documentation about when one last conducted cervical cancer screen- ing would be captured in the same file with all other HIV related information, it would be easier for health workers to remind clients about the appointment date for conducting the next cervical cancer screening procedure. “… when you are coming to pick drugs, they also test you from here is really so good because I went down there and they tested me. They told me to go back … but like how you said it at the beginning, I have never gone back. But if it is here and I have come to pick drugs or I have come for treatment, they can check in my file and remind me that on such a date, you are supposed to come and they test you.” (Respondent 1, FGD 4 _ Acceptable) Improved archiving of cervical cancer screening results HIV-infected women in some focus group discussions acknowledged that integration of cervical cancer screen- ing into routine HIV care would enhance improved record keeping of cervical cancer screening results. This was highly attributed to the fact that documenta- tion about when the client last conducted cervical can- cer screening would be captured in the same file with all other HIV related information. Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 11 of 16 Table 4 Codes, sub-themes and themes Codes Sub ‑Themes Theme - Receiving HIV drugs and cervical cancer screening ser- vices on the same scheduled date - Receiving HIV care and cervical cancer screening services from the same place - Reduced disturbance and movements seeking for HIV care and cervical cancer screening services from different clinics - Increased awareness about cervical cancer and the rationale to undergo cervical cancer screening services - Early detection and treatment of precancerous lesions - Compliance and adherence to undergo cervical cancer screening services annually - Increased opportunities to undergo cervical cancer screening services - Record keeping of cervical cancer screening results and HIV related information in one HIV patient’s file - Reduction in HIV stigma from the HIV uninfected women and non-HIV health workers at the cervical cancer screen- ing unit - Privacy of the HIV positive status - Free interaction with HIV health workers whom the HIV- infected women are used to - Fear to disclose the HIV status to non-HIV health workers - Ashamed to expose private parts to health workers who have known the HIV-infected women - Preference to open up and share experiences to health workers who don’t know them - Fear to continue interacting with the same health work- ers they have exposed their private parts for any other HIV related services - Increase on time spent at the HIV clinic to undergo cervi- cal cancer screening - Delay at the HIV clinic to receive both HIV and cervical cancer screening services Convenience to seek cervical cancer screening services Perceived benefits of integra- tion of cervical cancer screen- ing into routine HIV care Motivation to undergo cervical cancer screening Improved archiving of cervical cancer screening results Confidentiality of HIV patient information Preference to interact with HIV clinic health workers Shame to expose their privacy to HIV clinic health workers Increased waiting time Perceived challenges of integration of cervical cancer screening into routine HIV care “For me I see that the beauty about testing me from here is that everything will be done from here and all will be put in one file and anyone who gets your file will have to see all the diseases that you have.” (Respondent 5, FGD 5 _ Very Acceptable) HIV-infected women in few focus group discussions anticipated that undergoing cervical cancer screening from the HIV clinic would make it quite easier for follow up of cervical cancer screening results. “You see when we are examined from this our clinic, even when results don’t come there and then, …, when you come back, you know that you will find your results in your file and still the health worker will have the responsibility of telling you what the outcome of the results was, ….” (Respondent 3, FGD 5 _ Very Acceptable) Confidentiality of HIV patient information HIV-infected women in the majority of focus group dis- cussions claimed that the proposed intervention would promote confidentiality of HIV status and any other health related information due to interaction with only health workers and fellow HIV-infected women; as opposed to the current situation where HIV-infected women are forced to interact with other health work- ers and HIV uninfected women at the cervical cancer screening unit of Mbarara Regional Referral Hospital. Confidentiality was very much appreciated as a great attribute of integrated delivery of cervical cancer screen- ing comprehensively with all other routine HIV care ser- vices due to reduced stigma. “You see when they are sending us down there; they send us with our files and remember there are people who have come so early to test for cancer of the cer- vix. …, they see people bombarding them and then they start to ask, where are these ones coming from? Ahh they are HIV positive women. You see them Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 12 of 16 pointing fingers, ... But if it has come to our clinic and you know that if I have come to test for cancer of the cervix I will find services at the HIV clinic, I will also be confident wherever I will be seated because I will know that the people I am seated with are my fellows.” (Respondent 2, FGD 6 _ Very Acceptable) Preference to interact with HIV clinic health workers HIV-infected women in most focus group discussions appreciated integrated strategy whereby health work- ers from the HIV clinic will perform cervical cancer screening in addition to other HIV related services. HIV-infected women preferred receiving cervical cancer screening conducted by HIV clinic health workers to cur- rently practiced system where cervical cancer screening procedure is conducted by health workers at the cervical cancer screening unit of Mbarara Regional Referral Hos- pital. HIV-infected women revealed that they can freely share their experiences or any health-related issues with health workers at the HIV clinic due to strong relation- ships built over time as opposed to health workers at the cervical cancer screening unit of Mbarara Regional Referral Hospital. “..., me I think that if they bring them here, it will be easy for us …, not like down there because you find in most cases, when they would send us there, the health workers would look at you as if there is some- thing on you, they would all be neglecting to attend to you, because the people down there may not be as free to you as this one who can be here, you can explain everything to her ...” (Respondent 3, FGD 3 _ Acceptable) Perceived challenges of integration of cervical cancer screening into routine HIV care Shame to expose their privacy to HIV clinic health workers HIV-infected women reported that they would prefer receiving cervical cancer screening conducted by health workers at the cervical cancer screening unit of Mbarara Regional Referral Hospital as opposed to health work- ers at the HIV clinic. They argued that due to having interacted for quite a very long time, they would always be overwhelmed with fear and shame to continue inter- acting with the same health workers to whom they have exposed their private parts for any other HIV related services. They confessed that they would honestly prefer health workers at the cervical cancer screening unit of Mbarara Regional Referral Hospital because they rarely interact with them; hence no fear and shame to expose their private parts. “… ever since l started picking drugs from here, the nurse I found here is the one still there, the doctor I found there is the one still there, but there is a way we are created in our private parts, if someone has ever seen you, you feel ashamed as if she knows how you are exactly, as if she will tell others that so and so is like this and that, but down there where we go, sometimes you find they have changed them, you find that those who are always there are very dif- ferent. And even by the time you go there, they can’t be remembering you because you go there once, but here, we come every month, every after two months, someone knows how you are created down, but down there, someone cannot even be remembering that you are the one but here they know us as patients in and out … you say that now that he has seen me, he has remembered how I look like down there in my private parts...” (Respondent 5, FGD 1 _ Very Unac- ceptable and Acceptable) HIV-infected women anticipated that fear and shame to undergo cervical cancer screening conducted by health workers at the HIV clinic would even generate non-adherence to scheduled HIV clinic visits. “Sometimes, those people will start irregular treat- ment. The day they gave her, if she knows that she is going to be tested for cancer of the cervix, she won’t come or that day when she is to come, she will come when she is in her periods.” (Respondent 4, FGD 2 _ Neutral) Increased waiting time HIV-infected women in the majority of focus group dis- cussions argued that delivering cervical cancer screening comprehensively with other routine HIV care services during scheduled HIV clinic visits would be very time consuming hence increasing waiting time at the HIV clinic. With increased numbers of HIV-infected women enrolled into HIV care at the HIV clinic, participants were worried of potential delay at the HIV clinic attrib- uted to integration of cervical cancer screening into rou- tine HIV care. “By the time you get through with the line of pick- ing drugs, …, in fact you are already tired. And now imagine, you will also have to go the cancer room… don’t you think women will go back at night or even the next day [laughs] …” (Respondent 5, FGD 1 _ Very Unacceptable and Unacceptable) Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 13 of 16 Discussion This study assessed acceptability of integration of cervi- cal cancer screening into routine HIV care, associated factors and perceptions among HIV-infected women enrolled in the HIV clinic at Mbarara Regional Referral Hospital. Analysis of data showed that 64.5% of HIV- infected women enrolled in the HIV clinic at Mbarara Regional Referral Hospital accepted integration of cer- vical cancer screening into routine HIV care. This study finding is lower than acceptance among HIV-infected women to undergo cervical cancer screening procedure reported at 87.6% after incorporating Visual Inspection with Acetic acid into the routine clinical services offered at Family AIDS Care and Education Services clinics in Kisumu, Kenya [32]. Acceptability of the proposed intervention at 64.5% is lower than cervical cancer screening acceptance of 79.8% among HIV-infected women at the HIV treatment cen- tre, Nigerian Institute of Medical Research (NIMR), Lagos, Nigeria who participated in the study assessing acceptability of integration of cervical cancer screening into HIV care [8]. In addition, 64.5% is lower than 96.5% acceptability to undergo Visual Inspection with Ace- tic acid after integrating cervical cancer screening into HIV care and treatment services in a district hospital in Abuja, Nigeria [33]. This could be attributed to perceived challenges of integration of cervical cancer screening into routine HIV care among HIV-infected women enrolled at the HIV clinic. Results of multivariable regression analysis indicated that religion, perceived risk of developing cervical cancer and ever screened for cervical cancer were statistically significantly associated with acceptability of integra- tion of cervical cancer screening into routine HIV care among HIV-infected women. Despite the fact that Mus- lims accepted integration of cervical cancer screening into routine HIV care compared to Protestants, a study conducted among HIV-infected women reported that religion was not significantly associated with acceptabil- ity of integration of cervical cancer screening into routine HIV care [8]. Additionally, a study conducted in Ethiopia reported that religion differences were not significantly associated with acceptability and uptake of cervical can- cer screening among HIV-infected women at St. Paul’s and Zewditu Hospitals where cervical cancer screening had been integrated into HIV care and treatment services [34]. Regardless of the fact that this was a surprising find- ing, it is unclear why most Christians inclusive of Prot- estants consinder fatal illnesses like cervical cancer to be a punishment from God and believe that prevalence of cervical cancer is very low among children of God; hence the need for involvement of religious stakeholders in advocacy for integration of cervical cancer screening into routine HIV care. HIV-infected women with “Much Above Average” per- ceived risk of developing cervical cancer accepted inte- gration of cervical cancer screening into routine HIV care. This study finding highly correlates with findings from the systematic review conducted in Ethiopia where perceived susceptibility to acquiring cervical cancer (AOR = 3.26; 95% CI : 2.26, 4.26) was significantly associ- ated with cervical cancer screening acceptability among HIV-positive women [35]. However, a study conducted among HIV-infected women reported that perceived risk of acquiring cervical cancer was not significantly associated with acceptability of integration of cervical cancer screening into routine HIV care [8]. Nonethe- less, a study conducted in Ghana reported that perceived susceptibility to acquiring cervical cancer was not sig- nificantly associated with willingness and acceptability to undergo cervical cancer screening among HIV-infected women [36]. Discrepancies in the influence of perceived risk of cervical cancer to acceptability of integration of cervical cancer screening into routine HIV care could be explained by the gap of lack of awareness that HIV infected women are more at risk of developing cervical cancer compared to their counterparts, HIV negative women. It was also found out HIV-infected women who had not undergone cervical cancer screening were less likely to accept integration of cervical cancer screening into routine HIV care compared to HIV-infected women who had undergone cervical cancer screening. A study con- ducted among HIV-infected women at the HIV treat- ment centre, Nigerian Institute of Medical Research, Lagos, Nigeria reported that ever screened for cervical cancer was not significantly associated with acceptability of integration of cervical cancer screening into routine HIV care [8]. HIV-infected women who have previously undergone cervical cancer screening applaud integra- tion of cervical cancer screening into routine HIV care due to perceived benefits most especially convenience of integrated cervical cancer screening and HIV services. Despite the fact that HIV-infected women who had not undergone cervical cancer screening are highly expected to support integration of cervical cancer screening into routine HIV services, this might not necessarily be true because they could be ashamed of exposing their private parts to health workers whom they occasionally interact with in HIV clinics. Perceived benefits of the proposed intervention were: convenience to seek for cervical cancer screening, moti- vation to undergo cervical cancer screening, improved archiving of cervical cancer screening results, confi- dentiality of HIV patient information, and preference Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 14 of 16 to interact with HIV clinic health workers. In all FGDs, HIV-infected women acknowledged that integration of cervical cancer screening into routine HIV care would grant them an opportunity to receive both cervical cancer screening in addition to HIV services conveniently from the HIV clinic compared to currently practiced referral or client-initiated cervical cancer screening. This study finding is in agreement with findings from other stud- ies which reported that convenience of seeking for inte- grated HIV and cervical cancer screening services from one point without forgetting an added advantage of one off relief of anxiety was attributed to the integrated strat- egy compared to stand alone screening services [23, 24]. The intervention was regarded as a time saving strategy attributed to convenience through reducing frequency of health facility visits, necessity of several return journeys, travel time and return transportation costs [23, 24]. HIV- infected women anticipated that implementation of the proposed strategy of delivering cervical cancer screening would reduce disruption and movement from one clinic to another since all cervical cancer screening and HIV related services would be available and received under one roof in a single HIV clinic visit comprehensively. HIV-infected women in all FGDs reported that inte- gration of cervical cancer screening would motivate HIV infected women to undergo cervical cancer screen- ing services. Furthermore, HIV-infected women in the majority of FGDs alleged that integration of cervical cancer screening into routine HIV care would inspire HIV-infected women, targeted recipients for the pro- posed intervention, to comply and adhere to recom- mended annual cervical cancer screening. Researchers, community members, health care providers and policy makers were also optimistic that intervention would improve uptake of initial and annual cervical cancer screening among HIV-infected women [4, 15–24]. Inte- gration of cervical cancer screening into routine HIV services would reduce missed opportunities for HIV- infected women to undergo cervical cancer screening procedure; hence increasing uptake of cervical cancer screening among targeted recipients. HIV-infected women in the majority of FGDs claimed that the proposed intervention would promote confiden- tiality of HIV status and any other health related infor- mation due to interaction with only health workers and fellow HIV-infected women. This study finding was also reported in a study conducted in Uganda where women reported that the integrated model would promote con- fidentiality of HIV and any other health related informa- tion due to being accessed and confined to limited teams of health care practitioners in HIV clinics [24]. In most instances, HIV-infected women prefer confidentiality of their HIV status through interacting with only health workers in HIV clinics and archiving of cervical cancer screening information in the same file with all other HIV related information; due to fear of being stigmatised. Shame to expose their privacy to HIV clinic health workers and increased waiting time were the only per- ceived challenges of integration of cervical cancer screen- ing into routine HIV care among HIV-infected women. HIV-infected women in only “Very Unacceptable and Unacceptable” FGD reported that they would prefer undergoing cervical cancer screening conducted by health workers at the cervical cancer screening unit of Mbarara Regional Referral Hospital as opposed to HIV clinic health workers. Based on a study conducted in Uganda, discomfort and invasion of privacy expressed by HIV-infected women due to genital illnesses, menstrual periods, and fear to expose their private parts was one of the challenges to undergo cervical cancer screening integrated into routine HIV services at Mildmay Uganda [37]. HIV-infected women argued that they would be uncomfortable and ashamed to continue interacting with the same health workers to whom they have exposed their private parts for any other HIV related services. HIV-infected women in the majority of the FGDs argued that delivering cervical cancer screening com- prehensively with other routine HIV care services during scheduled HIV clinic visits would be very time consum- ing hence increasing waiting time at the HIV clinic. This study finding was consistent with findings from other studies where long waiting time was cited as one of the potential challenges HIV-infected women would have to endure through to undergo cervical cancer screen- ing integrated into comprehensive HIV services [23, 24, 37]. HIV-infected women in some FGDs expressed con- cern on how the integrated approach of cervical cancer screening and HIV treatment services in a single HIV clinic visit would be delivered to the ever-increasing numbers of recipients within the shortest time possible. The mixed methods study design using explanatory sequential approach increased rigor and internal validity of the study. Utilisation of systematic sampling method for selection of HIV-infected women to participate in the quantitative phase of the study was an added strength to the conduct of the study. However, the Theoretical Frame- work of Acceptability adapted to measure acceptability of integration of cervical cancer screening into routine HIV care had not yet been validated among Ugandan popula- tions. HIV-infected women who turned up on unsched- uled dates were excluded from participating in the study; increasing potential for information bias. Furthermore, this study was conducted in only one HIV clinic in the entire country due to limited funding; hence affect- ing generalisability of study findings to all HIV-infected women in Uganda. Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 15 of 16 Conclusion Study findings highlight the need to take advantage of this acceptability to prioritize implementation of inte- gration of cervical cancer screening into routine HIV care. Ministry of Health should prioritise implemen- tation of integration of cervical cancer screening into routine HIV care to increase uptake of cervical cancer screening among HIV-infected women along the contin- uum of HIV care and treatment services. Health work- ers delivering HIV care and treatment services should endeavour to conduct intensified health education and awareness about increased risk of developing cervi- cal cancer among HIV-infected women. HIV-infected women should be reassured of reduced waiting time, confidentiality and optimisation of privacy by HIV clinic health workers during provision of integrated HIV care and cervical cancer screening services. Furthermore, a nationally representative study should be conducted to assess acceptability of integration of cervical cancer screening into routine HIV care, associated factors, and perceptions among HIV-infected women enrolled in HIV clinics in Uganda. Acknowledgements Exceptional applaud to Dr. Joseph Kagaayi and Ms. Agnes Nyabigambo who have been very instrumental in inspiring the best of my potential in my research career. Authors’ contributions MN designed the proposal, coordinated quantitative and qualitative data col- lection, analysed data, presented study findings and compiled the manuscript. AN and JK provided great technical guidance through out every step of the proposal conceptualization, data analysis and manuscript preparation. The author(s) read and approved the final manuscript. Funding Not applicable. Availability of data and materials The datasets used and/or analysed during the study are available from the corresponding author on request. Declarations Ethics approval and consent to participate Approval was obtained from Makerere University School of Public Health Higher Degrees, Research and Ethics Committee to conduct the study. Administrative clearance was sought from the office of the Hospital Director of Mbarara Regional Referral Hospital before conducting data collection. Indi- viduals were informed of their right to agree to participate or withdraw from the study at any time without fear of any negative repercussions; and receipt of care was not dependent on participation in the study. After explaining the purpose and procedures of the study, a written informed consent was sought from study participants before administering questionnaires or conducting focus group discussions. All methods were performed in accordance with relevant guidelines and regulations in the Ethics Approval and Consent to Participate. Consent for publication Not applicable. Competing interests The authors do not have any competing interests. Author details 1 Department of Epidemiology and Biostatistics, School of Public Health, Makerere University, Kampala, Uganda. 2 Department of Community Health and Behavioural Sciences, School of Public Health, Makerere University, Kam- pala, Uganda. 3 Rakai Health Sciences Program, Kalisizo, Uganda. Received: 5 June 2022 Accepted: 22 March 2023 References 1. Bray F, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA. 2018;68(6):394–424. 2. WHO. Human papillomavirus (HPV) and cervical cancer. 2019. Available from: https:// www. who. int/ en/ news- room/ fact- sheets/ detail/ human- papil lomav irus- (hpv)- and- cervi cal- cancer. Accessed 5 Nov 2019. 3. WHO. Cervical cancer common amongst African women. 2019. Available from: https:// www. afro. who. int/ news/ cervi cal- cancer- common- among st- afric an- women. Accessed 5 Nov 2019. 4. Viviano M, et al. A review of screening strategies for cervical cancer in human immunodeficiency virus-positive women in sub-Saharan Africa. Int J Women’s Health. 2017;9:69. 5. Nakisige C, Schwartz M, Ndira AO. Cervical cancer screening and treat- ment in Uganda. Gynecol Oncol Rep. 2017;20:37–40. 6. Ghebre RG, et al. Cervical cancer control in HIV-infected women: Past, present and future. Gynecol Oncol Rep. 2017;21:101–8. 7. Bansil P, et al. Performance of cervical cancer screening techniques in HIV- infected women in Uganda. J Lower Genital tract Dis. 2015;19(3):215–9. 8. Ezechi OC, et al. The association between HIV infection, antiretroviral therapy and cervical squamous intraepithelial lesions in South Western Nigerian women. PLoS One. 2014;9(5):e97150. 9. Massad LS, et al. Incidence of cervical precancers among HIV-seropositive women. Am J Obstet Gynecol. 2015;212(5):606.e1-8. 10. Ngabo F, et al. Human papillomavirus infection in Rwanda at the moment of implementation of a national HPV vaccination programme. BMC Infect Dis. 2016;16(1):225. 11. Thorsteinsson K, et al. Incidence of cervical dysplasia and cervical cancer in women living with HIV in Denmark: comparison with the general population. HIV Med. 2016;17(1):7–17. 12. Thunga S, et al. Cervical cytological abnormalities and human papilloma virus infection in women infected with HIV in Southern India. J Obstet Gynaecol Res. 2016;42(12):1822–8. 13. WHO. WHO guidelines for screening and treatment of precancerous lesions for cervical cancer prevention. 2013. Available from: https:// apps. who. int/ iris/ bitst ream/ handle/ 10665/ 94830/ 97892 41548 694_ eng. pdf; jsess ionid 1. Accessed 10 Nov 2019. AB409 201BC 57FAF 0983E A5773 66A31 52? seque nce = = 14. Ministry of Health. Consolidated Guidelines for Prevention and Treatment of HIV in Uganda. Kampala: Republic of Uganda Ministry of Health; 2016. 15. UNAIDS. HPV, HIV, cervical cancer: Leveraging synergies to save women’s lives. UNAIDS. Published 2016 [cited 2019 5th November]; Available from: https:// www. unaids. org/ sites/ defau lt/ files/ media_ asset/ JC2851_ HPV- HIV- cervi calca ncer_ en. pdf. 16. UNAIDS. The little-known links between cervical cancer and HIV. 2019 [cited 2019 5th November]; Available from: https:// www. unaids. org/ en/ resou rces/ press centre/ featu resto ries/ 2019/ may/ 20190 531_ cervi cal- cancer- hiv. 17. Wanyenze RK, et al. Uptake and correlates of cervical cancer screening among HIV-infected women attending HIV care in Uganda. Global Health Act. 2017;10(1):1380361. 18. Coleman JS, et al. An insight into cervical cancer screening and treatment capacity in sub-Saharan Africa. J Lower Genital Tract Dis. 2016;20(1):31. 19. Denslow SA, et al. Incidence and progression of cervical lesions in women with HIV: a systematic global review. Int J STD AIDS. 2014;25(3):163–77. 20. Huchko MJ, et al. The time has come to make cervical cancer preven- tion an essential part of comprehensive sexual and reproductive health services for HIV-positive women in low-income countries. J Int AIDS Soc. 2015;18:20282. Ninsiima et al. BMC Health Services Research (2023) 23:333 Page 16 of 16 21. Mapanga W, et al. Prevention of cervical cancer in HIV-seropositive women from developing countries through cervical cancer screening: a systematic review. Systematic reviews. 2018;7(1):198. 22. Tchounga B, et al. Cervical cancer screening uptake and correlates among HIV-infected women: a cross-sectional survey in Côte d’Ivoire, West Africa. BMJ open. 2019;9(8):e029882. 23. Kumakech E, et al. Integration of HIV and cervical cancer screening perceptions of healthcare providers and policy makers in Uganda. BMC Public Health. 2014;14(1):810. 24. Kumakech E, et al. Integration of HIV and cervical cancer screening perceptions and preferences of communities in Uganda. BMC women’s health. 2015;15(1):23. 25. Sekhon M, Cartwright M, Francis JJ. Acceptability of healthcare interven- tions: an overview of reviews and development of a theoretical frame- work. BMC health services research. 2017;17(1):88. 26. Ronald M, et al. Cervical Cancer at Mbarara Regional Referral Hospi- tal: Magnitude, Trends, Stages at Presentation, Impact of Acetic Acid Screening and the Need for Radiotherapy Services. J Health Med Nur. 2016;27:126–39. 27. Aweke YH, Ayanto SY, Ersado TL. Knowledge, attitude and practice for cervical cancer prevention and control among women of childbearing age in Hossana Town, Hadiya zone, Southern Ethiopia: Community-based cross-sectional study. PloS One. 2017;12(7):e0181415. 28. Mengesha A, Messele A, Beletew B. Knowledge and attitude towards cervical cancer among reproductive age group women in Gondar town, North West Ethiopia. BMC public health. 2020;20(1):209. 29. Suresh K, Chandrashekara S. Sample size estimation and power analysis for clinical research studies. J Hum Reprod Sci. 2012;5(1):7. 30. Hajian-Tilaki K. Sample size estimation in epidemiologic studies. Caspian J Intern Med. 2011;2(4):289. 31. Semeere AS, et al. An assessment of implementation science research capacity in Uganda. Health Res Policy Syst. 2021;19(1):1–10. 32. Huchko MJ, Bukusi EA, Cohen CR. Building capacity for cervical cancer screening in outpatient HIV clinics in the Nyanza province of western Kenya. Int J Gynecol Obstet. 2011;114(2):106–10. 33. Odafe S, et al. Integrating cervical cancer screening with HIV care in a district hospital in Abuja, Nigeria. Niger Med J. 2013;54(3):176. 34. Kassahun E, Abebaw T-A, Abera A. "Role of awareness on cervical cancer screening uptake among HIV positive women in Addis Ababa, Ethiopia: A cross-sectional study." Women’s Health. 2021;17:17455065211017041. 35. Birye DM. "Cervical cancer screening uptake and associated factors among HIV-positive women in Ethiopia: a systematic review and meta- analysis." Adv Prev Med. 2020;2020. 36. Ebu NI, Ogah JK. Predictors of cervical cancer screening intention of HIV- positive women in the central region of Ghana. BMC Women’s Health. 2018;18(1):1–7. 37. Bukirwa A, et al. Motivations and barriers to cervical cancer screen- ing among HIV infected women in HIV care: a qualitative study. BMC Women’s Health. 2015;15(1):82. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12875-019-0972-1	Wadsworth et al. BMC Family Practice (2019) 20:97 https://doi.org/10.1186/s12875-019-0972-1 R E S E A R C H A R T I C L E Open Access Shared medical appointments and patient- centered experience: a mixed-methods systematic review Kim H. Wadsworth1* Adam S. Hoverman3 , Trevor G. Archibald1, Allison E. Payne1, Anita K. Cleary1, Byron L. Haney1,2 and Abstract Background: Shared medical appointments (SMAs), or group visits, are a healthcare delivery method with the potential to improve chronic disease management and preventive care. In this review, we sought to better understand opportunities, barriers, and limitations to SMAs based on patient experience in the primary care context. Methods: An experienced biomedical librarian conducted literature searches of PubMed, Cochrane Library, PsycINFO, CINAHL, Web of Science, ClinicalTrials.gov, and SSRN for peer-reviewed publications published 1997 or after. We searched grey literature, nonempirical reports, social science publications, and citations from published systematic reviews. The search yielded 1359 papers, including qualitative, quantitative, and mixed method studies. Categorization of the extracted data informed a thematic synthesis. We did not perform a formal meta-analysis. Results: Screening and quality assessment yielded 13 quantitative controlled trials, 11 qualitative papers, and two mixed methods studies that met inclusion criteria. We identified three consistent models of care: cooperative health care clinic (five articles), shared medical appointment / group visit (10 articles) and group prenatal care / CenteringPregnancy® (11 articles). Conclusions: SMAs in a variety of formats are increasingly employed in primary care settings, with no singular gold standard. Accepting and implementing this nontraditional approach by both patients and clinicians can yield measurable improvements in patient trust, patient perception of quality of care and quality of life, and relevant biophysical measurements of clinical parameters. Further refinement of this healthcare delivery model will be best driven by standardizing measures of patient satisfaction and clinical outcomes. Keywords: Shared medical appointment, Group visit, Cooperative health care clinic, Group prenatal care, Patient satisfaction, Patient experience, Health services, Primary care, Primary health care, Coproduction Background Shared medical appointments (SMAs), or group visits, are a healthcare delivery innovation arising from the changing demands of patient-centered medical home (PCMH) set- tings and the primary care context. The model emphasizes prompt access and improved service, increased doctor- patient contact time, greater patient education, enhanced prevention and disease self-management, closer attention to routine health maintenance and performance measures, * Correspondence: kim_ha@stanfordalumni.org 1Pacific Northwest University of Health Sciences, College of Osteopathic Medicine, Yakima, WA, USA Full list of author information is available at the end of the article and the central role of patient and clinician experience within the Triple Aim: enhancing patient experience, im- proving population health, and reducing costs [1–3]. More recently, Bodenheimer and Sinsky recommended that “the Triple Aim be expanded to a Quadruple Aim, adding the goal of improving the work life of health care providers, including clinicians and staff [4].” We chose SMA as the overarching term to encompass shared visit, group appointment, group medical appoint- ment, group visit (GV), group medical clinic, shared in- group medical appointment, group prenatal care (GPNC) and group-based antenatal care. SMAs prioritize the deliv- ery of care within interprofessional environments utilizing © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Wadsworth et al. BMC Family Practice (2019) 20:97 Page 2 of 13 peer-to-peer interactions [5]. Multiple standardized SMA delivery models have been established, from the drop-in group medical appointment, cooperative health care clinic (CHCC) and physicals shared medical appointment, to CenteringPregnancy® (CP) and parenting visits [3, 6]. These visits frequently emphasize the “coproduction” roles of patients as experts in their own circumstances and health professionals as facilitators rather than fixers, thus fostering a shared experience of illness and health to bet- ter inform, empower, and support [2]. SMAs have garnered a body of evidence in chronic disease management and preventive care. The various interpretations of the group clinical model have been ap- plied to a wide array of settings and a myriad of health promotion and disease-focused visits, including patients with diabetes, hypertension, congestive heart failure, chronic lung disease, asthma, arthritis, stroke, kidney disease, cancer, hearing impairment, and prenatal care, among other conditions [7–15]. Several systematic reviews summarize the effects of SMAs on healthcare delivery, economic factors, and bio- physical outcomes. Health systems have begun to em- brace the need for this transformative approach in achieving patient goals [2, 16–18]. In an era recognizing the role of patient-centeredness in improving healthcare quality, numerous authors have highlighted the need for a review that addresses the impacts of SMAs on patient experience of care [3, 7, 16, 17, 19]. This review aims to meet this need by examining the patient experience from the published literature alongside an assessment of SMAs to improve biophysical outcomes in the adult pri- mary care setting. Analyzing the existing body of evidence for shared medical appointments, we sought to understand the op- portunities, barriers, and limitations to SMAs based on self-reported patient experience, a notable component of the Triple Aim [2]. Specifically, our goal was to highlight effective approaches for patients participating in SMAs and determinants of effectiveness. Methods librarian conducted pre- An experienced biomedical planned literature searches of PubMed, Cochrane Li- brary, PsycINFO, Cumulative Index of Nursing and Allied Health Literature (CINAHL), Web of Science, ClinicalTrials.gov, and Social Science Research Network (SSRN) for peer-reviewed publications, using controlled vocabulary, keywords, and text words (see Additional file 1 for search strategy details). The search was limited to publications from 1997 or after. We also searched grey literature, non-empirical reports, social science publica- tions, and citations from published systematic reviews. The search yielded 1359 papers, including qualitative, quantitative, and mixed-methods studies. Case studies, pilot/feasibility studies, protocols, opinions, or advocacy articles were excluded. Eligibility criteria and methods of analysis were specified a priori. Two researchers independently reviewed citation titles, abstracts, and full-text articles to determine eligibility as well as extracted the data and performed quality and risk of bias assessment on included articles, as detailed below. Before general use, we pilot-tested the abstraction form templates on a sample of included articles and then re- vised accordingly to ensure that all relevant data elements were captured. Disagreements were resolved by consensus of the two reviewers or by obtaining a third investigator’s opinion when consensus could not be reached. Studies were required to meet five process (p) and out- come (o) criteria: clinical intervention (o), clinician-led visit (p), patient experience of care (o), primary care (p), and availability of individual clinical consultation (p), as detailed below. Studies were excluded if any participants were < 18 years of age. To limit potential bias, we ex- cluded studies involving addiction medicine, substance dependence / rehabilitation treatment, inpatient settings (both short and long term) or chronic care clinics that implemented multiple interventions, and SMAs requir- ing management by a specialist. We deemed SMAs to be clinician led if led by an inde- pendent licensed prescriber or clinician. This included medical doctors (MDs), doctors of osteopathy (DOs), ad- vanced registered nurse practitioners (ARNPs), certified nurse midwives (CNMs), and in some regions, nurse practitioners (NPs). We verified prescriptive authority and care responsibility by consulting organizational web- sites from the countries in which our identified studies were conducted [20–22]. Our review emphasized biophysical metrics of adult pa- tients in primary care environments. The study team in- cluded articles focused on SMAs that implemented a clinical intervention, such as vital sign measurements, lab checks (e.g., hemoglobin A1c, lipid panels), medication ad- justments, or physical exams. We excluded studies if the intervention was limited to patient education, facilitation, peer-facilitated support groups, or group talk therapy. We tracked confounders within targeted studies, such as participant inclusion/exclusion criteria, local barriers to implementation, reimbursement framework, types of SMA interventions, and patient characteristics including language, culture, and socioeconomic status. In our consideration of quantitative research, we in- cluded only those studies with a comparative control group. Studies with quantitative primary outcomes were evaluated using the modified Jadad score, which assesses the overall quality of the individual studies, including risk of bias, and has shown high inter-rater reliability [23–26]. To evaluate qualitative studies, our team used the “Trustworthiness of Qualitative Inquiry” framework to Wadsworth et al. BMC Family Practice (2019) 20:97 Page 3 of 13 assess credibility, transferability, dependability, and ob- jectivity [27]. Inter-rater reliability was assessed during the data ex- traction phase via two-way mixed measures intraclass cor- relation (ICC) value for average agreement presented [28]. In consideration of ENTREQ and PRISMA frameworks for this mixed-methods systematic review, categorization of the extracted data informed a thematic synthesis [29– 32]. We did not perform a formal meta-analysis. Results Thirteen quantitative controlled trials, 11 qualitative pa- pers, and two mixed methods studies met inclusion cri- teria. Three models were identified: CHCC (five articles), SMA / GV (10 articles) and GPNC / CP (11 articles). Figure 1 shows the Preferred Reporting Items for Sys- tematic Reviews and Meta-Analyses (PRISMA) flowchart for all included studies [32]. Summary of included studies SMA / GV is the most frequently mentioned model in quantitative studies whereas the GPNC / CP model is the most common in qualitative studies in this re- is the least represented in view. The CHCC model this review (Table 1). Table 2 breaks down the included articles into locale, healthcare system, reimbursement model, study design, single site or multiple sites, and study duration. Table 3 provides details of the typical configuration of the three models included in this review: CHCC, SMA / GV, and GPNC / CP. Generally, CHCC has a larger group size compared to SMA / GV and GPNC / CP. Physician- led intervention teams were cited in most SMA / GV studies, whereas certified nurse midwives were most often cited as leaders of the GPNC / CP visits. Per inclusion criteria, all 26 articles reported patient satisfaction and experience (Table 4). Only one article reported outcomes for all four aims [8]. Patient experience and satisfaction Methodologies for tracking patient experience and satis- faction were grouped by data collection method into the following five categories: One-on-One Interviews (via tele- phone or in person), Focus Group Style Interviews, Self- Efficacy / Participation / Satisfaction Questionnaires, Diabetes-Related Quality of Life (DQoL) Related Scales; and Primary Care Assessment Tool / Trust in Provider Outcomes (Table 5). When comparing the results of the patient experience / satisfaction data in these 26 articles, the following six Citations identified through database searches (n = 1537) Citations identified through grey literature searches (n = 22) Additional citations identified through bibliography sources (n = 73) Citations available for initial screening (n = 1632) Titles and abstracts screened, after duplicates removed (n = 1359) Full-text articles assessed for eligibility (n = 299) Citations included in mixed- methods systematic review (n = 26) Citations excluded at title / abstract level (n = 1060) Full-text articles excluded, with reasons (n = 273) Quantitative articles (n = 13) Qualitative articles (n = 11) Mixed methods articles (n = 2) Fig. 1 The PRISMA flowchart for all included studies Wadsworth et al. BMC Family Practice (2019) 20:97 Page 4 of 13 Table 1 List of 26 included articles in the primary care setting, categorized by model of group clinic and study type Model: CHCC SMA / GV GPNC / CP Quantitative (13 articles) X X X X X Beck, 1997 Clancy, 2007 Jafari F, 2010 Junling, 2015 Kennedy, 2011 Naik, 2011 Scott, 2004 Tandon, 2013 Trento, 2001 Trento, 2002 Trento, 2004 Trento, 2005 Trento, 2010 Qualitative (11 articles) Andersson, 2012 Andersson, 2013 Capello, 2008 Clancy, 2003 Herrman, 2012 Kennedy, 2009 McDonald, 2014 McNeil, 2012 Novick, 2011 Raballo, 2012 Wong, 2015 Mixed-methods (2 articles) Heberlein, 2016 Krzywkowski-Mohn,2008 Total no. of articles (26) 5 X X X X X X X X X X 10 X X X X X X X X X X X 11 Abbreviations: CHCC Cooperative health care clinic, CP CenteringPregnancy®, GPNC Group prenatal care, GV Group visit, SMA Shared medical appointment major themes emerged (also see Additional file 2 for more details). Patient-clinician dynamic Overall, data on the patient-clinician dynamic that emerged during SMAs were positive. SMAs saw quantita- tive advantages over individual visits in domains ranging from improved communication to overall satisfaction with the visit [7, 15, 33]. In SMA environments, more time was allotted to discuss healthcare issues with the clinician compared to traditional individual visits, and physicians were perceived as less hurried [7, 14]. One study indicated that SMA experiences resulted in markedly enhanced trust in one’s primary care physician [33]. Qualitative feedback similarly supported the patient- clinician dynamic as a notable aspect of SMAs. Inter- views with CP patients indicated that extra time with cli- nicians helped them to develop strong, supportive, and positive relationships with their healthcare clinicians, and reduced anxiety about potentially not being familiar with the practitioner who would oversee their obstetric deliveries [9–11, 34]. Feedback from patients indicated that room for further improvement of the patient-clinician dynamic in SMAs lies in the avoidance of a paternalistic, didactic style of communication from the clinician leader [12]. Patients appreciated being empowered by their clinicians and preferred a more encouraging and empowering commu- nication style within their groups. Overall quality of care Multiple studies demonstrated that patients participating in SMAs were significantly more satisfied with their care than those in individual models of care [7, 13–15]. When compared to patients receiving traditional individ- ual care, those participating in SMAs were more likely to describe their overall quality of care as excellent, to feel that their care was meeting all their needs, and to feel that their care was well coordinated [8, 35]. No studies showed significant decreases in patient percep- tions of quality of care in SMAs. Overall quality of care was not a direct theme ex- tracted from qualitative investigations of SMAs. How- interviews from Herrman’s research on the CP ever, program revealed that “multiparous women frequently commented that [SMAs were] far superior to their pre- vious experiences” [11]. Quality of life Trento’s research thoroughly addressed the theme of quality of life, using a modified version of the Diabetes Quality of Life Measure (DQoL) questionnaire consisting of 39 questions ranked along a 5-point Likert scale. This assessment scale was used across all five of Trento’s arti- cles, and demonstrated consistent results over 10 years of varied research on SMAs for patients with Diabetes Mellitus, Type 2 (T2DM). In all five of Trento’s studies discussed in this paper, DQoL scores significantly im- proved among group participants while worsening or remaining the same in control subjects [36–40]. Sense of community Patients in multiple studies reported that the feeling that they were not alone in their experience was central to the positive impact of SMAs and persisted whether the subject of the SMA was pregnancy, navigation of the VA Wadsworth et al. BMC Family Practice (2019) 20:97 Page 5 of 13 Table 2 Characteristics of included studies in the primary care setting Study characteristics No. of studies, by medical condition N studies (participants) Diabetes 10 (1881) Country United States Canada Europe (Italy, Sweden) Middle East (Iran) Asia (China) Healthcare system Govt (VA, FQHC, NHS, PHD) Private (HMO, MCO) University-affiliated clinic Healthcare payment model Public (Medicaid, Medicare, govt funded) Private (fee-for-service, managed care) Uninsured /underinsured Study design Randomized controlled trial Non-randomized controlled trial Observational / interviews / focus groups Mixed methods Sites Single Multisite Study duration < 6 months 6 months 7 to 11 months 12 to 18 months 24 months > 2 years 4 (426) 0 6 (1455) 0 0 3 (362) 1 (120) 6 (1399) 8 (1575) 0 2 (306) 9 (1848) 0 1 (33) 0 9 (1066) 1 (815) 1 (87) 1 (120) 0 2(219) 3 (1169) 3 (286) HTN 2 (1262) 1 (58) 0 0 0 1 (1204) 2 (1262) 0 0 2 (1262) 0 0 1 (1204) 0 1 (58) 0 1 (58) 1 (1204) 1 (1204) 1 (58) 0 0 0 0 MCC 3 (645) 2 (616) 1 (29) 0 0 0 1 (29) 2 (616) 0 3 (645) 0 0 2 (616) 0 1 (29) 0 1 (321) 2 (324) 0 0 0 2 (350) 1 (295) 0 Pregnancy 11 (2010) 6 (926) 2 (21) 2 (435) 1 (628) 0 8 (1908) 3 (102) 0 8 (1908) 3 (102) 0 4 (1591) 1 (268) 5 (122) 1 (29) 4 (84) 7 (1926) 0 0 11 (2010) 0 0 0 Abbreviations: FQHC Federally qualified health center, HMO Health maintenance organization, HTN Hypertension, MCC Multiple chronic conditions, MCO Managed care organization, NHS National health service, PHD Public health district, VA Veterans Administration system, or hypertension [6, 10, 12, 33, 41–44]. Creation of community via SMAs supported patients’ emotional health by providing validation and stemming the isola- tion often experienced when managing chronic condi- tions. This sense of community was viewed as a benefit, though one study referenced a member who reported that at times she avoided discussion of “disturbing topics for fear that it would negatively impact her cohort” [34]. Patient empowerment / role in healthcare This body of research suggests that a strength of SMAs over usual care is the ability to engage and empower pa- tients as active participants in their own healthcare. This empowerment bore out in both qualitative and quantita- tive research participants. Quantitatively, patients reported that they were more able to participate in their care and had significant improvements on scales of Coping Skills and Health Distress as compared to their counterparts [13, 14, 43]. In the realm of qualitative analyses, it was de- scribed that patients felt they were better able to interpret their medical data, thus making them more likely to dis- cuss their issues with their clinicians [42]. Within the CP model, patients reported feeling “reassured, prepared, less anxious, and confident,” and they felt that the group ses- sions made them more proactive with respect to their own health [9]. Raballo’s research also indicated that after Wadsworth et al. BMC Family Practice (2019) 20:97 Page 6 of 13 Table 3 Typical configuration of group models, as represented by included studies in the primary care setting Model (no. of articles) CHCC (5) Duration of each group session 90– 120 min Duration of individual consultation 5–10 min each at end of group session Group size 6–20 SMA / GV(10) 60– 90 min Optional 10 mins each or 24 mins total allotted at end of group session 5–15 GPNC / CPa(11) 90– 120 min 10 mins each at beginning of group session 8–12 Clinical intervention Nonclinical components Intervention team Disciplines (no. of articles) Size Vital signs Lab results review and medical records update Medication management Preventive measures Scheduling Medical-related paperwork requested by pts Brief 1:1 visits with physician, as necessary Vital signs Lab results review and medical records update Routine lab test orders 1:1 indiv consultation with physician, as necessary Health risk assessment Medication management Referrals, coordination of public health services Vital signs Physical exam Routine prenatal screening and labs Routine ultrasound Flu vaccine (seasonal) Postpartum visit Individual assessments prior to prenatal care within group setting Socialization Health education Group cohesion Orientation and socialization Interactive health education Group cohesion Self-monitoring Group discussion Medication compliance Group discussion, self-care, skills- building Active tracking of pregnancy changes (done by pts) Tour of birth unit, labor and delivery nurse Pediatric care resources Postpartum reunion 2–5 2–7 PCP (5) Nurse, RN or diabetes nurse educator (5) Clinical pharmacist (2) PT, OT (2) Dietitian (2) Community health worker (1) 1–2 physicians (9) Nurse, NP, RN (2) Diabetes educator/ RD (4) Clin psychologist, psychopedagogist (3) 1–2 postgraduate med students (1) Others (2) 2 + others invited 1–2 CNMs (8) NP (3) Medical asst (3) Physician (2) Health / perinatal educator (1) Others (1) Abbreviations: CHCC Cooperative health care clinic, CNM Certified nurse midwife, CP CenteringPregnancy®, GPNC Group prenatal care, GV Group visit, NP Nurse practitioner, OT Occupational therapist, PCP Primary care physician, PT Physical therapist, RD Registered dietitian, RN Registered nurse, SMA Shared medical appointment aWk 5–10: First visit w/ nurse. Wk 10–12: First visit with clinician. Wk 12–16: Start CP program experiencing SMAs, patients were significantly more likely to describe an internal locus of control for their health than those followed by usual care [45]. communication between clinicians, decreased waiting times, increased opportunities for learning throughout their visits, and improved administrative support [41, 42, 46]. Access / efficiency Several articles also establish benefits of SMAs with respect to access and efficiency. Quantitatively, participants re- ported that appointments were easily scheduled “as soon as [they liked]” and were more likely to report that visit waiting time was acceptable [8, 14]. Qualitatively, patients described experiencing “more comprehensive services,” smoother Biophysical outcomes Less than half of the included articles reported biophys- ical outcomes by health condition—either diabetes mellitus (DM) or hypertension (HTN)—as summarized in Table 6 [36–40, 42, 43, 45, 47, 48]. These studies claimed significant and non-significant improvements in biophysical metrics; however, heterogeneity of study Table 4 Quadruple aim reported in included studies Model (no. of articles) CHCC (5) SMA / GV (10) GPNC / CP (11) No. of articles Patient experience Population health 5 10 11 2 1 3 Cost 2 1 0 Clinician experience 3 3 1 Abbreviations: CHCC Cooperative health care clinic, CP CenteringPregnancy®, GPNC Group prenatal care, GV Group visit, SMA Shared medical appointment Wadsworth et al. BMC Family Practice (2019) 20:97 Page 7 of 13 Table 5 Methods used to collect patient experience data Method 1:1 phone or in-person interviewsa Focus group style interviewsa Self-efficacy / participation / satisfaction questionnaires Diabetes-related quality of life scales (DQoL) Primary care assessment tool & trust in clinician outcomes Total: aAndersson 2012 is double coded as it included both 1:1 and group interviews No. of articles 10 3 6 6 2 27 populations, methods and outcomes did not allow data across studies to be combined and analyzed. (one article) This data subset was categorized into quantitative (seven articles), qualitative (two articles), and mixed to include additional studies methods details (Table 7). Eight articles had a control comparator of usual care while two articles (one qualitative study and one mixed methods study) only compared pre- and post-group intervention. Only one article utilized the CHCC model while the remaining nine articles were SMAs / GVs. From the ten studies included in this sub- set, the reported biophysical profile data varied, keeping with previous systematic reviews on SMAs by Booth et al. and Edelman et al. [17, 18]. Barriers to implementation Few studies addressed barriers, as shown in Additional file 3. Prior reviews by Edelman et al., Booth et al., and Jones et al. cite several barriers to implementation of SMAs overall, including patient participation and at- tendance, group dynamic incompatibilities, cost-benefit concerns, and staff/facilities inadequacies [16, 17, 49]. Prior studies cited poor attendance at SMAs [7, 13, 33]. In tracking attendance and patient-centered out- comes through different group visit formats, durations and patient populations, a great variation of attendance rates was found, as shown in Additional file 4. Inter-rater reliability As shown in Additional file 5, the ICC(2,k) inter-rater reliability values are 0.956 for Jadad-modified score of quantitative studies, 0.923 for trustworthiness score of qualitative studies, and indeterminable for mixed method studies due to sample size of n = 2 studies. Values greater than 0.90 indicate excellent reliability [28]. Table 6 Overview of biophysical data from available studies, categorized by health condition (no. of articles = 10) HbA1c FBG Lipids BP BMI Body First author, year Diabetes X HDL, TG X HDL X HDL, TG X TC, HDL, TG X TC, LDL, HDL, TG X TC, HDL, TG X LDL X X X Trento, 2001 Trento, 2002 Trento, 2004 Trento, 2005 Trento, 2010 Naik, 2011 X X X X X X Raballo, 2012 X Krzywkowski- Mohn, 2008 X Hypertension Junling, 2015 Capello, 2008 wt X X X X X X X X X X X X X SBP X X X CV risk DM Rx dosage Kidney Eye Foot Physical activity X X X retinopathy X insulin X Cr X ACR X Cr X foot ulcers X retinal exam X foot exam X Abbreviations: ACR Albumin/Creatinine ratio, BMI Body mass index, BP Blood pressure, Cr Creatinine, CV Cardiovascular, DM Diabetes mellitus, FBG Fasting blood glucose, HbA1c Glycated hemoglobin, HDL High-density lipoprotein, LDL Low-density lipoprotein, Rx Prescription, SBP Systolic blood pressure, TC Total cholesterol, TG Triglycerides Wadsworth et al. BMC Family Practice (2019) 20:97 Page 8 of 13 Table 7 Biophysical data from available studies, categorized by research type (no. of articles = 10) First author, year Quantitative Model Health cond(s) Sample size (n) Biophysical measures Reported findings (with p-values) Junling, 2015 CHCC HTN 600 group, 604 control ● BP ● BMI SBP decreased significantly in both group (p < 0.001) and control (p = 0.001) from baseline to follow-up, although decreases in group > control. ● Physical activity DBP decreased significantly in group (p = 0.001) but did not decrease significantly in control. Trento, 2001 SMA / GV T2DM 56 group, 56 control ● HbA1c ● BMI ● HDL ● Fasting TG Trento, 2002 SMA / GV T2DM 56 group, 56 control ● Dosage of anti- hyperglycemic agents ● Body wt, BP and CV risk ● Metabolic control: - HbA1c - BMI - HDL - Retinopathy BMI did not change in both. Increases in physical activity in group (p < 0.001) more remarkable than in control. HbA1c stable in group, worsened in control (p < 0.002). Tendency toward lower BMI in group (p = 0.06). HDL cholesterol initially similar in both but later lower in group only (p < 0.05). Trend toward lower TG in group (p = 0.053). Dosage of hypoglycemic agents decreased (p < 0.001) among group compared to control. Body wt (p < 0.001) and BMI (p < 0.001) decreased in group but not in control. Similar reductions in BP and CV risk in group vs control, but diff significant only for DBP (p < 0.001). Significant decrease in HbA1c (p < 0.001) in group. HDL increased (p < 0.001) in group but not in control. Retinopathy progressed less in group (p = 0.009). Trento, 2004 SMA / GV T2DM (NIDDM) 56 group, 56 control ● HbA1c ● BMI ● HDL, TG ● Cr HbA1c remained stable in group but progressively increased among control (p < 0.001). BMI, HDL, TG and Cr improved over 5 yrs. in group, but not significantly different from control. Trento, 2005 SMA / GV T2DM 31 group, 31 control ● HbA1c HbA1c decreased in both, though not significantly. Trento, 2010 SMA / GV T2DM (NIDDM) 421 group, 394 control TC decreased in controls (p < 0.05), while HDL increased in group (p = 0.027). No significant modifications in other clinical variables monitored (body wt, BMI, FBG, insulin dosage, TG, ACR, foot ulcers). FBG, HbA1c, TC, TG, LDL cholesterol, SBP, DBP, and BMI decreased in group from baseline to year 4 compared to control (p < 0.001, for all measures). HDL increased in group (p < 0.001). Cr did not change significantly in group. BMI, HbA1c, TG, and Cr increased in control, whereas total, HDL, and LDL cholesterol and SBP did not change and DBP decreased. ● Lipids (TC, HDL, TG) ● Body wt, BMI ● FBG ● Insulin dosage ● ACR ● Foot ulcers ● FBG ● HbA1c ● TC, LDL, HDL, TG ● BP ● BMI ● Cr Naik, 2011 SMA / GV T2DM 45 group, 42 control ● HbA1c ● SBP ● BMI Significantly greater improvements in HbA1c immediately following active Intervention and persisted at 1-year follow-up (p = 0.05). SBP and BMI were only reported at baseline, but not significantly different between both. Wadsworth et al. BMC Family Practice (2019) 20:97 Page 9 of 13 Table 7 Biophysical data from available studies, categorized by research type (no. of articles = 10) (Continued) First author, year Qualitative Reported findings (with p-values) Biophysical measures Sample size (n) Health cond(s) Model Capello, 2008 SMA / GV HTN 58 group (no control) Raballo, 2012 SMA / GV T1DM, T2DM 121 group, 121 control Mixed Methods Krzywkowski- Mohn, 2008 SMA / GV T2DM 33 group (no control) ● BP Significant effects on SBP and DBP (p < 0.01). ● HbA1c ● Lipids (TC, HDL, TG) ● FBG ● BMI HbA1c lower in T1DM group than in control (p = 0.001) and not significantly so in T2DM (NS). Lower HDL in T1DM control (p = 0.002), but no other significant differences among both. Lower HbA1c after group intervention (p < 0.05). Diabetic clinical indicators: ● HbA1c ● LDL ● BP ● Retinal exam Increase in diabetic eye exams. ● Foot exam Lower LDL after 18 mos (p < 0.05). No significant diff. in SBP or DBP after 18 mos. No diff in diabetic foot exams (96.9% pre + post). Abbreviations: ACR Albumin/Creatinine ratio, BMI Body mass index, BP Blood pressure, Cr Creatinine, CV Cardiovascular, DBP Diastolic blood pressure, FBG Fasting blood glucose, HbA1c Glycated hemoglobin, HDL High density lipoprotein, HTN Hypertension, LDL Low density lipoprotein, NIDDM Non-insulin dependent diabetes mellitus, SBP Systolic blood pressure, T1DM Diabetes mellitus, type 1, T2DM Diabetes mellitus, type 2, TC Total cholesterol, TG Triglycerides Discussion This review limited SMA models to three general cat- egories: cooperative health care clinic, shared medical appointment / group visit, and group prenatal care / the focus on group CenteringPregnancy®. To meet in- intervention, we considered visits clinical cluded the following clinical components: review of labs, medication management, physical examination, or other medical interventions. From a strength of evidence perspective, 16 of the studies reflected a ran- domized controlled design and one non-randomized controlled design. The remaining nine studies were cohort and case study designs, with a median study duration of 12 months. that As SMAs are generalizable to primary care environ- ments, we prioritized reviews that included Internal Medicine, Obstetrics/Gynecology, Family Medicine, and Psychiatry. Though non-clinician-led SMAs have been applied in myriad ways in primary care settings, such as group-based acupuncture clinics, group psy- chotherapy for post-traumatic stress disorder and group interventions for disabled adults, we excluded them to evaluate SMAs as a variation of clinician-led primary care. To the best of our knowledge, our current review up- dates the evidence base to date and provides a necessary segue to patient-oriented outcomes. In the spirit of the Triple Aim, SMAs uniquely enhance patient-centered experience, thus we limited our review to settings that provide individual primary care consultation alongside the group visit. Individual consultation provides a re- served space for private concerns. This is an important distinction as privacy concerns have been a prominent drawback of the model identified by prior research [13, 15, 34]. We prioritized this element, recognizing the trust it fosters in the patient-clinician relationship. Summary of findings In sum, designing, promoting, and running SMAs from tested and proven formats proves to be vital for implemen- tation. Model and content fidelity demonstrate significant outcome improvement, most notably in the prenatal care and birth outcomes through the CenteringPregnancy® group process. Standardized training also improves facilita- tion of group care. Therefore, clinicians learning to facilitate group care are encouraged to receive training in facilitative leadership with emphasis on the role that a participatory at- mosphere has in improving outcomes [50]. Several models describe a physical design component to enhance the effect on patient experience or group process [3, 42, 51]. Some studies use displayed patient biophysical data for comparison and a visual aid for decision-making. Patient seating design has also been identified as a driver, both circular and U-shaped formats. Krzywokwski-Mohn stipulates that SMAs occur with participants seated around a circular conference table, with no one at the “head of the table,” balancing power and significantly influencing SMA participant outcomes [42]. Additionally, the emergence of the patient-centered medical home motivates improvement in patient educa- tion, experience of care, and measurable outcomes Wadsworth et al. BMC Family Practice (2019) 20:97 Page 10 of 13 without increasing clinical workload [3]. The interprofes- sional team plays a prominent role in SMAs across the lit- erature, including nurses, nutritionists, NPs, pharmacists, physical therapists, PAs, primary healthcare coordinators and nurse midwives [7, 8, 14, 34, 52]. Despite these reallo- cation of tasks, roles, and resources, SMAs demonstrate efficacy and feasibility across a wide range of healthcare systems [39, 53]. Despite SMAs objectively providing patients more time with their clinicians, the degree to which this affects satis- faction is unknown and patient characteristics and outside influences can affect satisfaction outcomes [7, 13, 49, 54]. Furthermore, evaluating and effectively responding to the social determinants of health requires improved identifica- tion of patient needs and outcomes assessment [55]. Nonetheless, our evaluation includes consideration of pa- tient experience fundamental for evaluating health-related quality of life, including disease-related health locus of control, health behaviors, self-efficacy, and other measures of patient perspective of care and quality of life. Lastly, studies emphasizing biophysical outcomes re- port statistically significant improvement in at least one biophysical metric, yet are too heterogeneous to com- pare across studies. Nonetheless, results are consistent with other systematic reviews by Booth et al., Edelman et al., and Jones et al. [17, 18, 49]. Limitations of review Our inclusion criteria and focus on the primary care context limited the number of articles that we evaluated in this review, which may impact the generalizability of our conclusions. Previous systematic reviews looked at a broader number of articles, though their approach also introduced more heterogeneity [17, 18, 49]. Single center studies, representing the majority for our included arti- cles on diabetes patients, may also limit generalizability. We also note that half of our included articles for the SMA / GV format were authored by the same researcher [36–40]. Other previous reviews have mentioned the im- possibility of blinding the participant and clinician / care team. Given that trials of SMA interventions cannot be designed in a traditional double-blinded manner, our quality assessment scores for quantitative studies could only receive a maximum of seven out of a total of eight points on the modified Jadad score. However, a few studies described minimizing performance bias by hav- ing the same clinician and care team manage the same intervention and control subjects and by measuring outcomes blindly for the treatment group. Furthermore, there may be sampling bias in nonrandomized con- trolled trials as well as focus groups and interviews due to the possibility that patients who are high frequency attenders may self-select to be included in the interven- subjects who have negative tion group; likewise, experiences with SMAs may decline to be interviewed or refuse to be randomized into the intervention group. Moreover, information bias may have appeared due to variation in attendance and/or completion of visits within our sample. Critiques exist concerning the evaluation of patient ex- perience through patient satisfaction measures. Aside from a lack of agreement on a converging definition of “satisfaction,” there are methodological challenges in re- liably and precisely measuring and interpreting percep- tions of the healthcare environment (survey content, mode and timing of survey administration, bias, con- founding, need for post-hoc adjustment, and subjective nature of interpersonal experiences, including patient- clinician communication as a unique dimension of qual- ity). Despite these challenges, patient experience has a meaningful role in quality improvement discussions and determination of perceived quality and sense of commu- nity [56]. Implications for practice, policy, and future research Improved resilience and coping skills, in concert with pa- tient agency and activation, are valuable outcomes of the spectrum of SMAs [34]. The primary care environment is an optimum setting to build the necessary trust, health lit- eracy, and awareness of health beliefs required for suc- cessful intersection with the broader healthcare system [35, 38]. Honoring adult learning strategies often requires nonclinical skill sets for interdisciplinary care clinicians [38]; yet, few studies focused on interprofessional practice despite widespread presence across differing SMA models. SMAs emphasize patient empowerment through peer ac- countability, socialization, and appreciation of local cul- tural context as well as patients’ familiarity and comfort with the setting [40, 43, 53]. Engaging group members in the design of these SMAs can maximize responsiveness to [43]. cultural context and acceptability of GPNC / CP have demonstrated efficacy in increasing health-related knowledge, social support, personal locus of control, emotional care, and self-care [52, 57]. the model In general, to improve quality and validity of report- ing patient experience as well as improved reporting of population health outcomes, we recommend longer duration of follow up in each study setting. We also recommend specific evaluation of team-based care, in- cluding perspectives of administrators and supporting clinical staff. As provision of healthcare is a service, measures of quality should include assessment of the extent to which patients and care teams reach a com- mon understanding of treatment course and health out- comes [2]. This intersection of shared well-being with health improvement warrants further evaluation to optimize healthcare delivery models, such as SMAs, to achieve the quadruple aim. Wadsworth et al. BMC Family Practice (2019) 20:97 Page 11 of 13 Conclusions Shared medical appointments are increasingly employed in primary care settings. This mixed-methods systematic review concludes that accepting and implementing this nontraditional approach by both patients and clinicians can yield measurable improvements in patient trust, pa- tient perception of quality of care and quality of life, and relevant biophysical measurements of clinical parameters. Compared to usual care, SMAs have a greater ability to engage and empower patients as active participants in their own healthcare while improving patient access and healthcare efficiency. The cumulative benefits of SMAs are most notable when implemented within a conducive environment such as a PCMH. No singular model of SMA best serves all settings. Similarly, there does not appear to be a priority set of their outcome measures nor consistent means for evaluation from our review. Our analysis indicates that both quantitative and qualitative methods are equally valid for evaluating patient experience. Further refine- ment of this healthcare delivery model will benefit from standardizing measures of patient satisfaction and clin- ical outcomes. Not surprisingly, critiques and cost-benefit concerns remain. Demonstration of global payment models result- ing in improved population health outcomes alongside economies of scale may be essential for wider acceptance of SMAs. We recommend further evaluation of the en- ablers and barriers to advance SMA integration in pri- mary care practice settings. We also recommend more thorough and longitudinal evaluations to better describe the consumer-minded approach for care delivery design and responsiveness to the voice of the customer to achieve the most efficient models possible. Additional files Additional file 1: Database search strategies (DOCX 27 kb) Additional file 2: Description of data: Reported significant findings related to patient experience and satisfaction, as reported in included articles (DOCX 31 kb) Additional file 3: Description of data: Barriers to implementation from available studies (no. of articles = 8) (DOCX 28 kb) Additional file 4: Description of data: SMA patient-centered variables vs. attendance and outcomes (no. of articles = 26) (DOCX 40 kb) Additional file 5: Inter-rater reliability of included articles using two-way mixed measures intraclass correlation (ICC) value for average agreement presented. (DOCX 29 kb) Abbreviations CHCC: Cooperative health care clinic; CP: CenteringPregnancy®; DQoL: Diabetes-Related Quality of Life; GPNC: Group prenatal care; GV: Group visit; PCMH: Patient centered medical home; SMA: Shared medical appointment Acknowledgments The authors thank Mary Giovanini for her help with full-text citations; Drs. Sean Cleary, MD, PhD, and Jennifer Best, MD, for their thorough edits of our manuscript; Dr. William Elliott, MD, PhD, for his critical suggestions on quality assessment of included articles; Dr. Bernadette Howlett, PhD, for her early in- put on our research methodology; Dr. Michele McCarroll, PhD, Carla S. Case and Anita Quintana, MA, for their kind assistance and support; and Tracy Dana, MLS, and Sarah Safranek, MLIS, for reviewing our literature search strategies. Authors’ contributions All listed authors significantly contributed to this project. KHW, AKC and ASH developed the study protocol. KHW, TGA and ASH conducted the title and abstract screening. KHW, TGA, AEP, and ASH conducted the full-text screen- ing, data extraction, quality assessments, and data synthesis. BLH provided the content expertise. AKC is the biomedical librarian who conducted the lit- erature search and managed the citations. All authors had access to the data, played a role in writing the manuscript, and read and approved the final manuscript. Funding This project was made possible with a Mapping the Landscape, Journeying Together grant from the Arnold P. Gold Foundation (APGF). The APGF did not have any role in the design of the study, collection, analysis and interpretation of data, nor writing the manuscript. Availability of data and materials Table 1 provides a list of the 26 included papers and Additional file 1 shows the database search strategy. Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Author details 1Pacific Northwest University of Health Sciences, College of Osteopathic Medicine, Yakima, WA, USA. 2Family Health Care of Ellensburg, Ellensburg, WA, USA. 3Multnomah County Health Department, Oregon Health and Science University–Portland State University School of Public Health, Portland, OR, USA. Received: 23 February 2018 Accepted: 31 May 2019 References 1. Berwick DM, Nolan TW, Whittington J. The triple aim: care, health, and cost. Health Aff Proj Hope. 2008;27:759–69. Batalden M, Batalden P, Margolis P, Seid M, Armstrong G, Opipari-Arrigan L, et al. Coproduction of healthcare service. BMJ Qual Saf. 2016;25:509–17. Noffsinger EB. Group visits -- the “secret sauce” of the medical home. Med Home News. 2013;5:1,6-8. Bodenheimer T, Sinsky C. From triple to quadruple aim: care of the patient requires care of the provider. Ann Fam Med. 2014;12:573–6. Noffsinger EB. Running group visits in your practice. New York: Springer; 2009. Novick G. CenteringPregnancy and the current state of prenatal care. J Midwifery Womens Health. 2004;49:405–11. Scott JC, Conner DA, Venohr I, Gade G, McKenzie M, Kramer AM, et al. Effectiveness of a group outpatient visit model for chronically ill older health maintenance organization members: a 2-year randomized trial of the cooperative health care clinic (structured abstract). J Am Geriatr Soc. 2004;52: 1463–70. Beck A, Scott J, Williams P, Robertson B, Jackson D, Gade G, et al. A randomized trial of group outpatient visits for chronically ill older HMO 2. 3. 4. 5. 6. 7. 8. Wadsworth et al. BMC Family Practice (2019) 20:97 Page 12 of 13 9. members: the cooperative Health care clinic. J Am Geriatr Soc. 1997;45: 543–9. Heberlein EC, Picklesimer AH, Billings DL, Covington-Kolb S, Farber N, Frongillo EA. Qualitative comparison of women’s perspectives on the functions and benefits of group and individual prenatal care. J Midwifery Womens Health. 2016;61:224–34. 10. McDonald SD, Sword W, Eryuzlu LE, Biringer AB. A qualitative descriptive study of the group prenatal care experience: perceptions of women with low-risk pregnancies and their midwives. BMC Pregnancy Childbirth. 2014;14:334. 11. Herrman JW, Rogers S, Ehrenthal DB. Women’s perceptions of centering pregnancy: a focus group study. MCN Am J Matern Nurs. 2012;37:19–28. 14. 13. 12. Andersson E, Christensson K, Hildingsson I. Parents’ experiences and perceptions of group-based antenatal care in four clinics in Sweden. Midwifery. 2012;28:442–8. Kennedy HP, Farrell T, Paden R, Hill S, Jolivet RR, Cooper BA, et al. A randomized clinical trial of group prenatal care in two military settings. Mil Med. 2011;176:1169–77. Tandon SD, Cluxton-Keller F, Colon L, Vega P, Alonso A. Improved adequacy of prenatal care and healthcare utilization among low-income Latinas receiving group prenatal care. J Women's Health. 2013;22:1056–61. Jafari F, Eftekhar H, Mohammad K, Fotouhi A. Does group prenatal care affect satisfaction and prenatal care utilization in Iranian pregnant women? Iran J Public Health. 2010;39:52–62. Edelman D, McDuffie JR, Oddone E, Gierisch JM, Nagi A, Williams JWJ. Shared medical appointments for chronic medical conditions: a systematic review. Washington (DC): Department of Veterans Affairs (US); 2012. https://www.ncbi. nlm.nih.gov/books/NBK99785/ 15. 16. 17. Booth A, Cantrell A, Preston L, Chambers D, Goyder E. What is the evidence for the effectiveness, appropriateness and feasibility of group clinics for patients with chronic conditions? A systematic review. Southampton: NIHR Journals Library; 2015. http://www.ncbi.nlm.nih.gov/books/NBK333454/. Accessed 9 Feb 2016 Edelman D, Gierisch JM, McDuffie JR, Oddone E, Williams JW. Shared medical appointments for patients with diabetes mellitus: a systematic review. J Gen Intern Med. 2015;30:99–106. 18. 19. Berwick DM. What “patient-centered” should mean: confessions of an extremist. Health Aff Proj Hope. 2009;28:w555–65. 20. Where Can Nurse Practitioners Work Without Physician Supervision? - Blog. 2016. https://onlinenursing.simmons.edu/nursing-blog/nurse-practitioners- scope-of-practice-map/. Accessed 5 Feb 2017. Survey of European Midwifery Regulators: Second Issue. 2010. http://www. ordre-sages-femmes.fr/wp-content/uploads/2015/11/Etude-NEMIR-2010-EN. pdf. Accessed 7 Feb 2017. 21. 22. CNM independent prescribing map \| NCSBN. https://www.ncsbn.org/index. 23. htm. Accessed 5 Feb 2017. Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJ, Gavaghan DJ, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials. 1996;17:1–12. 24. Oremus M, Wolfson C, Perrault A, Demers L, Momoli F, Moride Y. Interrater reliability of the modified Jadad quality scale for systematic reviews of Alzheimer’s disease drug trials. Dement Geriatr Cogn Disord. 2001;12:232–6. Johannsen M, Farver I, Beck N, Zachariae R. The efficacy of psychosocial intervention for pain in breast cancer patients and survivors: a systematic review and meta-analysis. Breast Cancer Res Treat. 2013;138:675–90. 25. 26. Greenland S. On the bias produced by quality scores in meta-analysis, and a hierarchical view of proposed solutions. Biostatistics. 2001;2:463–71. 27. Howlett B, Rogo E, Shelton TG. Implementation and evaluation in evidence based practice. In: Evidence based practice for Health professionals. Burlington: Jones & Bartlett Publishers; 2013. p. 309–58. Koo TK, Li MY. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J Chiropr Med. 2016;15:155–63. 28. 29. Voils CI, Sandelowski M, Barroso J, Hasselblad V. Making sense of qualitative and quantitative findings in mixed research synthesis studies. Field Methods. 2008;20:3–25. 30. Creswell JW, Klassen AC, Plano Clark VL, Smith KC, For The Office of Behavioral and Social Sciences Research. Best practices for mixed methods research in the health sciences: National Institutes of Health; 2011. https:// obssr.od.nih.gov/training/online-training-resources/mixed-methods- research/. Accessed 19 Oct 2015 Tong A, Flemming K, McInnes E, Oliver S, Craig J. Enhancing transparency in reporting the synthesis of qualitative research: ENTREQ. BMC Med Res Methodol. 2012;12:181. 31. 32. Moher D, Liberati A, Tetzlaff J, Altman DG. The PRISMA group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med. 2009;6:e1000097. 33. Clancy DE, Cope DW, Magruder KM, Huang P, Salter KH, Fields AW. Evaluating group visits in an uninsured or inadequately insured patient population with uncontrolled type 2 diabetes. Diabetes Educ. 2003;29:292–302. 34. Novick G, Sadler LS, Kennedy HP, Cohen SS, Groce NE, Knafl KA. Women’s experience of group prenatal care. Qual Health Res. 2011;21:97–116. 35. Clancy D, Yeager D, Huang P, Magruder K. Further evaluating the 36. 37. 38. 39. 40. acceptability of group visits in an uninsured or inadequately insured patient population with uncontrolled type 2 diabetes. Diabetes Educ. 2007;33:309–14. Trento M, Passera P, Bajardi M, Tomalino M, Grassi G, Borgo E, et al. Lifestyle intervention by group care prevents deterioration of type II diabetes: a 4- year randomized controlled clinical trial. Diabetologia. 2002;45:1231–9. Trento M, Passera P, Borgo E, Tomalino M, Bajardi M, Brescianini A, et al. A 3-year prospective randomized controlled clinical trial of group care in type 1 diabetes. Nutr Metab Cardiovasc Dis NMCD. 2005;15:293–301. Trento M, Passera P, Tomalino M, Bajardi M, Pomero F, Allione A, et al. Group visits improve metabolic control in type 2 diabetes - a 2-year follow- up. Diabetes Care. 2001;24:995–1000. Trento M, Gamba S, Gentile L, Grassi G, Miselli V, Morone G, et al. Rethink organization to iMprove education and outcomes (ROMEO): a multicenter randomized trial of lifestyle intervention by group care to manage type 2 diabetes. Diabetes Care. 2010;33:745–7. Trento M, Passera P, Borgo E, Tomalino M, Bajardi M, Cavallo F, et al. A 5- year randomized controlled study of learning, problem solving ability, and quality of life modifications in people with type 2 diabetes managed by group care. Diabetes Care. 2004;27:670–5. 42. 41. McNeil DA, Vekved M, Dolan SM, Siever J, Horn S, Tough SC. Getting more than they realized they needed: a qualitative study of women’s experience of group prenatal care. BMC Pregnancy Childbirth. 2012;12:17. Krzywkowski-Mohn SM. Diabetic control and patient perception of the scheduled in group medical appointment at the Cincinnati veterans administration medical center: University of Cincinnati; 2008. https://etd. ohiolink.edu/pg_10?0::NO:10:P10_ACCESSION_NUM:ucin1210103113. Accessed 9 Aug 2016 Junling G, Yang L, Junming D, Pinpin Z, Hua F. Evaluation of group visits for Chinese hypertensives based on primary health care center. Asia-Pac J Public Health Asia-Pac Acad Consort Public Health. 2015;27: NP350–60. Kennedy HP, Farrell T, Paden R, Hill S, Jolivet R, Willetts J, et al. ‘I wasn’t alone’--a study of group prenatal care in the military. J Midwifery Womens Health. 2009;54:176–83. 44. 43. 45. Raballo M, Trevisan M, Trinetta AF, Charrier L, Cavallo F, Porta M, et al. A study of patients’ perceptions of diabetes care delivery and diabetes: propositional analysis in people with type 1 and 2 diabetes managed by group or usual care. Diabetes Care. 2012;35:242–7. 46. Wong ST, Browne A, Lavoie J, Macleod MLP, Chongo M, Ulrich C. Incorporating group medical visits into primary healthcare: are there benefits? Healthc Policy Polit Santé. 2015;11:27–42. 47. Naik AD, Palmer N, Petersen NJ, Street RL, Rao R, Suarez-Almazor M, et al. Comparative effectiveness of goal setting in diabetes mellitus group clinics: randomized clinical trial. Arch Intern Med. 2011;171:453–9. 48. Capello J. An evaluation of the doctor interactive group medical Appointment : assessing changes in health behaviors attributed to an integrated healthcare model. 2008. https://repositories.lib.utexas.edu/ handle/2152/17781. Accessed 21 Jan 2016. Jones KR, Kaewluang N, Lekhak N. Group visits for chronic illness management: implementation challenges and recommendations. Nurs Econ. 2014;32:118–34 147. 49. 50. Novick G, Reid AE, Lewis J, Kershaw TS, Rising SS, Ickovics JR. Group prenatal care: model fidelity and outcomes. Am J Obstet Gynecol. 2013; 209:112.e1–6. 51. Northern Health. The group medical appointment manual first edition. 2007. http://www.hqontario.ca/Portals/0/documents/qi/learningcommunity/ roadmap%20resources/advanced%20access%20and%20efficiency/step%205/ pc-nha-group-medical-appointments-manual-en.pdf. Accessed 21 Jan 2016. 52. Andersson E, Christensson K, Hildingsson I. Mothers’ satisfaction with group antenatal care versus individual antenatal care – a clinical trial. Sex Reprod Healthc. 2013;4:113–20. Wadsworth et al. BMC Family Practice (2019) 20:97 Page 13 of 13 53. Campbell BB, Shah S, Gosselin D. Success with Men’s educational group appointments (MEGA): subjective improvements in patient education. Am J Mens Health. 2009;3:173–8. 54. Aharony L, Strasser S. Patient satisfaction: what we know about and what we still need to explore. Med Care Rev. 1993;50:49–79. 55. Asadi-Lari M, Tamburini M, Gray D. Patients’ needs, satisfaction, and health related quality of life: towards a comprehensive model. Health Qual Life Outcomes. 2004;2:32. 56. Manary MP, Boulding W, Staelin R, Glickman SW. The patient experience and health outcomes. N Engl J Med. 2013;368:201–3. 57. Baldwin K, Phillips G. Voices along the journey: midwives’ perceptions of implementing the CenteringPregnancy model of prenatal care. J Perinat Educ. 2011;20:210–7. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12875-021-01601-x	Wangler and Jansky BMC Family Practice (2021) 22:252 https://doi.org/10.1186/s12875-021-01601-x RESEARCH Open Access Prerequisites for providing effective support to family caregivers within the primary care setting – results of a study series in Germany Julian Wangler* and Michael Jansky Abstract Background: General Practitioners are considered to be well placed to monitor home-care settings and to respond specifically to family caregivers. To do this, they must be sensitive to the needs and expectations of caregivers. In order to determine the current status of GP care in terms of the support given to family caregivers, a series of studies were conducted to gather the perspectives of both caregivers and GPs. The results are used to derive starting points as to which measures would be sensible and useful to strengthen support offered to family caregivers in the primary care setting. Methods: Between 2020 and 2021, three sub-studies were conducted: a) an online survey of 612 family caregivers; b) qualitative interviews with 37 family caregivers; c) an online survey of 3556 GPs. Results: Family caregivers see GPs as a highly skilled and trustworthy central point of contact; there are many differ- ent reasons for consulting them on the subject of care. In the perception of caregivers, particular weaknesses in GP support are the absence of signposting to advisory and assistance services and, in many cases, the failure to involve family caregivers in good time. At the same time, GPs do not always have sufficient attention to the physical and psychological needs of family caregivers. The doctors interviewed consider the GP practice to be well suited to being a primary point of contact for caregivers, but recognise that various challenges exist. These relate, among other things, to the timely organisation of appropriate respite services, targeted referral to support services or the early identifica- tion of informal caregivers. Conclusions: GP practices can play a central role in supporting family caregivers. Caregivers should be approached by the practice team at an early stage and consistently signposted to help and support services. In order to support care settings successfully, it is important to consider the triadic constellation of needs, wishes and stresses of both the caregiver and the care recipient. More training and greater involvement of practice staff in the support and identifica- tion of caregivers seems advisable. Keywords: Caregivers, General practitioner, Ientification, Strain, Needs, Care, Support *Correspondence: julian.wangler@unimedizin-mainz.de Center for General Medicine and Geriatrics, University Medical Center of the Johannes Gutenberg University Mainz, Am Pulverturm 13, 55131 Mainz, Germany Background In the EU-27, over 20% of the population is already over 65 years old [1, 2]. This results in a growing need for care and support. In Germany, this need is documented on the basis of approx. 4.1 million peo- ple formally classed as needing care [3]. If informal unremunerated care and support activities are also © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Wangler and Jansky BMC Family Practice (2021) 22:252 Page 2 of 12 considered, this number increases to approximately 5.5 million who receive care or support [4, 5]. Informal care is predominantly provided in the home environment by private citizens, who bear a considera- ble share of the caregiving burden in caring for people close to them who are in need of care [6–8]. Accord- ing to representative data, more than 17% of 40- to 85-year-olds regularly support at least one person in coping with everyday life; of these, a good third pro- vide care in the stricter sense [9, 10]. Although research has shown that a caring role can provide a subjective sense of purpose [11, 12], it is associated with a greater health risk due to the physi- cal and mental strain involved [8, 10, 13–19]. If the consequences of the illness have not been considered in advance and precautionary measures have not been taken, it is not uncommon for caregivers to become burnt-out and exhausted [15, 20–22]. In order to avoid such crises and to promote the resilience of caregiv- ers, various support services have been established in Germany, including care support centres, outpatient psychiatric services and dementia networks [23]. How- ever, studies show that such services are only used by a proportion of caregivers [24–26]. Since they have provided ongoing care to the patient over many years and know them well, GPs are consid- ered to be well suited to provide support for home care settings and to respond to the particular concerns of family caregivers [6, 27–29]. Apart from diagnosing and treating health problems, GPs can provide infor- mation and advice to caregivers, offer psychosocial support and gain an overall picture of the care con- ditions so that needs can be addressed promptly. By referring patients to support and counselling services, GPs can set the course for successful long-term care and show caregivers ways to offset and relieve the bur- den of caregiving [24, 30]. In 2018, the National Association of Statutory Health Insurance Physicians (KBV) carried out a telephone survey of 6043 randomly selected citizens representa- tive of the German resident population. The study concluded that about 60% of family caregivers talk to their GP about their caring role [29]. Of these, around two-thirds had been made aware of concrete offers of help by their GP. Up to now, there has been a lack of reliable studies, especially for the German-speaking countries, which shed light on the status of GP support for the target group of family caregivers, but also on the practical challenges experienced, both broadly and from multiple perspectives, i.e. from the point of view of doctors and caregivers. Methods Overall study and research interest This paper wants to help determine the current status of German GP care in terms of the support given to fam- ily caregivers. By doing so, it summarises the results of a series of explorative studies conducted to gather the perspectives of both family caregivers and GPs, and com- pares the results with existing research. The study, which consists of three sub-studies, stands as an independent supplementary study in the broader context of an Innovation Fund model project on outpa- tient medical and nursing dementia care (DemStepCare) [31]. All three sub-studies have already been published or accepted for publication. The specific purpose of the pre- sent work was to bundle commonalities of the individual studies from an overarching perspective and to draw con- clusions in terms of an overall view of the study series. We are convinced that interconnecting the three studies in this way opens up a concentrated view and increases the informative value of all studies. The aim was to explore the attitudes, experiences and wishes of caregivers and GPs with regard to the support of caregivers provided by the GP setting. The focus was on the importance of GP support for caregivers and how GPs perceive their own remit as contact partners. One focus was to compare the needs of caregivers with the support they actually experience. Another aim was to identify the challenges for GP care. Against the backdrop of the joint consideration of all central results, the article aims to derive starting points as to which measures would be sensible and useful to strengthen support offered to family caregivers in the pri- mary care setting. In view of this focus, special attention is paid to weaknesses in the GP setting. Sub‑studies Initially, an online survey of 612 family caregivers [32] was conducted in spring 2020 to identify care needs and experiences in relation to GP care. The anony- mous survey was posted on 17 German-language Inter- net forums aimed at family care and family caregivers. The selected forums were usually embedded in general information portals on the subject of care. These web- sites are intended to support family caregivers across the board on a wide variety of questions relating to care in a domestic setting (no specific clinical pictures) and enable an exchange. Based on the registered number of members, the authors assume that the forums theo- retically reach up to 11,000 family caregivers in total. In order to obtain the broadest possible picture of the reality of care in Germany, the inclusion criterion was deliberately kept general; accordingly, the survey target Wangler and Jansky BMC Family Practice (2021) 22:252 Page 3 of 12 group included all kinds of family caregivers. The mean age of the respondents was 54 years, with 93% of the respondents being women. In a next step, we wanted to explore these results in more detail, a total of 37 family caregivers were recruited from the same online care forums and inter- viewed between autumn 2020 and spring 2021 [33]. In the respective Internet forums, a call was made in the form of a thread in which information was given on the general topic. People who were willing to be available for an interview could contact the given email address. As in the online survey, the interviews with family car- egivers essentially included all types of caregivers and care constellations. The inclusion criterion was that the caregivers had regularly cared for at least one relative, friend or neighbour in the last 12 months. In a next step, the attitudes and experiences of GPs with regard to the care of caregivers were gathered by means of an online survey [34]. In spring 2021, all 13,170 GPs in Baden-Württemberg, Hesse and Rhineland-Palatinate were invited to participate in the anonymised survey by post. In the one-time let- ter of invitation, the doctors were given, among other things, password-protected access to the online survey. Of the 3595 questionnaires processed, 3556 fully com- pleted questionnaires were included in the evaluation (response rate: 27%). This survey determined, inter alia, the priorities set by GPs when supporting caregivers and to what extent they use the available resources to make care more effective. The mean age of the GPs sur- veyed was 55 years, with about half of the doctors hav- ing their practice in rural regions. Incentives were not used in any of the three studies. Development of survey instruments Since the studies built on each other, there was a continu- ous learning process with regard to the design of the sub- sequent sub-study. In addition, the survey instruments developed were supported by other elements: • Preparations for the multi-part study series (includ- ing interviews with family caregivers in the context of DemStepCare, focus group with 8 GPs) • Further preliminary studies by the authors on dementia care by GPs (e.g., [35]) • General literature search (papers used here focused on caregivers and their support in the GP-based set- ting [12, 17, 24, 28, 30, 36, 37], including those by Geschke et al. [24], Greenwood et al. [28, 36] and Jol- ing et al. [17]. • Carrying out pre-tests in the run-up to data collec- tion The aim was to keep the instruments used to interview family caregivers and GPs mutually compatible. For this purpose, certain question models were adapted to facili- tate comparison of the results. Data analysis Data from the quantitative studies were evaluated using SPSS 23.0. Apart from the descriptive analysis, a T-test was applied to independent random samples in order to identify significant differences between two groups. In the case of the survey of family caregivers, binary logis- tic regression was used to identify possible influencing variables. Evaluation of the interview study as well as the open questions in the questionnaires is based on a quali- tative content analysis [38]. Results Figure 1 shows the starting points condensed from the analysis of the sub-studies with a view to more effective GP support for family caregivers. In the following, each of the dimensions presented will be discussed with ref- erence to the respective central findings and correlated with the existing research. Support, approach, communication As shown in the survey of family caregivers [32], car- egivers experience GPs as highly skilled and trustworthy central points of contact. Three out of four respondents (72%) talk to their GP about their caring role, with 54% doing so frequently. The way in which support is pro- vided is judged positively in important contexts, espe- cially the GP’s knowledge of the personal care situation, approachability on a wide range of problems and the attention given to the person in need of care. At the same time, the same survey shows that the wishes of caregivers with regard to an early approach by the GP practice are frequently not matched by their own experience. Fewer than one in two caregivers (42%) report having been promptly identified as such by their GP. One in five (18%) reports that the responsible GP did not wait for them to voice questions or problems regarding care but (pro-)actively approached the car- egiver. In the qualitative interviews [33], some of the car- egivers stated that they had initially felt uncertain about the extent to which their needs and problems should be a matter for GP support; there was hesitance, which sometimes deferred problematic situations. Accordingly, counselling sessions in the initial or preparatory phase of care are comparatively less frequent. Such findings from the interviews with family caregiv- ers are in line with the survey of GPs [34], which showed that the latter perceive it as a great challenge to system- atically identify informal caregivers in their daily practice Wangler and Jansky BMC Family Practice (2021) 22:252 Page 4 of 12 Fig. 1 Derived starting points for effective GP support for caregivers (own diagram) (59%). As other papers have pointed out, transitions to becoming informal caregivers are often fluid, so it can be difficult for the GP team to identify them [7, 27, 28, 36, 39]. Difficulties arise especially if the person receiv- ing care is not registered with the same practice as the caregiver [10]. Krug et al. [40] note that identification of relatives and their problems is often more likely to be in response to stress behaviours identified by the practice team. Because of this, caregivers are often not registered until stress or even decompensation processes are well advanced. In this respect, it is extremely important that GP responsibility for caregivers is explicitly signalled [37, 41, 42]. As the results of the sub-studies have shown, there is a comparatively large variation in the regularity and thus the interval between GP support consultations. In sur- veys and interviews [32, 33], caregivers take issue with a not infrequently rushed, irregular and sometimes rather casual treatment of their caring situation, which it is often taken up by other reasons for consultation (e.g., health check-ups, vaccinations). GPs often cite time constraints as a significant chal- lenge to providing adequate advice to caregivers in everyday practice (68%); in addition, many GPs find it challenging to ensure a regular exchange with caregiv- ers (43%), e.g. because the caregiver has a different GP. Linked with such barriers to support is the fact that GPs are often unable to adequately meet the need expressed by caregivers for a stabilising, psychosocial discussion [32, 33]. Care triad and needs of caregivers As already mentioned, the study results from both per- spectives reflect that GPs see themselves as contacts who are well acquainted with the situation of family caregiv- ers and have a generally accurate picture of their personal situation. Family caregivers are remarkably positive about the way in which GPs create insight on the part of the care recipient by offering explanations (85%) and involve them in decisions (82%). In contrast, slightly more than half of the caregivers interviewed report feeling that their views, needs and stresses were adequately considered by Wangler and Jansky BMC Family Practice (2021) 22:252 Page 5 of 12 the GP; 23% feel encouraged to address their own health situation [32]. advisory role when it comes to organising the framework conditions for care. The latter finding should be regarded with caution, since informal caregivers often trivialise their own com- plaints compared to the extent of the problems of the person cared for and consider potential complaints to be relative [39, 43]. Nevertheless, the results of the other studies also confirm that caregivers and their con- cerns generally receive far less consideration, given the time and resource constraints for GPs. In the interviews with caregivers [33], it was expressed that GPs are often mainly focused on the person being cared for, without considering the needs and stresses of the caregiver. On the other hand, 44% of GPs report that they find it chal- lenging to consider the needs and wishes of both the car- egiver and care recipient in their daily practice [34]. The research literature confirms such findings. Due to the often tardy and inconsistent identification of caregiv- ers and the somewhat sporadic contacts with them, GPs find it difficult to involve caregivers from the outset [32, 37, 38]. On the other hand, in the triadic constellation there is a tendency for GPs to perceive caregivers primar- ily in terms of their function relative to the person being cared for, so that psychosocial effects are marginalised [27, 36, 39, 43]. Against this background, the GP team should empathi- cally encourage caregivers to voice their own health con- cerns and offer support (consultations independently of the care recipient, where appropriate), as well as refer them to specific support services [37, 41, 42]. It is also important to involve caregivers in decision-making pro- cesses about adaptation of the care (organisation) [6, 15, 17]. Home visits can also help to better assess care and stress situations. The survey of family caregivers [32] has shown that it is not yet possible to adequately fulfil car- egivers’ wish to be visited by the GP team in their private premises. Information, advice, mediation In general, caregivers positively rate the information and advisory activities of GPs with regard to specific clini- cal pictures and courses of disease, as well as diagnostic and treatment options. One weakness identified in all the sub-studies is that GPs do not always provide referrals to counselling or support services. In the survey of fam- ily caregivers [32], for example, 60% report having been referred to support and care services by their GP at least once. The results of a regression analysis show that refer- rals to such services are an important factor influencing the feeling of being able to cope with the care situation. The interviews [33] also showed that a considerable pro- portion of caregivers would like the GP to play a greater Among the GPs surveyed [34], more than three quar- ters (79%) found it very challenging to point caregivers to the appropriate support and respite services in the area. 48% of doctors interviewed believe that they have made at least half of the family caregivers they have supported aware of concrete offers of help in recent years, day- care facilities or short-term care and care services being mentioned in particular. In response to open questions, doctors with rural practices in particular cite the lack of interprofessional structures (e.g. bridging care services, inpatient palliative care facilities) and bureaucratic hur- dles as the cause for limited mediation. In general, these results correspond with the finding in the research literature that GPs often do not have an ade- quate overview of external forms of support for caregiv- ers [24, 30] and are mostly not integrated into community health networks or (in)formal collaborative networks [43–46]. Another frequently encountered problem is the lack of availability of certain forms of assistance at short notice. For example, the GP survey [34] showed that 89% of all doctors experience the rapid availability of care or psychosocial respite services as a challenge. Use of resources In the course of the overall study, we were able to identify several practical resources which, if their use was manda- tory, can contribute to more effective support for family caregivers. One of these is the involvement of practice staff. Particularly when it comes to the identification of informal caregivers, it is important that this should not be seen as the exclusive task of GPs but, if it is to be effec- tive, as a task for the entire practice team [37, 42, 47]. Accordingly, it is important to make non-clinical practice staff (e.g. non-clinical practice assistants, primary care assistants) aware of the need to identify caregivers. Assuming that they genuinely receive appropriate training and are involved in the support of caregivers, non-clinical practice staff can also offer potential syner- gies when it comes to carrying out home visits. Even the assumption of advisory and coordinating roles (e.g. refer- ral to local support services) can relieve the burden on GPs and at the same time strengthen the mediator role of the GP practice [40]. Last but not least, practice staff can offer a lot of added value when it comes to providing (ini- tial) psychosocial support and, if necessary, linking this to referrals to stabilisation services. Practice management is particularly important when it comes to involving practice staff. Firstly, this is about creating the conditions that allow caregivers to be readily observed (e.g. rotation of staff between different duties) [37]. Secondly, obligatory and systematic arrangements Wangler and Jansky BMC Family Practice (2021) 22:252 Page 6 of 12 are required for documenting specific problems (e.g. ref- erences in the patient notes about caregiving role or signs of stress) [40, 47]. One problem is that, so far, GPs have only partially involved their practice staff in the identification and sup- port of caregivers. For example, 47% of the GPs surveyed [34] reported having members of their non-clinical prac- tice team who regularly support their own work in terms of identifying and supporting family caregivers. Similarly, only some GP practice staff are trained to undertake specific tasks associated with this. Similar findings have already been identified in several studies on the diagnosis of dementia in the primary care setting, where the gen- eral practice team has so far only been involved to a lim- ited extent in observing elderly patients and looking out for and/or documenting warning signs [35, 47]. Beyond the practice staff, another significant resource is the application of and compliance with evidence- based guidelines. The S3 guideline “Family caregivers” was published in Germany for GPs as early as 2005 and has since been updated and expanded [41]. With regard to the above-mentioned DEGAM guideline, 40% of the GPs surveyed report that they are aware of it. Of these, 55% reported using the guideline frequently or occasion- ally (44% rarely). Such results are consistent with other reports of the critical distancing of some GPs from guide- lines published by medical societies in particular [48–50]. Status quo and starting points for optimisation Overall, 68% of the caregivers surveyed who talk to their GP about care say they feel (very) well supported by the GP. 70% feel that their GP is usually good at helping them when they approach them with a care-related question. 47% of the GPs surveyed stated that there was a (very) good possibility of meeting the needs of family caregiv- ers in their everyday practice (53% less good or not good at all). The possibilities and structures that exist for GPs within the healthcare system to provide good support for caregivers are assessed positively by 44%, and rather negatively by 52%. In terms of an overall assessment, it appears that the vast majority of doctors (77%) consider the GP setting as the primary contact point for the needs of caregiv- ers. However, many respondents (56%) say that, when it comes to playing a more proactive role for this tar- get group, they are limited by the current framework conditions. In response to an open question, some of the doctors said that, in order to be able to better support caregivers in the future, they wanted to see better integration of GPs within local health and care structures or a closer col- laboration in the interprofessional network, so as to give them a better overview of existing services and the ability to make targeted referrals. In addition, they express the wish for the health insurance funds to systematically sup- port family caregivers, thereby assisting the work of GPs. Another suggestion is the creation of a low-threshold support programme, in which caregivers can be enrolled by GPs and which, on the basis of an individual risk stratification, guarantees them ongoing information and advice, as well as intervention measures when needed. Discussion Principal findings and comparison with prior work The study series was able to generate a broad picture of the current status of GP care with regard to support for family caregivers. Due to their position in the Ger- man health care system, GPs perform extensive primary care tasks. GPs are the first point of contact for patients and therefore often familiar with their patients and the patients’ family members for many years; there is a trust- ing doctor-patient relationship [6, 27–29]. The results obtained in the course of the sub-studies show that the GP setting has great potential to act as a central support for this group. Discussions with fam- ily caregivers about care (organisation) and care cir- cumstances are widespread in everyday practice and are based on a high level of trust on the part of caregivers. Especially the low-threshold accessibility for various problems, the familiarity with the personal circumstances as well as the attention to the person in need of care are experienced positively. This confirms previous studies which underline the major importance of GP support for the target group under consideration and see GPs as being in a position to make key contributions to the longer-term stabilisa- tion of home-care settings [6, 7, 14, 28–30, 51, 52]. Both caregivers and GPs believe that the primary care setting has great potential to address and deal with the problems of caregivers [7, 14, 29, 30, 52]. For example, a study con- ducted in Ireland highlights the priority role of the GP in developing longer-term coping and resilience strate- gies in home-care settings [53]. For their part, Green- wood and colleagues [30] were able to work out that the primary care setting can play a central role in support- ing and relieving the burden on caregivers and effectively coordinate further care. Nevertheless, the results of the present study also reveal weaknesses which mean that, despite being very aware of the need to support family caregivers, GPs are not always able to meet the needs of home-care situa- tions as part of their everyday practice [6, 51, 54]. This is true, for example, with regard to the role of GPs in identi- fying and anticipating care difficulties. Caregivers would also like the GP to play a greater advisory role when it comes to organising the framework conditions for care Wangler and Jansky BMC Family Practice (2021) 22:252 Page 7 of 12 and signposting them to help and support services. Addi- tionally, the sub-studies confirmed the findings from previous studies that GPs do not always consider the physical and emotional needs of family caregivers to the same extent as those of the person requiring care [30, 36, 37, 39, 42, 52]. In particular, the comparatively low level of GP referral activities and collaboration with support services in the provision of care results in restrictions and delays in the effective support and (preventive) stabilisation of caregiv- ers. As noted in various studies, GPs in Germany - espe- cially in rural regions - are often solitary providers and cannot access interprofessional networks and collabora- tions [24, 26, 30, 40, 43–46, 54]. The results of the sur- vey of family caregivers are confirmed, for example, by a Canadian study conducted by Parmar et al., who find that GPs fail to consistently address the need of caregivers and care recipients for early and regular signposting to respite services [45, 55]. When family caregivers are referred to such support services, they benefit from timely access to information on organising care [8, 52], which allows the caregiver to stay at home longer without care crises (e.g., hospitalisations) arising [24, 56]. Another issue is that the GP team does not always identify family caregivers in a timely and systematic way, making it harder to identify specific needs and antici- pate pressures. Overall, the results demonstrate the value of active communication by the GP team in relation to the family caregiver group. In the qualitative studies by Burridge et al. conducted in Australia, it is notable that caregivers do not always feel confident to voice their problems, if GPs do not signal to them that they see themselves as a point of contact [39, 57]. Against this backdrop, it makes sense to strengthen GPs’ conversa- tion skills in dealing with caring relatives through further training. If communication can be more open between both parties, family caregivers will be less reluctant to report feelings of burden, depression, and stress [51]. A systematic assessment of the caregivers’ general well- being, performed by the GP, is essential for the prompt adjustment of home care [58]. A fundamental problem not only of the German, but also of other health systems is fragmentation, meaning that the sectors are separated. As a result, primary care is often not integrated into multi-professional care, which also affects the care of family carergivers [59]. In Ger- many in particular, there is often a lack of staff who can relieve and supplement the GP, offer support to caregiv- ers and competently assign them to support services [30]. In this context, it is worth mentioning that only a proportion of GPs train non-clinical practice staff and involve them so that they can take on specific tasks such as identifying and supporting family caregivers [24, 30, 40, 47]. Studies like those by Krug et al. [40] show that the detection of exhaustion in caregivers is not system- atic among staff members, but rather a reaction to warn- ing signals that the caregivers show to the practice team. This problem is often related to a lack of knowledge and awareness [32, 35]. At the same time, various studies show that there is a great need for delegation in primary care since GPs are often overworked already in most countries [47]. Therefore, practice staff should be more systematically involved in the detection and support of family caregivers [35]. Staff members who have under- gone appropriate training can also take on referring and mediating activities to advisory and support networks. If the practice team is networked with other service provid- ers, this not only relieves caregivers, but also the prac- tices themselves; the mediator role of the GP’s practice can be strengthened. Requests made to the practice team could then be passed on to competent actors in the net- work. For example, closer cooperation with long-term care insurance funds, which GPs sometimes use in the context of care advice [40], and the local care support points could help relieve caregivers. Where such collabo- rative solutions exist in everyday practice, GPs also find it much easier to meet the needs of caregivers [51]. Practice management is of particular importance with regard to the involvement of the practice staff. On the one hand, prerequisites should be created under which it is possible to identify and observe caregivers (e.g. regularly changing work areas). On the other hand, it depends on system- atic arrangements with regard to the documentation of abnormalities (e.g. entering signs of stress in the patient file) [37, 42]. In order to stabilize home care settings, there is also the need for structured interdisciplinary forms of care that combine medical, nursing and further care offers in order to offer person-centered and evidence-based sup- port [60–62]. The lack of effective outpatient crisis inter- vention structures often leads to hospital admissions in crisis situations, which may result in serious complica- tions for patients [63]. There is some discussion on the introduction of case and support managers to assist GPs in supporting family care situations [52, 59, 64, 65]. Case managers offer the advantage that they are cross- sectorally networked and can act as a link between GPs and other care providers (e.g. care services, support net- works, emergency clinics), so that risk stratifications for those in need of care and carergivers can be carried out at an early stage [59]. Also important in care planning is the issue of ade- quate referral to care-supporting systems, networks and services. In this context, however, it has been found that GP teams often complain about inad- equate integration into professional care and advisory Wangler and Jansky BMC Family Practice (2021) 22:252 Page 8 of 12 networks [40]. A central lever for making GP support for family caregivers more effective is undoubtedly the closer integration of GPs into counselling and sup- port services [66]. To this end, it will be important to strengthen interdisciplinary communication, to estab- lish collaborative municipal networks in the area of health promotion [44, 58] and to provide GPs with a reliable knowledge of advisory services in their area in order to facilitate the straightforward referral of car- egivers. A systematic review by Plöthner et al. points out the importance of strengthening outpatient care structures [51]. The researchers draw the conclusion that establishing an outpatient care system, which sup- ports families and friends in providing (elderly) care while meeting the needs and wishes of informal car- egivers, is of high relevance. An important prerequi- site for this is to take into account family doctors with their own contractual elements, ensuring that they are appropriately remunerated when they take on advisory, mediating and caring activities for a caregiver network [21, 51, 56]. Scientifically supported model projects are already trying to strengthen the anchoring of GP-based care in regional advisory and support networks [30, 31, 59, 60]. The increased focus on evidence-based guidelines is also an important tool for better addressing the needs of caregivers. For example, manageable care plans derived from guidelines could help GPs tailor care management to the care needs of the caregiver and the patient [46, 49]. In doing so, the assessment of the care situation and its impact on the general well-being of the caregiver can approached in a structured way [66]. Clear and efficient guidelines from early diagnosis to adequate referrals can certainly improve the GP’s ability to support time- and energy-consuming home-care situations. Consequently, intervention trials focusing on the skills of GPs could be helpful in improving home-care outcomes regarding the family caregiver [32, 37]. Not only in Germany, but also internationally, there is a lack of longitudinal studies that include doctors, nurses (e.g. palliative care patients) and family caregiv- ers in order to support the development and effec- tiveness of family GP-related interventions [67] that maintain or increase the quality of life of patients and their relatives [68]. An exception is the implementa- tion of the Gold Standards Framework in Great Britain, in which family caregivers are explicitly included [69]. The caregivers‘perspectives and experiences were taken into account, e.g. the need for a professional coordinator [70] and the support of district nurses [71]. The extent to which such approaches can be adopted in the more frag- mented German health system is part of future research projects. Starting points The following starting points for effective GP support for family caregivers can be stated against the back- ground of the findings as well as the results from previ- ous studies: • Early identification, approach and involvement of (informal) caregivers is essential for providing good support [72]. For example, possible care activities can be consistently queried using anamnesis question- naire when new patients have initial contact with the practice. In addition, it seems worthwhile to design postings in the reception area of the practice and give advice from the practice team that family caregivers should identify themselves (if necessary, design in several languages). People in need of care should be asked who their informal caregivers are. In the case of new diagnoses that are known to be associated with a need for care, the practice team should ask about possible caregivers. Information with regard to care constellations can also be requested dur- ing home visits as well as informal caregivers can be identified. Patients with a presumed role as caregivers should be addressed about the issue [37, 42]. Moreo- ver, it would be beneficial if caregiving relatives also voluntarily point out their care activities and speak to GPs about this issue. This also requires health policy activities that emphasise how important it is for fam- ily caregivers to approach GPs on their own initiative and build a stable relationship [41, 42]. • In the context of early identification and crisis pre- vention, non-clinical practice staff could be more closely involved, and tasks could be delegated by the GP. To this end, GPs should invest more in spe- cial further training and in optimised practice man- agement. So far, practice teams do not systemati- cally record signs of stress and exhaustion in caring relatives. An entry in the patient file stating whether someone is a caring relative or which family mem- ber is mainly responsible for the care could pro- vide a remedy here and provide an initial indication of whom to pay particular attention to in terms of excessive demands from a care situation. The same applies to the observation and documentation of warning signals. Caregiving relatives could be pro- actively identified by the practice team during initial contact and during house calls, but also by actively asking the person in need of care [37, 51, 55, 65]. For internal communication and observation in the practice, a catalog of questions can be developed on perceptions with regard to contact with family car- egivers [41]. The systematic recording of burdens and resources of caring relatives and the networking with Wangler and Jansky BMC Family Practice (2021) 22:252 Page 9 of 12 other service providers as well as the knowledge of their offers facilitate appropriate intervention. • Family caregivers should be made aware that their support falls within the remit of the GP practice, so that any health concerns are articulated without delay. Similarly, it seems advisable not to wait for caregivers to raise problems but rather to proac- tively take the initiative (e.g. via opportunities such as health checks or vaccinations). Caregivers should be empathically encouraged to raise their own health concerns [37]. • GP practice teams should be made more aware that, within the triadic constellation, the needs, wishes and pressures of caregivers are key to the success of longer-term care [27, 32, 36, 39, 43]. If necessary, consultations should be arranged independently of the person being cared for and sufficient time should be made available, e.g. during a home visit [27, 46]. • The potential of practice staff can be used through targeted training, not only in identifying caregivers but also in advising caregivers and making home vis- its to better address care problems. In this context, psychosocial skills could be also expanded through further training. Such additional tasks taken over by members of the practice staff should be given greater consideration in the remuneration of the practice team [47, 65]. • It seems advisable to raise GPs awareness of the exist- ence and benefit of evidence-based guidelines, espe- cially with regard to supporting family caregivers [42, 72]. The Burden Scale for Family Caregivers (BSFC-s) should be used for the standardized recording of bur- dens [41]. • Family caregivers should be consistently involved in decisions with respect to the organisation of care right from the start. Caregivers too often feel bypassed when it comes to support of home care. In addition, studies reveal that interventions that are not previously discussed with the caregiver and which occur in an acute situation fail to achieve the expected result [66]. • Consistent and early mediation to concrete help and support services gives family caregivers timely access to information about the organisation of care; the risk of caregiver ‘burnout’ is significantly minimised [8, 10, 21, 24, 30, 45]. If family caregivers are suitably monitored, outpatient care can be arranged so that caregivers can stay at home longer [56, 64]. • The structural support for primary care as well as the intersectoral connection of GPs’ practices should be strengthened. In the role of multi-professional actors, case managers can mediate between GPs, patients and caregivers as well as other offers of help and, thereby, overcome the limits of a fragmented health system [52, 59, 61, 64]. A central lever for mak- ing GP support for family caregivers more effective is undoubtedly the closer integration of GPs into counselling and support services. To this end, it will be important to strengthen interdisciplinary com- munication, to establish collaborative networks in the area of health promotion [44, 58] and to provide GPs with a reliable knowledge of advisory services in their proximity in order to facilitate the straightfor- ward referral of specific caregivers (e.g. Parkinson- ism, Stroke, Dementia). A good knowledge of local conditions and effective networking of the practice team with other professional providers contribute to improved care for caregivers while strengthening their information and education as well as the pre- vention of care crises [21, 51, 56]. To this end, help and support services need to be systematised so that GPs have an overview and consultations can be structured and still be tailored to the individual needs of those affected. For some family caregivers, advice and written information will be sufficient; oth- ers will need more support and guidance. It could be worthwhile for GPs to take initiatives to improve their formal and informal cooperation with coun- seling and support actors in the field of community care. In that regard, e.g. doctor or practice networks offer great opportunities. However, this is primarily a task for structured municipal health promotion [44]. The establishment of health and prevention networks is associated with considerable advantages. Strengths and limitations This paper has helped in a comprehensive and multi- methodical way to identify information on the status quo of caregiver support in primary care. Because of the broad, heterogeneous and widely dispersed samples the results have national significance. However, the sub-studies fail to provide a representa- tive picture of opinion, due to the limited number of cases and the self-selection of respondents, since the surveys were conducted online. One has to allow for the fact that older people are less au fait with technology, so that older caregivers and GPs might be under-rep- resented in the sample. Accordingly, it can be assumed that the recruitment of caregivers in other settings (e.g. waiting room surveys in GP offices) would lead to poten- tially more generalisable statements about the population under consideration. Such studies should be conducted with a view to optimising primary care with regard to the needs of family caregivers. Wangler and Jansky BMC Family Practice (2021) 22:252 Page 10 of 12 It should also be borne in mind that caregivers were deliberately considered very broadly and that the spe- cific needs of different subgroups (e.g., those caring for dementia patients) could therefore not be considered separately. Conclusions GPs are very important to family caregivers for providing information on planning and organising care, as well as psychosocial support and reassurance. By responding to the needs of caregivers, GPs are able to stabilise home- care settings in the longer term and avert care crises. The results show that family caregivers see GPs as a highly skilled and trustworthy central point of contact. In the perception of caregivers, particular weaknesses in GP support are the absence of signposting to advisory and assistance services and, in many cases, the failure to involve family caregivers in good time. At the same time, GPs do not always have sufficient regard for the physical and psychological needs of caregivers. The doctors inter- viewed consider the GP practice to be well suited to being a primary point of contact for caregivers, but recognise that various challenges exist. These relate, among other things, to the timely organisation of appropriate respite services, mediation to appropriate assistance or the early identification of informal caregivers. Ideally, family caregivers – provided that the GP team is aware of their care activities – should be approached by the practice team at an early stage and consistently signposted to help and support services. To this end, it will be important to strengthen interdisciplinary commu- nication, to establish collaborative (municipal) networks in the area of health promotion and to provide GPs with a reliable knowledge of advisory services in their prox- imity. In order to support care successfully, it is impor- tant to consider the triadic constellation of needs, wishes and stresses of both the caregiver and the care recipient. More training and greater involvement of practice staff in the support and identification of caregivers seems advisable. Abbreviation GP(s): General Practitioner(s). Acknowledgements Not applicable. Authors’ contributions The authors alone are responsible for the content and the writing of the paper. JW prepared, coordinated and implemented the project. Both JW and MJ contributed to the project design, analysis of transcripts and drafting of the manuscript. Both authors read and approved the final manuscript. Funding Open Access funding enabled and organized by Projekt DEAL. Availability of data and materials All major data generated or analysed during this study are included in this published article. Additional information can be provided on request. Declarations Ethics approval and consent to participate All methods were carried out in accordance with relevant guidelines and regulations. Since this is an overview article on the status of the topic under discus- sion, the Ethics Commission of the State of Rhineland-Palatinate, Germany, informed us that approval by an ethics committee was not necessary. Written informed consent for participation was obtained from all participants before the start of the sub-studies [32–34]. The respondents received informa- tion about the aim and purpose of the respective study and were informed that it was an anonymous survey/interview study in accordance with the existing data protection standards. Furthermore, it was made clear that the data will only be used for scientific purposes. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Received: 14 September 2021 Accepted: 3 December 2021 References 1. Eurostat. Population structure and aging. 2021. Available from: URL: https:// ec. europa. eu/ euros tat/ stati stics- expla ined/ index. php? title ation_ struc ture_ and_ ageing. [Cited 2021 Sep 15]. = Popul 2. WHO Regional Office for Europe. Home Care in Europe. Copenhagen: 3. WHO/Europe; 2015. Statistisches Bundesamt [Federal Office of Statistics]. Pflegestatistik 2019 [Care statistics 2019]. Available from: URL: https:// www. desta tis. de/ DE/ Themen/ Gesel lscha ft- Umwelt/ Gesun dheit/ Pflege/ Publi katio nen/_ publi katio nen- innen- pfleg estat istik- deuts chland- ergeb nisse. html. [Cited 2021 Sep 15]. 4. Nowossadeck S, Engstler H, Klaus D: Pflege und Unterstützung durch Angehörige [Care and support by family members]. Report Altersdaten 1/2016. Berlin: Deutsches Zentrum für Altersfragen [German Centre of Gerontology]; 2016. 5. Klaus D, Tesch-Römer C. Pflege und Unterstützung bei gesundheitlichen Einschränkungen: Welchen Beitrag leisten Personen in der zweiten Leb- enshälfte für andere? [Care and support in health disabilities: What con- tribution do people in the second half of life give to others?]. In: Mahne K, Wolff J, Simonson J, et al., editors. Altern im Wandel: Zwei Jahrzehnte Deutscher Alterssurvey [Ageing in transition: The German Ageing Survey two decades on]. Wiesbaden: Springer VS; 2016. p. 185–200. 6. Connell CM, Boise L, Stuckey JC, et al. Attitudes toward the diagnosis and disclosure of dementia among family caregivers and primary care physi- cians. Gerontologist. 2004;44(4):500–7. PMID:15331807. https:// doi. org/ 10. 1093/ geront/ 44.4. 500. 7. DAK: DAK-Pflege-Report 2015 [DAK care report 2015]. Hamburg: DAK- Gesundheit; 2015. 8. Wuttke-Linnemann A, Henrici CB, Müller N, Lieb K, Fellgiebel A. Bouncing back from the burden of dementia: predictors of resilience from the perspective of the patient, the spousal caregiver, and the dyad — an exploratory study. GeroPsych. 2020;33(3):170–81. https:// doi. org/ 10. 1024/ 1662- 9647/ a0002 38. Linnemann A, Hilsenbek MM, Lelieveld I, Geschke K, Wolf D, Fellgiebel A. Comparison of psychosocial and medical characteristics of patients with dementia and their primary informal caregivers between inpatient and day clinic treatment. Dementia. 2020;19(3):606–17. PMID:29886778. https:// doi. org/ 10. 1177/ 14713 01218 781130. 9. Wangler and Jansky BMC Family Practice (2021) 22:252 Page 11 of 12 10. Schmidt M, Schneekloth U. Abschlussbericht zur Studie, Wirkungen 28. Greenwood N, Mackenzie A, Habibi R, Atkins C, Jones R. General des Pflege-Weiterentwicklungsgesetzes [Final report from the study on the effects of the law concerning further development of care]. Berlin: Bundesministerium für Gesundheit [Federal Ministry of Health]; 2011. 11. Bestmann B, Wüstholz E, Verheyen F. Belastung und sozialer Zusam- menhalt. Eine Befragung zur Situation von pflegenden Angehörigen. WINEGWissen 04. Hamburg: Techniker Krankenkasse; 2014. 12. O’Reilly D, Connolly S, Rosato M, Patterson C. Is caring associated with an increased risk of mortality? A longitudinal study. Soc Sci Med. 2008;67(8):1282–90. PMID:18667262. https:// doi. org/ 10. 1016/j. socsc imed. 2008. 06. 025. 13. Beach SR, Schulz R, Williamson GM, Miller LS, Weiner MF, Lance CE. Risk factors for potentially harmful informal caregiver behavior. J Am Geriatr Soc. 2005;53(2):255–61. PMID:15673349. https:// doi. org/ 10. 1111/j. 1532- 5415. 2005. 53111.x. 14. Cherry MG, Salmon P, Dickson JM, Powell D, Sikdar S, Ablett J. Factors influencing the resilience of carers of individuals with dementia. Rev Clin Gerontol. 2013;23(4):251–66. https:// doi. org/ 10. 1017/ S0959 25981 30001 30. 15. Schulz R, Sherwood P. Physical and mental health effects of family car- egiving. Am J Nurs. 2008;108(9 Suppl):23–7. PMID:18797217. https:// doi. org/ 10. 1097/ 01. NAJ. 00003 36406. 45248. 4c. 16. Dias R, Santos RL, Sousa MF, et al. Resilience of caregivers of people with dementia: a systematic review of biological and psychosocial determi- nants. Trends Psychiatry Psychother. 2015;37(1):12–9. PMID:25860562. https:// doi. org/ 10. 1590/ 2237- 6089- 2014- 0032. 17. Joling K, Windle G, Dröes R-M, et al. Factors of resilience in informal caregivers of people with dementia from integrative international data analysis. Dement Geriatr Cogn Disord. 2016;42(3–4):198–214. PMID:27669054. https:// doi. org/ 10. 1159/ 00044 9131. 18. Roepke SK, Mausbach BT, Patterson TL, et al. Effects of Alzheimer caregiv- ing on allostatic load. J Health Psychol. 2011;16(1):58–69. PMID:20709885. https:// doi. org/ 10. 1177/ 13591 05310 369188. 19. Laporte Uribe F, Heinrich S, Wolf-Ostermann K, et al. Caregiver burden assessed in dementia care networks in Germany: findings from the DemNet-D study baseline. Aging Ment Health. 2017;21(9):926–37. PMID:27171484. https:// doi. org/ 10. 1080/ 13607 863. 2016. 11817 13. 20. Laporte Uribe F, Gräske J, Grill S, et al. Regional dementia care networks in Germany: changes in caregiver burden at one-year follow-up and asso- ciated factors. Int Psychogeriatr. 2017;29(6):991–1004. PMID:28249632. https:// doi. org/ 10. 1017/ S1041 61021 70001 26. 21. Zwingman I, Hoffmann W, Michalowsky B, et al. Supporting family dementia caregivers: testing the efficacy of dementia care management on multifaceted caregivers’ burden. Aging Ment Health. 2018;22(7):889– 96. PMID:29156941. https:// doi. org/ 10. 1080/ 13607 863. 2017. 13993 41. practitioners and carers: a questionnaire survey of attitudes, awareness of issues, barriers and enablers to provision of services. BMC Fam Pract. 2010;11:100. PMID:21172001. https:// doi. org/ 10. 1186/ 1471- 2296- 11- 100. 29. Kassenärztliche Bundesvereinigung [National Association of Statutory Health Insurance Physicians]: Versichertenbefragung 2018. Ergebnisse einer repräsentativen Bevölkerungsumfrage [Results from a representa- tive survey]. Berlin; 2018. 30. Laux N, Melchinger H, Scheurich A, et al. Improving general practitioners guided dementia care. Dtsch Med Wochenschr. 2010;135(44):2175–80. https:// doi. org/ 10. 1055/s- 0030- 12674 94. 31. Gemeinsamer Bundesausschuss Innovationsausschuss [Federal Joint Committee, Innovation Committee]: DemStepCare – Hausarztbasierte Demenzversorgung mit koordinierter Kooperation und risikostratifizier- tem Einsatz spezialisierter Pflegekräfte [Dementia care from general prac- titioners with coordinated cooperation and risk-ratified use of specialised care]. [https:// innov ation sfonds. g- ba. de/ proje kte/ neue- verso rgung sform en/ demst epcare- hausa rztba sierte- demen zvers orgung- mit- koord inier ter- koope ration- und- risik ostra tifiz iertem- einsa tz- spezi alisi erter- pfleg ekrae fte. 279] Accessed on 15 Sep 2021. 32. Wangler J, Jansky M. Support, needs and expectations of family caregiv- ers regarding general practitioners - results from an online survey. BMC Fam Pract. 2021;22:47. PMID:33658009. https:// doi. org/ 10. 1186/ s12875- 021- 01381-4. 33. Wangler J, Jansky M: What prevention potential does the general prac- titioner setting offer for family caregivers? – findings from a qualitative interview study. Wien Med Wochenschr 2021 (in press). 34. Wangler J, Jansky M: General practitioners‘ attitudes, procedures and challenges towards supporting family caregivers – results of a survey of primary care physicians. Dtsch med Wochenschr 2021 [Online ahead of print]. PMID:34794181. https:// doi. org/ 10. 1055/a- 1671- 8621. 35. Wangler J, Fellgiebel A, Jansky M. Dementia diagnosis in general practitioner care – attitudes, procedures and challenges from the perspective of general practitioners in Rhineland-Palatinate. Dtsch Med Wochenschr. 2018;143(19):165–71. PMID:30231278. https:// doi. org/ 10. 1055/a- 0651- 1978. 36. Greenwood N, Mackenzie A, Harris R, Fenton W, Cloud G. Perception of the role of general practice and practical support measures for carers of stroke survivors: a qualitative study. BMC Fam Pract. 2011;12:57. PMID:21699722. https:// doi. org/ 10. 1186/ 1471- 2296- 12- 57. 37. Höppner C, Schneemilch M, Lichte T. Pflegende Angehörige und ihre Belastungen in Hausarztpraxen identifizieren – Hindernisse und Empfe- hlungen [Identifying Informal Carers and Their Burden in Family Practices – Barriers and Recommendations]. Z Allg Med. 2015;91(7/8):310–4. https:// doi. org/ 10. 3238/ zfa. 2015. 0310- 0314. 22. Hajek A, Brettschneider C, Ernst A, et al. Longitudinal predictors of 38. Mayring P. Qualitative content analysis. Basics and Techniques. Weinheim: informal and formal caregiving time in community-dwelling demen- tia patients. Soc Psychiatry Psychiatr Epidemiol. 2016;51(4):607–16. PMID:26498751. https:// doi. org/ 10. 1007/ s00127- 015- 1138-7. 23. Heinrich S, Laporte Uribe F, Wübbeler M, et al. Knowledge evaluation in dementia care networks: a mixed-methods analysis of knowledge evaluation strategies and the success of informing family caregivers about dementia support services. Int J Ment Health Syst. 2016;10(1):69. PMID:27777614. https:// doi. org/ 10. 1186/ s13033- 016- 0100-8. 24. Geschke K, Scheurich A, Schermuly I, Laux N, Böttcher A, Fellgiebel A. Effectivity of early psychosocial counselling for family caregivers in general practitioner based dementia care. Dtsch Med Wochenschr. 2012;137(43):2201–6. PMID:23076666. https:// doi. org/ 10. 1055/s- 0032- 13053 20. 25. Romero-Moreno R, Márquez-González M, Mausbach BT, Losada A. Variables modulating depression in dementia caregivers: a longitudinal study. Int Psychogeriatr. 2012;24(8):1316–24. PMID:22176670. https:// doi. org/ 10. 1017/ S1041 61021 10022 37. 26. Zwingmann I, Dreier-Wolfgramm A, Esser A, et al. Why do family dementia caregivers reject caregiver support services? Analyzing types of rejection and associated healthimpairments in a cluster-randomized controlled intervention trial. BMC Health Serv Res. 2020;20:121. https:// doi. org/ 10. 1186/ s12913- 020- 4970-8. 27. Bulsara CE, Fynn N. An exploratory study of gp awareness of carer emotional needs in Western Australia. BMC Fam Pract. 2006;7:33. PMID:16719930. https:// doi. org/ 10. 1186/ 1471- 2296-7- 33. Beltz; 2010. 39. Burridge LH, Mitchell GK, Jiwa M, Girgis A. Consultation etiquette in gen- eral practice: a qualitative study of what makes it different for lay cancer caregivers. BMC Fam Pract. 2011;12:110. PMID:21970440. https:// doi. org/ 10. 1186/ 1471- 2296- 12- 110. 40. Krug K, Bölter R, Ballhauen RA, Engeser P, Peters-Klimm F. Burden experi- enced by family caregivers of patients at the end of life: what do general practice teams offer? Gesundheitswesen. 2016;78(S 01):128–34. https:// doi. org/ 10. 1055/s- 0042- 111206. 41. Lichte T, Beyer M, Mand P, et al.: Pflegende Angehörige – DEGAM-Leitlinie Nr.6 [Family caregivers – DEGAM guideline No. 6]. Available from: URL: http:// www. degam. de/ leitl inien- 51. html. [Cited 2021 Sep 15]. 42. The princess Royal Trust for Carers and Royal College of general practi- tioners: supporting carers: an action guide for general practitioners and their teams. London: RCGP; 2011. 43. Bruce DG, Glenys AP, Underwood PJ, Roberts D, Steed D. Communication problems between general practitioners: effect on access to community support services. Med J Aust. 2002;177(4):186–8. PMID:12175321. https:// doi. org/ 10. 5694/j. 1326- 5377. 2002. tb047 29.x. 44. Prüfer F, Joos S, Milksch A. Die Rolle des Hausarztes in der kommunalen Gesundheitsförderung [the role of general practitioners in local health promotion]. Prävention und Gesundheitsförderung. 2015;10(2):180–5. https:// doi. org/ 10. 1007/ s11553- 015- 0486-1. 45. Parmar J, Anderson S, Abbasi M, et al. Family Physician‘s and Primary Care Team’s Perspectives on Supporting Family Caregivers in Primary Wangler and Jansky BMC Family Practice (2021) 22:252 Page 12 of 12 Care Networks. Int J Environ Res Public Health. 2021;18(6):3293. PMID:33806725. https:// doi. org/ 10. 3390/ ijerp h1806 3293. 2015;1(1):CD008345. PMID:25560977. https:// doi. org/ 10. 1002/ 14651 858. CD008 345. pub2. 46. Carduff E, Finuance A, Kendall M, et al. Understanding the barriers to 65. Gibson C, Goeman D, Pond D. What is the role of the practice nurse in the care of people living with dementia, or cognitive impairment, and their support person(s)?: a systematic review. BMC Fam Pract. 2020;21(1):141. PMID:32660419. https:// doi. org/ 10. 1186/ s12875- 020- 01177-y. 66. Schoenmakers B, Buntinx F, Delepeleire J. What is the role of the general practitioner towards the family caregiver of a community-dwelling demented relative? A systematic literature review. Scand J Prim Health Care. 2009;27(1):31–40. PMID:19040191. https:// doi. org/ 10. 1080/ 02813 43080 25889 07. 67. Grande G, Stajduhar K, Aoun S, et al. Supporting lay carers in end of life care: current gaps and future priorities. Palliat Med. 2009;23:339–44. 68. Jack B, O’Brien M. Dying at home: community nurses’ views on the impact of informal carers on cancer patients’ place of death. Eur J Cancer Care (Engl). 2010;19(5):636–42. PMID:20030704. https:// doi. org/ 10. 1111/j. 1365- 2354. 2009. 01103.x. 69. Thomas K. Caring for the dying at home. Oxford: Radcliffe: Publishing; 2003. 70. Boyd KJ, Worth A, Kendall M, et al. Making sure services deliver for people with advanced heart failure: a longitudinal qualitative study of patients, family carers, and health professionals. Palliat Med. 2009;23(8):767–76. PMID:19926645. https:// doi. org/ 10. 1177/ 02692 16309 346541. 71. Griffiths J, Ewing G, Rogers M. Early support visits by district nurses to cancer patients at home: a multi-perspective qualitative study. Palliat Med. 2013;27:349-357. PMID:22801979. https:// doi. org/ 10. 1177/ 02692 16312 451949. 72. Schneider N, Mitchell GK, Murray SA. Palliative care in urgent need of rec- ognition and development in general practice: the example of Germany. BMC Fam Pract. 2010;11:66. PMID:20843334. https:// doi. org/ 10. 1186/ 1471- 2296- 11- 66. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. identifying carers of people with advanced illness in primary care: trian- gulating three data sources. BMC Fam Pract. 2014;15:48. https:// doi. org/ 10. 1186/ 1471- 2296- 15- 48. 47. Wangler J, Fellgiebel A, Jansky M. The practice staff in primary care dementia recognition-is there an untapped potential? Z Gerontol Geriat. 2019;52(7):661–6. PMID:30478791. https:// doi. org/ 10. 1007/ s00391- 018- 01484-1. 48. Carlsen B, Bringedal B. Attitudes to clinical guidelines – do GPs differ from other medical doctors? BMJ Qual Saf. 2011;20(2):158–62. PMID:21209148. https:// doi. org/ 10. 1136/ bmjqs. 2009. 034249. 49. Carlsen B, Norheim OF. "what lies beneath it all?" – an interview study of GPs’ attitudes to the use of guidelines. BMC Health Serv Res. 2008;8:218. PMID:18945360. https:// doi. org/ 10. 1186/ 1472- 6963-8- 218. 50. Cranney M, Warren E, Barton S, et al. Why do GPs not implement evi- dence-based guidelines? A descriptive study. Fam Pract. 2001;18(4):359– 63. PMID:11477041. https:// doi. org/ 10. 1093/ fampra/ 18.4. 359. 51. Plöthner M, Schmidt K, de Long L, Zeidler J, Damm K. Needs and prefer- ences of informal caregivers regarding outpatient care for the elderly: a systematic literature review. BMC Geriatr. 2019;19(1):82. PMID:30866827. https:// doi. org/ 10. 1186/ s12877- 019- 1068-4. 52. Donath C, Gräßel E, Großfeld-Schmitz M, et al. Effects of general practi- tioner training and family support services on the care of home-dwelling dementia patients - results of a controlled cluster-randomized study. BMC Health Serv Res. 2010;10:314. PMID:21087474. https:// doi. org/ 10. 1186/ 1472- 6963- 10- 314. 53. Lane P, McKenna H, Ryan A, et al. The experience of the family caregiv- ers’ role: a qualitative study. Res theory Nurs Pract. 2003;17(2):137–51. PMID:12880218. https:// doi. org/ 10. 1891/ rtnp. 17.2. 137. 53173. 54. Pentzek M, Fuchs A, Abholz H-H, Wollny A. Awareness of local dementia services among general practitioners with academic affiliation. Aging Clin Exp Res. 2011;23(3):241–3. PMID:20530989. https:// doi. org/ 10. 1007/ BF033 37750. 55. Parmar J, Anderson S, Abbasi M, et al. Support for family caregivers: a scoping review of family physician’s perspectives on their role in sup- porting family caregivers. Health Soc Care Commun. 2020;28(3):716–33. PMID:31858674. https:// doi. org/ 10. 1111/ hsc. 12928. 56. Thyrian JR, Fiss T, Dreier A, et al. Life- and person-centred help in Mecklenburg-Western Pomerania, Germany (DelpHi): study protocol for a randomised controlled trial. Trials. 2012;13:56. PMID:22575023. https:// doi. org/ 10. 1186/ 1745- 6215- 13- 56. 57. Burridge LH, Mitchell G, Jiwa M, et al. Helping lay carers of people with advanced cancer and their GPs to talk: an exploration of Australian users’ views of a simple carer health checklist. Health Soc Care Commun. 2017;25(2):357–65. PMID:26694537. https:// doi. org/ 10. 1111/ hsc. 12312. 58. Kuske S, Graf R, Hartig M, Quasdorf T, Vollmar HC, Bartholomeyczik S. Dementia considered? Safety-relevant communication between health care settings: a systematic review. J Public Health. 2014;22(5):383–93. https:// doi. org/ 10. 1007/ s10389- 014- 0630-y. 59. Bablok I, Binder H, Graf E, Stelzer D, et al. Primary dementia care based on the individual needs of the patient: study protocol of the cluster randomised controlled study DemStepCare. BMC Geriatr. 2021;21(1):222. PMID:33794789. https:// doi. org/ 10. 1186/ s12877- 021- 02114-z. 60. Hermann K, Boelter R, Engeser P, et al. PalliPA: How can general practices support caregivers of patients at their end of life in a home-care setting? A study protocol. BMC Res Notes. 2012;5:233. PMID:22583663. https:// doi. org/ 10. 1186/ 1756- 0500-5- 233. 61. Mißlbeck A: Plädoyer für gemeinsame Behandlungspfade [A plea for common treatment pathways]. Available from: URL: https:// www. aerzt ezeit ung. de/ Polit ik/ Demenz- Plaed oyer- fuer- gemei nsame- Behan dlung spfade- 372175. html [Cited 2021 Sep 15]. 62. Radisch J, Baumgardt J, Touil E, et al. Dementia - treatment pathways for integrated outpatient care. Stuttgart: Kohlhammer; 2015. German 63. Wolf D, Rhein C, Geschke K, Fellgiebel A. Preventable hospitalizations among older patients with cognitive impairments and dementia. Int Psychogeriatr. 2019;31(3):383–91. PMID:30221613. https:// doi. org/ 10. 1017/ S1041 61021 80009 60. 64. Reilly S, Miranda-Castillo C, Malouf R, et al. Case management approaches to home support for people with dementia. Cochrane Database Syst Rev. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12875-020-01171-4	Pitt et al. BMC Family Practice (2020) 21:117 https://doi.org/10.1186/s12875-020-01171-4 R E S E A R C H A R T I C L E Open Access Sharing reports about domestic violence and abuse with general practitioners: a qualitative interview study Katherine Pitt1* Emma Williamson2 and Eszter Szilassy1 , Sandi Dheensa1, Gene Feder1, Emma Johnson1, Mei-See Man1, Jessica Roy2, Abstract Background: Domestic violence and abuse (DVA) is common and damaging to health. UK national guidance advocates a multi-agency response to DVA, and domestic homicide reviews consistently recommend improved information-sharing between agencies. Identification of patients experiencing DVA in general practice may come from external information shared with the practice, such as police incident reports and multi-agency risk assessment conference (MARAC) reports. The aim of this study was to explore the views of general practitioners (GPs) and the police about sharing reports about DVA with GPs. Methods: Qualitative semi-structured interviews were conducted with GPs, police staff and a partnership manager. Participants were located across England and Wales. Thematic analysis was undertaken. Results: Interviews were conducted with 23 GPs, six police staff and one former partnership manager. Experiences of information-sharing with GPs about DVA varied. Participants described the relevance and value of external reports to GPs to help address the health consequences of DVA and safeguard patients. They balanced competing priorities when managing this information in the electronic medical record, namely visibility to GPs versus the risk of unintended disclosure to patients. GPs also spoke of the judgements they made about exploring DVA with patients based on external reports, which varied between abusive and non-abusive adults and children. Some felt constrained by short general practice consultations. Some police and GPs reflected on a loss of control when information about DVA was shared between agencies, and the risk of unintended consequences. Both police and GPs highlighted the importance of clear information and a shared understanding about responsibility for action. Conclusion: GPs regarded external reports about DVA as relevant to their role, but safely recording this information in the electronic medical record and using it to support patients required complex judgements. Both GPs and police staff emphasised the importance of clarity of information and responsibility for action when information was shared between agencies about patients affected by DVA. Keywords: Domestic violence and abuse (DVA), Intimate partner violence (IPV), Information-sharing, Multi-agency, General practice, Primary care, Police, Electronic medical records (EMR) * Correspondence: kate.pitt@bristol.ac.uk 1Bristol Medical School (Population Health Sciences), University of Bristol, Canynge Hall, 39 Whatley Road, Bristol BS8 2PS, UK Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Pitt et al. BMC Family Practice (2020) 21:117 Page 2 of 10 Background Domestic violence and abuse (DVA) is a violation of hu- man rights that damages health, requiring a public health and clinical response. The UK Government de- fines DVA as: any incident of controlling, coercive or threatening be- haviour, violence or abuse between those aged 16 or over who are or have been intimate partners or family mem- bers, regardless of their gender or sexuality [1]. The 2018 Crime Survey for England and Wales found that an estimated 7.9% of women (1.3 million) and 4.2% of men (695,000) experienced DVA in the previous year [2]. Evidence suggests that the prevalence among pa- tients accessing health services is even higher. A cross- sectional survey in primary care waiting rooms found that 17% of women had experienced past year and 41% lifetime physical violence from a current/ former partner [3]. However, DVA remains under reported in general practice [3, 4]. DVA is associated with substantial morbidity and mor- tality [2, 5, 6]. Women are disproportionally affected as victims [2]. At the same time, male victims may experi- ence additional barriers to accessing help [7]. The phys- ical and mental health burden affecting victims of DVA is well evidenced [5, 6]. Furthermore, exposure to DVA is a form of child maltreatment [8]. Children affected by DVA are at higher risk of growth, developmental, and behavioural problems [9] and other forms of child mal- treatment [10]. Professional DVA costs health an estimated £2.3 billion, police £1.3 billion and housing £550 million annually [11]. The National Institute for Health and Care Excel- lence (NICE) advocates a coordinated multi-agency response [12]. Domestic homicide reviews and child serious case reviews have consistently advocated im- proved information-sharing between agencies [8, 13, that 14]. information-sharing between agencies may be justi- fied (in specific circumstances without consent) to safeguard children or prevent serious harm to DVA victims [15, 16]. Multi-agency structures are estab- lished in the UK to facilitate information-sharing, namely multi-agency risk assessment conferences (MARAC) hubs (MASHs) [17, 18]. and multi-agency safeguarding guidance indicates General practitioners (GPs) have an important role in the multi-agency response to DVA, given its prevalence and health impact [3, 5, 6]. Some DVA cases, involving children or adults with care and support needs, are also relevant to doctors’ safeguarding responsibilities [19, 20]. Professional guidance supports doctors’ role in identify- ing and responding to DVA [19, 21]. Research indicates that both men and women feel it appropriate for doctors to ask about DVA [22, 23]. Previous research with GPs has identified concerns about documenting DVA in the patient medical record [24]. Studies show that GPs use various approaches to documenting DVA [25] and that DVA remains under- recorded in the medical record [3, 4]. National guidance exists about safely recording DVA in electronic medical records (EMRs) [26]. Patient online access to EMRs has led to further concerns about unintended breaches of confidentiality [27, 28]. A randomised controlled trial of a general practice- based training and advocacy intervention - Identifica- tion and Referral to Improve Safety (IRIS) – reported increased rates of identification and referral to DVA specialist support of women affected by DVA [29]. An adaptation of IRIS, IRIS+, has extended this inter- vention to also respond to children and men. The first stage of the IRIS+ feasibility study was conducted in four general practices in one region of England over eleven months between 2017 and 2018. Follow- ing IRIS+ training, electronic medical records (EMRs) in the study practices were searched to measure DVA identifications and documented GP responses during the study period. from reports An evaluation of the IRIS+ feasibility study find- ings is underway. Briefly, two-thirds of DVA identifi- cations were shared with general practice from another agency incorporated into the EMR, not consultations with a clinician. A large ma- third-party notifications were police DVA jority of incident families with children or reports about MARAC reports about high-risk cases of DVA (Szi- lassy E, et al., Reaching everyone in general practice? Feasibility of an integrated domestic violence training and advocacy support intervention for primary care, unpublished). The importance of external reports as a source of DVA identifications in general practice was not anticipated in the IRIS+ study design. There was no documentation in the EMR to indicate how GPs were using or responding to this information. We explored attitudes towards these reports with four GPs who participated in IRIS+, and their views were variable. We wanted to explore professional at- titudes towards sharing reports about DVA with GPs more closely. This led to an IRIS+ interview sub- study to explore GP and police staff views about DVA information sharing, reported in this paper. A few studies explore information sharing between public sector organisations about DVA [30, 31]. However, little research explores professional atti- tudes towards sending reports about DVA to GPs. One study in Scotland found difficulties implement- ing information-sharing between the police and GPs in practice [32]. This study therefore addressed a gap in research. This study is timely, given national Pitt et al. BMC Family Practice (2020) 21:117 Page 3 of 10 advocating guidance inter-agency information-sharing and a specific general practice response to DVA [12, 33]. improved Methods Aims and objectives The aim of this qualitative interview study was to ex- plore the views of GPs and police staff about sharing in- formation concerning DVA with general practice, with a view to informing the interpretation of the IRIS+ feasi- bility study findings and the possible reconfiguration of the IRIS+ intervention. The objectives of the study in- cluded exploring perceptions about the benefits and problems associated with sending reports about DVA to GPs, how they were used to inform patient care, and what was incorporated into the EMR. Police staff and MARAC chairs were interviewed, in addition to GPs, be- cause most external reports about DVA identified in study practices were police incident reports or MARAC reports. Study setting The research team was based at the University of Bristol and participants were recruited from across England and Wales. Recruitment was not limited to the area covered by the IRIS+ feasibility study because DVA information- sharing is relevant to professionals working in other localities. Study design The study was a qualitative interview project. GPs were recruited by email advertisement to professional con- tacts, two regional clinical research networks and the IRIS network of practices [34]. The study was also pro- moted by two research posters at primary care confer- ences. Police staff were recruited from an advertisement on an electronic forum for a DVA charity. The sampling technique was predominantly convenience, based on re- sponse to the study invitation. All police staff who expressed interest were interviewed. Seven of the GPs were selected for known expertise in DVA or safeguard- ing. The final four GPs were purposively selected based on specific characteristics: prison service, region of the country, and male gender, with the aim of increasing the range of participants. In consequence, four participants who expressed interest towards the end of the study were not recruited. Potential participants were sent a participant informa- tion sheet and consent form in advance. Informed writ- ten or recorded verbal consent was then taken for all interviews. Participants were offered a certificate and £15 shopping voucher as a gesture of gratitude for their time. Interviews were conducted over a five-month period during 2018 by KP and SD. The topic guides were devel- oped based on interviews with GPs involved in the IRIS+ feasibility study (see topic guides in additional files). The interviews were all semi-structured, which means that we used topic guides flexibly, probing relevant aspects of participants’ responses. KP, SD, and EJ met regularly to discuss all aspects of the study, including topic guides. The topic guides were adjusted during the data collec- tion process in response to themes identified [35]. We added questions to the guide if we wanted more specific information about an issue. For example, as it emerged during the study that participants had different experi- ences of information-sharing, we added questions to the topic guide to explore participant perception of local information-sharing arrangements. The topic guide in- cluded the option of discussing a fictitious scenario in which a police report about an incident of DVA in a house- hold with a child was shared with a GP practice. Interviews were audio-recorded and transcribed verbatim [36]. By the final interviews, the researchers felt that they had obtained sufficient depth and variation in the main concepts discussed, indicating saturation of the main themes [37]. GP and police transcripts were analysed as a single data set using inductive thematic analysis by three researchers (KP, SD and EJ) [38]. Each transcript was double coded using NVivo software version 11. Dur- ing the analysis process, the team met to discuss how the codes identified could be organised into themes. KP led on reviewing the themes and designing a thematic structure. The final thematic structure was reviewed and agreed by SD and EJ. Ethics approval The study was approved by the South West – Frenchay NHS Research Ethics committee (Reference number: 17/ SW/0098). Results Interview participants consisted of 23 GPs, six police staff and one former DVA partnership manager (who worked closely with one constabulary). Interviews were conducted by telephone or face-to-face (four, university premises or workplace). The duration of interviews var- ied, with an average of 35 min. The demographic and professional characteristics of participants are summarised in Tables 1 and 2. GP participants worked in six different geographical re- gions of England and Wales. Police participants worked in five different geographical regions in Eng- land. GP participants were not from IRIS+ practice sites. Participants experience of information-sharing in relation to DVA: 47.8% of GPs interviewed had some experience of police reports varied in their Pitt et al. BMC Family Practice (2020) 21:117 Page 4 of 10 Table 1 participant characteristics (GPs) GP participants Number Age (years) 30–39 40–49 50–59 60–69 Gender F M General practice role* GP Registrar Locum GP Salaried GP GP Partner Retired GP Prison GP GP (unspecified) Specific DVA / safeguarding role* Current / former safeguarding lead IRIS (DVA) trainer 9 6 4 4 16 7 2 2 4 9 2 2 2 9 4 * Some participants had more than one professional role and / or MARAC reports being shared with general practice. GPs described receiving external reports about DVA from other agencies, namely children’s so- cial services as well as from Accident and Emergency departments. The police respondents had not shared reports directly with GPs; some described sharing Table 2 participant characteristics (Police) GP participants Number (%) Age (years) 30–39 40–49 50–59 Gender F M Police role* Detective inspector Sergeant Police constable Advanced practitioner Former partnership manager Specific DVA multi-agency role* MARAC chair *Some participants had more than one professional role 1 3 3 1 6 3 1 1 1 1 3 information with MARACs or MASHs, which in- cluded representatives from the health sector. Partici- pant attitudes did not vary consistently depending on demographic characteristic or professional experience. Researchers identified four overarching themes during the process of analysis: [1] the relevance and value of ex- ternal reports about DVA to GPs [2]; managing compet- ing priorities in the EMR [3]; exploring DVA with patients - professional judgement and system con- straints; and [4] the challenge of coordinating action be- tween agencies. Theme 1: relevance and value of external reports about DVA to GPs Notifications about DVA were considered valuable by GPs because they brought to light hidden issues of DVA and helped them to address the health consequences and safe- guard children. These subthemes are explored below. (1a) identify DVA that might otherwise remain hidden Participants described some patients as reluctant to dis- close DVA due to fear and mistrust of professionals. Some GPs described DVA as thereby under-identified, and external reports as key to addressing this. ‘It’s really important for us to know this stuff. Actu- ally, for a lot of these patients if it wasn’t for the po- lice reports, we might not know it, we might not have asked about it, it might not have come up. It is valu- able information.’ [GP 01] (1b) address health consequences of DVA GPs highlighted the adverse health consequences of DVA, and their role in addressing these. ‘I think in terms of your question about what GPs can bring, I think we're absolutely central, because as we know with domestic abuse, the impact on your mental and physical health is massive. Therefore, it's really important that we know about it because it impacts on your overall wellbeing.’ [GP 02] GPs varied in how broadly they defined their role, from specifically addressing the health consequences to pro- viding holistic support. (1c) safeguard children GPs regarded DVA to be a risk to children’s welfare and child safeguarding as a key responsibility. Participants information-sharing be- discussed the importance of tween agencies and professionals in child protection. GPs valued external reports about DVA involving chil- dren, because they helped to contribute to a better un- derstanding of potential risks to the child. Pitt et al. BMC Family Practice (2020) 21:117 Page 5 of 10 ‘I realise when we do our serious case reviews every time that there’s a child death and there seemed to have been domestic violence in the house and differ- ent agencies haven’t known what’s been happening.’ [GP 19] Theme 2: managing competing priorities in the EMR While GPs regarded notifications about DVA as valu- able, some raised dilemmas about how to record this in- formation in patients’ EMR. These dilemmas are explored below. (2a) balancing the visibility of information against the risk of unintended disclosure GPs felt that DVA should be visible to GPs in the EMR, particularly when caring for victims and children. How- ever, the visibility of information about DVA was bal- anced against the risk of unintended disclosure to patients, such as perpetrators. ‘…there’s huge concern about recording and how to record and when to record, and how to redact, and when it’s confidential.’ [GP 05] (2b) ownership of the EMR and labelling Participants described a tension between the patient’s ownership and access to the EMR, versus its role as a re- pository of information assisting the safeguarding re- sponsibilities of GPs. ‘…there’s also the issue of these are patients’ notes, normally when we code something, like diabetes or something, we’d be having a discussion with the pa- tient. Obviously, this is a bit sensitive, and I think, “Well, you don’t want to be paternalistic.” I’m still a bit trying to figure out what the best way is.’ [GP 12] Some GPs were concerned about assigning labels within the EMR. A few participants felt that DVA could be regarded as stigmatising by patients (if they accessed the EMR). They also described the complexity of DVA and the risk of making assumptions based on informa- tion in external reports. ‘…we don’t always know the full extent of the story. Perhaps it risks putting something, which can give somebody stigma, in their notes without having the full benefit of the facts.’ [GP 18] Theme 3: exploring DVA with patients - professional judgement and system constraints GPs described the judgements they made about explor- ing DVA with patients, with additional complexities when information came from an external report. These judgements, and the constraints of the general practice consultation described by participants, are explored below. (3a) navigating consultations with victims Some participants reported that patient knowledge of, and consent for, information-sharing would influence how they responded. Without this, some GPs were con- cerned that enquiry based on third party reports may be intrusive or upset patients. ‘If the police were to send the report to the GP, I think it’s important for those patients that they get their consent. I presume they are going to consent patients, to say, “Look, we are forwarding this information on to the GP, who may choose to contact you. Or you may choose to contact the GP directly.” I think if that con- versation happens then that’s very good.’ [GP 09] Participants described variation in patients’ recognition of abuse and readiness for support. GPs explained that the patients would often have a different agenda when they consulted, such as addressing a physical health complaint. Given this, some GPs favoured an incremen- tal or indirect approach to asking about DVA, providing patients with the space to disclose DVA themselves. GPs talked of the value of building a trusting relationship with patients by supporting them with medical problems more specific to the GP role. ‘There is something about […] relationships. I think we have a role in being able to address their prob- lems, so actually it's making sure you've got effective contraception. You're really worried about your physical health. Yes, well, if that's one less worry for you. It's not always about talking about domestic abuse all the time or abuse in general. It's about holistic care, isn't it?’ [GP 02] let's deal with that, Some participants reflected on the autonomy of adult victims of DVA. In contrast, others described the vulner- ability of victims of DVA, and how coercion limited their freedom and capacity to make decisions. This tension could lead to uncertainty for professionals supporting victims of DVA about how to help, particularly when formal safeguarding procedures did not apply. ‘…if there are kids or vulnerable adults, that’s fine, we know what we’re doing. It’s that grey area where survivors are falling into… people assume because safeguarding has been affected, that somehow the as- sociated issues, be it the sexual violence or domestic abuse, substance misuse, have also been addressed and that’s absolutely not the case. [Police 02]’ Pitt et al. BMC Family Practice (2020) 21:117 Page 6 of 10 (3b) responding to perpetrators: risk and boundaries to the GP role GP participants expressed concerns about causing an es- calation of violence if they attempted to discuss perpet- ration. feared Some GP and police participants undermining the doctor-patient relationship. (4a) losing control when information is shared with another agency A few participants feared that sharing information with external agencies or other practitioners could have un- certain consequences, particularly if the recipient’s confi- dence in responding to DVA was unknown. ‘…actually GPs are about care, aren’t they? GPs are about providing care for someone who is in need of physical and sometimes mental health from a med- ical professional. If a perpetrator rocks up, who is perfectly entitled to his or her medical care, you run the risk of changing the perspective of that GP to the point where their decision making is affected because of that, and they no longer are giving perhaps the impartial service that they might have done.’ [Police 05] Some participants described the GP role as supporting perpetrators with health problems that might exacerbate DVA, namely poor mental health and substance misuse. (3c) constraints of the GP consultation with children The constraints of the GP consultation were raised, par- ticularly in relation to speaking to young children identi- fied in external reports. Some participants highlighted the importance of hearing the perspective of children in households affected by DVA. ‘…the reality is that we probably don’t ask the children enough and don’t hear their voice enough.’ [GP 01] However, many participants felt that speaking to young children required sensitivity, which depended on time and rapport, whilst consultations were described as time-limited and mostly taken up dealing with unrelated problems. Participants also discussed the complexity of managing a consultation with a child brought in by their parent / carer, and difficulty seeing the child alone. ‘Whereas in a paediatric setting or with specialists or safeguarding specialists often the infrastructure is there to allow them to see the children individually, to ask them questions and to build that rapport, but we just don’t get that opportunity really.’ [GP 21] Theme 4: challenge of coordinating action between agencies While GP participants acknowledged the value of exter- nal reports about DVA, both GP and police participants described difficulty coordinating action across organisa- tional boundaries. These difficulties, pertaining to losing control when information is shared and clarity about re- sponsibility for action, are explored below. ‘I’m not confident that if we shared the information with the GP, what the GP would actually do with it. That would range from nothing to potentially in- appropriately sharing information’ [Police 03] (4b) clarity of information and responsibility for action Some police and GP participants argued that sending in- formation about DVA to GPs could be conflated with inappropriately transferring responsibility for action. ‘It would unfair to just say, “Well, we’ve sent it to you, your problem.”’ [Police 01] Participants advocated thresholds to govern when in- formation about DVA was sent to GPs, describing the high workload and limited capacity in primary care. They discussed the importance of a shared understand- ing between agencies about expectations when informa- tion about DVA was shared. ‘I think we owe it to them that if we burden them with information around domestic abuse it’s only fair that we let them know what our expectations are around that.’ [Police 05] GPs too highlighted the importance of the clarity and content of the report itself. ‘What we need is a way of highlighting domestic vio- lence saying “This is what has been done, this is what is going to be done, this is what we need you to do as a GP”.’ [GP 21] Discussion Study findings This study explored the views of GPs and the police about sharing reports about DVA with GPs. The GPs interviewed regarded external reports about DVA as relevant to their role. Participants perceived GPs as a po- tential source of support for people affected by DVA. However, there were barriers to directly addressing DVA in the GP consultation, particularly based on informa- tion from an external report. Patient knowledge of and consent to information-sharing was an important deter- minant influencing the response of GPs. Some partici- pants were concerned about addressing perpetration in general practice, due to fears about undermining the Pitt et al. BMC Family Practice (2020) 21:117 Page 7 of 10 doctor-patient relationship and causing an escalation of violence. Speaking to children about DVA was particu- larly challenging in time-limited consultations. Despite the difficulties with addressing DVA directly, GPs argued that information about DVA informed the care that they provided to patients. Recording this information within the EMR required GPs to balance visibility to clinicians against the risk of unintended disclosure to patients. Both GPs and police staff described the practical chal- lenges of inter-agency information-sharing, and the im- portance of guidance for the recipient and clarity about responsibilities. Participant experiences of information- sharing varied. However, the variation in attitudes did not of correspond information-sharing or job role. consistently experience to Comparison to existing literature GPs’ view of DVA as relevant to health care profes- sionals reflects professional guidance [19, 21]. A previous study about DVA police reports sent to GPs in Scotland found similar perceptions [32]. Although participants in this study described GPs as a potential source of on- going support, research with survivors indicates mixed experiences of gaining support from GPs, with some sur- vivors concerned about GPs’ skill or experience asking about and responding to DVA [39, 40]. This echoes con- cern by participants in this study about information- sharing if the recipient’s competence in responding to DVA is unknown. Concerns about discussing DVA mirror previous stud- ies (not based on external reports), include time con- straints, perceived patient reluctance to discuss DVA and stigma, and fear of making things worse [41–44]. GPs in this and previous studies discussed the role of trust in enabling discussions about DVA [41, 45]. In this study, GPs discussed the additional complexity of ex- ploring DVA based on information in an external report. Interestingly, GPs in this study and in previous studies described strategies to frame inquiry that were often in- direct and incremental [41, 45]. As with the current study, fears about discussing per- petration have been raised in previous research with GPs [46, 47]. Evidence indicates limited engagement by other professions with DVA perpetrators, for example social care [48, 49]. Some GPs described the complex dynamics of DVA, which may be unclear in an external report. GPs in other research studies have found it difficult to differentiate between perpetrators and victims of DVA in practice [46]. GPs in this study consistently highlighted the impli- cation of DVA for child safeguarding. This link was identified variably in previous interview studies with GPs [42, 50]. GPs highlighted the importance of inter-agency information-sharing to building a picture of a child’s welfare. This reflects national policy guid- ance and learning from child serious case reviews [51]. Some GPs emphasised the importance of under- standing children’s own perspective, an approach that national guidance supports [51, 52]. GPs also de- scribed this as difficult, especially given the topic’s sensitivity and time constraints [50, 53]. GPs described making complex judgements about re- cording DVA reports in the EMR. As with previous re- search with GPs about recording DVA and potentially stigmatising information in the medical record, some were reticent to record third-party information in al- leged perpetrators’ EMRs, citing risk of unintended dis- closures and concerns about labelling [25, 54]. GPs discussed tensions between their and their patients’ ownership of the EMR, and heterogeneous approaches to recording sensitive information [25, 55]. Participants reflected national policy in advocating a multi-agency response to DVA [12]. However, police staff were concerned about sharing information with professionals outside their organisation. As with a previ- ous study, they were unsure how GPs would respond to reports about DVA [32], which may reflect GPs’ limited participation in multi-agency DVA collaborations [24]. Participants described the large workload and limited capacity in general practice. Previous research about po- lice reports shared with children’s social services also de- scribed the workload implication [49]. Participants specified the need for clear expectations, when informa- tion was shared between agencies, reflecting previous re- search about DVA partnerships [48, 56]. Implications for training and professional practice General practice DVA training (IRIS) currently focuses on direct disclosures of DVA to a clinician. Since GPs also receive information from other agencies, they may benefit from advice about how to manage this informa- tion. This has been incorporated into GP training in the next stage of the IRIS+ feasibility study. Any guidance will need to reflect local multi-agency arrangements. We are also contributing to the development of guidelines about recording DVA from external reports in patient facing EMRs that balances visibility to clinicians against the risk of unintended disclosures to patients. Strengths and limitations This study adds to the existing literature about how GPs respond to families affected by DVA and manage the EMR. It also complements existing research into the fac- tors influencing the success of information-sharing and multi-agency working in the response to DVA and child safeguarding. The study benefited from having the per- spectives of both GPs and police staff, and GPs with ex- pertise in safeguarding and DVA and GPs without a Pitt et al. BMC Family Practice (2020) 21:117 Page 8 of 10 specific role in this area. Participants were based in dif- ferent regions of England and Wales and described dif- ferent experiences of information-sharing. The study had several limitations. Qualitative research does not aim to be generalisable or representative, so the perspectives of the participants cannot be presented as representative of GPs or the police more generally. The case scenario involved a child aged six, which may have shaped respondents’ comments about speaking to chil- dren. The interview study did not include other profes- sional groups pertinent to the multi-agency response to DVA, such as social workers and health representatives on multi-agency safeguarding forums. Future studies should aim to explore the views of other professional agencies about DVA information-sharing. The police staff themselves had not shared reports directly with GPs (although some discussed sharing information with health representations on multi-agency forums). Some GPs had not received MARAC reports or police incident reports. While this means that their views were based on hypothetical scenarios, they were still able to share valu- able insight the issues might be around information-sharing with GPs, based on, for example, their previous experience working with other services and agencies. Research is also needed to explore the atti- tude of people affected by DVA to information-sharing between agencies. into what Conclusions GPs participants felt that notification about DVA from third parties helped them to treat the health impacts of DVA and safeguard vulnerable patients. However, in- corporating this information into the EMR and using it to inform patient care in the consultation required care- ful professional judgement. Both police staff and GPs de- scribed the importance of clarity about expectations and responsibility for action when information about DVA was shared between agencies. GPs should be supported by colleagues with expertise in DVA and safeguarding in responding to this information. Training and guidance for GPs about DVA should explicitly address the chal- lenge of recording and responding to information re- ceived from other agencies. Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s12875-020-01171-4. Additional file 1. Interview topic guide (GPs). Abbreviations DVA: domestic violence and abuse; EMR: electronic medical record; GPs: general practitioners; IRIS: Identification and Referral to Improve Safety; IRIS+: Enhanced Identification and Referral to Improve Safety; MARAC: multi- 2. agency risk assessment conference; MASH: multi-agency safeguarding hub; NICE: National Institute for Health and Social Care Excellence Acknowledgements Authors would like to thank the professionals who gave their time to take part in the interviews. Authors’ contributions GF, MM, JR, ES, EW and KP were instrumental to the conception and design of the study. SD and KP conducted the interviews. SD, EJ and KP coded the interview transcripts. SD, EJ and KP contributed to the thematic analysis. KP drafted the manuscript. All authors (KP, SD, GF, EJ, MM, JR, EW, ES) revised the drafts and contributed to the final version of the manuscript. The authors read and approved the final manuscript. Funding IRIS+ is part of the REPROVIDE programme (Reaching Everyone Programme of Research On Violence in diverse Domestic Environments, an independent research programme funded by the National Institute for Health Research (Programme Grants for Applied Research) (RP-PG-0614-20012). KP was a General Practice Academic Clinical Fellow funded by the National Institute for Health Research. The views expressed in this publication are those of the authors and not necessarily those of the National Institute for Health Research or the Department of Health and Social Care. The funding bodies were not involved in the design of the study, the collection, analysis or interpretation of data, or writing of the manuscript. Availability of data and materials Anonymised transcript data will be stored on the University of Bristol’s Research Data Service repository. Bona fide researchers will be able to access this data subject to a data access agreement and following approval from the University of Bristol Data Access Committee. Ethics approval and consent to participate The study was approved by the South West – Frenchay NHS Research Ethics committee (Reference number: 17/SW/0098) including a protocol to obtain written and verbal consent from study participants. Participants were sent a copy of the participant information sheet and consent form in advance, had the opportunity to discuss the study and answer questions prior to consenting to participate, and gave informed written or audio-recorded verbal consent to participate in the interview study. After discussion with EW, Faculty of Social Science and Law Research Ethics Officer, University of Bristol, it was agreed that audio-recorded in- formed verbal consent, based on participants having read the written con- sent form, was sufficient for telephone interviews. Consent for publication Written or verbal consent for publication of anonymised excerpts of transcripts was gained from all participants. Competing interests None. Author details 1Bristol Medical School (Population Health Sciences), University of Bristol, Canynge Hall, 39 Whatley Road, Bristol BS8 2PS, UK. 2School for Policy Studies, University of Bristol, Bristol, UK. Received: 18 August 2019 Accepted: 21 May 2020 References 1. The Crown Prosecution Service. Domestic abuse. https://www.cps.gov.uk/ domestic-abuse. Accessed 18 May 2020. Office for National Statistics. Domestic abuse: findings from the Crime Survey for England and Wales: year ending March 2018. https://www.ons. gov.uk/peoplepopulationandcommunity/crimeandjustice/articles/ domesticabusefindingsfromthecrimesurveyforenglandandwales/ yearendingmarch2018. Accessed 18 May 2020. Pitt et al. BMC Family Practice (2020) 21:117 Page 9 of 10 3. 5. 4. 6. Richardson J, Coid J, Petruckevitch A, Chung W, Moorey S, Feder G. Identifying domestic violence: cross sectional study in primary care. BMJ. 2002;324:74. Hester M, Ferrari G, Jones SK, et al. Occurrence and impact of negative behaviour, including domestic violence and abuse, in men attending UK primary care health clinics: a cross-sectional survey. BMJ Open. 2015;5: e007141. https://doi.org/10.1136/bmjopen-2014-007141. Chandan JS, Thomas T, Bradbury-Jones C, et al. Female survivors of intimate partner violence and risk of depression, anxiety and serious mental illness. Br J Psychiatry. 2019;1–6. https://doi.org/10.1192/bjp.2019.124. Coker AL, Smith PH, Bethea L, King MR, McKeown RE. Physical health consequences of physical and psychological intimate partner violence. Arch Fam Med. 2000;9(5):451–7. https://doi.org/10.1001/archfami.9.5.451. Huntley AL, Potter L, Williamson E, et al. Help-seeking by male victims of domestic violence and abuse (DVA): a systematic review and qualitative evidence synthesis. BMJ Open. 2019;9:e021960. https://doi.org/10.1136/ bmjopen-2018-021960. Sidebotham P, Brandon M, Bailey S, et al. Pathways to harm, pathways to protection: a triennial analysis of serious case reviews 2011 to 2014. Department of Education. https://www.gov.uk/government/publications/ analysis-of-serious-case-reviews-2011-to-2014. Accessed 18 May 2020. 9. McFarlane J, Symes L, Binder BK, Maddoux J, Paulson R. Maternal-child dyads of functioning: the intergenerational impact of violence against women on children. Matern Child Health J. 2014;18(9):2236–43. https://doi. org/10.1007/s10995-014-1473-4. 7. 8. 10. Gilbert R, Kemp A, Thoburn J, et al. Recognising and responding to child maltreatment. Lancet. 2009;373(9658):167–80. https://doi.org/10.1016/S0140- 6736(08)61707-9. 11. Oliver R, Alexander B, Roe S, Wlasny M. The economic and social costs of domestic abuse; 2019. https://www.gov.uk/government/publications/the- economic-and-social-costs-of-domestic-abuse. Accessed 18 May 2020. 12. National Institute for Health and Care Excellence. Domestic violence and abuse: multi-agency working. NICE; Public health guideline [PH50]. 2014. https://www.nice.org.uk/guidance/ph50. Accessed 18 May 2020. 13. Home Office. Domestic homicide reviews: key findings from analysis of domestic homicide reviews. 2016. https://www.gov.uk/government/ publications/domestic-homicide-review-lessons-learned. Accessed 18 May 2020. Sharp-Jeffs N, Kelly L. Domestic Homicide Review (DHR): Case Analysis. Report for Standing Together. Standing Together. London Metropolitan University; 2016. http://www.standingtogether.org.uk/news/domestic- homicide-review-case-analysis-report. Accessed 18 May 2020. 14. 15. UK Council of Caldicott Guardians. 'Striking the Balance': practical guidance on the application of Caldicott Guardian principles to domestic violence and MARACs (Multi Agency Risk Assessment Conferences). Department of Health. 2012. https://www.gov.uk/government/publications/striking-the- balance-practical-guidance-on-the-application-of-caldicott-guardian- principles-to-domestic-violence-and-maracs-multi-agency-risk-assessment- conferences. Accessed 18 May 2020. 16. HM Government. Information sharing: advice for practitioners providing 23. Morgan K, Williamson E, Hester M, Jones S, Feder G. Asking men about domestic violence and abuse in a family medicine context: help seeking and views on the general practitioner role. Aggress Violent Behav. 2014; 19(6):637–42. 24. Williamson E. Domestic violence and health. The response of the medical profession. Bristol: Policy Press; 2000. 25. Drinkwater J, Stanley N, Szilassy E, Larkins C, Hester M, Feder G. Juggling confidentiality and safety: a qualitative study of how general practice clinicians document domestic violence in families with children. Br J Gen Pract. 2017;67(659):e437–44. https://doi.org/10.3399/bjgp17X689353. Feder G, Potter L, Johnson M. Guidance on recording domestic violence and abuse information in general practice medical. records; 2017. https:// www.rcgp.org.uk/clinical-and-research/resources/toolkits/adult-safeguarding- toolkit/section-3-practice-resources.aspx. Accessed 18 May 2020. 26. 27. Woodman J, Sohal AH, Gilbert R, Feder G. Online access to medical records: finding ways to minimise harms. Br J Gen Pract. 2015;65(635):280–1. https:// doi.org/10.3399/bjgp15X685129. 28. Blake S. Electronic Personal Health Records. There are many benefits but 29. 30. what about the risks?; 2018. Available from: https://sarahblake.co.uk/2018/ 06/29/electronic-personal-health-records/. Accessed 18 May 2020. Feder G, Davies RA, Baird K, et al. Identification and Referral to Improve Safety (IRIS) of women experiencing domestic violence with a primary care training and support programme: a cluster randomised controlled trial. Lancet. 2011;378(9805):1788–95. https://doi.org/10.1016/S0140- 6736(11)61179-3. Stanley N, Humphreys C. Multi-agency risk assessment and management for children and families experiencing domestic violence. Child Youth Serv Rev. 2014;47:78–85. 31. Robinson AL. Reducing repeat victimization among high-risk victims of domestic violence: the benefits of a coordinated community response in Cardiff, Wales. Violence Against Women. 2006;12(8):761–88. https://doi.org/ 10.1177/1077801206291477. 32. Mackenzie M, Bradley L, Stanley N, et al. What might Normalisation Process Theory bring to policy implementation studies? Learning lessons and uncovering questions through a case study of the profound implementation failure of a new policing policy. Social Policy & Administration. 2018:1–15. https://doi.org/10.1111/spol.12467. 33. Royal College of General Practitioners. Domestic abuse. https://www.rcgp. org.uk/clinical-and-research/safeguarding/domestic-abuse.aspx. Accessed 18 May 2020. Identification and Referral to Improve Safety (IRIS) Interventions. https://irisi. org/. Accessed 18 May 2020. 34. 35. Rapley T. Interviews. In: Seale C GG, Gubrium J, Silverman D, editor. Qualitative 36. Research Practice. London: SAGE Publications Ltd; 2011. p. 16–34. King N, Horrocks C. Interviews in qualitative research. London: Sage Publications; 2012. 37. Corbin J, Strauss A. Basics of Qualitative Research. 4th Edition ed: Sage Publications; 2015. 38. Braun V, Clarke V. Using thematic analysis in psychology. Qual Res Psychol. 2006;3(2):77–101. Evans MA, Feder GS. Help-seeking amongst women survivors of domestic violence: a qualitative study of pathways towards formal and informal support. Health Expect. 2016;19(1):62–73. https://doi.org/10.1111/hex.12330. safeguarding services to children, young people, parents and carers. 2018. https://www.gov.uk/government/publications/safeguarding-practitioners- information-sharing-advice. Accessed 18 May 2020. 39. 17. Home Office. Multi agency working and information sharing project: final report. 2014. https://www.gov.uk/government/publications/multi-agency- working-and-information-sharing-project. Accessed 18 May 2020. 18. Home Office Violent and Youth Crime Prevention Unit (VCYU) and Research and Analysis Unit (RAU). Research into Multi-Agency Risk Assessment Conferences (MARACs). Cordis Bright Consulting; 2011. 19. BMA Board of Science. Domestic abuse. 2007 (updated 2014). https:// archive.bma.org.uk/collective-voice/committees/board-of-science/ publications. Accessed 18 May 2020. The Care Act 2014. Section 42. http://www.legislation.gov.uk/ukpga/2 014/23/section/42/enacted. Accessed 18 May 2020. 20. 21. National Institute for Health and Care Excellent. Domestic violence and 22. abuse. NICE 2016. Quality Standard [QS116]. https://www.nice.org.uk/ guidance/qs116. Accessed 18 May 2020. Feder GS, Hutson M, Ramsay J, Taket AR. Women exposed to intimate partner violence: expectations and experiences when they encounter health care professionals: a meta-analysis of qualitative studies. Arch Intern Med. 2006;166(1):22–37. https://doi.org/10.1001/archinte.166.1.22. 40. Mackenzie M, Gannon M, Stanley N, Cosgrove K, Feder G. 'You certainly don't go back to the doctor once you've been told, "I'll never understand women like you." Seeking candidacy and structural competency in the dynamics of domestic abuse disclosure. Sociology of Health and Illness. 2019:1–16. 41. Yeung H, Chowdhury N, Malpass A, Feder G. Responding to domestic violence in general practice: a qualitative study on perceptions and experiences. Int J Family Med. 2012;2012:960523. https://doi.org/10.1155/ 2012/960523. Taft A, Broom DH, Legge D. General practitioner management of intimate partner abuse and the whole family: qualitative study. BMJ. 2004;328(7440): 618. https://doi.org/10.1136/bmj.38014.627535.0B. 42. 43. Gutmanis I, Beynon C, Tutty L, Wathen CN, MacMillan HL. Factors influencing identification of and response to intimate partner violence: a survey of physicians and nurses. BMC Public Health. 2007;7:12. Published 2007 Jan 24. https://doi.org/10.1186/1471-2458-7-12. 44. Waalen J, Goodwin MM, Spitz AM, Petersen R, Saltzman LE. Screening for intimate partner violence by health care providers. Barriers and Pitt et al. BMC Family Practice (2020) 21:117 Page 10 of 10 interventions. Am J Prev Med. 2000;19(4):230–237. https://doi.org/10.1016/ s0749-3797(00)00229-4. 45. Gerbert B, Caspers N, Bronstone A, Moe J, Abercrombie P. A qualitative analysis of how physicians with expertise in domestic violence approach the identification of victims. Ann Intern Med. 1999;131(8):578–84. https://doi. org/10.7326/0003-4819-131-8-199910190-00005. 46. Williamson E, Jones SK, Ferrari G, Debbonaire T, Feder G, Hester M. Health professionals responding to men for safety (HERMES): feasibility of a general practice training intervention to improve the response to male patients who have experienced or perpetrated domestic violence and abuse. Prim Health Care Res Dev. 2015;16(3):281–8. https://doi.org/10.1017/ S1463423614000358. 47. Miller D, Jaye C. GPs' perception of their role in the identification and management of family violence. Fam Pract. 2007;24(2):95–101. https://doi. org/10.1093/fampra/cmm001. 49. 48. Peckover S, Golding B, Cooling P. Multi-Agency Working in Domestic Abuse and Safeguarding Children: Part of the Problem or Part of the Solution? University of Huddersfield; 2013. Stanley N, Miller P, Foster H, Thomson G. A stop-start response: social services' interventions with children and families notified following domestic violence incidents. Br J Soc Work. 2010;41(2):296–313. Szilassy E, Drinkwater J, Hester M, et al. Making the links between domestic violence and child safeguarding: an evidence-based pilot training for general practice. Health Soc Care Community. 2017;25(6):1722–32. https:// doi.org/10.1111/hsc.12401. 50. 51. HM Government. Working together to safeguard children: a guide to inter- agency working to safeguard and promote the welfare of children. 2018. https://www.gov.uk/government/publications/working-together-to- safeguard-children--2. Accessed 18 May 2020. 52. General Medical Council. Protecting children and young people: the 53. responsibilities of all doctors. 2012. Larkins C, Drinkwater J, Hester M, Stanley N, Szilassy E, Feder G. General practice clinicians' perspectives on involving and supporting children and adult perpetrators in families experiencing domestic violence and abuse. Fam Pract. 2015;32(6):701–5. https://doi.org/10.1093/fampra/cmv070. 54. Dossa A, Welch LC. GPs' approaches to documenting stigmatising information: a qualitative study. Br J Gen Pract. 2015;65(635):e372–e378. https://doi.org/10.3399/bjgp15X685273. 55. Woodman J, Allister J, Rafi I, et al. A simple approach to improve recording of concerns about child maltreatment in primary care records: developing a quality improvement intervention. Br J Gen Pract. 2012;62(600):e478–86. https://doi.org/10.3399/bjgp12X652346. 56. McAughtrie A. Supporting multi-agency communication in working with domestic abuse in families with children: Evidence from published research. NHS Education for Scotland. The National Evidence Search and Summary Service; 2016. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12913-021-07120-w	Saldarriaga et al. BMC Health Services Research (2021) 21:1092 https://doi.org/10.1186/s12913-021-07120-w R E S E A R C H Open Access The Annual costs of treating genital warts in the Public Healthcare Sector in Peru Enrique M Saldarriaga1,2, Cesar P. Cárcamo1, Joseph B. Babigumira2,3 and Patricia J. García1,3* Abstract Objectives: To estimate the cost of six different techniques used to treat Genital Warts and the annual average cost of treating a typical GW patient in Peru. To estimate the annual economic burden diagnosing and treating GW in the Peruvian public healthcare system. Methods: We developed a prevalence-based, cost-of-illness study from the provider’s perspective, the healthcare facilities under the purview of Peruvian Ministry of Health. We used an activity-based costing approach. We conducted primary data collection in three regions in Peru and supplemented it with governmental data. Uncertainty of the costing estimates was assessed via Monte Carlo simulations. We estimated the average cost and associated confidence intervals for six treatment options – three topical and three surgical – and the overall cost per patient. Results: The average treatment cost per patient was 59.9USD (95 %CI 45.5, 77.6). Given a population of 18.4 million adults between 18 and 60 years of age and a GW prevalence of 2.28 %, the annual cost of treating GW was 25.1 million USD (uncertainty interval 16.9, 36.6). Conclusions: This study provides the first quantification of the economic burden of treating genital warts in Peru and one of the few in Latin America. The costing data did not include other healthcare providers or out-of-pocket expenditures, and hence we present a conservative estimate of the COI of GW in Peru. Our findings bring attention to the financial burden of treating GW, a vaccine-preventable disease. Keywords: Genital warts, Cost-of-illness, Micro-costing, Peru Introduction Genital warts (GW) are the most common viral sexually transmitted infection (STI) globally [1]. They are mani- festations of anogenital human papillomavirus (HPV). In particular, HPV-subtypes 6 and 11 are causative agents of the disease [2–4]. GW present as external skin lesions of the vulva, penis, anus and scrotum, and mucosal le- sions of the vagina, cervix, and urethra [5]. * Correspondence: patricia.garcia@upch.pe 1Epidemiology, STD and HIV Unit, School of Public Health and Administration, Universidad Peruana Cayetano Heredia, Lima, Peru 3Department of Global Health, School of Public Health, University of Washington, Seattle, Washington, USA Full list of author information is available at the end of the article Data related to the incidence and prevalence of GW varied significantly within countries [1]. To date, the only study that has estimated the prevalence of GW in Peru was conducted by García et al [6]. The authors conducted a survey among 100 physicians from public facilities to quantify the frequency of GW cases detected, as well as diagnosis practices and patients’ characteris- tics. The prevalence of GW among all Peruvian adults between 18 and 60 years of age was estimated in 2.28 % (95 % Confidence Interval (CI) 2.02, 2.56). Among males the prevalence was 5.25 % (95 %CI 4.46, 6.13) and among females was 1.35 % (95 %CI 1.13, 1.61) [6]. While reports suggest that most GW cases are asymp- tomatic, location, size, and number of the lesions usually including pain, determines the presence of symptoms, © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Saldarriaga et al. BMC Health Services Research (2021) 21:1092 Page 2 of 10 pruritus, and bleeding [7, 8]. While GW can be self- limiting, several patients require topical treatment or surgical procedures. The choice of treatment usually de- pends on the clinical assessment of the lesion, cost of the procedure, patients’ characteristics, and physicians’ preferences [9]. Although the treatment of GWs has been associated to increased individual and healthcare costs [10–12], only few countries in Latin America have studies assessing the costs and the economic burden of GW in the population [13–15]; Peru is not one of them. Peru has a mixed healthcare system with public and private providers and insurers. While most of the private institutions are specialized in either service, public insti- tutions are upstream integrated and therefore offer in- surance [16]. The most important provider and insurer is the Ministry of Health (MoH) through the comprehensive health insurance (SIS by its acronym in Spanish – Seguro Integral de Salud) that covers 44.4 % of the population [17]. This structure has two implications. First, access to care its restricted by insurance membership; e.g., only holders of the SIS can be treated by the MoH. Second, most institutions are both a health services providers and payers. There- fore, the MoH, as the most important provider in the country, is expected to bear the biggest proportion of the costs associated to diagnose and treat GW in the country. and healthcare services The MoH provides vaccination free of charge to all in the national for all vaccines Peruvian citizens immunization scheme. Peru introduced the bivalent HPV vaccine, that confers protection against the high- risk subtypes, 16 and 18 [18], for the first time in 2011 for girls from 9 to 14 years old. Just in 2016 the national vaccination scheme changed to the quadrivalent vaccine that also protects against the HPV-types 6 and 11 [19]. the study was The objective of the population-level costs of GW diagnosis and treatment, to present the economic burden of a disease, that could be prevented with a gender-neutral vaccine. to address Methods Study design We conducted a cost-of-illness (COI) study aiming to estimate the total healthcare expenditures used to diag- nose and treat people with GW. We used an activity- based (micro-costing) technique to estimate the cost of diagnosis and each treatment option from the provider’s perspective, the facilities under the purview of the Peru- vian Ministry of Health (MoH). The micro-costing technique decompounds each ser- vice (i.e., diagnosis and treatment options) into the in- puts and quantity required to provide it. We then find the best price for each input and multiply it by the inputs provides an amount needed. The sum of all estimate of the cost per service. Since we used the pro- vider’s perspective for the costing analysis, only direct medical costs (e.g. drugs, materials, equipment, and phy- sicians’ and nurses’ wages) were included [20]. Since we used a prevalence-based approach, the COI is determined by the product of the prevalent cases and the average treatment cost [21, 22]. The prevalence was obtained from a previous study conducted by Garcia et al [6]. In the following sections we describe how each treatment’s technique cost was estimated, as well as how we arrived at the overall average treatment cost. Materials We leverage the results found by Garcia et al [6] regard- ing the prevalence of GW, providers’ preferences for GW diagnostic methods, and distribution of cases across gender and type of case. Cases were categorized by phy- sicians into “new” – no history of previous diagnosis, “resistant” – episode lasting longer than six months des- pite treatment, and “recurrent” – new case that appears within 12 months of previous episode. That study in- cluded physicians from six specialties: primary care phy- sicians and family medicine doctors), gynecologists, urologists, dermatolo- gists, and infectious disease specialists. (including general practitioners To identify resource use for a typical visit, we devel- oped a flow map of key activities completed during a visit (Fig. 1). Then we conducted a review of national guidelines [23] to identify the materials used in each ac- tivity according to protocol. Additionally, we updated and improve this information with eight in-depth inter- views with physicians that participated in Garcia’s study [6]. These interviews were used to get further informa- tion about treatment practices, preferences for specific treatments, materials and equipment used in each pro- cedure, duration of each procedure, and validation of the treatment algorithms. In 2016, we conducted primary costing-data collection in Lima (coastal city and capital of Peru), Ayacucho (An- dean region), and Iquitos (Jungle). The selection of sites was purposive. Each site represents a major region in Peru and therefore it allowed us to collect the most het- erogeneous costing data, resources utilization (i.e., quan- tity of the resources used), and clinical practices to create robust estimates. In addition, it is coherent with the study design of Garcia, et al., so it preserves internal consistency. We interviewed a total of nine administrative and lo- gistics officers that provided the unitary costs of all drugs, materials, medical supplies, and equipment used for each treatment option. From each interview we ob- tained purchasing data that contained, for each input, volume of purchase and price paid, or directly unitary cost. The unit cost of disposable inputs (e.g., cotton, Saldarriaga et al. BMC Health Services Research (2021) 21:1092 Page 3 of 10 Fig. 1 Clinical flow-map for genital warts treatment at a public healthcare facility in Peru needles) relies on the assumption that in every session the entire input is used (i.e., no partition for reuse). To estimate the unitary cost of durable inputs (e.g., equip- ment, medical instruments) we used the depreciation method [24]. From the interviews with administrative of- ficers, we obtained the total cost of the good, enquire about the rotation period (e.g., how often an equipment is changed, or infrastructure renovated), from which we obtained the useful life, which we finally used to esti- mate the depreciation cost per minute. Thus, the unitary costs were in the same unit as the duration of each activity. Regarding human resources, we used the opportunity cost of the paid-time of the healthcare workers(HCW) [25]. We obtained salary data from the National Registry of Healthcare Personnel (INFORHUS) from the MoH to improve the precision of the wage estimates. We used this information to estimate the cost per minute per type of HCW and the estimated time per activity reported by the interviewed physicians to calculate the attributable costs to each treatment. We obtained the costs for six key HCW: receptionist, file staff, cashier, nurse, phys- ician, pharmacist, whose regular activities are fairly dif- ferentiated and therefore we minimized the risk of overlapping. We used the information from the validate flow map (Fig. 1) to match activities with the HCW that most likely will perform them. From this information we estimated mean and stand- ard deviation (SD) of the unitary cost of each input. In all cases outliers were excluded if a value was a at least 3 times bigger than its peers. While a few inputs have a large variability (see supplemental material sheet “Cost- ing data”) we decided in favor of the mean, instead of quantile-based metrics such as the median, because it a more easier communicate, allows for to it straightforward implementation of the probability-based sensitivity analysis, and the impact of each individual in- put is too small to bias the results. The SD captures the variability of the unitary cost given regional differences, purchase preferences, and others. Analysis First, we estimated the cost per session for a diagnosis appointment and each treatment technique. This is the sum-product of the resources’ amount needed to provide a service and its unitary costs. To account for the vari- ability of the unitary costs we performed a Monte Carlo simulation [26]. We used a random number generator to obtain 1,000 estimates of each unitary cost based on a gamma distribution – the recommended distribution for cost data [27] – parameterized using the mean and standard deviation. Each draw from the distribution for all parameters is a simulation of the costing data, and therefore we obtained 1,000 simulations of the data. We estimated the final average costs 1,000 times and we were able to obtain the 2.5 % and 97.5 % percentiles of the distribution. Some parameters had no sample vari- ability when only one source of information was ob- tained. In those cases, we assumed a SD of 1e-9 to conduct the simulation. Following the same process, we decomposed the cost per session into categories of costs: human resources, infrastructure, equipment, drugs, med- ical instruments, disposable materials, and public ser- vices. We present both the cost per session and the cost per cost category. Second, in our study, the GW are compartmentalized by the combination of four groups given by the bio- logical gender of the patient (2 categories), the type of case (3 categories), the physician’s specialty (5 categor- ies), and treatment technique received (6 categories), Saldarriaga et al. BMC Health Services Research (2021) 21:1092 Page 4 of 10 resulting in 180 possible combinations. Each combin- ation is called a compartment. The probability across compartments is not homogeneous and hence we sought to find the specific distribution of cases for each one. We used the information reported by Garcia et al [6] to estimate the distribution of cases given the combination of patient’s gender, type of case, and physicians’ spe- cialty. The probability of no receiving treatment care once the warts have been detected vary across specialties but in all cases is negligible (see Garcia et al. [6]). The in-depth interviews provided us with the probability of choosing each treatment by physician specialty. We use these probabilities to estimate the final probability of each compartment. We report all the probabilities used for this analysis. Third, we estimated the annual cost of treatment per patient as the product of (a) the cost of each session plus the cost of the diagnosis appointment, (b) the number of times the treatment was applied to each patient, dependent upon patients’ characteristics, type of case (new, recurrent, or resistant), and physicians’ prefer- ences, and (c) the number of episodes within a year for recurrent cases. Thus, we obtained the annual cost per patient in each compartment. We report the average number of sessions per treatment, the overall number of episodes in recurrent cases by gender, and the final cost by treatment. Fourth, we obtained an estimate of the average cost of diagnosing and treating a typical patient in one year as the sum-product of the annual cost per patient in each com- partment and the associated compartment’s probability. Hence, this corresponds to a weighted average, where the weights are the probabilities of observing each combination of patient’ characteristics and physicians’ preferences. Given the distribution estimated for the cost per session (first point), we can estimate 95 % confidence interval for the an- nual cost of treating a typical case of GW. Fifth, the COI of diagnosis and treating GW, is the product of the average treatment cost and the preva- lence of GW. We calculate the point-estimate and range of feasible values of the COI. The point-estimate is the product of the mean values of prevalence and the aver- age treatment costs. The lower bound is the product of the 95 %CI’s lower-bound for both the prevalence of genital warts and the cost of treatment; conversely, the upper-bound uses the upper-bound of the 95 %CI for both metrics. Finally, the number of GW cases is based on the most recent estimation of population by age and gender, in 2017 the population of Peruvians between 18 and 60 years old was 18.4 million, 9.3 of them are males and 9.1 females [28]. All the costing data was collected in 2016 Peruvian Soles (PEN), but the results are expressed in 2019 US Dollars (USD) using a fixed conservative exchange rate of 3.3 PEN for each USD, and a yearly inflation of 2.5 %. Results The most used techniques to treat GW are three topical (podophyllin, imiquimod topical, trichloroacetic acid – TCA) and three surgical (cryotherapy, electro surgery, and surgical excision). Table 1 shows the estimated cost for each treatment in a single session and the distribu- tion among categories of resource input. Among treat- ment options, the most expensive was the surgical excision 24USD (95 %CI 12, 57), and the cheapest was TCA with a cost per session of 11USD (95 %CI 6, 18). For all treatment options, the category representing the largest share was human resources, whose cost per- session accounted for 70–95 %. The second most im- portant category is disposable materials varying from 1 to 21 % of the total cost per session. This included all goods that are used just once and then disposed of. Sup- plemental Material contains all costing data used to make these estimations. Across all treatments, on average, a new case received 2.5 (min: 1, max: 4) sessions of treatment, while a resist- ant case received 3.4 (min: 1, max: 6). For recurrent cases, the average number of episodes in a year is 1.7 for males and 1.6 for females. Using this information, we es- timated the annual average cost per treatment. The most expensive treatment was cryotherapy at 78USD per pa- tient, followed by podophyllin at 58USD, and electrosur- gery at 55USD. In contrast, surgical excision was the cheapest treatment at 36USD per patient. (Table 2) For detailed information on the distribution of probabilities across all compartments, please refer to Supplemental material. (40 %), After estimating the probabilities for all possible com- binations, podophyllin was the most frequently chosen followed by electrosurgery (20 %), treatment cryotherapy (17 %), TCA (16 %), imiquimod (5 %), and fi- nally surgical excision (1 %). (Table 2) Fig. 2 shows the distribution of treatment choice within physician spe- cialty and type of case. While there is a lot of variability across physicians’ specialty, we observe some patterns. For instance, there was a preference for podophyllin by general practitioners which represents the biggest pro- portion of cases, and dermatologists preferred cryother- apy in any circumstance. We estimated the average treatment cost in 59.9USD. From the Monte Carlo simulation we estimate the 95 % credibility interval in 45.5 to 77.6USD. The distribution of the simulations behaves as normal with skewness 0.3 and Kurtosis 3.1. Considering the distribution of cases across gender in the sample, we estimated the annual average cost of treating a male in 61.3USD and a female in 58.9USD. Saldarriaga et al. BMC Health Services Research (2021) 21:1092 Page 5 of 10 Table 1 Cost for each treatment in a single session, by categories of cost Diagnosis Appointment 6.04 (2.5, 11.8) Podophyllin Imiquimod TCA Cryotherapy Electro Surgery Surgical Excision 12.08 (7.4, 19.3) 10.94 (6.2, 18.3) 10.84 (6.1, 18.2) 15.4 (8.1, 26.8) 17.44 (9.4, 29.3) 24.59 (12.3, 57.6) Cost per session (USD) Category of cost (USD) Human Resources 6.03 (2.4, 11.8) 9.21 (4.5, 16.7) 9.21 (4.5, 16.7) 9.21 (4.5, 16.7) 13.39 (6.2, 24.7) 13.39 (6.2, 24.7) 13.86 (6.2, 26.4) Infrastructure 0.08 (0.08, 0.08) 0.09 (0.09, 0.09) 0.09 (0.09, 0.09) 0.09 (0.09, 0.09) 0.28 (0.27, 0.7 (0.69, 0.71) 0.44 (0.43, 0.44) 0.29) Equipment Drugs Medical instruments - - - - - - 0.52 (0.4, 0.7) 1.08 (0.01, 6.5) - 1.06 (1.06, 1.06) 1.44 (1.44, 1.44) 0.86 (0.86, 0.86) - - - - - - 0.96 (0.4, 1.86) 0.19 (0.18, 0.19) 0.69 (0.68, 0.7) Disposable Materials 0.05 (0, 0.25) 2.01 (1.91, 2.28) 0.46 (0.26, 0.79) 0.93 (0.73, 1.29) 1.47 (0.69, 2.3 (0.46, 6.82) 9.23 (2.04, 40.58) 4.82) Water and Electricity 0.07 (0.07, 0.07) 0.08 (0.08, 0.08) 0.08 (0.08, 0.08) 0.08 (0.08, 0.08) 0.22 (0.22, 0.32 (0.32, 0.32) 0.19 (0.19, 0.19) 0.22) USD: United States Dollars; TCA: trichloroacetic acid The “cost per session” row represent the cost in which the payer incurs every time a physician applies a given treatment. Usually, each treatment is applied more than once depending on the physician’s assessment, who considers type of case (new, recurrent, or resistant), and treatment characteristics Cells marked with “-“ indicates that the treatment did not employ those resources Given the size of the population of the prevalence, we estimate 333,709 annual cases of GW among males and is 85,811 among females. Thus, 25.1 million USD, with a feasible range of 16.9 to 36.5 million USD. Given that the estimated prevalence of GW in males is 3.9 higher than in females [6]. Con- sidering this, the estimated COI for females is 5.1USD million, while the estimated COI for males is 20 million. the estimated COI Discussion This is the first study to estimate the economic burden of treating GW in Peru and one the few in Latin Amer- ica. Our study leverages information from multiple sources to estimate the cost of several techniques to treat GW, and the associated annual COI for Peruvian healthcare sector. Our statistical approach explicitly in- corporates the key factors that determined which treat- the ments assessment of usage probabilities, as well as the uncer- tainty of the costing data. are used under what conditions via We found that human resources account for 70–95 % of the cost per treatment. While these values might look surprising, they are a consequence of the use of diagno- sis and treatment techniques that are more intensive in time from healthcare workers than in equipment or other resources. In the selection and implementation of the methods for this study, we accounted for the known issues related to attribution of shared costs [29], that can lead, among others, to overestimation of human re- sources. We followed several procedures to ensure an Table 2 Average number of sessions, final cost, and probability of usage for each treatment Average Sessions per treatment New (Min, Max) Resistant (Min, Max) Annual average cost per treatment (USD) Probability of treatment usage Podophyllin Imiquimod 3.2 (2.5, 3.7) 3.0 (2.5, 4.0) Trichloroacetic Acid 2.5 (1.6, 3.3) Cryotherapy 3.4 (2.8, 4.0) Electro Surgery 2.0 (1.1, 4.0) Surgical Excision 1.0 (1.0, 1.0) 4.0 (4.0, 4.0) 2.7 (2.4, 3) 5.3 (2.0, 8.0) 5.1 (3.0, 6.2) 2.1 (2.0, 2.3) 1.0 (1.0, 1.0) 58.6 46.5 52.7 78.2 55.5 36.1 39.6 % 5.3 % 16.4 % 17.4 % 20.3 % 0.9 % USD: United States Dollars “Average sessions per treatment” indicates the number of sessions until episode resolution. “Probability of treatment usage” column shows the proportions of cases in which each treatment was used. “Annual average cost per treatment” includes cost of diagnosis and control sessions, information of number of sessions per treatment, and proportion of recurrent cases treated with each technique This table presents summary data. Hence, the sum-product of the last two columns will differ from the reported annual average cost of treatment per person (59.9USD) Saldarriaga et al. BMC Health Services Research (2021) 21:1092 Page 6 of 10 Fig. 2 Distribution of treatment preferences by type of case, physicians’ specialty, and gender of the patient. Upper panel (A) corresponds to female patients, and lower panel (B), to female patients. Sections with no label represent 1 % accurate estimation of costs. First, the attribution of healthcare worker by activity was based on a validated flow map. Second, we selected healthcare workers whose activities would less likely overlap to prevent mismatch- ing. Third, the costs estimation was based on govern- mental data and cleaned from outliers. Our estimation of 25.1USD million for the COI of GW represents the total expenditures that the Peru- vian MoH would face in diagnoses and treatment of GW in a year. There are few caveats that are import- ant to highlight regarding this estimation. First, the prevalence used was based on formal care; it does not Saldarriaga et al. BMC Health Services Research (2021) 21:1092 Page 7 of 10 include self-treated cases and the cost in which these patients incur are not considered in our estimation of the COI. Second, per the structure of the Peruvian heath sys- tem, only holders of the SIS are eligible to be treated by the MoH [16]. An important proportion of people with no insurance would seek and receive treatment in the MoH because it is cheaper than other pro- viders. However, they would need to pay out-of- for all services received. Given the demand pocket size and associated negotiation power, the MoH prices are lower than other providers, making our COI estimate rather conservative of the total cost of treating GW in the public sector. On the other hand, stock-outs in MoH operated-facilities and other con- ditions have been associated with out-of-pocket ex- an penditures opposite effect over the COI as some of the MoH costs could be transferred to individuals. [30] which creates in SIS-holders Third, our analysis implicitly assumes that all pa- tients with detected GW would receive one of the six studied treatment options. According to Garcia et al [6], the proportion of cases that are left untreated are 0.16 % for females and 0.17 % for males. Considering the small number of cases that do not receive treat- ment once are detected, we did not impose any cor- rection on the prevalence to estimate the COI. On the other hand, in both, Garcia et al. study and our the six treatments we analyzed in-depth interviews, here were the most frequently used ones. Although other treatments were reported, such as interferon or fluorouracil, the proportion of usage was very small and out of the scope of our priorities. We do not have evidence that the inclusion of these treatments would importantly change our results, especially con- sidering the small variability across the cost per ses- sion of topical treatments. Despite these caveats, the estimated COI represent an important amount of money. According to the Peruvian Ministry of Economy and Finance [31], in 2019 the budget for individual health, which includes all actions that aim to treat and rehabilitate people, was around 5.2 billion USD (15.6 billion PEN; the total budget for the MoH was 20.9 billion PEN), equivalent to 145 USD per capita. Thus, the cost to diagnose and treat GW, as conservative as it is and without considering out-of- pocket expenditures, represents 0.16 % of the total insti- tutional budget. Further, the average per person cost of GW, represents over a third of the per capita value des- tined to individual health. Few studies in Latin America have estimated the cost of diagnosis and treatment of GW. For ease of compari- son across countries we converted our estimates into international dollars (intl.D) using the purchase parity pawer (PPP) factor published by the World Bank [32]. Considering a PPP factor of 1.74, we estimate the annual average cost of treatment per person in 104.2 intl.D (95 %CI 79.2, 135), and the COI in 43.7intl.D million (uncertainty interval 29.4, 63.5). An Ecuadorian study used a societal perspective using expert consultation to determine clinical practices. They found that the average cost of treatment varies from 205.4 to 251.7intl.D (PPP factor 0.52), depending on the treatment used [14]. Our results cannot be compared to these results because the authors used a societal per- spective, which includes a broader range of costs than to our analysis and includes private practices. Our study was based on the payer’s perspective and only from pub- lic facilities which tend to be cheaper than private options. In Mexico, the average cost for diagnosis and treat- ment of GW was found to be 1,326.8 1,418.5intl.D (PPP factor of 9.15) for men and women, respectively [15]. This study used the healthcare system perspective, and the information was obtained through specialists’ inter- views. Our estimates are lower than these results, due to the cost of a diagnosis. According to Garcia et al [6]. the diagnosis in Peruvian facilities is made using visual in- spection in more than 95 % of the cases; a very cheap technique for which the diagnosis visit is 10.4intl.D. In the Mexican study, consulted physicians reported using laboratory test to diagnose GW. Hence, the average diagnosis appointment alone was 803.4intl.D. GW is a vaccine-preventable disease and therefore the costs associated to its diagnosis and treatment can be mitigated [33, 34]. In the United States, quadrivalent vaccine was introduced for females in 2006, and subse- quently to males in 2009 [35, 36]. Although the vaccine achieves its highest efficacy in HPV non-exposed indi- viduals (i.e. before sexual initiation, usually younger than 13 years old), the vaccine has been recommended for everybody up to 26 years of age [37]. As a result, there has been a reduction of 0.8 % of genital warts between 2007 and 2014 [38, 39]. Australia is probably an even better example. The country was the first to introduce the quadrivalent HPV vaccine implemented through a national program, targeting girls aged 12 and 13 years, with an additional two catch-up campaigns from 2007 to 2009 targeting women up to 26 years old [40]. In 4 years, the proportion of women under 21 years old diag- nosed with GW was reduced from 11.5 % to 2007 to 0.85 % in 2011. Similar declines were observed in men under 21 years of age and men and women of 21 to 30 years old [41, 42]. Under the current immunization scheme, only girls are eligible for the quadrivalent HPV vaccine [19]. there is evidence that vaccinating boys However, level, by could also be beneficial at the individual Saldarriaga et al. BMC Health Services Research (2021) 21:1092 Page 8 of 10 reducing incidence of associated cancers and GW, and at the community level, by reducing the spread of the HPV [43, 44]. Considering the high concentra- tion of GW cases in males [6] and the estimated higher cost of providing them care, a gender-neutral HPV vaccination approach could potentially save mil- lions of dollars by preventing GW in Peru and im- prove health outcomes for both males and females [45]. Further, according to the pricelist of the Pan American Health Association (PAHO) for the two-dose quadrivalent HPV-vaccine is 19.9 USD, almost a third of the average cost of treating a GW. Although a study would need to determine how many GW infections the quadrivalent vaccine can prevent in Peru under specific uptake scenarios and target population, two studies in Latin America have found that it can significantly reduce the incidence of GW in up to 80 % [46, 47]. the cost This study was not without limitations. First, we did not collect costs for other healthcare providers besides the MoH and therefore the estimated COI is a conserva- tive number compared to the true cost of treating GW bore by the Peruvian healthcare system. Second, this study uses the prevalence values and physicians’ re- ported preferences for treatment presented by Garcia et al [6] and hence subject to the same limitations and potential sources of bias as presented in that study. Third, the probabilities used to estimate the distribution of cases across combinations of patients’ and physicians’ characteristics were obtained through interviews and hence subject to recollection and social-desirability biases. Nonetheless, the application of our instruments followed best-practices for data collection [48] and we are confident on the accuracy of our results within its limitations. Conclusions To our knowledge, this is the first study that analyzes the economic burden of GW in Peru, and one of the few in Latin America. We used a micro-costing technique with data collection in multiple settings to account for the regional price variability. Our fieldwork collection was supplemented with governmental data, which im- prove the precision of our estimates. Finally, we expli- citly assess uncertainty in our estimates by including confidence intervals and performing a Monte Carlo simulation to estimate credible intervals for the average cost of treatment. We hope our findings bring attention to the economic consequences of diagnosing and treating GW, and the burden that it represents to the MoH and the Peruvian Healthcare sector at large and becomes another reason to make the decision of moving forward into gender- neutral vaccination. Supplementary information The online version contains supplementary material available at https://doi. org/10.1186/s12913-021-07120-w. Additional file 1 Acknowledgements Jonathan Augurto and Sandra Solis, University Cayetano Heredia, provided excellent assistant with the data collection and cleaning. The authors would like to thank the journal reviewers for their excellent comments. Authors’ contributions Concept and Design: ES, PG, CC; Acquisition of Data: ES; Analysis and interpretation of data: ES, PG, CC, JB; Drafting of the manuscript: ES; Critical revision of the paper: ES, PG, CC, JB; Statistical analysis: ES; Obtaining funding: PG, CC. All authors have read and approved the manuscript. Funding Financial support for this study was provided Merck & Co., Inc., Kenilworth, NJ, USA. Availability of data and materials The authors have made available all data and materials in the Supplemental file. Declarations Ethics approval and consent to participate The protocol and instruments were reviewed and approved by the IRB of the Universidad Peruana Cayetano Heredia (# 65616). All protocols are carried out in accordance with relevant guidelines and regulations. Participants were aware of the study and understood the benefits and risks involved. Informed consent was obtained from all participants. Consent for publication Not applicable. Competing interests Merck & Co., Inc., produces Gardasil 4-valent HPV vaccine however, the funder had no role in the design and conduct of the study; collection, man- agement, analysis, and interpretation of the data; preparation, review, or ap- proval of the manuscript; and decision to submit the manuscript for publication. The authors have no competing interest to declare. Author details 1Epidemiology, STD and HIV Unit, School of Public Health and Administration, Universidad Peruana Cayetano Heredia, Lima, Peru. 2The Comparative Health Outcomes, Policy, and Economics (CHOICE) Institute, University of Washington, Seattle, Washington, USA. 3Department of Global Health, School of Public Health, University of Washington, Seattle, Washington, USA. Received: 1 June 2021 Accepted: 1 October 2021 References 1. 2. 4. Patel H, Wagner M, Singhal P, Kothari S. Systematic review of the incidence and prevalence of genital warts. BMC Infect Dis. 2013 Jan 25;13(1):39. Garland SM, Steben M, Sings HL, James M, Lu S, Railkar R, et al. Natural history of genital warts: analysis of the placebo arm of 2 randomized phase III trials of a quadrivalent human papillomavirus (types 6, 11, 16, and 18) vaccine. J Infect Dis. 2009 Mar 15;199(6):805–14. 3. McQuillan G, Kruszon-Moran D, Markowitz LE, Unger ER, Paulose-Ram R. Prevalence of HPV in Adults Aged 18–69: United States, 2011–2014. NCHS Data Brief. 2017 Apr;(280):1–8. Satterwhite CL, Torrone E, Meites E, Dunne EF, Mahajan R, Ocfemia MCB, et al. Sexually transmitted infections among US women and men: prevalence and incidence estimates, 2008. Sex Transm Dis. 2013 Mar;40(3): 187–93. Saldarriaga et al. BMC Health Services Research (2021) 21:1092 Page 9 of 10 5. 6. 7. 8. Lynde C, Vender R, Bourcier M, Bhatia N. Clinical features of external genital warts. J Cutan Med Surg. 2013 Dec;17 Suppl 2:S55-60. García PJ, Carcamo CP, Valderrama M, La Rosa S, James C, Gutiérrez R, et al. Burden of genital warts in Peru: an observational study. Int J STD AIDS. 2019 Mar;30(3):264–74. Kilciler M, Bedir S, Erdemir F, Coban H, Erten K, Ors O, et al. Condylomata acuminata of external urethral meatus causing infravesical obstruction. Int Urol Nephrol. 2007;39(1):107–9. Chae JY, Bae JH, Yoon CY, Park HS, Moon DG, Lee JG, et al. Female Urethral Condyloma Causing Bladder Outlet Obstruction. Int Neurourol J. 2014 Mar; 18(1):42–4. 9. Workowski KA, Bolan G. Sexually Transmitted Diseases Treatment Guidelines, 2015 [Internet]. Centers for Disease Control and Prevention - CDC; 2015. Report No.: 64(RR2); 1–137. Available from: https://www.cdc.gov/mmwr/ preview/mmwrhtml/rr6403a1.htm 10. Woodhall SC, Jit M, Soldan K, Kinghorn G, Gilson R, Nathan M, et al. The impact of genital warts: loss of quality of life and cost of treatment in eight sexual health clinics in the UK. Sex Transm Infect. 2011 Oct;87(6):458–63. 11. Park YJ, Kim JM, Lee BR, Kim TH, Lee EG. Annual prevalence and economic burden of genital warts in Korea: Health Insurance Review and Assessment (HIRA) service data from 2007 to 2015. Epidemiol Infect. 2018;146(2):177–86. 12. Coles V a. H, Chapman R, Lanitis T, Carroll SM. The costs of managing genital warts in the UK by devolved nation: England, Scotland, Wales and Northern Ireland. Int J STD AIDS. 2016 Jan;27(1):51–7. 13. Domenech-Viñolas M, León-Maldonado L, Ramírez-Palacios P, Flores YN, Granados-García V, Brown B, et al. Incidence, psychosocial burden, and economic impact of genital warts in Mexico. Salud Publica Mex. 2018 Nov; 60(6):624–32. 14. Roldos M, Bustamante V. Consensus of Clinical Practices And Associated Costs To Diagnose And Treat Genital Warts Caused By Human Papilloma Virus (Hpv) In Ecuador: Results From A Panel Of Experts. Value Health. 2015 Nov 1;18:A833. 15. Reyes-Lopez A, Pérez Bolde-Villarreal C, Pastor-Martinez V. Uso de recursos y costos asociados al diagnóstico y tratamiento de las verrugas genitales en instituciones públicas de salud en México. Rev Mex Urol. 2015 Mar 1;75:72–81. 16. Alcalde-Rabanal JE, Lazo-González O, Nigenda G. [The health system of Peru]. Salud Publica Mex. 2011;53 Suppl 2:s243-254. 17. Durand-Carrion D. [Population with any kind of health insurance] [Internet]. Lima: Instituto Nacional de Estadística e Informática - INEI; 2018 Nov. Available from: https://www.inei.gob.pe/media/MenuRecursivo/publica ciones_digitales/Est/Lib1587/libro01.pdf 18. Murillo R, Almonte M, Pereira A, Ferrer E, Gamboa OA, Jerónimo J, et al. Cervical cancer screening programs in Latin America and the Caribbean. Vaccine. 2008 Aug 19;26 Suppl 11:L37-48. 19. MINSA. Resolución Ministerial 651-2016-MINSA. Ministerio de Salud - MINSA; 2016 Agosto. 20. Garrison LP, Pauly MV, Willke RJ, Neumann PJ. An Overview of Value, Perspective, and Decision Context—A Health Economics Approach: An ISPOR Special Task Force Report [2]. Value Health. 2018 Feb;21(2):124–30. 21. Akobundu E, Ju J, Blatt L, Mullins CD. Cost-of-Illness Studies. 22. PharmacoEconomics. 2006 Sep 1;24(9):869–90. Jo C. Cost-of-illness studies: concepts, scopes, and methods. Clin Mol Hepatol. 2014 Dec;20(4):327–37. 23. DGSP, DAIS. [National Guidelines for integral care of Sexual and 24. reproductive health] [Internet]. Lima, Peru: Ministry of Health; 2004. Available from: http://bvs.minsa.gob.pe/local/dgsp/63_guiasnac.pdf Lau F. Chapter 14 Methods for eHealth Economic Evaluation Studies [Internet]. Handbook of eHealth Evaluation: An Evidence-based Approach [Internet]. University of Victoria; 2017 [cited 2021 Jun 28]. Available from: https://www.ncbi.nlm.nih.gov/books/NBK481607/ 25. Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes [Internet]. Oxford, UNITED KINGDOM: Oxford University Press; 2015 [cited 2021 May 31]. Available from: http://ebookcentral.proquest.com/lib/washington/deta il.action?docID=4605509 26. Briggs A. Probabilistic Analysis of Cost-Effectiveness Models: Statistical Representation of Parameter Uncertainty. Value Health. 2005;8(1):1–2. 27. Briggs AH, Weinstein MC, Fenwick EAL, Karnon J, Sculpher MJ, Paltiel AD, 28. INEI. [Total population at June 30 of every year, by sex and age groups] [Internet]. Instituto Nacional de estadistica e Informatica - INEI; ND. Available from: https://www.inei.gob.pe/estadisticas/indice-tematico/poblacion-y- vivienda/ 29. Rosen AB, Aizcorbe A, Highfill T, Chernew ME, Liebman E, Ghosh K, et al. Attribution of Health Care Costs to Diseases: Does the Method Matter? In: Measuring and Modeling Health Care Costs [Internet]. University of Chicago Press; 2016 [cited 2021 Aug 2]. p. 173–210. Available from: https://www. nber.org/books-and-chapters/measuring-and-modeling-health-care-costs/a ttribution-health-care-costs-diseases-does-method-matter 30. Pavone MP, Sánchez EJ. Determinants of out-of-pocket spending on health among the poor population served by public health services in Peru, 2010– 2014. Rev Panam Salud Pública. 2018 Oct 18;42:e20. 31. Peruvian Ministry of Economic and Finance. [Friendly Search (Monthly). Public Expenditure Execution Search] [Internet]. Economic Transparency. [cited 2020 Jan 15]. Available from: http://apps5.mineco.gob.pe/transpa rencia/mensual/ The World Bank. PPP conversion factor [Internet]. [cited 2021 Apr 5]. Available from: https://data.worldbank.org/indicator/PA.NUS.PPP 32. 33. Garland SM, Hernandez-Avila M, Wheeler CM, Perez G, Harper DM, Leodolter S, et al. Quadrivalent vaccine against human papillomavirus to prevent anogenital diseases. N Engl J Med. 2007 May 10;356(19):1928–43. FUTURE II Study Group. Quadrivalent vaccine against human papillomavirus to prevent high-grade cervical lesions. N Engl J Med. 2007 May 10;356(19): 1915–27. 34. 35. Centers for Disease Control and Prevention (CDC). FDA licensure of quadrivalent human papillomavirus vaccine (HPV4, Gardasil) for use in males and guidance from the Advisory Committee on Immunization Practices (ACIP). MMWR Morb Mortal Wkly Rep. 2010 May 28;59(20):630–2. 36. Markowitz LE, Dunne EF, Saraiya M, Lawson HW, Chesson H, Unger ER, et al. Quadrivalent Human Papillomavirus Vaccine: Recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep Morb Mortal Wkly Rep Recomm Rep. 2007 Mar 23;56(RR-2):1–24. 37. Meites E. Human Papillomavirus Vaccination for Adults: Updated Recommendations of the Advisory Committee on Immunization Practices. MMWR Morb Mortal Wkly Rep [Internet]. 2019 [cited 2019 Oct 25];68. Available from: https://www.cdc.gov/mmwr/volumes/68/wr/ mm6832a3.htm 38. Oliver SE, Unger ER, Lewis R, McDaniel D, Gargano JW, Steinau M, et al. Prevalence of Human Papillomavirus Among Females After Vaccine Introduction—National Health and Nutrition Examination Survey, United States, 2003–2014. J Infect Dis. 2017 Sep 1;216(5):594–603. 39. Markowitz LE, Liu G, Hariri S, Steinau M, Dunne EF, Unger ER. Prevalence of HPV After Introduction of the Vaccination Program in the United States. Pediatrics. 2016 Mar 1;137(3):e20151968. 40. Garland SM, Skinner SR, Brotherton JML. Adolescent and young adult HPV vaccination in Australia: Achievements and challenges. Prev Med. 2011 Oct 1;53:S29–35. 41. Ali H, Donovan B, Wand H, Read TRH, Regan DG, Grulich AE, et al. Genital warts in young Australians five years into national human papillomavirus vaccination programme: national surveillance data. BMJ. 2013 Apr 18;346:f2032. 42. Patel C, Brotherton JM, Pillsbury A, Jayasinghe S, Donovan B, Macartney K, et al. The impact of 10 years of human papillomavirus (HPV) vaccination in Australia: what additional disease burden will a nonavalent vaccine prevent? Eurosurveillance [Internet]. 2018 Oct 11 [cited 2019 Oct 25];23(41). Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6194907/ 43. Bogaards JA, Wallinga J, Brakenhoff RH, Meijer CJLM, Berkhof J. Direct benefit of vaccinating boys along with girls against oncogenic human papillomavirus: bayesian evidence synthesis. BMJ. 2015 May 12;350:h2016. Elfström KM, Lazzarato F, Franceschi S, Dillner J, Baussano I. Human Papillomavirus Vaccination of Boys and Extended Catch-up Vaccination: Effects on the Resilience of Programs. J Infect Dis. 2016 Jan 15;213(2): 199–205. 44. 45. Hintze JM, O’Neill JP. Strengthening the case for gender-neutral and the nonavalent HPV vaccine. Eur Arch Oto-Rhino-Laryngol Off J Eur Fed Oto- Rhino-Laryngol Soc EUFOS Affil Ger Soc Oto-Rhino-Laryngol - Head Neck Surg. 2018 Apr;275(4):857–65. et al. Model parameter estimation and uncertainty analysis: a report of the ISPOR-SMDM Modeling Good Research Practices Task Force Working Group- 6. Med Decis Mak Int J Soc Med Decis Mak. 2012 Oct;32(5):722–32. 46. Ortiz AP, Ortiz-Ortiz KJ, Ríos M, Laborde J, Kulkarni A, Pillsbury M, et al. Modelling the effects of quadrivalent Human Papillomavirus (HPV) vaccination in Puerto Rico. PLoS ONE. 2017 Nov 30;12(11):e0184540. Saldarriaga et al. BMC Health Services Research (2021) 21:1092 Page 10 of 10 47. Bardach AE, Garay OU, Calderón M, Pichón-Riviére A, Augustovski F, Martí SG, et al. Health economic evaluation of Human Papillomavirus vaccines in women from Venezuela by a lifetime Markov cohort model. BMC Public Health. 2017 Feb 2;17:152. Frick KD. Micro-Costing Quantity Data Collection Methods. Med Care. 2009 Jul;47(7 Suppl 1):S76–81. 48. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12870-023-04198-8	Ramazan et al. BMC Plant Biology (2023) 23:183 https://doi.org/10.1186/s12870-023-04198-8 BMC Plant Biology Comparative protein analysis of two maize genotypes with contrasting tolerance to low temperature Salika Ramazan1, Nelofer Jan1 and Riffat John1* Abstract Background Low temperature (LT) stress is one of the major environmental stress factors affecting the growth and yield of maize (Zea mays L.). Hence, it is important to unravel the molecular mechanisms behind LT stress tolerance to improve molecular breeding in LT tolerant genotypes. In the present study, two maize genotypes viz. Gurez local from Kashmir Himalaya and tropical grown GM6, were dissected for their LT stress response in terms of accumulation of differentially regulated proteins (DRPs). Leaf proteome analysis at three-leaf stage of maize seedlings subjected to LT stress of 6 °C for a total of 12 h duration was performed using two dimensional gel electrophoresis (2D-PAGE) followed by subsequent identification of the proteins involved. Results After MALDI-TOF (Matrix-assisted laser desorption/ionization-time of flight) and bioinformatics analysis, 19 proteins were successfully identified in Gurez local, while as 10 proteins were found to get successful identification in GM6. The interesting observations from the present investigation is the identification of three novel proteins viz. threonine dehydratase biosynthetic chloroplastic, thylakoidal processing peptidase 1 chloroplastic, and nodulin-like protein, whose role in abiotic stress tolerance, in general, and LT stress, in particular, has not been reported so far. It is important to highlight here that most of LT responsive proteins including the three novel proteins were identified from Gurez local only, owing to its exceptional LT tolerance. From the protein profiles, obtained in both genotypes immediately after LT stress perception, it was inferred that stress responsive protein accumulation and their expression fashion help the Gurez local in seedling establishment and withstand unfavorable conditions as compared to GM6. This was inferred from the findings of pathway enrichment analysis like regulation of seed growth, timing of floral transition, lipid glycosylation, and aspartate family amino acid catabolic processes, besides other key stress defense mechanisms. However, in GM6, metabolic pathways enriched were found to be involved in more general processes including cell cycle DNA replication and regulation of phenylpropanoid metabolism. Furthermore, majority of the qRT-PCR results of the selected proteins demonstrated positive correlation between protein levels and transcript abundance, thereby strengthening our findings. Conclusions In conclusion, our findings reported majority of the identified proteins in Gurez local exhibiting up-regulated pattern under LT stress as compared to GM6. Furthermore, three novel proteins induced by LT stress *Correspondence: Riffat John riffatminhaj@kashmiruniversity.ac.in Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessPage 2 of 20 were found in Gurez local, requiring further functional validation. Therefore, our results offer more insights for elucidating the molecular networks mediating LT stress tolerance in maize. Keywords 2D-PAGE, Maize, Low temperature, Protein abundance, Stress Background Low temperature (LT) tolerance is a principal agro- nomic attribute of temperate grown crops. As changing climatic conditions threaten the global food security [1], therefore, screening out LT tolerant crops is critical for adapting climate smart agricultural practices [2]. Maize (Zea mays L.), a critically important food resource, and tropical in origin, is inherently susceptible to low temper- atures. At any crucial stage in its life cycle, a suboptimal temperature for longer duration can cause remarkable decline in the overall growth and crop yield [3]. Ironi- cally, the present agricultural output trend of leading food crops including maize are not adequate to meet the future requirements [4]. Hence, LT tolerance of maize must be improved by developing and selecting LT toler- ant genotypes employing various traditional breeding protocols and high-throughput genomic approaches [2]. For many years now, the biological mechanisms behind LT-susceptibility of maize have been analyzed [5]. From the very first physio-morphological studies, including our previous investigations, the research revealed the impact of LT stress mostly on photosynthetic apparatus, trans- port processes, water relations, and plant stature [6, 7]. Additionally, the metabolomic studies have also shown that both primary and specialized metabolites accu- mulate within the crop in response to LT stress, which include diverse compatible solutes, antioxidant com- pounds, and biomass precursors [8, 9]. The availability of the complete maize genome sequence has made it possible to explore the crop’s hidden genetic potential using a variety of molecular tools and techniques [3]. It is evident from various transcriptomic studies based on the response of maize to LT stress [10, 11]. These inves- tigations unveiled several genes coding for various tran- scription factors, DNA/RNA binding proteins, chromatin condensation, chloroplast functioning, circadian rhythm, and phytohormone signaling among other attributes in maize [10]. In addition to above, proteomics is a potent tool fur- nishing the summary of different cellular and molecular changes occurring in plants under stress conditions [12]. With respect to maize, impressive progress in the field of sequencing technologies have led to a large and fast increasing proteomic datasets (i.e., amino acid sequences and protein interaction networks). A powerful insight of these protein sequences finds many applications, such as environmental stress resistance, yield and grain quality improvement, and so on [13]. One of the versatile pro- teomic techniques is two-dimensional polyacrylamide gel electrophoresis (2D-PAGE) which has been widely used in differential proteomic analyses of wheat [14] and rice [15] under stress induced conditions. In case of maize, 2D-PAGE has been employed: to establish the proteome map of endosperm [16]; to analyze biotic and abiotic stress induced protein changes [17, 18]; and to study the proteomic differences in transgenic and non-transgenic maize [19]. In the above context, the present study was carried out in two maize genotypes viz. LT tolerant Gurez local and LT susceptible Gujarat-Maize-6 (G-M-6), differing in their geographical origins (temperate and tropical, respectively). The genotypes were selected to elucidate the comparative protein framework underlying the LT stress tolerance in maize seedling leaf cells, that too in very first hours of stress perception. Owing to our previ- ous experiments on same genotypes [7, 9], interestingly, the temperate grown ‘Gurez local’ was found to exhibit exceptional LT tolerance providing a promising resource for identification of novel LT induced proteins. The pro- teomic (2D-PAGE and MALDI-TOF-MS) investigations revealed the altered expression of some remarkable LT responsive proteins besides identifying the novel four proteins from ‘Gurez local’. Tough subjected to further characterization and functional validation, our findings provide a comprehensive understanding of LT tolerance in maize and reveal an important role of newly identified LT regulatory proteins. Results Inventory of maize seedling leaf proteins identified by 2D-PAGE Gel analysis revealed that in Gurez local, 119 spots were obtained through database comparison and analysis, among which 80 spots were found to show up-regulated and 34 down-regulated expression. Among all these spots, only 20 spots of interest were selected based on their fold change value for MALDI-TOF analysis. Simi- larly, in case of GM6, total of 42 spots were found to exhibit significant changes in expression levels through- out the stress time-points, however, only 14 spots of interest based on their fold change value (p < 0.05) were selected and subjected to further analysis Fig. 2. In the end, 19 spots showed positive identification in case of Gurez local while 10 spots were identified in GM6 based on the MALDI-TOF analysis and subsequent MASCOT search as shown in Venn Diagram Fig. 3. With respect to their expression level, in Gurez local, 17 spots (pI 5–10/ Ramazan et al. BMC Plant Biology (2023) 23:183 Page 3 of 20 Fig. 1 Maize seedlings (A-E) Gurez local (F-J) GM6 with their corresponding protein profiles at various LT (6 °C) stress time-points. Labels A-E, and F-J indicate the stress time points corresponding to 0 h, 2 h, 6 h, 8 and 12 h for Gurez local and GM6, respectively. The 2-DE gel is the ‘Master gel’ from the three gel replicates Fig. 2 2-DE representative gel from leaves of A) Gurez local B) GM6. The numbers in the Gel images represent the protein spots with their ID’s, gener- ated in the Image Master Gel Analysis Software. The red ellipses indicate up-regulated proteins, while the black ones indicate down-regulated proteins MW 3000-280000 Da) exhibited up regulated expres- sion while only two were found to show down regu- lated expression (Spot ID 84; GINS complex protein under NCBI Accession no. AQK40686.1, and spot ID 26; F10K1.23 under NCBI Acc No. AQK81609.1). How- ever, in GM6, 5 proteins exhibited up and 5 showed down regulated expression (pI 6–10/MW 3000–35,000 Da) (Table 1). Meanwhile, identified proteins in both genotypes under LT stress conditions revealed differ- ences in the expression patterns as compared to control grown plants, depicted in Heatmap clustering Fig. 4. Moreover, PCA Biplot analysis also revealed that in LT tolerant Gurez local majority of the proteins contribute towards LT stress tolerance at different time intervals of Ramazan et al. BMC Plant Biology (2023) 23:183 Page 4 of 20 Fig. 3 Venn diagram showing total number of identified proteins in Gurez local and GM6 through 2D-PAGE. The number in the middle of larger circles indicates total number of proteins for each genotype. The overlapping region depicts the proteins common to both genotypes induced under LT conditions. The number within smaller circles on the topside of figure represent up-regulated proteins, while the same on the downside indicate down- regulated proteins the treatment as compared to control grown maize seed- lings as well as in comparison to GM6 Fig. 5. It can also be assumed from PCA Biplot analysis that all these pro- teins get expressed on exposure to LT stress only in case of LT tolerant Gurez local as compared to GM6. Three novel proteins were isolated from the protein inventory in the Gurez local (Table 1). These proteins include Thre- onine dehydratase biosynthetic chloroplastic, Thylakoidal processing peptidase 1 chloroplastic, and Nodulin-like protein, whose further study may offer more information about their predicted roles in LT tolerance in maize. Generation of 3D models and their stereochemical validation The three-dimensional models (3D) necessary for visual- ization and better understanding of secondary structural features of all identified proteins were successfully gen- erated with > 80% residues modelled at ≥ 90% confidence (Fig. S1). In addition, the PDB files obtained from Phyre2 (http://www.sbg.bio.ic.ac.uk/phyre2/html/page. server cgi?id=index) uploaded at PDBsum (https://www.ebi. ac.uk/thornton-srv/databases/cgi-bin/pdbsum/GetPage. pl?pdbcode=index.html) gave us the stereochemical vali- dation of all protein models in the form of Ramachandran Plots (RC plots) (Fig. S2). The RC plot statistics (Table S1) showing the distribution of main Ф-ψ angles in relation to ‘core’ (red) and ‘allowed’ (brown) regions, with resi- dues falling in the ‘generously allowed’ (dark yellow) and ‘disallowed’ (pale yellow) regions plotted as red squares were successfully obtained for all the proteins. Classification of LT responsive proteins in both genotypes Online servers viz. CLAP and Blast2GO (B2G) were used to classify the identified proteins respectively based on an alignment-free local sequence similarity computing approach and the functional annotations Gene ontol- ogy functional analysis revealed that although most of the GO terms are shared between the two genotypes, but the differential regulation patterns (DRPs) vary con- siderably in Gurez local from that of GM6 with respect to GO distribution level in terms of metabolic processes (GO: 0008152), biological regulation (GO: 0065007), and localization (GO: 0051179). However, 3 terms were found to be unique to Gurez local including the response to stimulus (GO: 0050896), signaling (GO: 0023052), and detoxification (GO: 0098754) under ‘biological pro- cess’ category. With regards to molecular functions, ATP dependent activity (GO: 0140657) and antioxidant activ- ity (GO: 0016209) were specifically enriched in Gurez local only while rest other terms were shared between the two genotypes (Fig. 6). Furthermore, in CLAP (Classification of Proteins), results in the form of Hieracrchiel clustering for both genotypes the protein sequences labelled as ‘NCBI Accession numbers’ were efficiently clustered into groups exhibiting high domain architectural similarities (Fig. 7). Ramazan et al. BMC Plant Biology (2023) 23:183 Table 1 Differentially expressed proteins in maize genotypes in response to LT NCBI Accession Number Protein name Spot ID Zea mays (Gurez local) 57 ONM31447.1 167 153 ONM51210.1 AQK63454.1 18 AQK47553.1 AQK43300.1 AQK55089.1 ONM53258.1 NP_001349277.1 ONM07168.1 ONM04429.1 AQK40686.1 RNA-metabolising metallo-beta-lactamase family protein [Zea mays] rough sheath1 [Zea mays] Threonine dehydratase biosynthetic chloroplastic [Zea mays] Thylakoidal processing peptidase 1 chloroplastic [Zea mays] farnesylated protein 3 [Zea mays] NHL domain-containing protein [Zea mays] Aquaporin PIP2-1 [Zea mays] Auxin response factor 12 [Zea mays] IND1(iron-sulfur protein required for NADH dehydrogenase)-like [Zea mays] Splicing factor 3B subunit 5/RDS3 complex sub- unit 10 [Zea mays] GINS complex protein [Zea mays] Translation initiation factor IF-2 [Zea mays] Threonine ammonia-lyase OS = Zea mays XP_008668879.1 A0A1D6GJ70\|A0A1D6GJ70 A0A1D6NDN6\|A0A1D6NDN6 Rx_N domain-containing protein ACG43269.1 XP_008670600.1 ONM27097.1 149 165 23 75 64 32 84 26 52 63 148 81 82 159 122 NP_001266439.2 Zea mays (GM6) AQK72491.1 220 1964 XP_008645909.1 1972 NP_001169298.1 310 ONM20750.1 180 AQK78980.1 3349 NP_001149881.1 1279 ONM25687.1 5275 ONM30654.1 178 ONM53258.1 6241 ONL95273.1 Nodulin-like protein [Zea mays] Peroxidase 2-like [Zea mays] Tetratricopeptide repeat (TPR)-like superfamily protein [Zea mays] Calcium-dependent protein kinase substrate protein [Zea mays] LETM1-like protein [Zea mays] Mitochondrial proton/calcium exchanger protein [Zea mays] Flowering time control protein FCA-like [Zea mays] RNA binding [Zea mays] Mediator of RNA polymerase II transcription subunit 11 [Zea mays] Glycosyltransferase family 28 C-terminal domain containing protein [Zea mays] Nuclear pore complex protein NUP98A [Zea mays] Importin subunit alpha-2, partial [Zea mays] Aquaporin PIP2-1 [Zea mays] Nuclear pore complex protein NUP155 [Zea mays] Page 5 of 20 pI MW (Da) 6.03 23,819 MAS- COT Score 40 Fold Change Regulated Type 10.7891 Up regulated 5.64 9.68 285,450 46 51 30,407 14.5208 Up regulated 41.2574 Up regulated 10.06 20,281 37 3.54261 Up regulated 40,895 5.25 44,718 6.69 3203 6.52 5.95 9099.3 10.17 10,534 6.52 3203 38 40 23 48 31 23 4.45764 Up regulated 4.11602 Up regulated Up regulated 3.0303 2.32466 Up regulated 4.20944 Up regulated 4.16235 Up regulated 8.64 2293.28 56 2.49025 Down regulated 10.94 3584.84 57 32,881 5.21 36 3102.67 49 9.51 36 48,806 8.87 42 29,956 5.24 36 42,195 6.29 regulated 14.2597 Up regulated 11.5956 Up regulated 6.37876 Up regulated 10.6314 Up regulated 7.93804 Up regulated 33.5755 Up regulated 9.93 16,833 37 2.0599 Up regulated 9.94 6.05 4361.72 11 11 8611.6 9.12254 Up regulated 8.17725 Down regulated 9.08 7925.87 38 4.02113 Down regulated 9.24 25,210 38 4.22351 Down regulated 6.70 6426 45 5.22440 Down regulated 7.63 4766.59 50 2.94550 Up regulated 10.88 17,084 34 3.20661 Up regulated 10.14 35,762 6.25 9.62 3203 25,430 51 23 43 5.98263 Up regulated 2.08445 Up regulated 2.56641 Down regulated AQK81609.1 F10K1.23 [Zea mays] 9.86 21,278 40 4.38033 Down In Gurez local, proteins viz. Threonine dehydratase bio- synthetic chloroplastic (Spot ID 153; NCBI Acc No. AQK63454.1) and Translation initiation factor IF-2 (Spot ID 52; NCBI Acc No. XP_008668879.1) formed a biologi- cally meaningful group of two proteins while rest 17 pro- teins clustered together into another group. The later was found to form 4 clusters each containing closely related set of proteins separated at distinct nodes. Similarly, in GM6 for clustering illustrations, 3 groups of proteins were obtained. In the first two groups, 2 pro- teins were clustered together, however, the third group splitted further into another two distinct clusters based on the set of closely connected proteins. Ramazan et al. BMC Plant Biology (2023) 23:183 Page 6 of 20 Fig. 4 Heatmap Clustering analysis of differentially regulated proteins (DRPs) in (A) Gurez local (B) GM6. The scale bar indicates up-regulated (red) and down-regulated (blue) DRPs Enriched metabolic pathways of differentially regulated proteins (DRPs) in Gurez local and GM6 To further analyze the functional fates of DRPs, KEGG enrichment analysis using ShinyGO (http://bioinformat- ics.sdstate.edu/go/) was performed. It was observed that higher protein numbers were constituted in enriched pathways in case of Gurez local as compared to the GM6 (Fig. 8), Also, the two genotypes were found to exhibit a remarkable divergence in metabolic pathway responses to LT stress. Using hyper geometric test, metabolic path- ways with –log10 (FDR) value greater than 1.2 were considered to be significantly affected by LT stress. We observed that in case of LT tolerant genotype Gurez local, regulation of seed growth, regulation of timing of transition from vegetative to reproductive phase, lipid glycosylation and aspartate family amino acid catabolic processes were significantly enriched (Fig. S3). Contrast- ingly, in GM6, cell cycle DNA replication initiation, and regulation of phenylpropanoid metabolic processes were remarkably enriched. Apart from two enriched meta- bolic processes (aspartate family amino acid catabolic processes and threonine catabolic processes) in the case of Gurez local (Fig. S4), most metabolic processes also clustered together in a network showing an interactive fashion of metabolic pathways. However, GM6 displayed a single cluster network of enriched pathways apart from the regulation of phenylpropanoid metabolic pro- cess (Fig. S5). Though additional studies are required for empirical observations of the metabolic events, and the growth and developmental transitions in plants (includ- ing maize) where the identified proteins apparently play a role, following are some of the enriched pathways which are pertinent to mention in Gurez local: Regulation of seed growth The leaf proteome revealed that under LT stress condi- tions, for regulation of seed growth, a number of proteins change their expression fashion. These include upregu- lated proteins viz. NHL domain-containing protein (Spot ID 165: NCBI Acc No. AQK55089.1) modulating the seed germination under abiotic stress, RNA-metabolising metallo-beta-lactamase family protein (Spot ID 57: NCBI Acc. No. ONM31447.1) involved in DNA/RNA metabo- lism, and downregulated proteins, such as, GINS com- plex (Spot ID 84: NCBI Acc No. AQK40686.1) protein and F10K1.23 (Spot ID 26: NCBI Acc No. AQK81609.1) playing a critical role in chromosomal DNA replication. All these proteins regulate the seed growth and aid in seedling emergence and its establishment. Regulation of timing of transition from vegetative to reproductive phase Developmental transitions from embryonic to post embryonic mode (seed germination), and vegetative to flowering phase, are regulated by the environmental cues [27]. The transition is also controlled by a complex genetic network including miRNAs, various transcrip- tional factors, and altered protein accumulation [28]. In the present study, proteins found in higher abun- dance under LT treatment include: ‘Thylakoidal pro- cessing peptidase 1 chloroplastic’ (Spot ID 18: NCBI Acc No. AQK47553.1) required for choloroplast protein Ramazan et al. BMC Plant Biology (2023) 23:183 Page 7 of 20 Fig. 5 Principal Component Biplot analysis (PCA) of identified proteins with their NCBI Accession no.’s in (A) Gurez local, and (B) GM6. The scale on the right depicts contribution of each component (protein) at different LT stress time-points. Proteins with their corresponding IDs’ in Gurez local and GM6 are: Threonine dehydratase biosynthetic chloroplastic (AQK63454.1), Translation initiation factor IF-2 (XP_008668879.1), NHL domain-containing protein (AQK55089.1), Peroxidase 2-like (XP_008670600.1), Calcium-dependent protein kinase substrate protein (NP_001266439.2), Auxin response factor 12 (NP_001349277.1), Aquaporin PIP2-1 (ONM53258.1), Calcium-dependent protein kinase substrate protein (NP_001266439.2), Tetratricopeptide repeat (TPR)-like superfamily protein (ONM27097.1),, F10K1.23 (AQK81609.1), Splicing factor 3B subunit 5/RDS3 complex subunit 10 (ONM04429.1), farnesylated protein 3 (AQK43300.1), rough sheath1 (ONM51210.1), Thylakoidal processing peptidase 1 chloroplastic (AQK47553.1), GINS complex protein (AQK40686.1), Threonine dehydratase biosynthetic chloroplastic (AQK63454.1), RNA-metabolising metallo-beta-lactamase family protein (ONM31447.1) & IND1(iron- sulfur protein required for NADH dehydrogenase)-like (ONM07168.1), Threonine ammonia-lyase (A0A1D6GJ70\|A0A1D6GJ70), Rx_N domain-containing protein (A0A1D6NDN6\|A0A1D6NDN6) and Flowering time control protein FCA-like (NP_001169298.1), RNA binding (ONM20750.1), LETM1-like protein (AQK72491.1), Mitochondrial proton/calcium exchanger protein (XP_008645909.1), Aquaporin PIP2-1 (ONM53258.1), Mediator of RNA polymerase II tran- scription subunit 11 (AQK78980.1), Importin subunit alpha-2, partial (ONM30654.1), Nuclear pore complex protein NUP155 (ONL95273.1), Glycosyltrans- ferase family 28 C-terminal domain containing protein (NP_001149881.1), Nuclear pore complex protein NUP98A (ONM25687.1), respectively Fig. 6 GO functional categorization based on biological process, molecular function and cellular component of identified proteins in (A) Gurez local (B) GM6 seedling leaves. The scale above represents the number of proteins contributing in each function degradation/chloroplast development/proper thylakoidal development in photosynthetic tissues; ‘rough sheath1’ (Spot ID 167: NCBI Acc No. ONM51210.1) neces- sary for cell fate/cell division; ‘translation initiation fac- tor IF-2’ (Spot ID 52: NCBI Acc No. XP_008668879.1) controlling different translational processes during veg- etative and reproductive growth; ‘calcium-dependent protein kinase substrate protein’ (Spot ID 122: NCBI Acc No. NP_001266439.2) critical for flowering; and ‘splicing factor 3B subunit 5/RDS3 complex subunit 10’ (Spot ID Ramazan et al. BMC Plant Biology (2023) 23:183 Page 8 of 20 Fig. 7 Hieracrchiel clustering exhibiting high domain architectural similarities of identified proteins in (A) Gurez local (B) GM6, protein sequences labelled as NCBI Accession numbers. In Gurez local, Threonine dehydratase biosynthetic chloroplastic (AQK63454.1) and Translation initiation fac- tor IF-2 (XP_008668879.1) formed a biologically meaningful group of two proteins while rest 17 proteins clustered together into another group. The later was found to form 4 clusters each containing closely related set of proteins [I: NHL domain-containing protein (AQK55089.1) & Peroxidase 2-like (XP_008670600.1)]; [II: Calcium-dependent protein kinase substrate protein (NP_001266439.2), Auxin response factor 12 (NP_001349277.1), Aquaporin PIP2-1 (ONM53258.1)]; [III: Calcium-dependent protein kinase substrate protein (NP_001266439.2), Tetratricopeptide repeat (TPR)-like superfamily pro- tein (ONM27097.1),, F10K1.23 (AQK81609.1), Splicing factor 3B subunit 5/RDS3 complex subunit 10 (ONM04429.1), farnesylated protein 3 (AQK43300.1), rough sheath1 (ONM51210.1) & Thylakoidal processing peptidase 1 chloroplastic (AQK47553.1)] and [IV: GINS complex protein (AQK40686.1), Threonine dehydratase biosynthetic chloroplastic (AQK63454.1), RNA-metabolising metallo-beta-lactamase family protein (ONM31447.1) & IND1(iron-sulfur protein required for NADH dehydrogenase)-like (ONM07168.1)] separated at distinct nodes. In GM6, 3 groups of proteins were obtained. In the first two groups, 2 proteins each [(Flowering time control protein FCA-like (NP_001169298.1), RNA binding (ONM20750.1) & LETM1-like protein (AQK72491.1), Mitochondrial proton/calcium exchanger protein (XP_008645909.1)] were clustered together, however, the third group splitted further into another two distinct clusters of three proteins each [Aquaporin PIP2-1 (ONM53258.1), Mediator of RNA polymerase II transcription subunit 11 (AQK78980.1), Importin subunit alpha- 2, partial (ONM30654.1) & Nuclear pore complex protein NUP155 (ONL95273.1), Glycosyltransferase family 28 C-terminal domain containing protein (NP_001149881.1), Nuclear pore complex protein NUP98A (ONM25687.1) based on the set of closely connected proteins 32: NCBI Acc No. ONM04429.1) for checking the growth of terminal buds. All the above mentioned proteins accu- mulate in higher proportions in stress response in Gurez local and play a critical roles in floral transitions. Lipid glycosylation Glycosylation is one of the fundamental post-transla- tional modifications modulating various developmen- tal processes including stress response in plants [29]. Remarkably, under LT stress conditions, glycerolipids serve as signaling molecules and also protect organeller membranes from stress induced damage [30]. Interest- ingly, one of the novel proteins found in Gurez local viz. Nodulin-like protein is believed to have a probable role in transport of various carbohydrate meoties or other sol- utes that may be indirectly involved in lipid glycosylation process under LT stress conditions, though subject to further investigations. Aspartate family amino acid catabolic processes and threonine catabolic processes The co-regulation activity of aspartate coupled with other amino acids belonging to ‘Asp family amino acids’ is beneficial for plant stress adaptation [31]. Hence, the up-regulation of Threonine dehydratase biosynthetic chloroplastic (Spot ID 153: NCBI Acc No. AQK63454.1) and Threonine ammonia-lyase (Spot ID 63: Uniprot ID. A0A1D6GJ70) proteins in Gurez local reveal their potent roles in amino acid metabolism associated LT stress tol- erance in maize. Other pathways involved in stress response in Gurez local Under environmental stress induced conditions, mul- tiple signaling pathways converge to regulate stress induced genes that in turn produce proteins and enzymes required for stress metabolism in maize [32]. The present study revealed that most of the proteins accumulated in Ramazan et al. BMC Plant Biology (2023) 23:183 Page 9 of 20 Fig. 8 KEGG pathway enrichment analysis in DRPs in (A) Gurez local (B) GM6. The size of point represents the number of proteins enriched in a particular pathway and the X-axis represents fold enrichment value. KEGG, as developed by Kanehisa Laboratories (https://doi.org/10.1093/nar/28.1.27), was used in the imagery higher levels under LT stress in Gurez local are involved in defense mechanism against LT stress. The accu- mulation of various abiotic stress responsive proteins include farnesylated protein (Spot ID 149: NCBI Acc. No. AQK43300.1), auxin response factor 12 (Spot ID 75: NCBI Acc. No. NP_001349277.1), IND1 (iron-sulfur pro- tein required for NADH dehydrogenase)-like (Spot ID 64: NCBI Acc. No. ONM07168.1), translation initiation factor IF-2 (Spot ID 52: NCBI Acc. No. XP_008668879.1), and tetratricopeptide repeat (TPR)-like superfamily pro- tein (Spot ID 159: NCBI Acc. No. ONM27097.1), acting as a negative regulator in cold stress signaling. Other proteins exhibiting upregulated expression and having an indispensable role in both abiotic and biotic stress tol- erance constitute NHL domain-containing protein (Spot ID 165: NCBI Acc. No. AQK55089.1), aquaporin PIP2-1 (Spot ID 23: NCBI Acc. No. ONM53258.1) and peroxi- dase 2-like (Spot ID 82: NCBI Acc. No.XP_008670600.1). Ramazan et al. BMC Plant Biology (2023) 23:183 Page 10 of 20 However, another protein viz. Rx_N domain-containing protein (Spot ID 148: NCBI Acc. No. A0A1D6NDN6) which gets remarkably induced under LT treatment in Gurez local maize is well-known for its critical role in biotic stress tolerance in plants [33]. Enriched metabolic processes in GM6 as follows Cell cycle DNA replication initiation, RNA metabolism and nucleo-cytoplasmic trafficking Protein activity generated after mRNA translation and other post-transcriptional/translational modifications involving a plethora of molecular machinery, directly or indirectly contribute to adaptation at cellular level under stress induced conditions [34]. As evident from the protein list obtained in the study (Table 1), few pro- teins playing a vital role in nuclear metabolism which include ‘RNA binding protein’ (Spot ID 310: NCBI Acc No. ONM20750.1) and Mediator of RNA polymerase II transcription subunit 11 (Spot ID 180: NCBI Acc. No. AQK78980.1) display downregulated expression in LT sensitive genotype. Contrastingly, the other two impor- tant nuclear proteins responsible for nuclear transport i.e., nuclear pore complex protein NUP98A (Spot ID: 1279: NCBI Acc. No. ONM25687.1) and importin sub- unit alpha-2, partial (Spot ID: 5275: NCBI Acc. No. ONM30654.1) accumulate in higher proportion in GM6 in response to the LT stress. Other stress responsive pathways Numerous stress factors affect flowering time in plants [35], transmembrane water movement [36] and exchange of other small molecules across mitochondria [37]. The proteins found in significant abundance in above men- tioned stress signaling pathways in GM6 include aquapo- rin PIP2-1 (Spot ID 178: NCBI Acc. No. ONM53258.1) and LETM1-like protein (Spot ID 220: NCBI Acc. No. AQK72491.1). On the other hand, flowering time control protein FCA-like (Spot ID 1972: NCBI Acc. No. NP_001169298.1) and mitochondrial proton/cal- cium exchanger protein (Spot ID: 1964: NCBI Acc. No. XP_008645909.1) were found in lower abundance with LT treatment progression in GM6. Other metabolic pro- cesses involving transfer of sugar moieties onto a variety of small molecules involve the glycosyltransferase family 28 C-terminal domain containing protein (Spot ID 3349: NCBI Acc. No. NP_001149881.1) which exhibited rela- tively higher levels in LT sensitive genotype. Protein-protein interaction networks are associated with LT stress response Using STRING 10.5 database (http://www.string-db. org/, accessed on 18 April 2022) protein-protein interac- tion and functional relationships among the differentially regulated proteins was predicted with the confidence score greater than 0.7 in both genotypes of maize. In case of Gurez local, single group constituting 9 interact- ing proteins was identified in the network. These include splicing factor 3B subunit 5/RDS3 complex subunit 10 (GRMZM2G121942_P02, pco072231b), GINS complex (GRMZM2G049536_P02, pco124429), NHL protein domain-containing protein (GRMZM2G326783_P01, thx29), DNA helicase protein (GRMZM2G139894_ P01, mcm7), GINS complex protein with a predicted functional partner ‘Flowering time control protein FY’ (GRMZM5G881296_P03), GINS complex protein (GRMZM2G076128_P01, TTN10), Protein spa-1 related 4 isoform X1(GRMZM2G061602_P01), DNA replication complex protein (GRMZM2G134295_P03, pco089553b), and PHD finger-like domain containing protein 5 A (GRMZM2G047018_P01) (Fig. 9). All these proteins take part in chromosomal DNA replication. On the other side, GM6 also exhibited a single cluster of 7 proteins in the interaction network. These include mediator of RNA polymerase II transcription subunit 11 (GRMZM2G180815_P01), nuclear pore complex protein NUP98A (GRMZM2G348675_P02), nuclear pore com- plex protein NUP155 (GRMZM2G057853_P01, IDP351), importin subunit alpha-2, partial (GRMZM2G059015_ P01, cl23032_-2), component of nuclear pore complex (GRMZM2G058498_P01), and RNA binding protein (GRMZM5G881296_P03). All these proteins together aid in nucleocytoplasmic transport of proteins and RNAs (Fig. 9). These results indicate that interaction among various proteins especially involved in a particular meta- bolic pathway is essential in responding to the low tem- perature stress in maize. Expression levels of genes encoding DRPs in response to LT in maize genotypes Expression analysis employing qRT-PCR was used to analyze the transcriptional activities of fifteen randomly selected genes including the three main novel proteins (10 from Gurez local and 5 from GM6) in order to cor- roborate our proteomic findings. In general, in case of Gurez local, the results showed that 5 up-regulated [RNA-metabolising metallo-beta-lactamase family pro- tein; rough sheath1; threonine dehydratase biosynthetic chloroplastic (reasonably higher expression at 6 and 8 h of LT treatments); thylakoidal processing peptidase 1 chloroplastic; and nodulin-like protein], and 2 down- regulated proteins (GINS complex protein and F10K1.23 showing down-regulation post 6 h LT stress) coincided well with the patterns of the transcript levels of corre- sponding coding genes at different LT stress time points. However, other 3 proteins (auxin response factor 12, splicing factor 3B subunit 5/RDS3 complex subunit 10, and threonine ammonia-lyase) showed an exponential increment in their corresponding mRNA levels in the Ramazan et al. BMC Plant Biology (2023) 23:183 Page 11 of 20 Fig. 9 Protein interaction networks of identified proteins in (A) Gurez local (B) GM6. The network was generated with the help of STRING (https:// string-db.org/) program at a confidence score greater than 0.7. Nodes (colored circles) indicate proteins and the thickness of lines connecting the nodes denotes the strength of supplementary data. Different types of interactions between nodes are represented by colored lines. Red lines indicate the fusion of genes, green lines neighborhood of genes, blue lines co-occurrence across species, purple lines experimental evidence, yellow lines text mining of abstracts from literature, light blue lines databases, and black lines co-expression in the same of other species first hours of stress, which later on remarkably decreased with the progression of stress treatment (Fig. 10). The fascinating part of the story is that the proteins including threonine dehydratase biosynthetic chloroplastic, thyla- koidal processing peptidase 1 chloroplastic, and nodulin- like protein believed to have a novel and potent role in LT stress tolerance in maize as identified in Gurez local were consistent with the proteomic findings, hence strength- ening the basis for the set hypothesis. In case of GM6, among the 5 selected proteins, 3 pro- teins (mitochondrial proton/calcium exchanger pro- tein, nuclear pore complex protein NUP98A, and RNA binding) were observed to to replicate the qRT-PCR approach while considering the general trend of their transcript expression. Hence, overall expression patterns of the studied genes were found to corroborate with our proteomic findings. However, rest 2 proteins viz. glyco- syltransferase family 28 C-terminal domain containing protein, and aquaporin PIP2-1 exhibited opposite trends with their mRNA homologues (Fig. 11). Correlation analysis of abundance of selected proteins with their corresponding transcript levels in maize genotypes The correlational study between the levels of selected proteins’ transcripts and their abundance in the two gen- otypes demonstrated the consistency of gene expression at various stress time points (Fig. 12). Evidently, pro- teins with uniprot IDs A0A1D6MR20, A0A1D6HRV5, AOA1D6GJ70, K7TTX2, C0PL36, and B6TE00 exhib- ited a positive correlation with their mRNA levels in the Gurez local genotype. However, the trend was reverse in the rest four proteins bearing IDs K7TV72, A0A1D6IXQ3, A0A1D6LQ45, and B6U1N6. Similarly, in case of GM6, the correlation analysis revealed positive (for protein with uniprot ID of A0A1D6ELI2), slightly negative (uniprot ID: A0A1D6HCR5), considerably nega- tive (uniprot IDs: A0A1D6F2P9 and NUP98A), and no (uniprot ID: B6TJ06) corroboration between the mRNA and protein levels. Discussion LT stress severely affects early vigor and production of maize [38] which in turn signifies the importance of exploring and understanding peculiar molecular net- works behind cold and other abiotic stresses. Therefore, unravelling the stress-driven proteome changes in plants holds remarkable significance since proteins, unlike tran- scripts, are direct effectors of plant responses to stress conditions [39]. In this context, we, therefore, demon- strated the role of various LT induced proteins conferring exceptional LT tolerance to the Gurez local maize geno- type from Kashmir Himalaya compared to its susceptible counterpart, GM6. Interestingly, a few proteins showing Ramazan et al. BMC Plant Biology (2023) 23:183 Page 12 of 20 Fig. 10 Validation of proteomic data by qRT-PCR in Gurez local genotype. The Y axis represent the ‘relative expression’ and X axis the ‘stress time points’ for all the selected genes from DRPs. Data show the mean ± SD of three replicates, and significant differences between treatment and control samples were indicated by letters at p ≤ 0.05. The internal control used was alpha tubulin and ΔΔCt was calculated using 0 h as control significant expression levels in Gurez local under LT treatment are reported for the first time for their potent role in conferring LT stress tolerance to maize, however, the hypothesis warrants further investigations. Overall, our comparative proteomic analysis serves as an impor- tant exploratory method to determine the impact of envi- ronment modulated gene expression under LT stress. Differentially regulated proteins in Gurez local We identified 19 proteins in Gurez local in response to LT stress, among which only two were found to get down-regulated, and rest 17 exhibited remarkable abun- dance with the progression of LT stress including a few novel proteins. Pertinently, numerous stress responsive proteins with some novel LT-induced ones have been well documented from different species earlier as well [40]. Here, majority of the proteins identified in Gurez local were found to be involved in regulation of seed germination and seedling growth, timing of floral tran- sition, lipid glycosylation, amino acid metabolism, and defense mechanisms in response to LT stress. Based on the involvement of identified proteins in different pro- cesses, it may be concluded that Gurez local has devel- oped strong molecular strategies in due course of time, to stand with the harsh environmental cues especially at sprouting stage. Nevertheless, and in light of the limited scope of present investigation, further experimentation is needed to validate the role of identified proteins in differ- ent metabolic machineries and through various develop- mental stages of the plant. Proteins related to seedling establishment in LT tolerant Gurez local Studies have shown that crop quality and yield is primar- ily affected by seed germination and seedling establish- ment [41]. Importantly, our study was also focused on the stress-responsiveness of maize seedlings at three-leaf stage. As a response against LT stress, we found four pro- teins involved in the process of seedling growth from the LT tolerant genotype of this staple crop. Among these, Ramazan et al. BMC Plant Biology (2023) 23:183 Page 13 of 20 Fig. 11 Validation of proteomic data by qRT-PCR for selected genes from DRPs in GM6 maize genotype. Data show the mean ± SD of three replicates, and significant differences between treatment and control samples were indicated by letters at p ≤ 0.05 Fig. 12 Correlational analysis of selected proteins with their corresponding mRNA levels in (A) Gurez local (B) GM6. The red color in the scale denotes positive correlation while blue denotes inverse correlation between the two ‘NHL domain-containing protein (NHL1)’, with estab- lished role against pathogen induced stress condition in Glycine max at seedling stage [42], was supposed to exhibit possible relation with the seedling establishment in Gurez local as well. Other proteins include ‘GINS complex protein’ and ‘RNA-metabolising metallo-beta- lactamase family protein’. The former has been found to play a role in cell cycle processes in Arabidopsis and rice [43] while the members of later in early stages of plant development in Arabidopsis [44]. These proteins Ramazan et al. BMC Plant Biology (2023) 23:183 Page 14 of 20 subsequently result in the overall vegetative growth of the plant through coordinated interaction of cellular cycle and cell expansion machinery. Finally, ‘F10K1.23’, belonging to the family of UDP glycosyltransferases is also believed to promote the seedling growth associated LT tolerance in maize. Very recently, this protein was found to play a critical role in regulating grain size and abiotic stress tolerance with associated metabolic redi- rection flux in rice [45]. Proteins involved in floral transition under LT stress As soon as the corn seedlings establish themselves towards final vegetative phase (V5) a crucial develop- mental change in plant’s lifecycle is marked and transi- tion to flowering begins. The flowering time is critical factor in determining the crop yield in maize, and impor- tantly alteration in flowering timing is a strategy used to survive abiotic stresses [46]. In our study, we identified five proteins with a potent role in transition to flowering stage in maize. For instance, the novel protein identified under LT stress viz. ‘Thylakoidal processing peptidase 1 chloroplastic’ is believed to develop the photosynthetic tissues to mark plants’ growth towards flowering stage owing to its documented role in chloroplast development under salt and osmotic stress in wheat seedlings [47]. Similarly, the up-regulation of ‘rough sheath1’ protein known to play an indispensable role in cell division and expansion in maize leaf [48], is also supposed to cause expansion of maize leaves and leaf initiation, heading the plant towards the flowering stage. The ‘translation ini- tiation factor IF-2’ conferring abiotic stress tolerance to Tamarix hispida [49] and salt stress tolerance to yeast and plants [50], plays an important role in regulating the protein synthesis necessary for switching the plant from vegetative to reproductive phase. The plant signaling key factor ‘calcium-dependent protein kinase’, otherwise involved in calcium signaling while regulating flower- ing time in Arabidopsis [51], may also regulate the floral transition in Gurez local under LT stress. Additionally, the ‘splicing factor 3B subunit 5/RDS3 complex subunit 10’ regulating the fall of alfalfa terminal buds [52] might also contribute in the terminal bud growth thereby allow- ing the transition of Gurez local genotype from juvenile to adult phase. All these proteins, in one way or the other, help the Gurez local to grow and develop under unfavor- able LT stress. Transport of sugar moieties and essential amino acids is a vital strategy of LT tolerance in Gurez local One of the interesting and novel proteins that we iden- tified in Gurez local is ‘nodulin-like protein’ that was found highly abundant under LT treatment in terms of both protein and mRNA content. As the phloem load- ing/unloading is highly responsive to environmental changes, it is believed to be involved in transport related activities in phloem under LT stress conditions. In fact, previous findings show its potent role in transport of sugar moieties wherein this protein was found to get accumulated in the sieve element plasma membrane of Arabidopsis [53]. Interestingly, in non-nodulating plant species including maize, nodulin like proteins have also been found with their special role in transporter activity throughout the plant developmental stages [54]. Hence, in non-leguminous plants like maize, the functions of ‘nodulin like protein’ are emerging and require further investigations to validate its role in abiotic stress associ- ated phloem transport. Apart from sugars, majority of the amino acids have a central role in plant environmental stress response [55]. Proteins related to amino acid metabolism, particularly, ‘Asp family amino acids’ that were found in increased abundance in Gurez local were one among the novel pro- teins viz. ‘Threonine dehydratase biosynthetic chloro- plastic’ and ‘Threonine ammonia-lyase’. Besides playing primary role in transport of nitrogen at different stages of development in plants, the ‘Asp family amino acids’ have essential roles in plant abiotic stress response as well [31]. This was reported in a recent study wherein exog- enous application of some amino acids on maize under high temperature stress showed positive effect on growth and development of the crop plant [56]. However, the direct involvement of ‘Threonine dehydratase biosyn- thetic chloroplastic protein’ in stress responsiveness in plants has not been confirmed so far and requires further explorations. Proteins in relation to ‘response to stimuli’ under LT stress induced conditions Series of defense mechanisms in response to environ- mental cues operate within an organism necessary for its survival [57], and specific proteins in response to stress per se are produced in huge quantities through a variety of mechanisms [58]. In our study, proteins asso- ciated with LT stress tolerance found in higher abun- dance include the farnesylated protein, auxin response factor 12, the role of which has been highlighted in case of stressed tomato [59] and banana [60]; IND1 (iron- sulfur protein required for NADH dehydrogenase)-like reported earlier in abiotic stress response in rice [61]; Tetratricopeptide repeat (TPR)-like superfamily protein expressed under osmotic stress responses in Arabidop- sis [62]; NHL domain-containing protein, the overex- pression of which provides resistance to biotic stress in soyabean [42]; Aquaporin PIP2-1 having a full-fledged role in abiotic stress tolerance in plants [63]; and Peroxi- dase 2-like conferring insect resistance in maize kernels [64]. The present study provides additional evidence that all these proteins contribute to LT tolerance in maize, Ramazan et al. BMC Plant Biology (2023) 23:183 Page 15 of 20 hence, our findings provide a potent genetic resource for enhancing LT tolerance in this demanding crop plant. Differentially regulated proteins in GM6 A differential ability of LT tolerant and LT sensitive maize genotypes to cope with stress conditions is evident from the changes in abundance of different stress protective proteins in Gurez local as compared to its counterpart. The tolerant genotype does not suffer more from energy metabolism disruption and withstands the harsh envi- ronmental scenario which has been reported earlier also [65] as compared to the stress-sensitive genotype. The reduced levels of most of the stress protective proteins and alteration in the expression of proteins involved in basic metabolic processes in GM6 reveals the ineffi- cient stress tolerance mechanism in the stress sensitive genotype. Regulation of mRNA/protein nucleo-cytoplasmic trafficking in plant stress response The nuclear pore complex (NPC) family components not only play key roles in general growth and develop- ment of plants but also in response to different stressors affecting the plant survival [66]. The NPC family consist- ing of different proteins named as nucleoporins (Nups) regulate the molecular trafficking between nucleus and cytoplasm [67]. In the present study, two proteins asso- ciated with NPC family were found in abundance under LT stress in GM6. These include nuclear pore complex protein NUP98A, and importin subunit alpha-2, partial. On the other hand, the down-regulation of third protein viz. nuclear pore complex protein NUP155 in LT suscep- tible genotype depicts the alteration in relative protein abundance due to unfavorable stress conditions. All these findings demonstrate the role of these proteins in NPC- mediated plant stress responses. Furthermore, the down-regulation of RNA binding protein and mediator of RNA polymerase II transcription subunit 11 by GM6 in response to LT treatment depicts the overall susceptibility of the genotype towards stress conditions, hence revealing the developmental arrest. Similar reports on decreased abundance of most of the stress responsive proteins in stress-sensitive species have been well documented [65] previously. Otherwise, RNA binding proteins associated with post transcriptional regulation of RNA metabolism play central roles in stress responses besides helping the plant growth and develop- ment [68]. Altered abundance of other critical stress sensitive proteins in GM6 Aquaporins belonging to the class of membrane proteins facilitate water transport and transport of other solutes thereby playing a vital role in cell signaling, nutrient acquisition and stress response [69]. In GM6, aquapo- rin PIP2-1 was found in higher abundance to ensure the water availability to the plant under stressful conditions, coherent with the previous findings on maize geno- types under drought stress [69]. On the other side, leu- cine zipper/EFhandcontaining transmembrane protein 1 (LETM1) like protein has a remarkable role in mitochon- drial translation in early seed development as studied in Arabidopsis thaliana [70]. The higher levels of this pro- tein in GM6 depicts the struggle of this genotype under LT stress required for the re-establishment of mitochon- drial homeostasis and repairing of stress induced molec- ular damage. Interestingly, the reduced levels of another mitochondria associated protein viz. mitochondrial pro- ton/calcium exchanger protein are concomitant with the former one. Actually, one of the death inducing mecha- nisms in plants under adverse environmental conditions is the mitochondrial permeability transition which is characterized by collapsing of electrochemical gradient across the inner mitochondrial membrane [71]. Hence the probable reason behind the reduction in the protein levels is to prevent the GM6 genotype from mitochon- drial damage caused due LT stress, as no special adaptive mechanisms operate in this stress sensitive genotype to withstand LT conditions. In addition to the above, it is well known fact that ambient temperature profoundly affects the flowering time in plants [72]. So, in case of LT sensitive GM6, the decreased abundance of flowering time control protein FCA-like clearly indicates the delayed flowering tran- sition due to LT stress thereby diminishing the plant growth. Furthermore, enhanced transfer of sugar moi- eties under abiotic stress conditions has gained much interest during the past decade for their roles in ROS scavenging, signaling and osmoprotectants [73]. In our study also, the elevated levels of glycosyltransferase fam- ily 28 C-terminal domain containing protein are believed to have a potential role in transfer of sugar moieties to alleviate the cellular levels of sugars that in turn help the GM6 genotype to survive under adverse environment. Our study provides additional theoretical basis as well as practical significance for further exploration of molecu- lar mechanisms operating in the golden crop, maize, in response to LT stress. Positive correlation between proteins and their transcript abundance confirms the reliability of our data Most of the proteins among those which were selected for qRT-PCR analysis coupled well with their corre- sponding transcript levels especially the three novel proteins from Gurez local viz. hreonine dehydratase biosynthetic chloroplastic, thylakoidal processing pepti- dase 1 chloroplastic, and nodulin-like protein identified in our study. The observation was also confirmed by the Ramazan et al. BMC Plant Biology (2023) 23:183 Page 16 of 20 correlational analysis between the two. However, few proteins whose abundance did not match with transcript levels from both genotypes may be due to differences in the synthesis and decay rates of mRNA and proteins in addition to microRNA regulated protein synthesis [74]. Conclusion In the present study, a comprehensive comparative approach was applied to decipher the molecular basis of LT tolerance in Gurez local as compared to GM6. We successfully identified 19 DRPs in Gurez local and only 10 in GM6 in response to LT stress of 12 h duration at early seedling stage. Most of the proteins (17) in Gurez local exhibited up-regulated expression under LT stress while only 2 were found to get down-regulated with the progression of LT treatment. The proteins were associ- ated with the regulation of seed germination and seed- ling growth, floral transition timing, lipid glycosylation, amino acid metabolism, and defense adaptations in response to LT stress, necessary for the survival under extreme environments during early growth phases. Con- trary to this, in LT susceptible GM6, 5 proteins were found in higher abundance and 5 got down-regulated under LT conditions. These identified proteins influence the basic metabolic activities like cell cycle regulation, nucleo-cytoplasmic trafficking and a few in stress defense mechanisms. Most notably, the present study identified the ‘three’ novel proteins from Gurez local, which include threonine dehydratase biosynthetic chloroplastic, thyla- koidal processing peptidase 1 chloroplastic, and nodulin- like protein, the roles of which has not been established in plant stress defense so far. The results show that Gurez local may serve as a repository for exploring more LT responsive genes important in molecular breeding of LT tolerant maize genotypes. Furthermore, for most of the proteins identified in Gurez local, the changes in pro- tein abundance were consistent with their correspond- ing transcript levels, hence, substantiating our proteomic findings. In nutshell, our investigations clarified the strat- egies employed by temperate grown ‘Gurez local maize’ to withstand adverse low temperatures of Kashmir Hima- layas comparative to its tropical grown counterpart and also elucidated the basic molecular networks and meta- bolic pathways associated with LT tolerance in maize. Materials and methods Plant material, lt treatment and sampling Two maize genotypes exhibiting differential temperature tolerance were selected for the experiment. LT tolerant genotype viz., Gurez local (native to Kashmir Himalayas) and LT sensitive, Gujarat-Maize-6 (GM6-native to Guja- rat) lines were obtained from IACRP-Maize (SKAUST-K) Srinagar Centre. Seeds were sown in soil: sand mixture (v/v 3:1) in pots (volume 300 ml, diameter 8 cm and height 12 cm) in a completely randomized experimen- tal design and a single seedling was maintained per pot. The seeds were germinated and allowed to grow up to three-leaf stage in a growth chamber (Blue Star, ICo No: NKL-750) maintained at a temperature of 25 ± 2 °C, pho- toperiod of 16 h light/8 h dark under a relative humidity of 70%. After two weeks, maize seedlings were subjected to LT stress of 6 °C for a period of 12 h wherein 0 h served as the control. Based on our preliminary physio- biochemical investigations on the same samples [9], sig- nificant results were seen at some selected hours of stress time points viz., 2 h, 6 h, 8 and 12 h (Fig. 1). Hence, for carrying out the proteomic analysis and quantitative real time PCR (qRT-PCR), plant leaf samples were har- vested on these selected hours of stress. So, a total of 5 time-points were selected for carrying out the proteomic analysis in both the genotypes. All methods were carried out in accordance with relevant guidelines. Protein sample preparation and 2D-PAGE Extraction of protein samples was done following the methodology [20] subjected to certain modifications. Briefly, for each time point, 3–5 seedlings (at three-leaf stage) were collected and their leaves (three) were har- vested. Latter were pooled and 1 g leaf tissue was ground to fine powder in liquid nitrogen and then dissolved in 10 ml chilled homogenization buffer [sucrose (40%), 50 mM HEPES-KOH (pH, 7.5), β mercaptoethanol (1%), 1 mM EDTA (pH, 7.5), 60 mM sodium fluoride]. After vor- texing the homogenates, 15 ml tris-equiliberated phenol was added to each sample. Sample solutions were kept on rocker for 30´ at 4 °C, and then centrifuged (5000 x g at 4 °C) for 15 min. Supernatants (phenol phase) were transferred to clean and pre-chilled tubes, and 0.1 M ammonium acetate in methanol was added for protein precipitation at -20 °C overnight. The samples were cen- trifuged (10,000 x g at 4 °C) for 25 min and pellets were washed three times with 80% acetone and then resus- pended in 500 µl rehydration buffer on ice. Prior to isoelectric focusing (IEF), protein samples (500 µg) were subjected for active rehydration overnight at 25 °C into a 24 cm GE Healthcare Strip Holder. First, the protein samples were diluted with 2-D rehydration buffer [8 M urea, 2 M thiourea, 2% CHAPS (w/v), 20 mM DTT, 0.5% pharmalyte (v/v, pH 4 − 7) and 0.05% bromo- phenol blue (w/v)], and 450 µL of the diluted proteins were used to rehydrate the strips passively. The rehy- drated strips (24 cm, pH 4–7) were subjected to IEF using Ettan IPGphor system (GE Healthcare, USA) under the following standardized 18 h program: 200 v for 1 h, 500 v for 1 h, 1000 v for 2 h each for gradient and step, and 10,000 v for 2 h gradient with 8 h step. The focused strips were equilibrated in 15 mL of equilibration buffer [6 M urea, 50 mM Tris-HCl (pH 8.8), glycerol (30%, v/v), and Ramazan et al. BMC Plant Biology (2023) 23:183 Page 17 of 20 SDS (2%, w/v)] first by reduction with DTT, followed by alkylation with iodoacetamide (2.5%, w/v) in the same buffer, each for 10 min. The second dimension was conducted by loading the IPG strips onto the 12% SDS polyacrylamide gels. On the top of 1 mm thick 2D gel, the equilibrated strips were sealed using 0.5% low-melting agarose in SDS- electro- phoresis buffer (25 mM Tris, free base, 200 mM glycine, and 0.1% SDS). The proteins were allowed to get resolved till the bromophenol blue front reached the gel end on Ettan Dalt-6 electrophoresis unit (GE Healthcare, USA) at a constant voltage of 90 V. Gel image analysis and proteome profile evaluation Gel staining was done using Coomassie Brilliant Blue (CBB) G-250 (Bio-Rad, USA) in a gel staining solution (10% GAA, 45% methanol with 0.3% CBB). Proteins were visualized after 24 h by destaining the 2D gels. Digital images of all the gels were obtained using ImageMaster 2D Platinum (GE Healthcare, USA). Three replicate gel images were combined to create a ‘master gel’ for each stress time-point. The matchset was generated by com- paring the master gels from each time-point obtained through pairwise comparisons. After background subtraction, and spot detection, matched spots were normalized with the help of total density index of the gel images. The proteins show- ing a change > 2 folds were treated as differentially accumulated. Spot excision and In-Gel trypsin digestion After proper washing of gel slabs, the selected spots were excised using clean scalpel. The gel pieces were cut into roughly 1 mm3 cubes and put in clean 1.5 ml eppendorf tubes. After that, de-staining of gel slices was carried out using water followed by 100 µl of 100 mM ammo- nium bicarbonate. Later on, reduction and alkylation of proteins was completed with 10 mM DTT and 55 mM iodoacetamide, followed by drying of samples in vacuum centrifuge. Trypsin digestion of all the samples was executed by rehydrating the gel pieces with 30 µl of digestion buffer containing 50 mM NH4HCO3 and 20 ng/µl of trypsin for 30 min at 4 °C. After required quantity of digestion buffer was absorbed by the proteins, the excess gel enzyme solu- tion was removed. Then, 100 µl 50 mM NH4HCO3 buf- fer was added to cover the gel and incubated overnight at 37 °C. After overnight incubation, the digest was recov- ered to a new tube and formic acid was added to the each tube to stop enzymatic reaction. The final extraction of peptides from gels was carried out and sample peptides were resolubilized in 10–20 µl of 0.1% of formic acid to proceed for MALDI-TOF/TOF analysis, until then dried samples were stored at -20 °C. Protein species identification by MALDI-TOF MS All protein digest samples (protein digests) in lyophi- lized condition were dissolved with 10 µl dissolving sol- vents (70% H2O, 30% acetone and 0.1% TFA). A suitable matrix solution of ά-cyano-4-hydroxy- cinnamic acid was prepared (6.0 mg of CHCA, 50% water, 50% aceto- nitrile, and 0.1% TFA). Both matrix and sample solutions were mixed in a 1:1 ratio and were spotted in MALDI plate (Opti-TOF™ 384 Well Insert (123 × 81 mm), MDS Sciex) in dried droplet method. Co-crystallized matrix and sample were allowed to air dry and then subjected for MS/MS analysis by 4800 Plus MALDI-TOF/TOF Analyzer (AB Sciex Pte Ltd.). MALDI-MS and MS/ MS spectrum were obtained using standard operating software 4000 Series Explorer™ Remote Client. Instru- ment was calibrated (plate model) for both MS and MS/MS in reflectron mode against the standard pro- cedures using calibration mixture of 6 known peptides [des-Arg1-Bradykinin, Angiotensin 1, Glu 1-Fibrino- peptide, ACTH(1–17), ACTH(18–39), ACTH III(7–38 )]. MALDI-Spectrum was acquired using laser Nd-YAG which operates at 355 nm to ionize samples. A fixed laser intensity of 6000 and 7000 Hz was used for MS and MS/ MS analysis, respectively to ionize the samples. All raw mass spectrometry data in the form of LC-MS/MS files were transformed into mgf (mascot generated files) using Mascot distiller software (www.matrixscience.com). Pro- tein identification was done through maize protein data- base (UP7305_Z_ mays (AA); accessed on 02 December 2021) using the Mascot search engine (version 2.3.02; Matrix Science, London, UK). Following search param- eters were set while performing the identification: taxon- omy as ‘other green plants’, cleavage enzyme used trypsin, 1 maximum missed cleavages allowed, carbamidomethyl (C) as fixed modifications and oxidation (M) as variable modifications. The results were filtered at a significance threshold of P < 0.05 combining evidence from individual peptides generated through Mascot. The identified pep- tide sequences were queried against BLASTp-NCBI (pro- tein-protein BLAST with E-value cut-off of 1e-10) under RefSeq protein of Zea mays (https://blast.ncbi.nlm.nih. gov/Blast.cgi?PAGE=Proteins; Zea mays taxid: 4577) for further identification and annotation of identified maize proteins. Bioinformatic analysis of differentially expressed proteins Beginning with protein structure prediction, 3D models for each identified protein was developed using Phyre2 web portal for protein modeling, prediction and analysis. [21]. This algorithm applies hidden Markov mod- els (HMMs) for classification of proteins based on their amino acid sequence, hence predicting the presence of a specific protein domain. For analysis of stereochemi- cal quality of each protein, PDBsum web server was Ramazan et al. BMC Plant Biology (2023) 23:183 used [22]. Furthermore, enrichment analysis of all pro- teins was carried out through ShinyGO graphical tool. ShinyGO provides in-depth analysis of gene lists and their characteristics, with graphical visualization of enrichment, pathway and protein interactions [23]. For understanding the system-level of cellular processes under LT stress, protein-protein interaction was stud- ied using STRING database (Search Tool for Retrieval of Interacting Genes/Proteins; Version 11.0; [24]. In addi- tion, online web server CLAP, a useful protein-clustering tool, was employed for classification of identified protein sequences [25]. Blast2GO (now OmicsBox), a research tool for performing high quality functional annota- tion [26] and for classification of proteins on the basis of molecular functions, biological processes and cellular components was exercised to generate gene ontology representation. Heatmap for visualization of hierarchal clustering of proteomics data and correlation analysis of transcript levels and corresponding protein abundance was generated using R (www.r-project.org/). RNA extraction and qRT-PCR Total RNA extraction from all the selected LT stress time points was performed using Trizol reagent (Invitrogen, Carlsbad, USA) and then reverse transcribed using Revert Aid First Strand cDNA Synthesis kit (Thermo-Fisher Scientific) as per the manufacturers’ recommendations. Primers for selected candidate proteins were designed using Primer 3 (http://bioinfo.Ut.ee/primer3-0.4.0/) soft- ware and were purchased from Integrated DNA Technol- ogies (IDT), USA (Table S1). The subsequent qRT-PCR analysis was done as previously described [9] while using 0 h as control against the different treatment time points including 2 h, 6 h, 8 h, and 12 h. Statistical analysis The experimental design followed in the study was com- pletely randomized Design (CRD) with three replica- tions. The qRT-PCR data are presented as mean ± SD of the three biological samples with three technical repli- cates for each and analysed by GraphPad Prism for Win- dows Version 5.0 (Graph pad Software San Diego, CA, USA) and Statistix 10 (Analytical software, 2013) fol- lowing one-way ANOVA. Post-hoc test employed was Tukey’s multiple comparison tests among the means. Results were considered to be significant at p values less than 0.05 (p < 0.05) and the significant differences between control and treated samples were indicated by letters. Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12870-023-04198-8. Page 18 of 20 Supplementary Material 1 Acknowledgements Not applicable. Author Contribution SR did all the experimental work and drafted the manuscript; NJ helped in proteomics work; RJ conceived the idea, supervised the work and edited the manuscript. All authors contributed to the article and approved the submitted version. Funding Authors acknowledge the funding support by Science Engineering and Research Board (CRG/2020/006169). Data Availability The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (http://www.ebi.ac.uk/pride) partner repository with the dataset identifier PXD036246 (project accession) and https://doi.org/10.6019/PXD036246 (project DOI). Note: The data is currently private, and can be accessed with reviewer account that has been created (Username: reviewer_pxd036246@ebi.ac.uk; Password: OYD2kgKf ). It is only after paper is accepted, the PRIDE is to be notified and data will be public. Declarations Ethics approval and consent to participate Gujarat-Maize-6 (GM6-native to Gujarat) and Gurez Local lines were obtained from Dry land Agriculture Research Station of Sheri Kashmir University of Agricultural Science and Technology (SKAUST), Srinagar and all the procedures for procurement of seeds were adhered to. Consent for publication Not applicable. Competing interests The authors declare that they have no conflicts of interest. Author details 1Plant Molecular Biology Lab, Department of Botany, University of Kashmir, Srinagar, Kashmir 190 006, India Received: 5 August 2022 / Accepted: 28 March 2023 References 1. Raza A, Razzaq A, Mehmood SS, Zou X, Zhang X, Lv Y, Xu J. Impact of climate change on crops adaptation and strategies to tackle its outcome: a review. Plants. 2019;8(2):34. Zeng R, Li Z, Shi Y, Fu D, Yin P, Cheng J, et al. Natural variation in a type- A response regulator confers maize chilling tolerance. Nat Commun. 2021;12(1):1–13. 2. 4. 5. 3. Waqas MA, Wang X, Zafar SA, Noor MA, Hussain HA, Azher Nawaz M, Farooq M. Thermal stresses in maize: effects and management strategies. Plants. 2021;10(2):293. Long SP, Ort DR. More than taking the heat: crops and global change. Curr Opin Plant Bio. 2010;13(3):240–7. Leipner J, Stamp P. Chilling stress in maize seedlings. Handbook of maize: its biology. 2009; (pp. 291–310).Springer, New York, NY. Sobkowiak A, Jończyk M, Adamczyk J, Szczepanik J, Solecka D, Kuciara I, Sowiński P. Molecular foundations of chilling-tolerance of modern maize. BMC Genomics. 2016;17(1):1–22. Ramazan S, Bhat HA, Zargar MA, Ahmad P, John R. Combined gas exchange characteristics, chlorophyll fluorescence and response curves as 6. 7. Ramazan et al. BMC Plant Biology (2023) 23:183 Page 19 of 20 selection traits for temperature tolerance in maize genotypes. Photo Res. 2021a;150(1):213–25. Lamari N, Zhendre V, Urrutia M, Bernillon S, Maucourt M, Deborde C, Moing A. Metabotyping of 30 maize hybrids under early-sowing conditions reveals potential marker-metabolites for breeding. Metabolomics. 2018;14(10):1–15. Salika, R., Riffat, J. Abiotic stress responses in maize: a review. Acta Physiol Plant 2021;43:130. https://doi.org/10.1007/s11738-021-03296-0 8. 9. 32. Bhusal B, Poudel MR, Rishav P, Regmi R, Neupane P, Bhattarai K, Acharya S. A review on abiotic stress resistance in maize (Zea mays L.): effects, resistance mechanisms and management. J Biol Todays World. 2021;10(2):1–3. 33. Hao W, Collier SM, Moffett P, Chai J. Structural basis for the interaction between the potato virus X resistance protein (rx) and its cofactor ran GTPase-activating protein 2 (RanGAP2). J Biol Chem. 2013;288(50):35868–76. 34. Owttrim GW. RNA helicases and abiotic stress. Nucleic acids Res. 10. Sowiński P, Fronk J, Jończyk M, Grzybowski M, Kowalec P, Sobkowiak A. Maize 2006;34(11):3220–30. response to low temperatures at the gene expression level: a critical survey of transcriptomic studies.Front Plant Sci. 2020;1488. 11. Yu T, Zhang J, Cao J, Cai Q, Li X, Sun Y, Duan Y. Leaf transcriptomic response mediated by cold stress in two maize inbred lines with contrasting tolerance levels. Genomics. 2021;113(2):782–94. 12. Li H, Yang M, Zhao C, Wang Y, Zhang R. Physiological and proteomic analyses revealed the response mechanisms of two different drought-resistant maize varieties. BMC Plant Biol. 2021;21(1):1–15. 13. Zhou G, Wang J, Zhang X, Guo M, Yu G. Predicting functions of maize proteins using graph convolutional network. BMC Bioinform. 2020;21(16):1–16. 14. Cheng L, Wang Y, He Q, Li H, Zhang X, Zhang F. Comparative proteomics illustrates the complexity of drought resistance mechanisms in two wheat (Triticum aestivum L.) cultivars under dehydration and rehydration. BMC Plant Biol. 2016;16(1):1–23. 15. Xu E, Chen M, He H, Zhan C, Cheng Y, Zhang H, Wang Z. Proteomic analysis reveals proteins involved in seed imbibition under salt stress in rice. Front Plant Sci. 2017;7:2006. 16. Méchin V, Balliau T, Château-Joubert S, Davanture M, Langella O, Négroni L, Damerval C. A two-dimensional proteome map of maize endosperm. Phyto- chem. 2004;65(11):1609–18. 17. Zhang XL, SI BW, Fan CM, LI HJ, Wang XM. Proteomics identification of differ- entially expressed leaf proteins in response to Setosphaeria turcica infection in resistant maize. J Integr Agric. 2014;13(4):789–803. 18. Dworak A, Nykiel M, Walczak B, Miazek A, Szworst-Łupina D, Zagdańska B, Kiełkiewicz M. Maize proteomic responses to separate or overlapping soil drought and two-spotted spider mite stresses. Planta. 2016;244(4):939–60. 19. Vidal N, Barbosa H, Jacob S, Arruda M. Comparative study of transgenic and non-transgenic maize (Zea mays) flours commercialized in Brazil, focusing on proteomic analyses. Food chem. 2015;180:288–94. 20. Subba P, Barua P, Kumar R, Datta A, Soni KK, Chakraborty S, Chakraborty N. Phosphoproteomic dynamics of chickpea (Cicer arietinum L.) reveals shared and distinct components of dehydration response. J Proteome Res. 2013;12(11):5025–47. 21. Kelley LA, Mezulis S, Yates CM, Wass MN, Sternberg MJ. The Phyre2 web portal for protein modeling, prediction and analysis. Nat. 2015;10(6):845–58. 22. Laskowski RA, Jabłońska J, Pravda L, Vařeková RS, Thornton JM, PDBsum. Structural summaries of PDB entries. Protein Sci. 2018;27(1):129–34. 23. Ge SX, Jung D, Yao R. ShinyGO: a graphical gene-set enrichment tool for animals and plants. Bioinform. 2020;36(8):2628–9. 24. Szklarczyk D, Gable AL, Lyon D, Junge A, Wyder S, Huerta-Cepas J, Mering CV. STRING v11: protein–protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic acids Res. 2019;47(D1):D607–13. 25. Gnanavel M, Mehrotra P, Rakshambikai R, Martin J, Srinivasan N, Bhaskara RM. CLAP: a web-server for automatic classification of proteins with special refer- ence to multi-domain proteins. BMC Bioinform. 2014;15(1):1–6. 26. Conesa A, Götz S, García-Gómez JM, Terol J, Talón M, Robles M. Blast2GO: a universal tool for annotation, visualization and analysis in functional genom- ics research. Bioinform. 2005;21(18):3674–6. 35. Kazan K, Lyons R. The link between flowering time and stress tolerance. J Exp Bot. 2006;67(1):47–60. 36. Vishwakarma K, Mishra M, Patil G, Mulkey S, Ramawat N, Pratap Singh V, Sharma S. Avenues of the membrane transport system in adaptation of plants to abiotic stresses. Crit Rev Biotechnol. 2019;39(7):861–83. 37. Dourmap C, Roque S, Morin A, Caubrière D, Kerdiles M, Béguin K, Couée I. Stress signalling dynamics of the mitochondrial electron transport chain and oxidative phosphorylation system in higher plants. Ann Bot. 2020;125(5):721–36. 38. Zeng R, Li Z, Shi Y, Fu D, Yin P, Cheng J, Yang. Natural variation in a type-A response regulator confers maize chilling tolerance. Nat Commun. 2021;12(1):1–13. 39. Dong A, Yang Y, Liu S, Zenda T, Liu X, Wang Y, Duan H. Comparative pro- teomics analysis of two maize hybrids revealed drought-stress tolerance mechanisms. Biotechnol Biotechnol Equip. 2020;34(1):763–80. 40. Gao F, Zhou Y, Zhu W, Li X, Fan L, Zhang G. Proteomic analysis of cold stress-responsive proteins in Thellungiella rosette leaves. Planta. 2009;230(5):1033–46. 41. Tian Y, Guan B, Zhou D, Yu J, Li G, Lou Y. Responses of seed germination, seedling growth, and seed yield traits to seed pretreatment in maize (Zea mays L.).Sci World J. 2014. 42. Maldonado A, Youssef R, McDonald M, Brewer E, Beard H, Matthews B. Overexpression of four Arabidopsis thaliana NHL genes in soybean (Glycine max) roots and their effect on resistance to the soybean cyst nematode (Heterodera glycines). Physiol Mol Plant Pathol. 2014;86:1–10. 43. Shultz RW, Tatineni VM, Hanley-Bowdoin L, Thompson WF. Genome-wide analysis of the core DNA replication machinery in the higher plants Arabi- dopsis and rice. Plant Physiol. 2007 Aug;144(4):1697–714. 44. Xu R, Ye X, Li QQ. AtCPSF73-II gene encoding an Arabidopsis homolog of CPSF 73 kDa subunit is critical for early embryo development.Gene. 2004 Jan7;324:35–45. 45. Dong NQ, Sun Y, Guo T, Shi CL, Zhang YM, Kan Y, Lin HX. UDP-glucosyl- transferase regulates grain size and abiotic stress tolerance associated with metabolic flux redirection in rice. Nat Commun. 2020;11(1):1–16. 46. Song K, Kim HC, Shin S, Kim KH, Moon JC, Kim JY, Lee BM. Transcriptome analysis of flowering time genes under drought stress in maize leaves. Front Plant Sci. 2017;8:267. 47. Zhu D, Luo F, Zou R, Liu J, Yan Y. Integrated physiological and chloroplast proteome analysis of wheat seedling leaves under salt and osmotic stresses. J Proteom. 2021;234:104097. 48. Schneeberger RG, Becraft PW, Hake S, Freeling M. Ectopic expression of the knox homeo box gene rough sheath1 alters cell fate in the maize leaf. Genes Dev. 1995;9(18):2292–304. 49. Yang G, Yu L, Wang Y, Wang C, Gao C. The translation initiation factor 1A (TheIF1A) from Tamarix hispida is regulated by a Dof transcription factor and increased abiotic stress tolerance. Front Plant Sci. 2017;8:513. 50. Rausell A, Kanhonou R, Yenush L, Serrano R, Ros R. The translation initiation factor eIF1A is an important determinant in the tolerance to NaCl stress in yeast and plants. Plant J. 2003;34(3):257–67. 27. Bäurle I, Dean C. The timing of developmental transitions in plants. Cell. 51. Shi S, Li S, Asim M, Mao J, Xu D, Ullah Z, Liu G, Wang Q, Liu H. The Arabidopsis 2006;125(4):655–64. 28. Fan T, Li X, Yang W, Xia K, Ouyang J, Zhang M. Rice osa-miR171c mediates phase change from vegetative to reproductive development and shoot api- cal meristem maintenance by repressing four OsHAM transcription factors. PLoS ONE. 2015;10(5):e0125833. 29. Strasser R, Seifert G, Doblin MS, Johnson KL, Ruprecht C, Pfrengle F, Estevez JM. Cracking the “Sugar Code”: A snapshot of N-and O-glycosylation path- ways and functions in plants cells.Front Plant Sci. 2021; 12. 30. Aslam M, Fakher B, Ashraf MA, Cheng Y, Wang B, Qin Y. Plant low-temperature stress: signaling and response. Agronomy. 2022;12(3):702. 31. Han M, Zhang C, Suglo P, Sun S, Wang M, Su T. L-Aspartate: an essen- tial metabolite for plant growth and stress acclimation. Molecules. 2021;26(7):1887. calcium-dependent protein kinases (CDPKs) and their roles in plant growth regulation and abiotic stress responses.International journal of molecular sciences. 2018 Jun28;19 (7):1900. 52. Du H, Shi Y, Li D, Fan W, Wang Y, Wang G, Wang C. Proteomics reveals key proteins participating in growth difference between fall dormant and non- dormant alfalfa in terminal buds. J Proteom. 2018;173:126–38. 53. Khan JA, Wang Q, Sjolund RD, Schulz A, Thompson GA. An early nodulin-like protein accumulates in the sieve element plasma membrane of Arabidopsis. Plant Physiol. 2007;143(4):1576–89. 54. Denancé N, Szurek B, Noël LD. (2014). Emerging functions of nodulin-like pro- teins in non-nodulating plant species. Plant Cell Physiol. 2014; 55(3): 469–474. 55. Trovato M, Funck D, Forlani G, Okumoto S, Amir R. Amino Acids in Plants: Regulation and Functions in Development and Stress Defense.Front Plant Sci. 2021; 12. Ramazan et al. BMC Plant Biology (2023) 23:183 Page 20 of 20 56. Matysiak K, Kierzek R, Siatkowski I, Kowalska J, Krawczyk R, Miziniak W. Effect 66. Yang Y, Wang W, Chu Z, Zhu JK, Zhang H. (2017). Roles of nuclear pores and of exogenous application of amino acids L-arginine and glycine on maize under temperature stress. Agronomy. 2020;10(6):769. nucleo-cytoplasmic trafficking in plant stress responses. Front Plant Sci. 2017; 8: 574. 57. Nejat N, Mantri N. Plant immune system: crosstalk between responses to 67. Xu XM, Meier I. The nuclear pore comes to the fore. Trends Plant Sci. biotic and abiotic stresses the missing link in understanding plant defence. Curr Issues Mol Biol. 2017 Jul;23(1):1–6. 2008;13(1):20–7. 68. Lee K, Kang H. Emerging roles of RNA-binding proteins in plant growth, 58. Zhang H, Zhu J, Gong Z, Zhu JK. Abiotic stress responses in plants. Nat Rev development, and stress responses. Mol cells. 2016;39(3):179. Genet. 2022 Feb;23(2):104–19. 59. Bouzroud S, Gouiaa S, Hu N, Bernadac A, Mila I, Bendaou N, Zouine M. Auxin response factors (ARFs) are potential mediators of auxin action in tomato response to biotic and abiotic stress (Solanum lycopersicum). PLoS ONE. 2018;13(2):e0193517. 60. Hu W, Zuo J, Hou X, Yan Y, Wei Y, Liu J, Jin Z. The auxin response factor gene family in banana: genome-wide identification and expression analyses dur- ing development, ripening, and abiotic stress. Front Plant Sci. 2015;6:742. 61. Liang X, Qin L, Liu P, Wang M, Ye H. Genes for iron–sulphur cluster assem- bly are targets of abiotic stress in rice, Oryza sativa. Plant Cell Environ. 2014;37(3):780–94. 69. Ovrutska II, Kordyum EL. PIP2; 1 aquaporin gene expression in maize hybrids different for drought tolerance to water deficit. Доповіді НАН Украïни; 2019. 70. Zhang B. Functional characterization of mitochondrial stress response in Arabidopsis thaliana (Doctoral dissertation, University of Western Australia). 2012. 71. Gutiérrez-Aguilar M. Mitochondrial calcium transport and permeability 72. transition as rational targets for plant protection. Biochim et Biophys Acta (BBA)-Bioenergetics. 2020;1861(12):148288. Jung JH, Seo PJ, Ahn JH, Park CM. Arabidopsis RNA-binding protein FCA regulates microRNA172 processing in thermosensory flowering. J Biol Chem. 2012;287(19):16007–16. 62. Rosado A, Schapire AL, Bressan RA, Harfouche AL, Hasegawa PM, Valpuesta V, Botella MA. The Arabidopsis tetratricopeptide repeat-containing protein TTL1 is required for osmotic stress responses and abscisic acid sensitivity. Plant Physiol. 2006;142(3):1113–26. 73. Gangola MP, Ramadoss BR. Sugars play a critical role in abiotic stress toler- ance in plants. Biochemical, physiological and molecular avenues for com- bating abiotic stress tolerance in plants. 2018; (pp. 17–38).Academic Press. 74. Payne SH. The utility of protein and mRNA correlation. Trends Biochem Sci. 63. Afzal Z, Howton TC, Sun Y, Mukhtar MS. The roles of aquaporins in plant stress 2015;40(1):1–3. responses. J Develop Biol. 2016;4(1):9. 64. López-Castillo LM, González-Leyzaola A, Diaz-Flores-Rivera MF, Winkler R, Wielsch N, García-Lara S. Modulation of aleurone peroxidases in kernels of insect-resistant maize (Zea mays L.; Pob84-C3R) after mechanical and insect damage.Front Plant Sci. 2020;781. 65. Kosová K, Vítámvás P, Prášil IT. Proteomics of stress responses in wheat and barley—search for potential protein markers of stress tolerance. Front Plant Sci. 2014;5:711. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Ramazan et al. BMC Plant Biology (2023) 23:183
10.1186_s12913-021-06331-5	Bousmah et al. BMC Health Services Research (2021) 21:313 https://doi.org/10.1186/s12913-021-06331-5 R E S E A R C H A R T I C L E Open Access Free access to antiretroviral treatment and protection against the risk of catastrophic health expenditure in people living with HIV: evidence from Cameroon Marwân-al-Qays Bousmah1,2* , Marie Libérée Nishimwe1 , Christopher Kuaban3 and Sylvie Boyer1 Abstract Background: To foster access to care and reduce the burden of health expenditures on people living with HIV (PLHIV), several sub-Saharan African countries, including Cameroon, have adopted a policy of removing HIV-related fees, especially for antiretroviral treatment (ART). We investigate the impact of Cameroon’s free antiretroviral treatment (ART) policy, enacted in May 2007, on catastrophic health expenditure (CHE) risk according to socioeconomic status, in PLHIV enrolled in the country’s treatment access program. Methods: Based on primary data from two cross-sectional surveys of PLHIV outpatients in 2006–2007 and 2014 (i.e., before and after the policy’s implementation, respectively), we used inverse propensity score weighting to reduce covariate imbalances between participants in both surveys, combined with probit regressions of CHE incidence. The analysis included participants treated with ART in one of the 11 HIV services common to both surveys (n = 1275). Results: The free ART policy was associated with a significantly lower risk of CHE only in the poorest PLHIV while no significant effect was found in lower-middle or upper socioeconomic status PLHIV. Unexpectedly, the risk of CHE was higher in those with middle socioeconomic status after the policy’s implementation. Conclusions: Our findings suggest that Cameroon’s free ART policy is pro-poor. As it only benefitted PLHIV with the lowest socioeconomic status, increased comprehensive HIV care coverage is needed to substantially reduce the risk of CHE and the associated risk of impoverishment for all PLHIV. Keywords: HIV, Catastrophic health expenditure, Costs, Treatment, Free antiretroviral treatment, Policy evaluation, Poverty, Cameroon * Correspondence: marwan-al-qays.bousmah@ird.fr 1Aix Marseille University, INSERM, IRD, SESSTIM, Sciences Economiques & Sociales de la Santé & Traitement de l’Information Médicale, Marseille, France 2Centre Population et Développement (Ceped), Institut de recherche pour le développement (IRD) & Université de Paris, Inserm ERL 1244, 45 rue des Saints-Pères, 75006 Paris, France Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Bousmah et al. BMC Health Services Research (2021) 21:313 Page 2 of 7 Introduction Universal health coverage is a major target for health re- form in many countries and a priority of the Sustainable Development Goals. The development of health insur- ance schemes and the removal of user fees for specific populations and/or key health services are considered core actions to achieve this target [1]. People living with HIV (PLHIV) constitute an espe- cially vulnerable population as they bear particularly large disease-related health expenses which may severely affect their household welfare [2]. Catastrophic health expenditure (CHE) - which assess the financial hardship caused by out-of-pocket payments for health on house- hold welfare - is very high in this population, although some studies suggest that access to antiretroviral treat- ment (ART) and longer time on ART may decrease its risk [3, 4]. To foster access to care and reduce the bur- den of health expenditures on PLHIV, several sub- Saharan African countries, including Cameroon, have adopted a policy of removing ART-related fees. How- ever, evidence for this policy’s success is limited. In par- ticular, little is known about the effects of free ART on CHE according to PLHIV socioeconomic status, and es- pecially on the financial risks faced by the poorest groups in this vulnerable population. This information is crucial for decision makers, to identify the strengths and weaknesses of current policy and tailor future policy accordingly. The Cameroonian health system is mainly financed by private health expenditures through out-of-pocket pay- ments, the latter accounting for 71% of current health expenditures in 2017 [5]. Following a recent multi- country study which estimated CHE incidence in 2014 in Cameroon’s general population at 3% - using the common thresholds of 10% of total consumption or 40% of non- food consumption - the country was classified in the sec- ond highest quintile of CHE incidence worldwide [6]. Furthermore, Cameroon is one of the countries most affected by the HIV epidemic in West and Central Af- rica, with a prevalence estimated at 3.6% in 2018 in adults aged 15–49 years [7]. In 2001, the government launched a national treatment access program based on a strategy of HIV care decentralization, whereby ART was delivered at the district level [8]. This provided healthcare facilities a relatively large degree of autonomy in terms of the design of user fee schemes, resulting in price differentiation in healthcare facilities across the country. However, ART prices were fixed and were sub- sidized by the government [9]. A study conducted in 2006–2007 showed that, overall, decentralization had a protective effect on the risk of CHE in PLHIV [9]. How- ever, 39.6% of this population risked CHE at the thresh- old of 20% of household income, with an even higher the individuals. risk for the poorest In May 2007, national authorities decided to provide ART for free to facilitate access to treatment and remove financial bar- riers. However, user fees continued to apply to all other including consultations and biological HIV services, tests. This study aimed to explore the differential effects of the free ART policy on the risk of CHE according to so- cioeconomic status in PLHIV enrolled in Cameroon’s ART access program. Data Study design We used data from two cross-sectional surveys con- ducted in 2006–2007 and 2014 in PLHIV outpatients at- tending HIV services participating in the Cameroonian ART access program. The first survey, entitled EVAL (ANRS-12116), included 3151 HIV-positive outpatients consulting in 27 HIV services in 6 of the country’s 10 re- gions [8]. The second survey, entitled EVOLCam (ANRS-12288), included 2130 participants consulting in 19 HIV services in the Centre and Littoral regions. Eleven of the latter also participated in the 2006–2007 survey [10]. In order to study evolutions in the program over time, both surveys used the same design and proce- dures, which are described in detail elsewhere [11]. Briefly, outpatients diagnosed HIV-positive at least 3 months previously and aged ≥21 years were randomly selected for inclusion. All participants provided written consent before data collection. They answered a face-to- face questionnaire documenting socioeconomic charac- teristics, household consumption, healthcare use and re- lated expenditures. Blood samples were also collected and analyzed in a reference laboratory in Yaoundé for viral load and CD4 cell count measurements. Clinical data were obtained from both clinical examination and retrospective medical files using a standardized medical questionnaire. The EVAL (ANRS-12116) and the EVOLCam (ANRS- 12288) surveys were both approved by the Ministry of Public Health in Cameroon and the Cameroonian Na- tional Ethics Committee. Study population The study population for the present analysis included participants treated with ART for at least 1 month either in 2006–2007 (EVAL survey) or in 2014 (EVOLCam sur- vey) in one of the 11 HIV services participating in both surveys i.e., 615 and 660 participants, respectively). (n = 1275, Variables The outcome of interest was a binary variable measuring CHE incidence. Using a standard method, CHE was de- fined as out-of-pocket payments for health accounting Bousmah et al. BMC Health Services Research (2021) 21:313 Page 3 of 7 for 40% or more of a household’s capacity to pay (i.e., the income remaining after subsistence needs are met) [12]. The following socioeconomic and clinical variables were included in the analysis as determinants of CHE: wealth level (which depending on the model was assessed either by i) the log of the monthly equivalized consumption expenditure (continuous variable) or ii) the monthly equivalized consumption expenditure in deciles (categorical variable)), age, gender, level of formal educa- tion, marital status, being the head of the household, having an economic activity, time (in months) since ART initiation and CD4 cell count at the time of the survey (< 200 versus ≥ 200cells/mm3). We also included i) a decentralization variable to indicate each HIV ser- vice’s decentralization level (i.e., central or district) and ii) the survey period (2006–2007, i.e., before the free ART policy’s implementation versus 2014, i.e., after its implementation). A full description of the variables is presented in Supplementary Table S1. Methods We implemented an inverse probability of treatment weighting (IPTW) technique in order to reduce covariate imbalances between ‘treated’ and ‘untreated’ individuals (i.e., participants in the 2014 survey who benefited from the free ART policy and participants in the 2006–2007 survey who did not, respectively). This methodology, which is part of the general framework developed by Rosenbaum & Rubin [13] for estimating average treat- ment effects (ATE), allowed us to reduce any possible bias when estimating the policy’s causal effect on CHE. IPTW is widely used for estimating ATE in the presence of selection bias related to sample selection or stratifica- tion based on covariates [14]. More specifically, it con- sists in weighting the ‘treated’ group by the inverse of the propensity score. The rationale behind using the in- verse propensity score lies in the so-called “double ro- bustness” result, implying that the ATE is consistent if at least one of the conditional mean functions of the re- sponse or the propensity score model is correctly speci- fied [15]. First, we estimated the propensity score of the treat- ment (i.e. benefitting from the policy) using a probit model including the explanatory variables listed in the Variables subsection above. Balance properties were ana- lyzed using a standard approach [16]. We then estimated a probit model weighted by the inverse probability of treatment to identify the determinants of CHE and more specifically the effect of the free ART policy. The latter analysis was restricted to the region of common support. Finally, two alternative model specifications were consid- ered to investigate the impact of the policy on the risk of CHE according to socioeconomic status. The first speci- equivalized fication continuous included the consumption expenditure variable and an interaction of this variable with the free ART policy variable, to test whether the policy had pro-poor effects (i.e., decreased the probability of CHE more in poorer individuals than in others). The second specification included the equiva- lized consumption expenditure variable in deciles and an interaction of this variable with the free ART policy vari- able, to test whether the effect of the policy differed ac- cording to socioeconomic status. As the data source consisted of two cross-sectional surveys, one before and one after the policy’s implemen- tation in 2007, an ‘omitted variable’ bias due to poten- tially unobserved factors cannot be ruled out. To limit this bias, we adjusted the different models for the level of HIV service decentralization, as a previous study con- ducted in the first phase of Cameroon’s ART access pro- gram showed that linkage to care at the district level was associated with a lower risk of CHE [9]. Furthermore, as the policy may have influenced CHE in ways other than direct cost reduction - for instance, by improving access to care which in turn resulted in fewer adverse health events - we also controlled for health-related variables (e.g., CD4 cell count at the time of the survey). Results Main characteristics of the study population according to each survey period CHE incidence was 22% in 2006–2007 but only 15% in 2014 (see Supplementary Table S2). The mean monthly equivalized consumption expenditure was very similar in both periods (US$91 in 2006–2007 and US$90 in 2014). Mean age was 39 and 42 years in 2006–2007 and 2014, respectively. Finally, women accounted for 68 and 71% of the study population, respectively. Inverse probability of treatment weighting ATE estimation Ninety-eight participants of the total 1275 participants were outside the region of common support and thus were excluded from the current analysis. The balancing property of the propensity score was satisfied (see Sup- plementary Table S3). Details on the inverse probability weighting estimation of the ATE are provided in Supple- mentary Table S4 and balancing results in Supplemen- tary Table S5. Results from the overidentification test for covariate balance (p = 0.41) showed that most of the bias between ‘treated’ and ‘untreated’ individuals was re- moved. Our methodological choice was justified, as ATE estimation without addressing covariate imbalances be- tween ‘treated’ and untreated” individuals yielded a sig- nificant but biased ATE of − 0.0509 (p = 0.047). Using the IPTW estimator, the ATE of the free ART policy on CHE was not significant. Bousmah et al. BMC Health Services Research (2021) 21:313 Page 4 of 7 Effect of the free ART policy according to socioeconomic groups Results are displayed in Fig. 1 and full regression results are provided in Supplementary Table S6. Panel 1.A shows the predicted CHE probabilities for the ‘untreated’ and ‘treated’ groups obtained in the first regression, with solid lines representing significant group differences (p < 0.05). Average marginal effects of the policy across the wealth distribution are displayed in Panel 1.B (negative effects for a specific consumption expenditure indicate a lower prob- ability of CHE at that level). Figure 1 also shows that the policy had a significant negative average marginal effect on CHE for relatively low levels of wealth, specifically for values of the log monthly equivalized consumption ex- penditure between US$6/month and US$50/month. These amounts were below Cameroon’s official minimum wage in 2014 (US$62/month) [17]. The results obtained in the second regression are illus- trated in Panels 1.C and 1.D which show, for each wealth decile, the predicted probabilities of CHE for both the ‘untreated’ and ‘treated’ groups and the average marginal effects of the free ART policy, respectively. Detailed results, including predicted probabilities, average marginal effects, and tests of the equality of marginal effects across wealth deciles, are summarized in Table 1. Results obtained in the first regression were confirmed. More specifically, the free ART policy signifi- cantly reduced the probability of CHE by 23.7 percent- age points (from 37.2 to 13.5%) and 10.8 percentage points (from 22.3 to 11.5%) in individuals in the lowest respectively. and second lowest wealth deciles, In addition, results showed a significant positive average marginal effect of the policy - a 7.4 percentage point in- crease in CHE probability - in individuals in the fifth wealth decile. No significant effect was found for the other wealth deciles. Discussion Our findings suggest that the free ART policy intro- duced in Cameroon in 2007 had a different impact on the risk of CHE depending on PLHIV socioeconomic status. More precisely, PLHIV in the lowest and second- lowest wealth deciles benefited from a 23.7 and 10.8 per- centage point decrease in CHE risk, respectively. Ac- cordingly, the policy was pro-poor. Conversely, PLHIV in the fifth wealth decile had a 7.4 percentage point in- creased CHE risk, while for all other wealth deciles the risk of CHE remained relatively stable before and after the introduction of the policy. These findings might be explained by the fact that concurrently with the policy’s adoption, out-of-pocket spending for certain health goods and services increased. Our data showed that ex- penditures for biological tests almost doubled between 2006 and 2007 and 2014 (p < 0.001), while other spend- ing categories (consultation fees, hospital bills, other medications, and traditional medicine) remained stable (see Supplementary Table S2). However, this increase (whether demand- or supply-driven) was mostly ob- group. The served in the middle socioeconomic Fig. 1 Predicted probabilities of catastrophic health expenditure and average marginal effects of the free ARV policy across wealth levels Bousmah et al. BMC Health Services Research (2021) 21:313 Page 5 of 7 Table 1 Probability of facing catastrophic health expenditure: average marginal effects of the free ARV program across wealth deciles Untreated Treated 1st decile 2nd decile 3rd decile 4th decile 5th decile 6th decile 7th decile 8th decile 9th decile 10th decile 0.372 0.223 0.227 0.155 0.090 0.177 0.171 0.145 0.118 0.147 0.135 0.115 0.106 0.125 0.164 0.239 0.187 0.161 0.179 0.135 Average marginal effects of the program −0.237* −0.108* −0.121 −0.030 0.074* 0.062 0.016 0.016 0.061 −0.012 Tests for the equality of marginal effects 4th–10th 5th–6th, 9th 5th–6th 1st 1st-3rd 1st-3rd 1st 1st 1st-2nd 1st The tests for the equality of marginal effects report which marginal effects are significantly different (p < 0.05, two-tailed tests) across wealth deciles *p < 0.05, two-tailed tests significant increase in CHE risk observed in PLHIV in the fifth wealth decile may also be related to a phenomenon revealed by Wagstaff & Lindelow [18] in China. In that study, the authors highlighted that health insurance substantially increased the risk of high and catastrophic health spending, as individuals received more sophisticated and expensive medical care once insured. Our findings have important implications for HIV healthcare policy. They suggest that removing ART fees is a pro-poor measure. CHE incidence significantly re- duced in poorer PLHIV in Cameroon between the two study time points. Although much higher in this popula- tion in 2006–2007, it was similar or even lower to that observed in wealthier PLHIV in 2014. However, the free ART policy alone seemed to be insufficient to effectively protect all PLHIV against the financial risk related to their infection, as suggested by the continued high level of CHE incidence in 2014 across all socioeconomic groups (from 11.5% for the second lowest decile to 23.9% for the 6th decile). Increased comprehensive HIV care coverage is necessary to significantly lower the risk of CHE and related impoverishment in the PLHIV popu- lation in Cameroon and elsewhere. In 2019, the coun- try’s health authorities took an important step in this direction by providing free comprehensive care, includ- ing consultations, biological tests and prophylaxis drugs, to all PLHIV [19]. Further research is needed to inform policymakers about the impact of the roll-out of the free HIV care policy on socioeconomic inequalities. The removal of user fees for HIV care, initiated in Cameroon in 2007 with the adoption of free ART and pursued in 2019 with the removal of user fees for all other HIV services, raised the question of the funding of these measures and, in turn, of the budget reforms required to compensate providers for lost fee revenue [20]. Not com- pensating for this loss of revenue may impede the quality of care and lead to adverse effects such as drug shortages [21], which may result in increasing access to private and costly care and, in turn, increase out-of-pocket payments for health care. However, some studies suggested that the increase of public health expenditures required to fund free care may be not financially sustainable for national governments in the long term [22]. This issue is part of a more general discussion on the financing of UHC policies in developing countries. A recent study suggested that a fi- nancing of UHC through consumption taxation was the best policy option, in terms of both fiscal sustainability and intergenerational equity [23]. Our main study limitation is the risk of ‘omitted vari- able’ bias in the estimation of the free ART policy effect. Although we used the IPTW method to address issues re- lated to sample selection bias and stratification, and ad- justed the models for other factors that may have influenced CHE, we cannot fully exclude it. We therefore acknowledge that the estimated differences presented here may not be entirely attributable to the free ART policy. In conclusion, Cameroon’s free ART policy seems to have significantly reduced CHE incidence for the poorest PLHIV after its introduction in 2007. The constant, and in some cases, increased spending for certain healthcare costs items such as biological tests, might explain why other socioeconomic groups did not benefit from this policy and remained at a high risk of CHE. Increased comprehensive coverage of HIV care is needed to sub- stantially reduce the risk of CHE and associated impoverishment. Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12913-021-06331-5. Additional file 1: Table S1. Description of variables. Table S2. Descriptive statistics according to survey year. Table S3. Propensity score Bousmah et al. BMC Health Services Research (2021) 21:313 Page 6 of 7 model: inferior bounds and number of untreated and treated individuals in each block. Table S4. Average treatment effect of the free ART program on catastrophic health expenditure. Table S5. Inverse probability of treatment weighting: covariate balance. Table S6. Results of the inverse probability of treatment weighting probit model for catastrophic health expenditure. Acknowledgments We are grateful to all study participants, to the staff at the HIV services involved, and to the members of the ANRS 12288 EVOLCam study group. We also thank the French Agency for Research on AIDS and Viral Hepatitis (ANRS) for its support to the study, Gwenaëlle Maradan for the monitoring of the data collection, and Jude Sweeney for the English revision and editing of the manuscript. Code availability Available from the authors upon reasonable request. Authors’ contributions CK was one of the principal investigators of the EVOLCam project. SB was responsible for the overall economic study lead. MQB and MLN performed the econometric analysis. The manuscript was drafted by MQB and SB. All authors made critical comments on the manuscript draft, approved the final version of the manuscript for submission and agreed to be responsible for all aspects of the work. Funding The EVAL (ANRS-12116) and EVOLCam (ANRS-12288) surveys were financed by the French Agency for Research on AIDS and Viral Hepatitis (ANRS). Availability of data and materials The manuscript has data included as electronic supplementary material. More complete data is available from the authors upon request (contact: Marwân-al-Qays Bousmah. CEPED (UMR 196), Université de Paris, Campus Saint-Germain, 45 Rue des Saints-Pères, 75006 Paris, France. Tel.: + 33643521166. E-mail: marwan-al-qays.bousmah@ird.fr). Declarations Ethics approval and consent to participate The EVAL (ANRS-12116) and EVOLCam (ANRS-12288) surveys were approved by the Ministry of Public Health in Cameroon and the Cameroonian National Ethics Committee. All methods were carried out in accordance with relevant guidelines and regulations. All experimental protocols were approved by the Ministry of Public Health in Cameroon and the Cameroonian National Ethics Committee. Informed consent was obtained from all subjects or, if subjects were under 18, from a parent and/or legal guardian. Consent for publication Not applicable. Competing interests We declare that we have no conflict of interest. Author details 1Aix Marseille University, INSERM, IRD, SESSTIM, Sciences Economiques & Sociales de la Santé & Traitement de l’Information Médicale, Marseille, France. 2Centre Population et Développement (Ceped), Institut de recherche pour le développement (IRD) & Université de Paris, Inserm ERL 1244, 45 rue des Saints-Pères, 75006 Paris, France. 3Faculty of Medicine and Biomedical Sciences, University of Yaoundé, Yaoundé, Cameroon. Received: 3 December 2020 Accepted: 24 March 2021 References 1. McIntyre D, Obse AG, Barasa EW, Ataguba JE. Challenges in financing universal health coverage in sub-Saharan Africa. In: Oxford research encyclopedia of economics and finance. New York: Oxford University Press USA; 2018. https://doi.org/10.1093/acrefore/9780190625979.013.28. 2. Whiteside A. Poverty and HIV/AIDS in Africa. Third World Q. 2002;23(2):313– 3. 32. https://doi.org/10.1080/01436590220126667. Beaulière A, Touré S, Alexandre P-K, Koné K, Pouhé A, Kouadio B, et al. The financial burden of morbidity in HIV-infected adults on antiretroviral therapy in Côte d’Ivoire. PLoS One. 2010;5(6):e11213. https://doi.org/10.1371/journal. pone.0011213. Dickerson S, Baranov V, Bor J, Barofsky J. Treatment as insurance: HIV antiretroviral therapy offers financial risk protection in Malawi. Health Policy Plan. 2020;35(6):676–83. https://doi.org/10.1093/heapol/czaa023. 5. WHO. Global Health Observatory data repository. Geneva: World Health 4. Organization; 2020. https://www.who.int/data/gho 6. Wagstaff A, Flores G, Hsu J, Smitz M-F, Chepynoga K, Buisman LR, et al. 8. 7. Progress on catastrophic health spending in 133 countries: a retrospective observational study. Lancet Glob Health. 2018;6(2):e169–79. https://doi.org/1 0.1016/S2214-109X(17)30429-1. UNAIDS. UNAIDS data 2019. Geneva: Joint United Nations Programme on HIV and AIDS; 2019. https://www.unaids.org/en/resources/documents/201 9/2019-UNAIDS-data Boyer S, Eboko F, Camara M, Abé C, Nguini MEO, Koulla-Shiro S, et al. Scaling up access to antiretroviral treatment for HIV infection: the impact of decentralization of healthcare delivery in Cameroon. AIDS. 2010;24(Suppl 1): S5–15. https://doi.org/10.1097/01.aids.0000366078.45451.46. Boyer S, Abu-Zaineh M, Blanche J, Loubière S, Bonono R-C, Moatti J-P, et al. Does HIV services decentralization protect against the risk of catastrophic health expenditures?: some lessons from Cameroon. Health Serv Res. 2011; 46(6pt2):2029–56. https://doi.org/10.1111/j.1475-6773.2011.01312.x. Tong C, Suzan-Monti M, Sagaon-Teyssier L, Mimi M, Laurent C, Maradan G, et al. Treatment interruption in HIV-positive patients followed up in Cameroon’s antiretroviral treatment programme: individual and health care supply-related factors (ANRS-12288 EVOLCam survey). Tropical Med Int Health. 2018;23(3):315–26. https://doi.org/10.1111/tmi.13030. 11. Boyer S, Clerc I, Bonono C-R, Marcellin F, Bilé P-C, Ventelou B. Non- 10. 9. adherence to antiretroviral treatment and unplanned treatment interruption among people living with HIV/AIDS in Cameroon: individual and healthcare supply-related factors. Soc Sci Med. 2011;72(8):1383–92. https://doi.org/10.1 016/j.socscimed.2011.02.030. 12. Xu K, Evans DB, Kawabata K, Zeramdini R, Klavus J, Murray CJL. Household catastrophic health expenditure: a multicountry analysis. Lancet. 2003; 362(9378):111–7. https://doi.org/10.1016/S0140-6736(03)13861-5. 13. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41–55. https:// doi.org/10.1093/biomet/70.1.41. 14. Wooldridge JM. Inverse probability weighted estimation for general missing data problems. J Econ. 2007;141(2):1281–301. https://doi.org/10.1016/j. jeconom.2007.02.002. 15. Robins JM, Rotnitzky A. Semiparametric efficiency in multivariate regression models with missing data. J Am Stat Assoc. 1995;90(429):122–9. https://doi. org/10.1080/01621459.1995.10476494. Imai K, Ratkovic M. Covariate balancing propensity score. J R Statist Soc B. 2014;76(1):243–63. https://doi.org/10.1111/rssb.12027. 16. 17. Ministère du Travail et de la Sécurité Sociale. Décret n° 2014/2217/PM du 24 juillet 2014 portant revalorisation du salaire minimum interprofessionnel garanti (SMIG). Yaoundé, Cameroon: Ministère du Travail et de la Sécurité Sociale; 2014. http://ilo.org/dyn/natlex/natlex4.detail?p_lang=fr&p_isn=9692 5. Accessed 11 June 2020 18. Wagstaff A, Lindelow M. Can insurance increase financial risk? The curious case of health insurance in China. J Health Econ. 2008;27(4):990–1005. https://doi.org/10.1016/j.jhealeco.2008.02.002. 19. Ministère de la Santé Publique. Décision n°0498 D/MINSANTE/SG/CNLS/ GTC/SP du 04 avril 2019 fixant les modalités d’accès et de suivi des populations aux services de dépistage et prise en charge du VIH dans les formations sanitaires publiques et leur organisations à base communautaire affiliées. Yaoundé, Cameroon: Ministère de la Santé Publique; 2019. James CD, Hanson K, McPake B, Balabanova D, Gwatkin D, Hopwood I, et al. To retain or remove user fees? Appl Health Econ Health Policy. 2006;5(3): 137–53. https://doi.org/10.2165/00148365-200605030-00001. 20. 21. Ridde V, Robert E, Meessen B. A literature review of the disruptive effects of user fee exemption policies on health systems. BMC Public Health. 2012; 12(1):289. https://doi.org/10.1186/1471-2458-12-289. 22. Awawda S, Abu-Zaineh M. An operationalizing theoretical framework for the analysis of universal health coverage reforms: first test on an archetype Bousmah et al. BMC Health Services Research (2021) 21:313 Page 7 of 7 developing economy. AMSE Working Papers. 2019:WP 2019-Nr 04 https://ha lshs.archives-ouvertes.fr/halshs-02009858/document. Accessed 4 Mar 2021. 23. Abu-Zaineh M, Awawda S, Ventelou B. Who bears the burden of universal health coverage? An assessment of alternative financing policies using an overlapping-generations general equilibrium model. Health Policy Plan. 2020;35(7):867–77. https://doi.org/10.1093/heapol/czaa041. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12902-019-0370-7	Bahri et al. BMC Endocrine Disorders (2019) 19:47 https://doi.org/10.1186/s12902-019-0370-7 R E S E A R C H A R T I C L E Open Access Overtime trend of thyroid hormones and thyroid autoimmunity and ovarian reserve: a longitudinal population study with a 12- year follow up Sara Bahri1, Fahimeh Ramezani Tehrani1* Maryam Vasheghani5 and Fereidoun Azizi6 , Atieh Amouzgar2, Maryam Rahmati3, Maryam Tohidi4, Abstract Background: Ovarian reserve, vital for reproductive function, can be adversely affected by thyroid diseases. Despite alternations of thyroid hormones with ageing, data on interactions between the overtime trend of thyroid functions and ovarian reserve status has rarely been reported. We aimed to examine the overtime trend of thyroid hormones, thyroid peroxidase antibody (TPO Ab) and their associations with ovarian reserve status, identified by levels of age specific anti-mullerian hormone (AMH) in reproductive aged women, who participated in 12-year cohort of Tehran Thyroid Study (TTS). Methods: Reproductive age women(n = 775) without any thyroid disease or ovarian dysfunction were selected from the Tehran Thyroid Study cohort. Participants were divided into four age specific AMH quartiles (Q1-Q4), Q1, the lowest and Q4, the highest. AMH was measured at the initiation of study and thyroid stimulating hormone (TSH), free T4 (FT4), and TPO Ab were measured at baseline and at three follow up visits. Results: At baseline, there was no statistically significant difference in thyroid hormones between women of the four quartiles, although TPO Ab levels were higher in women of Q1. During the follow ups, FT4 was decreased in all quartiles (p < 0.05), whereas TPO Ab increased in Q1 (p = 0.02). Odds ratio of overall TPO Ab positivity in women of Q1 was 2.08 fold higher than those in Q4. (OR: 2.08, 95%CI: 1.16, 3.72; p = 0.01). Conclusion: Women with the lowest ovarian reserves had higher levels of TPO Ab, with a positive trend of this antibody overtime in comparison to other quartiles, indicating that this group may be at a higher risk of hypothyroidism over time. Keywords: Anti-mullerian hormone, Ovarian reserve, Tehran thyroid study (TTS), Thyroid autoimmunity, Thyroid hormones Background Ovarian reserve is vital for reproductive function, and women with reduced ovarian follicles are at an increased risk of premature ovarian failure (POF) [1]. The status of ovarian by anti-Müllerian hormone (AMH), which is secreted by the precisely measured reserve can be * Correspondence: fah.tehrani@gmail.com 1Reproductive Endocrinology Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran Full list of author information is available at the end of the article granulosa cells of ovarian follicles [2]. Among the various factors assumed to be associated with POF [2, 3], thyroid function status and thyroid autoimmunity have in particu- lar, been suggested [4]. Thyroid dysfunction influences the reproductive system directly by affecting oocytes, via thy- roid hormone receptors on the surface of these cells [4], and indirectly through increasing prolactin secretion and disruption of GnRH function [5]. In addition it has been shown that thyroid autoimmune disease may be associ- ated with a general autoimmunity, leading to premature ovarian failure; however the main pathophysiology linking © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Bahri et al. BMC Endocrine Disorders (2019) 19:47 Page 2 of 9 ovarian reserves with TPO antibody and thyroid hor- mones has not been elucidated [6–8]. In a cross sectional study of 5000 Belgian women (mean age: 32.0, SD: 5.5 years), there were no significant association, between ovar- ian reserve, thyroid hormones and thyroid autoimmunity [9]. Several studies also show an inverse association between ovarian reserves, hypothyroidism and thyroid autoimmunity in humans and animals [10, 11]; however to the best of our knowledge, the interaction between changes in thyroid functions and ovarian reserves status has not yet been reported and it is unclear whether or not women with poor ovarian reserves are at increased risk of occurrence of hypothyroidism. We aimed to assess the trend of changes in thyroid hormones and TPO Ab among reproductive aged women by various ovarian reserve statuses, identified by age- specific AMH levels in a population based cohort study, the Tehran Thyroid Study, with an average 12 years of follow up. Methods Study population We used data from the participants of the Tehran Thy- roid Study (TTS) [12], a population-based cohort study being conducted within the framework of the Tehran Lipid and Glucose Study (TLGS) [13]. In the TLGS, 15005 people, aged ≥3 years, living in district No. 13 of Tehran, were selected by a multistage cluster random sampling method; Detailed descriptions of the TLGS have been published elsewhere [14]; of these 4174 indi- viduals (male and female), aged ≥20 were included in the Tehran Thyroid Study. The TTS aimed at determining the prevalence and natural courses of thyroid disease, and to demonstrate the relationship between thyroid disease and cardiovas- cular risk factors, ischemic heart disease, and all-cause mortality in the population of Tehran. In the current prospective study, participants were followed every 3 years interval over a 12 year follow up. Data on smoking status, history of radioiodine expos- ure, physical activity levels, history of thyroid surgery, and using levothyroxine or anti-thyroid medications were obtained during face to face interviews every 3 years by trained staff. At each follow up, participants were also asked about their reproductive history, includ- ing marital status, regularity of menstrual cycle, parity, and the current and previous contraceptive methods used. If their menstrual cycles had ceased, the date of the last cycle was recorded. With exception of serum AMH, all biochemical variables including thyroid hor- mones, TPO Ab and TSH were measured at each follow-up, every 3 years. For the purpose of present study, we selected women, aged 20–50 years at baseline who met eligibility criteria, which included: Having regular and predictable men- struation at initiation of study and documented natural fertility; exclusion criteria were: history of endometriosis, Poly Cystic Ovarian Syndrome (PCOS), hysterectomy, oophorectomy or any other kind of ovarian surgery, using hormonal replacement for menopausal symptoms at the initiation of the study or during follow-up; and usage of hormonal contraception for ≥3 months before entering the study. Those with history of thyroidectomy, radio iodine treatment, and those using of levothyroxine, amiodarone, glucocorticoids or anti thyroid dugs were also excluded. Participants were divided into four age specific AMH quartiles (Q1-Q4); (Q1: the lowest and Q4: the highest). Women included had to have complete data and hormonal profiles for all follow-ups. All partici- pants signed written informed consent forms and the study was approved by the ethics committee of the Re- search Institute for Endocrine Sciences, RIES affiliated, to the Shahid Beheshti University of Medical Sciences. IR ENDOCRINE,REC.1396.416. Measurements All measurements were conducted at baseline and each of the three follow ups. Anthropometric measures in- cluding weight and waist circumference (WC), were measured at the umbilicus level using an unstretched tape meter, without any pressure to body surface; Body Mass Index (BMI) was calculated as weight in kilograms (kg) divided by height in squared (m2). Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured twice using a standard mercury sphygmoman- ometer, in a seated position after a 15-min rest period, and the mean of these two measurement was considered as the subject’s blood pressure. Blood samples were taken from the subjects between 7:00 AM and 9:00 AM while sitting, using vacutainer tubes after a 12-h overnight fast. Samples were centri- fuged within 30–45 min of collection and stored at -80C. At baseline the serum concentration of anti-Müllerian hormone (AMH) was measured by the two-site enzyme immunoassay (EIA) method; the kit was the Gen II kit Inc., CA, USA) and intra- and (Beckman Coulter, inter-assay CVs were 1.9 and 2.0%, respectively and the reader was Sunrise ELISA reader (Tecan Co. Salzburg, Austria). All AMH measurements were taken simultan- eously at the same laboratory. AMH Gen II controls A79766 were used at two levels of concentration to monitor accuracy of the assay. Serum TSH and free T4 concentrations were measured by the electro the chemi luminescence immunoassay (ECLIA) method, using commercial kits on the Cobase e - 411 analyzer (Roche Diagnostics GmbH, Mannheim, Germany). The inter- and intra-assay CVs were 3.3 and 1.3% for free T4 and respectively. 4.6 and 1.4% for TSH determination, Bahri et al. BMC Endocrine Disorders (2019) 19:47 Page 3 of 9 Sensitivities of free T4 and TSH measurement were 0.28 pmol/L and 0.005 mU/L, respectively. The reference values for TSH were (0.32–5.06) mU/L and the refer- ence values for T4 were (11.58–19.95) pmol/L. TPOAb was assayed by immune enzymometric assay (IEMA-- Monobind, Costa Mesa, CA, USA), using the Sunrise ELISA reader (Tecan Co., Salzburg, Austria); inter- and intra-assay CVs were 4.2 and 3.9%, respectively. The sen- sitivity of the assay was 0.92 IU/mL. Serum AMH level was measured one time at the beginning of study, while all other biochemical measurements were conducted at baseline and also at each of the three follow ups. Definitions According to normal-based methodology of Altman and Chitty [15] and Royston and Wright [16], age-specific AMH percentiles were estimated based on AMH levels. With AMH centiles, age-specific AMH was calculated, using the exponential–normal 3-parameter model; the cut-off values for age-specific quartiles have previously been described in detail [17]. Individuals were initially placed in one of percentiles and eventually in the quar- tile) categories of age specific AMH level. Cut off values calculated in Tehran Thyroid Study [18] were used for definition of various thyroid dysfunctions as follows: Hypothyroidism TSH > 5.06 mU/L and FT4 < 11.58 pmol/L; subclinical hypothyroidism TSH > 5.06 mU/L and normal FT4 (11.58–19.95) pmol/L; hyperthyroidism TSH < 0.32 mU/L and FT4 > 19.95 pmol/L and subclin- ical hyperthyroidism TSH < 0.32 mU/L and normal FT4 [12]; TPO Ab > 35 IU/mL was considered as TPO Ab positivity [18]. Smoking was considered as a binomial variable as: ever smokers: (using ≥1 cigarette per day) and never smokers. Statistical analysis Baseline characteristics are presented as mean (standard deviation) for numerical variables and number (percent- age) for the categorical measures. For numerical vari- ables with skewed distribution, median (inter quartile range) was calculated. Differences in descriptive baseline characteristics of women of various AMH quartiles were explored using analysis of variance. The Kruskal-Wallis test was applied to compare baseline values of variables with skewed dis- tributions. Chi square test was used to compare the prevalence of thyroid disorders in the quartiles of AMH. To analyze the person-years incidence rate of thyroid disorders the following formula was used: Number of new events of the condition cases ð Þ during the study time Sum of person−time person (cid:2) year ð Þat risk among the study participants The hazard ratio, which is derived from the Cox pro- portional hazards model, provides a statistical test of thyroid disorders comparing incidence rates of in age-specific quartiles of AMH. The hazard ratio is equivalent to the odds that an individual in the group with higher hazard reaches the endpoint first [19]. After excluding those with history of thyroidectomy, radio iod- ine treatment, and those using of levothyroxine or anti thyroid dugs during the follow ups, the generalized esti- mation equation (GEE) method was used as an estima- tion method for marginal, i.e. population-averaged modeling of repeated data [20] to investigate the secular longitudinal trends of TPO Ab, FT4, and TSH. Models for examination of time trends were fitted separately for the first and fourth quartiles of AMH and P values for trend were reported in each quartile. This model in- cluded the following predictors: Time (follow-up years), AMH quartile status, and an interaction term of these two (follow-up years × AMH quartile status) and was adjusted for BMI, smoking, parity, and education level. This interaction shows how the effect of AMH quartile status on thyroid functions changed overtime. Statistical analysis was performed, using the software package STATA (version 12; STATA Inc., College sta- tion, TX, USA); significance level was set at P < 0.05, and CI as 95%. Results Figure 1 illustrates the study flowchart; of 4174 partici- pants of the Tehran Thyroid Study, there were 1502 women, aged between 20 and 50 years; among these, 775 met our eligibility criteria. At baseline, 203, 181, 201 and 190 subjects were in the first (Q1 or lowest level), (Q2) second, (Q3) third and fourth (Q4 or highest level) quar- tiles of age-specific AMH, respectively; their baseline characteristics are presented in Table 1. Initially, there was no statistically significant difference in the an- thropometric measures, diastolic blood pressures, parity, smoking status, educational level, and thyroid hormone levels among participants of various quartiles. Women in the first quartile had higher levels of TPO Ab; 7.85(3.69, 23.3) IU/ml in comparison with those in the 2nd, 3rd and 4thquartiles 5.59 (3.10, 18.66), 5.14 (3.01, 11.35), 6.32 (3.37, 13.89), respectively; p-value = 0.02. Table 2 demonstrates the prevalence of various types of thyroid dysfunction among participants with different AMH quartiles at initiation of the study; there was no significant difference between study groups in terms of thyroid dysfunction. According to GEE analysis (Table 3), after adjustment for BMI, parity, smoking status, and educational level, there was a statistically significant annual decrease in mean changes of FT4 in all AMH quartiles; these decreases were [− 0.28(95%CI: -0.43, − 0.1); p < 0.001], [− 0.18, (95%CI: -0.34, − 0.01); p = 0.02], [− 0.25 (95%CI: -0.42, − 0.10); p = 0.001], [− 0.25 (95%CI: -0.41, − 0.09) Bahri et al. BMC Endocrine Disorders (2019) 19:47 Page 4 of 9 Fig. 1 Flow chart of the study pmol/L; p = 0.002] for 1st, 2nd, 3rd and 4th quartiles of AMH respectively. There was a statistically significant annual increase in TPO Ab in women of 1st quartile of AMH (10.66 ± 3.60 IU/ml), after adjustment for men- tioned variables; the odds of TPO Ab positivity (TPO Ab+) in these women increased by 8% (95% CI: 0.06, 0.16) per year; these significant changes in TPO Ab was not observed in other quartiles of AMH (Table 3). There was no statistically significant difference in mean changes of TSH in all age-specific AMH quartiles (Table 3). Based on GEE analysis, overall odds of TPO Ab+ in the first quartile of AMH was 2.05 fold higher than those of the fourth [95% CI:(1.31–3.74); p = 0.01], while the overall mean changes of TSH (p-value = 0.79), Table 1 Baseline characteristics of study subjects according to the age-specific anti-Mullerian hormone (AMH) quartiles Variable Age(years) BMI (kg/m2) WC (cm) HC (cm) WHR SBP (mm/Hg) DBP (mm/Hg) Parity (n) FT4 (pmol/L) TSH (mu/L) TPO Ab (IU/ml) Smoking Status Never Smokers n (%) Ever Smokers n (%) Educational Level Upper diploma n (%) Diploma/Under diploma n(%) 1st quartile n = 203 38.3 ± 6.7 26.8 ± 4.1 85.3 ± 9.8 104.0 ± 7.9 0.816 ± 0.6 113.3 ± 13.9 75.3 ± 9.92 3.3 ± 1.7 14.67 ± 2.13 2.05 (1.03,3.13) 7.85 (3.69,23.3) 7 (3.4%) 196(96.6%) 21(10.6%) 182(89.4%) 2nd quartile n = 181 37.3 ± 6.5 26.8 ± 4.4 84.7 ± 11.3 104.1 ± 8.5 0.810 ± 0.7 108.9 ± 12.9 73.9 ± 9.2 3.0 ± 1.4 23.94 ± 4.50 1.86 (1.06,3.72) 5.59 (3.10,18.66) 8 (4.4%) 173(95.6%) 24(13.5%) 157(86.5%) 3rd quartile n = 201 36.4 ± 6.8 27.1 ± 4.4 85.4 ± 10.8 104.3 ± 9.1 0.816 ± 0.6 111.9 ± 14.9 75.3 ± 10.1 3.1 ± 1.6 21.36 ± 2.57 1.66 (1.00,2.75) 5.14 (3.01,11.35) 6 (3.0%) 195(97%) 23(11.6%) 178(88.4%) 4th quartile n = 190 37.2 ± 6.7 27.4 ± 4.5 86.0 ± 11.2 104.6 ± 9.0 0.820 ± 0.7 111.4 ± 12.4 75.2 ± 8.9 3.0 ± 1.7 21.11 ± 4.50 1.64 (0.92,2.82) 6.32 (3.37,13.89) 8 (4.2%) 182(95.8%) 19(10.0%) 171(90%) P- value 0.42 0.44 0.70 0.91 0.64 0.01ª 0.38 0.73 0.36 0.26 0.02ª 0.87 0.73 Values are expressed as mean (SD), median (IQR) or number (percentage). IQR, inter quartile range; BMI, body mass index; WC, waist circumference; HC, hip circumference; WHR, waist circumference to hip circumference ratio; FT4, free thyroxin level; TSH, thyroid stimulating hormone; TPO Ab, Thyroid peroxidase anti body Bahri et al. BMC Endocrine Disorders (2019) 19:47 Page 5 of 9 Table 2 Prevalence of thyroid disorders at the initiation of study according to the age-specific anti-Mullerian hormone (AMH) quartiles Variables N (%) TPO Ab+ Subclinical Hypothyroidism Overt Hypothyroidism Subclinical Hyperthyroidism Overt Hyperthyroidism 1st quartile n = 203 41 (20.2%) 14 (6.9%) 7 (3.4%) 11 (5.4%) 3 (1.5%) 2nd quartile n = 181 28(15.5%) 8(4.4%) 5 (2.8%) 11 (6.1%) 4 (2.2%) 3rd quartile n = 201 28(13.9%) 4(2%) 5 (2.5%) 8 (4%) 7 (3.5%) 4th quartile n = 190 25(13.4%) 8(4.3%) 6 (3.2%) 8 (4.3%) 6 (3.2%) Values are expressed as number (percent); TPO Ab+: Thyroid peroxidase antibody positivity P- value 0.22 0.12 0.94 0.76 0.56 showed no statistically difference between these two quartiles, after adjustment for BMI, parity, smoking, and education level (Table 4). The interaction (follow-up year × AMH quartile status) in women in the first and the fourth quartiles of AMH was not statistically significant for any of the thyroid parameters (p-value > 0.05). Figure 2 a, b, and c illustrate trends of TPO Ab, TSH and FT4 in women in the first and fourth quartiles of AMH, adjusted for BMI, parity, smoking, and education level. TSH had a significant increase in women in the first quartile (mean TSH = 2.4 mu/L in the first visit vs. 5.2 mu/L in the last visit, Ptrend = 0.04); these values were 2.25 and 3.25 mu/L, Ptrend < 0.001 for the first and last visits in the fourth quartile of AMH. However, mean changes of TSH in women in the first quartile were not significantly different from women in the fourth quartile (Pinteraction = 0.30) Although FT4 showed a significant decrease in women in the fourth quartile (mean FT4 = 15.5 pmol/L in the first visit vs. 14.28 pmol/L the last visit, < 0.001), FT4 of women in the first quartile showed no significantly different value overtime (Ptrend = 0.14). Interaction between follow-up years and AMH quartile demonstrate that as time progresses, changes in thyroid hormones of women in the first quartile of AMH do not differ or become significantly worse compared to those in the fourth quartile (Fig. 2). There was no significant difference in incidence of subclinical hypothyroidism [HR:1.34 (95% CI:0.52,2.12)], overt hypothyroidism [HR:0.78(95% CI:0.34,1.80)] sub- clinical hyperthyroidism [HR:1.97(95% CI:0.59,6.63)], overt hyperthyroidism [HR:0.63(95% CI: 0.14,2.86)] and TPO Ab positivity [HR:1.49 (95% CI:0.73,3.03)] between those of the 1st versus those of the 4th quartile. Inci- dence rates of subclinical hypothyroidism in the 1st, 2nd, 3rd and 4th quartiles were [20(95% CI:10,27), 15(95% CI:10,23), 5 (95% CI:3,10); p = 0.002] and 14(95% CI: 10,21) per1000 person-years, respectively. This inci- dence in the 3rd quartile was significantly lower than in others, (p = 0.002). Discussion The long term prospective population-based Tehran Thyroid Study provides a unique opportunity to investi- thyroid gate associations between overtime trend of functions and the ovarian reserve status. In the current study, we found that women with lower ovarian reserves had higher levels of TPO Ab at baseline; moreover a positive trend of this antibody was observed in these groups of women compared to those ones with better ovarian reserve status; there were more incidents of thy- roid autoimmunity in the former group. There are several studies with controversial results on association between hypothyroidism and thyroid auto- immunity with low ovarian reserve [9, 11, 21, 22]; how- ever effects of low ovarian reserve on the overtime trend of thyroid function have not been investigated yet. Table 3 Annual changes in various parameters according to age-specific Anti-Mullerian hormone (AMH) quartiles using generalized estimating equation (GEE) Variables FT4 (pmol/L) TSH (mu/L) 1st quartile n = 203 −0.28 ± 0.07ª 0.58 ± 0.30 2nd quartile n = 181 −0.18 ± 0.07ª 0.23 ± 0.32 3rd quartile n = 201 −0.25 ± 0.07ª 0.05 ± 0.30 4th quartile n = 190 −0.25 ± 0.07ª 0.08 ± 0.31 10.66 ± 3.60ª TPO Ab (IU/ml) TPO Ab+b ªP < 0.05 interaction form (follow-up year ×age-specific AMH quartile) (GEE); FT4, free thyroxin level; TSH, Thyroid stimulating hormone; TPO Ab, Thyroid peroxides anti body; TPO Ab+, Thyroid peroxidase antibody positivity, Data were adjusted for BMI, parity, smoking status, educational level bValues are expressed as odds ratio and confidence interval 1.08ª (1.0006,1.16) 0.95(0.87,1.04) 1.06(0.98,1.15) 6.49 ± 3.66 1.78 ± 3.68 5.37 ± 3.94 1.05(0.97,1.14) Bahri et al. BMC Endocrine Disorders (2019) 19:47 Page 6 of 9 Table 4 Parameter estimates of GEE model in women in 1st quartile of AMH compared to those of women in the 4th quartile Variable TSH (mu/L) FT4 (pmol/L) TPO Ab+ Parameter AMHQ1 AMHQ4 Time Time×AMHQ1 Time×AMHQ4 AMHQ1 AMHQ4 Time Time×AMHQ1 Time×AMHQ4 AMHQ1 AMHQ4 Time Time×AMHQ1 Time×AMHQ4 β coef. −0.48 Reference 0.33 0.63 Reference −1.15 Reference 0.38 0.29 Reference 2.05 Reference 1.10 0.93 Reference SE 1.8 0.44 0.62 0.38 −0.11 0.16 0.62 0.05 0.06 95% confidence interval (−4.16,3.19) (−0.53,1.21) (−0.58,1.85) (−2.05,-0.12) (−0.64,-0.12) (−0.10,6.34) (1.31, 3.74) (1.004,1.22) (0.81,1.06) P-Value 0.79 0.45 0.30 0.02ª < 0.001ª 0.06 0.01ª 0.04ª 0.28 AMHQ1 AMH (Anti- mullerian hormone) in women of 1st quartile, AMHQ4 AMH in women of 4th quartile, BMI body mass index, FT4 free thyroxin level, TSH thyroid stimulating hormone, TPO Ab Thyroid peroxidase antibody, TPO Ab+ Thyroid peroxides antibody positivity, × indicates interaction In agreement with our findings, Chen et al. (2017) in a cross-sectional study of 1044 infertile Chinese women demonstrated that idiopathic low ovarian reserve with lower serum concentration of AMH was associated with more frequent positive TPO Ab rather than thyroid function or Tg Ab positivity [21]. Saglam et al. in a case control study found that even after age adjustment, autoimmune thyroid disease(AITD) was independently associated with AMH, which was lower in AITD women than in controls [22]. In contrast, Polyzos et al. (2015) in a large retrospective cross-sectional analysis reported that thyroid autoimmunity was not associated with low ovarian reserve [9]. In an animal study, hypothyroidism in adult female rats was induced by an iodide deficient diet in combination with perchlorate supplementation for inhibition of iodide uptake by the thyroid. This con- dition in adult females negatively affects the ovarian fol- licular reserve and the size of the growing follicle population [11]. Moreover, some studies report that ele- vated serum concentration of TSH is associated with decreased serum level of AMH [4, 23, 24]. Weghofer et al. in a study of 225 infertile women reported that even after adjustment for thyroid autoimmunity and age, TSH < 3.0μIU/mL was associated with significantly bet- ter ovarian reserve and higher AMH than TSH ≥3.0μIU/ mL [23]. In another study among 23 pairs of infertile and fertile women, it was found that both TSH levels and age were negatively correlated with AMH levels in infertile participants with standardized partial regression coefficient (β) of − 0.534 and − 0.361, respectively, but fertile ones [4]. This finding is in not in normal agreement with ours as we also found no association be- tween serum concentration of TSH and ovarian reserve status among participants who were mostly fertile; since thyroid autoimmunity precedes thyroid dysfunction [25], it may need a longer follow up to cause significant elevation in TSH levels among women with thyroid autoimmunity. The underlying pathophysiological mechanism of the association between thyroid autoimmunity and ovarian reserve status is not completely understood. In the Mon- teleone et al. study [26], for the first time the presence of thyroid antibodies in ovarian follicular fluid was dem- onstrated and a significantly lower oocyte fertilization and percentage of grade A embryos was found in infer- tile women with thyroid autoimmunity undergoing IVF compared to controls. Although this data are prelimin- ary, but it may be assumed that TPO Ab that passes through the blood follicle barrier during follicular evolu- tion may result in the destruction and damaging of growing follicles and oocytes via thyroid hormone recep- tors, on these cells [26]; a possible mechanism which needs to be confirmed by further comprehensive studies. Except for autoimmunity, thyroid dysfunction may influ- ence the reproductive system via thyroid hormone receptors on the surface of oocytes [4], or through dis- ruption of GnRH function due to increasing prolactin secretion [5]. Furthermore, in the present study, regardless of ovar- ian reserve status, we found that serum concentration of TSH had an incremental trend by age. In cross-sectional individuals without thyroid disease, serum studies of Bahri et al. BMC Endocrine Disorders (2019) 19:47 Page 7 of 9 concentrations of free T4 and TSH, as well as on the negative feedback set point [31, 32]. of the We incidence showed that subclinical hypothyroidism in women of the 3rd AMH quartile was lower than those in other quartiles, even the 4th; this may partly be explained by the including of women with subclinical ovarian dysfunction in the 4th AMH quartile, who had higher AMH levels and regular menstrual cy- cles, but had polycystic morphology in sonography and subclinical anovulation [33, 34] and also higher rates of thyroid autoimmunity and hypothyroidism [35], despite our effort to recruit fertile women with regular men- strual cycles for the purpose of the present study. The main strength of the present study is its method- ology; it is the first study that has investigated the tem- poral trend of changes in thyroid hormones, TPO Ab and thyroid disorders among a general population of women with different levels of age specific AMH; in addition the follow up time was 12 years, which seems a sufficient duration for the purpose of the current study. instead of crude We used age-specific AMH levels, AMH levels because this method provides more accur- ate and precise assessment of ovarian reserve status [36, 37]. The intra-assay and inter-assay in our data is likely thyroid assessments and to be minimal because all AMH assays were performed in the same laboratory by the same expert. This study had some limitations as well: 1) other methods for estimation of ovarian reserve including: FSH, antral follicular count, inhibin B were not assessed in the study; however AMH has been introduced to be the most reliable [15, 16, 36]. 2) Age specific AMH was identified according to a single AMH measurement at initiation of study and we did not evaluate ovarian reserve over time; however it has been shown that AMH level remains almost constant from one cycle to another and has a high interclass correlation coefficient, as a re- sult of which only one measurement provides a reliable estimate of age specific level [38–40]. 3) Subjects were not assessed for the subclinical ovarian dysfunction, which may have affected the characteristics of women with higher ovarian reserves [41, 42]. 4) We did not measure Tg Ab in this study and thyroid autoimmunity was evaluated based only on serum TPO Ab measure- ment; however the presence of TPO Ab in serum is more frequent than Tg Ab [43]; furthermore a large population based study demonstrated a similar preva- lence in both antibodies [44] and finally our study assessed only thyroid autoimmunity in relation to age specific AMH levels and autoimmunity in general was not assessed. Previous studies showed that ovarian reserve is closely linked to autoimmunity [22, 37]. In an- other study premature ovarian failure (POF) was signifi- cantly associated with adrenal cortex autoantibodies Fig. 2 Mean of changes within follow-ups in the first and fourth quartiles of age specific AMH assuming the interaction between time and study group after adjusting for BMI, parity, smoking, and education level. a TPO Ab+ (Thyroid peroxidase antibody positivity), b FT4 (free thyroxin level), c TSH (thyroid stimulating hormone). First quartile of AMH (Q1), fourth quartile of AMH (Q4) TSH concentrations increased with age [27–29], al- though this age-related TSH increase was not accom- panied by a fall in free T4, suggesting an alteration in the TSH regulatory system [28, 30]. In the present study, the decline in FT4 among women with higher ovarian reserves is clinically unimportant. Additionally, some indicate hereditary and genetic effects on studies Bahri et al. BMC Endocrine Disorders (2019) 19:47 Page 8 of 9 [45], which may be due to general autoimmunity that af- fects multiple organs and TPO Ab is a part of this process that has destructive effect on oocytes [26]. Large long-term studies are needed to evaluate both the causality between low ovarian reserve and appear- ance of thyroid autoimmunity, and the culprits that lead to the autoimmune process in women with limited ovar- ian reserve. Conclusion In summary this study demonstrates that women with low ovarian reserves had both higher baseline and increasing levels of TPO Ab over a12 year follow. They had also a higher incidence of TPO Ab positivity over time. A pro- spective comprehensive study with several time points measurement of AMH and thyroid assessment is recom- mended to define the evolution of the association between ovarian reserve status and thyroid function. Abbreviations AMH: Anti-mullerian hormone; BMI: Body Mass Index; DBP: Diastolic Blood Pressure; EIA: Enzyme Immunoassay; FT4: Free T4; GEE: Generalized Estimation Eq.; HC: Hip Circumference; IEMA: Immune Enzymometric Assay; IRES: Research Institute for Endocrine Sciences; PCOS: Poly Cystic Ovarian Syndrome; POF: Premature ovarian failure; SBP: Systolic Blood Pressure; TLGS: Tehran Lipid and Glucose Study; TPO Ab: Thyroid peroxidase antibody; TPO Ab+: TPO Ab positivity; TSH: Thyroid stimulating hormone; TTS: Tehran Thyroid Study; WC: Waist Circumference; WHR: Waist Circumference to Hip Circumference Ratio Acknowledgments This article has been extracted from the thesis written by Ms. Sara Bahri, School of Medicine Shahid Beheshti University of Medical Science. (Registration No: 132). We wish to acknowledge the laboratory personnel of the IRES for their assistance. The authors would like to acknowledge Ms. Niloofar Shiva for critical editing of English grammar and syntax of the manuscript. Funding This study was funded by the Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences. Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Authors’ contributions SB. contributed to the study design and execution, data analysis, manuscript drafting and critical discussion. FRT. contributed to the study design and execution, data analysis, manuscript drafting and critical discussion. AA. contributed to manuscript drafting and critical discussion. MR. contributed to the data analysis and manuscript drafting. MT. contributed to the laboratory testing and manuscript drafting. MV. contributed to manuscript drafting. FA. contributed to the study design and execution and manuscript drafting. All authors read and approved the final manuscript. Ethics approval and consent to participate All participants signed written informed consent forms and the study was approved by the ethics committee of the Research Institute for Endocrine Sciences, RIES affiliated to the Shahid Beheshti University of Medical Sciences. IR ENDOCRINE, REC.1396.416.. Consent for publication Not applicable. Competing interests The authors report no Competing interest. The authors alone are responsible for the content and writing of the article. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Author details 1Reproductive Endocrinology Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran. 2Endocrine Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran. 3Department of Epidemiology and Biostatistics, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran. 4Prevention of Metabolic Disorders Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran. 5Chronic Respiratory Diseases Research Center, National Research Institute of Tuberculosis and Lung Diseases (NRITLD), Shahid Beheshti University of Medical Sciences, Tehran, Iran. 6Endocrine Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran. Received: 17 November 2018 Accepted: 11 April 2019 References 1. Webber L, Davies M, Anderson R, Bartlett J, Braat D, Cartwright B, et al. 2. 3. 4. 5. 6. 7. 8. 9. European Society for Human Reproduction and Embryology (ESHRE) guideline group on POI. ESHRE guideline: management of women with premature ovarian insufficiency. Hum Reprod. 2016;31(5):926–37. Buyuk E, Seifer DB, Illions E, Grazi RV, Lieman H. Elevated body mass index is associated with lower serum anti-mullerian hormone levels in infertile women with diminished ovarian reserve but not with normal ovarian reserve. Fertil Steril. 2011;95(7):2364–8. La Marca A, Sighinolfi G, Papaleo E, Cagnacci A, Volpe A, Faddy MJ. Prediction of age at menopause from assessment of ovarian reserve may be improved by using body mass index and smoking status. PLoS One. 2013; 8(3):e57005. Kuroda K, Uchida T, Nagai S, Ozaki R, Yamaguchi T, Sato Y, et al. Elevated serum thyroid-stimulating hormone is associated with decreased anti- Müllerian hormone in infertile women of reproductive age. J Assist Reprod Genet. 2015;32(2):243–7. Krassas G, Poppe K, Glinoer D. Thyroid function and human reproductive health. Endocr Rev. 2010;31(5):702–55. Aoki Y, Belin RM, Clickner R, Jeffries R, Phillips L, Mahaffey KR. Serum TSH and total T4 in the United States population and their association with participant characteristics: National Health and nutrition examination survey (NHANES 1999–2002). Thyroid. 2007;17(12):1211–23. Bremner AP, Feddema P, Leedman PJ, Brown SJ, Beilby JP, Lim EM, et al. Age-related changes in thyroid function: a longitudinal study of a community-based cohort. J Clin. Endocrinol Metab. 2012;97(5):1554–62. Vadasz Z, Haj T, Kessel A, Toubi E. Age-related autoimmunity. BMC Med. 2013;11(1):94. Polyzos NP, Sakkas E, Vaiarelli A, Poppe K, Camus M, Tournaye H. Thyroid autoimmunity, hypothyroidism and ovarian reserve: a cross-sectional study of 5000 women based on age-specific AMH values. Hum Reprod. 2015; 30(7):1690–6. 10. Hiraoka T, Wada-Hiraike O, Hirota Y, Hirata T, Koga K, Osuga Y, et al. The impact of elevated thyroid stimulating hormone on female subfertility. Reprod Med Biol. 2016;15(2):121–6. 11. Meng L, Rijntjes E, Swarts HJ, Keijer J, Teerds KJ. Prolonged hypothyroidism severely reduces ovarian follicular reserve in adult rats. J Ovarian Res. 2017; 10(1):19. 12. Azizi F, Amouzegar A, Delshad H, Tohidi M, Mehran L. Natural course of thyroid disease profile in a population in nutrition transition: Tehran thyroid study. Arch of Iran Med. 2013;16(7):418. 13. Azizi F, Salehi P, Etemadi A, Zahedi-Asl S. Prevalence of metabolic syndrome in an urban population: Tehran lipid and glucose study. Diabetes Res Clin Pract. 2003;61(1):29–37. Bahri et al. BMC Endocrine Disorders (2019) 19:47 Page 9 of 9 38. de Vet A, Laven JS, de Jong FH, Themmen AP, Fauser BC. Antimüllerian hormone serum levels: a putative marker for ovarian aging. Fertil Steril. 2002;77(2):357–62. van Rooij IA, Broekmans FJ, Scheffer GJ, Looman CW, Habbema JDF, de Jong FH, et al. Serum antimüllerian hormone levels best reflect the reproductive decline with age in normal women with proven fertility: a longitudinal study. Fertil Steril. 2005;83(4):979–87. 39. 40. Hosseinpanah F, Barzin M, Erfani H, Serahati S, Ramezani Tehrani F, Azizi F. Lipid accumulation product and insulin resistance in Iranian PCOS prevalence study. Clin Endocrinol. 2014;81(1):52–7. 41. Malcolm CE, Cumming DC. Does anovulation exist in eumenorrheic 42. women? Obstet Gynecol. 2003;102(2):317–8. Tehrani FR, Simbar M, Tohidi M, Hosseinpanah F, Azizi F. The prevalence of polycystic ovary syndrome in a community sample of Iranian population: Iranian PCOS prevalence study. Reprod Biol Endocrinol. 2011;9(1):39. 43. Bryhni B, Aanderud S, Sundsfjord J, Rekvig O, Jorde R. Thyroid antibodies in northern Norway: prevalence, persistence and relevance. J Intern Med. 1996; 239(6):517–23. 44. Pedersen IB, Knudsen N, Jørgensen T, Perrild H, Ovesen L, Laurberg P. Thyroid peroxidase and thyroglobulin autoantibodies in a large survey of populations with mild and moderate iodine deficiency. Clin Endocrinol. 2003;58(1):36–42. 45. Bakalov VK, Anasti JN, Calis KA, Vanderhoof VH, Premkumar A, Chen S, et al. Autoimmune oophoritis as a mechanism of follicular dysfunction in women with 46, XX spontaneous premature ovarian failure. Fertil Steril. 2005;84(4): 958–65. 14. Azizi F, Ghanbarian A, Momenan AA, Hadaegh F, Mirmiran P, Hedayati M, et al. Prevention of non-communicable disease in a population in nutrition transition: Tehran lipid and glucose study phase II. Trials. 2009;10(1):5. 15. Altman DG, Chitty LS. Charts of fetal size: 1. Methodol BJOG. 1994;101(1):29–34. 16. Royston P, Wright E. How to construct ‘normal ranges’ for fetal variables. 17. Ultrasound Obstet Gynecol. 1998;11(1):30–8. Tehrani F, Mansournia M, Solaymani-Dodaran M, Azizi F. Age-specific serum anti-Müllerian hormone levels: estimates from a large population-based sample. Climacteric. 2014;17(5):591–7. 18. Amouzegar A, Mehran L, Takyar M, Abdi H, Azizi F. Tehran thyroid study 19. (TTS). Int J Endocrinol Metab. 2018;16(4 Suppl). Spruance SL, Reid JE, Grace M, Samore M. Hazard ratio in clinical trials. Antimicrob Agents Chemother. 2004;48(8):2787–92. 20. Zeger SL, Liang K-Y. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986:121–30. 21. Chen C-W, Huang Y-L, Tzeng C-R, Huang R-L, Chen C-H. Idiopathic low 22. ovarian reserve is associated with more frequent positive thyroid peroxidase antibodies. Thyroid. 2017;27(9):1194–200. Saglam F, Onal ED, Ersoy R, Koca C, Ergin M, Erel O, et al. Anti-Müllerian hormone as a marker of premature ovarian aging in autoimmune thyroid disease. Gynecol Endocrinol. 2015;31(2):165–8. 23. Weghofer A, Barad DH, Darmon S, Kushnir VA, Gleicher N. What affects 24. functional ovarian reserve, thyroid function or thyroid autoimmunity? Reprod Biol Endocrinol. 2016;14(1):26. Kucukler FK, Gorkem U, Simsek Y, Kocabas R, Guler S. Evaluation of ovarian reserve in women with overt or subclinical hypothyroidism. Arch Med Sci. 2018;14(3):521. 25. Vanderpump M, Tunbrldge W, French J, Appleton D, Bates D, Clark F, et al. The incidence of thyroid disorders in the community: a twenty-year follow- up of the Whickham survey. Clin Endocrinol. 1995;43(1):55–68. 26. Monteleone P, Parrini D, Faviana P, Carletti E, Casarosa E, Uccelli A, et al. Female infertility related to thyroid autoimmunity: the ovarian follicle hypothesis. Am J Reprod Immunol. 2011;66(2):108–14. 27. Hadlow NC, Rothacker KM, Wardrop R, Brown SJ, Lim EM, Walsh JP. The 28. relationship between TSH and free T4 in a large population is complex and nonlinear and differs by age and sex. J Clin Endocrinol Metab. 2013;98(7): 2936–43. Kahapola-Arachchige KM, Hadlow N, Wardrop R, Lim EM, Walsh JP. Age- specific TSH reference ranges have minimal impact on the diagnosis of thyroid dysfunction. Clin Endocrinol. 2012;77(5):773–9. Surks MI, Hollowell JG. Age-specific distribution of serum thyrotropin and antithyroid antibodies in the US population: implications for the prevalence of subclinical hypothyroidism. J Clin Endocrinol Metab. 2007;92(12):4575–82. 30. Nazarpour S, Ramezani Tehrani F, Simbar M, Tohidi M, Minooee S, Rahmati M, Azizi F. Effects of levothyroxine on pregnant women with subclinical hypothyroidism, negative for thyroid peroxidase antibodies. J Clin Endocrinol Metab. 2018;103(3):926–35. 29. 31. Hansen PS, Brix TH, Iachine I, Sørensen TI, Kyvik KO, Hegedüs L. Genetic and environmental interrelations between measurements of thyroid function in a healthy Danish twin population. Am J Physiol Endocrinol Metab. 2007; 292(3):E765–E70. 32. Panicker V, Wilson S, Spector T, Brown S, Falchi M, Richards J, et al. 33. Heritability of serum TSH, free T4 and free T3 concentrations: a study of a large UK twin cohort. Clin Endocrinol. 2008;68(4):652–9. Tal R, Seifer DB, Khanimov M, Malter HE, Grazi RV, Leader B. Characterization of women with elevated antimüllerian hormone levels (AMH): correlation of AMH with polycystic ovarian syndrome phenotypes and assisted reproductive technology outcomes. Am J Obstet Gynecol. 2014;211(1):59. e1–8. 34. Dumont A, Robin G, Catteau-Jonard S, Dewailly D. Role of anti-Müllerian 35. 36. hormone in pathophysiology, diagnosis and treatment of polycystic ovary syndrome: a review. Reprod Biol Endocrinol. 2015;13(1):137. Tuten A, Hatipoglu E, Oncul M, Imamoglu M, Acikgoz AS, Yilmaz N, et al. Evaluation of ovarian reserve in Hashimoto's thyroiditis. Gynecol Endocrinol. 2014;30(10):708–11. Tehrani FR, Mansournia MA, Solaymani-Dodaran M, Steyerberg E, Azizi F. Flexible parametric survival models built on age-specific antimüllerian hormone percentiles are better predictors of menopause. Menopause. 2016;23(6):676–81. 37. Broer S, Eijkemans M, Scheffer G, Van Rooij I, De Vet A, Themmen A, et al. Anti-Müllerian hormone predicts menopause: a long-term follow-up study in normoovulatory women. J Clin Endocrinol Metab. 2011;96(8):2532–9.
10.1186_s12913-019-3983-7	Zheng et al. BMC Health Services Research (2019) 19:155 https://doi.org/10.1186/s12913-019-3983-7 R E S E A R C H A R T I C L E Open Access One-year costs of medical admissions with and without a 30-day readmission and enhanced risk adjustment , Amresh Hanchate2 and Michael Shwartz3 Sarah Zheng1* Abstract Background: To overcome the limitations of administrative data in adequately adjusting for differences in patients’ risk of readmissions, recent studies have added supplemental data from patient surveys and other sources (e.g., electronic health records). However, judging the adequacy of enhanced risk adjustment for use in assessment of 30-day readmission as a hospital quality indicator is not straightforward. In this paper, we evaluate the adequacy of risk adjustment by comparing the one-year costs of those readmitted within 30 days to those not after excluding the costs of the readmission. Methods: In this two-step study, we first used comprehensive administrative and survey data on a nationally representative Medicare cohort of hospitalized patients to compare patients with a medical admission who experienced a 30-day readmission to patients without a readmission in terms of their overall Medicare payments during 12 months following the index discharge. We then examined the extent to which a series of enhanced risk adjustment models incorporating code-based comorbidities, self-reported health status and prior healthcare utilization, reduced the payment differences between the admitted and not readmitted groups. Results: Our analytic cohort consisted 4684 index medical hospitalization of which 842 met the 30-day readmission criteria. Those readmitted were more likely to be older, White, sicker and with higher healthcare utilization in the previous year. The unadjusted subsequent one-year Medicare spending among those readmitted ($56,856) was 60% higher than that among the non-readmitted ($35,465). Even with enhanced risk adjustment, and across a variety of sensitivity analyses, one-year Medicare spending remained substantially higher (46.6%, p < 0.01) among readmitted patients. Conclusions: Enhanced risk adjustment models combining health status indicators from administrative and survey data with previous healthcare utilization are unable to substantially reduce the cost differences between those medical admission patients readmitted within 30 days and those not. The unmeasured patient severity that these cost differences most likely reflect raises the question of the fairness of programs that place large penalties on hospitals with higher than expected readmission rates. Keywords: 30-day readmissions, Utilization, Disease severity * Correspondence: szheng@uvic.ca 1University of Victoria Gustavson School of Business, 3800 Finnerty Rd, Victoria, BC V8P 5C2, Canada Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Zheng et al. BMC Health Services Research (2019) 19:155 Page 2 of 10 Background Studies over several decades have emphasized the inad- equacy of administrative data-based risk adjustment models like that used in the US by the Centers for Medi- care and Medicaid Services (CMS) in its 30-day hospital readmission profiling and penalty program, largely because administrative databases include only limited information on patient severity and disease burden [1–7]. Despite the criticism, there is broad consensus on the preventability of some readmissions [8] and evidence of reductions in read- missions from targeted interventions [9]. Thus, the CMS models and others like it continue to be in routine use to obtain publicly reported hospital measures of quality and performance-based penalties [10–13]. As an indication of the importance of these models, CMS imposed penalties estimated at $528 million on 78% of US hospitals in 2017 as part of the Hospital Readmissions Reduction Program [12] because of excess readmissions over those predicted by the risk adjustment model. In recent years, research papers have described models using additional variables from administrative databases (e.g., race/ethnicity and socio-economic status) as well as enhancing adminis- trative data with self-reported and medical chart data, which capture previously unmeasured patient risk indi- cators such as health behavior, mental health status, functional health, socioeconomic vulnerabilities, and family and social support [2–4, 14–24]. However, the ex- tent to which previously unmeasured disease burden and severity is captured in the enhanced models is unclear. The value of 30-day readmissions as a performance measure depends upon the extent to which risk adjust- ment is able to “make comparable” at the time of the index hospital admission those that are readmitted within 30 days and those that are not. For, only if the two groups are comparable at the time of admission is it reasonable to penalize the hospital for the readmission. To evaluate the adequacy of risk adjustment when analyzing 30-day re- admission rates, in this paper we take a novel approach that as far as we know has not been used before: specific- ally, we examine the longer term costs (one-year costs) of those readmitted within 30 days and those not after ex- cluding the costs of the readmission. If risk adjusted longer-term costs are same in the readmitted and not re- admitted groups, it suggests the groups are comparable with the exception of the readmission. This finding would provide strong support for the validity of 30-day readmis- sions as a performance measure. If longer-term costs differ between the two groups, there are two possible ex- planations: 1) a hospital error or deficiency in practice (e.g., inadequate discharge planning) has long-term cost implications; or, 2) there are still important unmeasured risk factors (e.g., unmeasured patient severity). To the ex- tent costs associated with the first explanation tend to be of a short-term nature, longer-term differences in costs are an indication of inadequate risk adjustment. Also, whatever the cause of the difference in risk-adjusted longer-term costs between the admitted and not readmit- ted groups, if there is a significant difference, it becomes increasingly unreasonable to estimate the dollar savings from a reduction in readmissions as the costs of the re- admission “prevented.” [25] It seems much more likely that due to the increased severity of the readmitted group, the “prevented” 30-day readmission was not really pre- vented but just shifted to a later time after 30 days. Risk prediction and risk adjustment models Before turning to our specific study, we briefly place our work in a somewhat broader 30-day readmission risk mod- eling context. The literature on models that predict the likelihood of hospital readmissions can be distinguished based on purpose, which to a large extent dictates the vari- ables available for modeling and thus, how well the model is likely to perform. Risk prediction models that attempt to identify patients at high risk for readmission during the course of their hospitalization can use variables whose values become available while the patient is in the hos- pital (e.g., days in the ICU); risk prediction models that prioritize patients for post-discharge interventions can use information available only at the time of discharge (e.g., length of stay). Much of the literature focusing on in-hospital and post-discharge interventions to reduce readmission risk has evaluated the extent to which la- boratory data and vital signs, plus additional data from electronic health records, can improve the ability of the model to better identify high risk patients and target them for interventions. Though risk adjustment models pre- dict risk for individual patients, that is not their goal. It is to control (or adjust) for differences in patient characteris- tics, essentially “leveling the playing field,” so that when the outcome of one group of patients (e.g., those treated at particular hospital or those receiving a particular inter- vention) is compared to another group, it is under the as- sumption that both groups are similar prior to event of interest (e.g., admission to the hospital or receipt of the intervention). When the models are used for provider pro- filing or incentive programs, which are usually undertaken by large administrative units (e.g., states, provinces or countries), the models are limited to data from adminis- trative databases in which data elements are collected in standardized ways across a large number of provider. In addition, these models do not use data that may create perverse incentives, e.g., prior utilization or cost. However, when risk adjustment models are used to evaluate an the important variables included in the intervention, model are confounders, i.e., variables related both to re- ceipt of the intervention and, independent of the interven- tion, to the outcome. If important confounders are not controlled for, it is impossible to know if an outcome Zheng et al. BMC Health Services Research (2019) 19:155 Page 3 of 10 following an intervention is due to the intervention or the uncontrolled for confounders. For example, prior utilization may increase the likelihood a patient receives a particular intervention and, because prior utilization may be associated with increased illness burden, it is likely related to the outcome whether or not the person receives the intervention. In this paper, we initially consider a risk adjustment model with independent variables similar to the ones used by CMS to predict 30-day readmissions. These models, which include age, sex and comorbidities from CMS administrative databases, have c statistics (a stand- ard measure of model performance when predicting a binary outcome variable) in the low 0.60 range (often considered below the “acceptable discrimination” thresh- old of 0.70) [26]. Models using prior utilization and data available at the time of hospital discharge to predict 30-day readmissions can have c statistics above 0.80 [3]. In what follows, we sequentially do the following: 1) Compare one-year subsequent healthcare utilization (measured as Medicare payments) between 30-day re- admitted and non-readmitted medical admission pa- tients, excluding payments for the 30-day readmission stay of the readmitted patients; 2) Examine the extent to which the large differences in healthcare utilization (our outcome) between the readmitted and non-readmitted group (essentially, the intervention, which is passive and sorts patients into 2 groups) could be “explained” by in- cluding additional variables (potential confounders) in the risk adjustment model; and 3) Finally, in the Discus- sion, pull together information from the analyses that provides support for the hypothesis that unmeasured disease burden (an unmeasured confounder) is the most likely factor accounting for the still large differences in costs that remain after controlling for a wide range of fac- tors. In these analyses, we used longitudinal healthcare utilization data from CMS’ Medicare Current Beneficiary Survey (MCBS), which includes both administrative and survey data and therefore permits evaluation of a range of patient risk factors beyond those identified in administrative data [3, 15]. Methods Data and analytic sample We used 2000–2011 MCBS Cost and Use files. The MCBS is a weighted stratified, multistage, area probabil- ity sample of Medicare enrollees (community and facility dwellers) drawn from the Medicare enrollment file [27]. The sample cohort consists of three rotating panels, each followed for 3 years, with one panel replaced each year. Medicare claims data are supplemented with indi- vidual surveys of demographics, health status, health be- havior, healthcare utilization and Medicare payments. There were 38,059 enrollees aged 65 and older with claims data for 3 years, or until death during the study period. We excluded those enrolled in any Medicare Ad- vantage plan during their three-year survey period (n = 391) (since claims data are unavailable for this group), who were residents of Puerto Rico (n = 36), and with missing key measures (n = 195). Further details on exclusions are in the Additional file 1. In identifying index hospitalizations, we included all non-surgical hospitalizations, identified by Diagnostic Related Groups (DRG) designation of “medical” [28]. We look at 30-day readmissions occurring after all these non-surgical hospitalizations. We wanted to ensure at least 12 months of follow-up healthcare utilization after all index hospitalizations. Therefore, we selected the first hospitalization during the second year of the follow-up period of the 3-year MCBS cohort as index hospitaliza- tions; those without a hospitalization in the second year were excluded from the analysis. Following previous work on evaluation of 30-day readmissions as a quality indicator, we excluded index hospitalizations that in- volved patient death within 30-days of index discharge, transfer to another acute care facility, discharge against medical advice or discharge to hospice [29]. Study design To examine if hospitalized patients with a 30-day readmis- sion (“readmitted patients”) had different healthcare spend- ing and utilization patterns compared to those without a 30-day readmission (“non-readmitted patients”), we used a retrospective study design to compare total healthcare spending following an index hospitalization (excluding costs associated with a readmission) between patients with and without a 30-day readmission. Outcomes Our main outcome measure was one-year subsequent Medicare spending ($), defined as the total Medicare spending for all inpatient and outpatient care during one year after the index hospitalization admission date. For clarity in comparison, we excluded Medicare payments for the index inpatient stay as well as the payment of re- admission stay for the subgroup with a 30-day readmis- sion; all other hospitalizations during the 12-month follow-up period were included. We used Medicare payments reported in each claim record as the measure of healthcare spending. For comparability of spending over time, we applied the national Consumer Price Index to express all dollar values in terms of 2011 dollars [30]. To limit the influence of outliers, we top-coded in- dividual annual spending at the 95th percentile (with larger spending values reset to the 95th percentile level, which is $200,000). Zheng et al. BMC Health Services Research (2019) 19:155 Page 4 of 10 As secondary outcomes, we examined Medicare spend- ing and utilization one year following discharge from the index admission by type of service: acute inpatient care spending; number of days in acute inpatient care; out- patient care spending; had an outpatient care visit within 30 days of index discharge. Utilization and payments asso- ciated with the readmission stay (for those readmitted) were not included in any measure. Independent variables The main independent variable of interest was 30-day readmission status (Yes = 1, No = 0, indicated by re- admission to any hospital for any admission condition within 30 days after the index discharge date. We in- cluded a range of other independent variables, clinical identified as risk factors for 30-day and non-clinical, readmission in prior work [3, 7, 15]. These included pa- tients’ sex, age and race/ethnicity (non-Hispanic Whites, non-Hispanic Blacks, Hispanics, Others). As our study inpatient admissions, we used the cohort includes all Charlson Comorbidity Index conditions, coded as indica- tor variables [31]. Comorbidity condition status was based on (a) all secondary diagnosis codes in the index ad- mission and (b) all diagnosis codes in the inpatient and outpatient records one-year prior to the index ad- mission date. Following prior studies using supplemental data, we also identified a range of self-reported patient health behavior and other risk factors [3, 7, 15]: ever smoking; overweight and obesity (body mass index > 25); living type (communi- ty-alone, community-two people, community-more than two and facility); education (< 12 years of education, high school, college); income categorized based on quar- tiles as lowest, second lowest and top two quartiles; marital status; and Medicaid coverage in addition to Medicare (dual coverage). We included three measures of healthcare utilization dur- ing the one-year period prior to the index hospitalization: number of hospitalizations; total days of inpatient stay; and overall Medicare spending. To control for secular trends in health care costs other than due to inflation (which we ad- justed for using the Consumer Price Index), we adjusted for calendar year of the index hospitalization. To adjust for sys- tematic regional variation in healthcare utilization, we used enrollee’s residence location which, based on the Dart- mouth hospital referral region-level measure of Medi- care spending per person, was assigned to quintile of per-person spending [32]. We also adjusted for the fol- lowing admission conditions: heart failure, pneumonia, pulmonary disease, digest disorder, gastrointestinal bleed- ing, septicemia, psychoses, intracranial hemorrhage/cere- bral infarction, kidney & urinary tract infections, circulatory disorders, and other conditions. Statistical analysis We performed bivariate comparisons of the aforemen- tioned covariates between readmitted and non-readmitted patients using regression models – binary logit, multi- nomial logit and ordinary least squares, depending on the covariate measure – in order to most easily adjust for sur- vey weights. For our core analysis of the comparison of the main outcome (subsequent one-year Medicare spend- ing) between readmitted vs. non-readmitted patients, we estimated ordinary least squares (OLS) models of subse- quent Medicare spending including readmission status and covariates as independent variables, again using sur- vey weights. To adjust for skewness in the outcome meas- ure, and following prior studies, we also estimated a generalized linear model (GLM) with a gamma distribu- tion and log link; as estimates from both models were similar, to facilitate direct interpretation of coefficients, we have reported the OLS estimates as our preferred results (GLM estimates are in the Additional file 1). We esti- mated four models with different combinations of covari- ates. Model 1 patient sex, age, race/ethnicity, smoking status, overweight status, comorbidities, dual coverage, Hospital Referral Region-level Medicare spending (quin- tiles), index admission condition, census region (N = 9) and index year (2001–2010). Model 2 also included pa- tient education, income, living type, and marital status. Model 3 included only indicators of prior year patient care (i.e., hospitalizations days, number of hospitalizations, overall medical spending in the prior years). Model 4, the most comprehensive model, extended Model 2 by includ- ing prior year patient care measures. All model estimates were based on heteroskedasticity-consistent robust stand- ard errors. All analyses were conducted using Stata 13 (StataCorp, College Station, Texas). In addition to the GLM models, we performed other sensitivity analyses (all reported in the Additional file 1). First, because systematic differences in one or more of the model covariates between readmitted vs. non-readmitted patients may potentially influence final estimates, we re- peated the analysis with a propensity score matched sam- ple of readmitted and non-readmitted cases. To create the matched samples, we used the following approach sug- gested by Austin [33]: randomize all readmitted cases; set as caliper 0.5 of the standard deviation of the logit of the propensity score (this was the smallest caliper size that en- sured at least two matched observations for each readmit- ted case); select in order of the randomized readmitted cases the first two non-readmitted cases within the caliper distance of the readmitted cases without replacement; run an ordinary least square regression model with these matched groups. Standard errors that included error in es- timation of the propensity score were used. To evaluate the success of matching, we compared covariate balance between readmitted and matched non-readmitted cases. Zheng et al. BMC Health Services Research (2019) 19:155 Page 5 of 10 Table 1 Sample Characteristics of Studied Population (N = 4684) N % Readmission No Readmission Readmission No Readmission P values All Female Age 65–74 75–84 85+ Race/Ethnicity Whites Blacks Hispanics Charlson comorbidity categories Acute myocardial infarction (AMI) Congestive heart failure (CHF) Peripheral vascular disease (PVD) Cerebrovascular disease Dementia Chronic Obstructive Pulmonary Disease (COPD) Rheumatoid Disease Peptic Ulcer Mild liver disorder Diabetes Diabetes + Complications Hemiplegia or Paraplegia Renal disease Cancer Moderate/severe Liver disease Metastic Cancer Acquired immune deficiency syndrome (AIDS) Ever-smoker Overweight/Obese Education < 12 years of education High school College and above Income, Household Poorest quartile Second quartile Top two quartiles Marital Status Married Widowed Divorced/separated Never married 842 483 297 405 140 682 87 34 100 302 183 268 62 341 42 44 17 330 92 29 137 192 9 60 1 499 496 299 428 115 172 249 421 385 358 74 25 3842 2204 1564 1588 691 3280 294 134 491 1111 697 1002 242 1377 232 176 24 1295 399 64 483 721 15 140 3 2220 2360 1252 2040 550 699 1099 2044 1849 1552 316 125 18.0 57.4 35.3 48.1 16.6 81.0 10.3 4.0 11.9 35.9 21.7 31.8 7.4 40.5 5.0 5.2 2.0 39.2 10.9 3.5 16.2 22.8 1.1 7.1 0.1 59.3 59.0 35.5 50.8 13.6 20.5 29.6 50.0 45.8 42.5 8.8 3.0 82.0 57.4 40.7 41.3 18.0 85.4 7.7 3.5 12.8 28.9 18.1 26.1 6.3 35.8 6.0 4.6 0.6 33.7 10.4 1.7 12.6 18.8 0.4 3.6 0.1 57.8 61.4 32.6 53.1 14.3 18.2 28.6 53.2 48.1 40.4 8.2 3.2 0.988 0.064 0.003 0.551 < 0.001 0.032 0.001 0.220 0.025 0.272 0.463 0.001 0.002 0.662 0.002 0.011 0.006 0.024 < 0.001 0.736 0.458 0.204 0.175 0.104 0.301 Zheng et al. BMC Health Services Research (2019) 19:155 Page 6 of 10 Table 1 Sample Characteristics of Studied Population (N = 4684) (Continued) N % Readmission No Readmission Readmission No Readmission Living Type Community-Alone Community-Two people Community-More than two Facility Dual (Medicare-Medicaid) coverage Death During Follow-up Utilizations by hospital referral regions First quintile Second quintile Third quintile Fourth quintile Fifth quintile Previous year utilization Days of stay Number of hospitalizations Overall Medicare spending ($) Medicare spending 1-year following index discharge 281 401 131 29 159 278 100 137 137 255 213 4.6 0.7 1285 1847 530 180 583 740 610 625 700 1057 851 2.4 0.4 33.4 47.6 15.6 3.5 18.9 33.0 11.8 16.3 16.2 30.3 25.4 4.6 0.7 33.4 48.1 13.8 4.7 15.2 19.3 15.9 16.3 18.2 27.5 22.1 2.4 0.4 39,004.5 56,855.5 25,103.7 35,464.5 39,004.5 56,855.5 25,103.7 35,464.5 P values 0.952 0.006 < 0.001 0.009 < 0.001 < 0.001 < 0.001 < 0.001 Notes. Index condition, region and year are not included in this table When there are survey weights, there are a number of un- resolved issues when matching on propensity scores [33]. Thus, we ran propensity score analysis without adjusting for survey weights. As an indirect test of the sensitivity of estimates to survey weight adjustment, we compared our main OLS estimates with and without survey weights. Second, systematic differences in patient death between readmitted vs. non-readmitted patients may influence sub- sequent healthcare utilization. To test the sensitivity of re- sults to this possibility, we estimated out main model (OLS) for the subgroup of patients who did not die during the 3-year survey period. Finally, it may be that overall re- admitted patients have higher one-year spending than non-readmitted patients, but the difference is driven by the higher proportion of very expensive cases among the readmitted patients. To evaluate this possibility, we exam- ined the difference in one-year costs of readmitted and non-readmitted patients who had spending that was below different dollar thresholds. Results Our analytic cohort consisted of 4684 index hospitaliza- tions of which 842 met the 30-day readmission criteria. Al- though similar in some characteristics (see Table 1), those readmitted were more likely than the non-readmitted to be older, White, sicker (Charlson comorbidity), covered by Medicaid (dual coverage), and with higher healthcare utilization in the previous year. The average subsequent one-year Medicare spending was $39,314 overall; among the readmitted, average spending was $56,856, over 60% higher than among the non-readmitted, $35,465. Adjusted for risk factors in Model 1, subsequent one-year Medicare spending among the readmitted patients was $17,726 (50%) higher than that among the non-readmitted (Table 2). Age group 65–74, Black race, and comorbidities of renal disorders, diabetes, and peripheral vascular disease were associated with higher spending. Adjusting further for patients’ nonclinical factors led to no sizable change in this difference (Model 2, Table 2). An alternate model which only adjusted for indicators of inpatient care utilization in the previous year, led to an adjusted difference of $18,163 (51%) in spending between those with and without readmission (Model 3, Table 2). Our final model that adjusted for all aforementioned covariates indi- cated a $16,516 (47%) difference in spending (Model 4, Table 2 and Additional file 1: Table S2). Analogous comparison of secondary outcomes (using Model 4) indicated that the readmitted, compared to the non-readmitted, spent $13,191 more on acute inpatient care, were in inpatient care 12.6 more days, spent $3325 more on outpatient care, and had 17% higher likelihood of having an outpatient care visit within 30 days of index discharge (Table 3). Our results were robust to alternative model specifica- tion (Additional file 1: Table S3). Using a generalized linear Zheng et al. BMC Health Services Research (2019) 19:155 Page 7 of 10 Table 2 Models of Medicare spending 1-year following index discharge (N = 4684) Readmission 17726* 17670* 18163* 16516* Model 1 Model 2 Model 3 Model 4 Female Age 65–74 75–84 85+ Race Whites Blacks Hispanics Others Dual (Medicare-Medicaid) coverage Ever Smoker Overweight+Obese Education < 12 years of education High school Bachelor and above Living Type 524 894 413 Reference − 872 Reference −520 − 8009* − 7460 Reference 799 − 4510^ Reference Reference Reference 10750* 10079* 6619^ − 2463 − 1403 5677 − 3586 − 1836 2703^ − 2252 2687^ − 2432 Reference − 1433 − 4562^ 8768 6288 − 2035 − 2188 2689^ − 1194 Reference − 1534 − 4630^ Community-Alone Reference Reference Community-Two people Community-More than two Facility Income Poorest quartile Second quartile Top two quartiles Marital Status Married Widowed Divorced/separated Never married Previous year utilization Hospitalizations days Number of hospitalizations Overall Medicare spending ($) 1143 5162 1228 Reference − 2059 149 Reference − 824 559 − 5076 1411 4767 2200 Reference − 2083 332 Reference 38 1697 − 3735 − 441^ − 343 2113 712 0.255* 0.198* Constant 6054 7821 29190* 8961 0.14 R-squared 0.13 Notes. Other covariates included in models 1, 2, and 4 were Charlson comorbidity index conditions, HRR-level Medicare spending (quintiles), index admission condition, census region (N = 9) and index year (2001–2010). All spending are in dollars ^ p < 0.10, p < 0.01, * p < 0.001 0.17 0.11 model with gamma distribution (instead of OLS model for Model 4, Table 2), subsequent one-year Medicare spending was $17,281 higher than among the non-readmitted. Matching readmitted with the non-readmitted based on propensity scores led to better balancing of all characteris- tics, including the proportion of Blacks and Medicaid cov- ered, and previous year inpatient care (Additional file 1: Table S1). Regression estimation of the propensity score matched cohort indicated that spending among the readmitted was $18,337 higher than among those not readmitted (Additional file 1: Table S3). Also, the afore- mentioned excess spending estimate of $16,516 in Table 2 (Model 4) was not sensitive to use of survey weights; re-estimation without weights resulted in $16,281 excess spending (Additional file 1: Table S3). Our results were also robust after limiting the sample only to those who were three-year observation window alive throughout our ($19,945 difference in cost, Additional file 1: Table S4). As shown in Table 4, though the ratio of readmitted to non-readmitted spending declines as the threshold (i.e., the dollar amount below which patients are included in the average) is reduced from $130,000 to $15,000 (approxi- mately the median spending of non-readmitted patients), low-cost readmitted patients still have between 28 to 50% higher costs than low-cost non-readmitted patients. Discussion Across a number of different models and sensitivity ana- lyses, we consistently found that medical admission pa- tients readmitted within 30 days have approximately 50% higher one-year costs than those not readmitted. In par- ticular, enhanced risk adjustment had no major effect on the cost differences between the two groups. This gives rise to the question “what accounts for the difference?” In some cases a serious medical error in the initial hospitalization of those readmitted within 30 days may have led to substantially higher utilization and costs over the subsequent year. However, there is nothing in the lit- erature of which we are aware that suggests most 30-day readmissions are due to serious medical errors with long-term cost implications at the index hospitalization. It is also possible that the higher death rate among those re- admitted combined with high end-of-life utilization pat- terns accounts for the difference in payments between the readmitted and not readmitted groups. However, as noted, when we reran the models including only those who sur- vived for the full 3-year period, we still found substantial differences between those readmitted and those not ($19,945 difference in costs, Additional file 1: Table S4). Another possibility is that there may be systematic dif- ferences in provider practices between the two cohorts in terms of the risk of admission and readmission; for in- stance, some geographic areas may have lower thresholds, in terms of patient severity, in admitting and readmitting Zheng et al. BMC Health Services Research (2019) 19:155 Page 8 of 10 Table 3 Models of Medicare spending 1-year following index admission by type of service (N = 4684) Average observed value Model-predicted value of the difference in Medicare utilization associated with patients with 30-day readmission N r2 * p < 0.001 (1) (2) (3) (4) Acute inpatient care spending ($) Acute inpatient care, days Outpatient care spending ($) Had an outpatient care visit within 30 days of index discharge 24,633 13,191* 4684 0.14 7.10 12.59* 4684 0.23 13,784 3325* 4684 0.19 0.95 0.17*** 4684 0.04 patients [34]. Note, however, that we have controlled for patients’ healthcare spending in the previous year. To il- lustrate the implications of this, consider two patients with equally high health care spending in the year prior to their index admission (which the model allows us to do, since it includes prior health care spending as an independent variable). Assume that in terms of unmeasured patient se- verity the two patients are similar to the average patients in the sample and that their high spending is due to the fact that the patients live in areas with an equally high propensity to consume health care resources (i.e., low threshold to provide services). One of the patients has a 30-day readmission and the other patient does not. As- sume one-year costs of the readmitted patient (excluding the readmission) are 50% higher than the non-readmitted patient. It is certainly possible that the providers for the readmitted patient have increased even further their pro- pensity to provide services from the pre-period whereas those for the non-readmitted patient have not. However, the more likely hypothesis is that the readmitted patient became sicker than the non-readmitted patient and this increased illness burden is the reason for the readmission and the higher one-year costs. Several facts support our hypothesis: 1) In addition to controlling for one-year prior health care spending, we have also controlled for the aver- age annual Medicare spending per enrollee in the patient’s hospital referral region; 2) Readmitted patients not only had 61.0% higher inpatient care spending but 25.6% higher adjusted outpatient care spending in the subsequent one year period following the index hospitalization; and they were 17% more likely to have an outpatient visit within 30 days of the index hospitalization discharge; 3) Observed risk factors were more prevalent among the readmission cohort; in particular, readmitted patients had higher prevalence of heart failure, cerebrovascular disease, chronic obstructive pulmonary disease, demen- tia, renal disease and cancer; and finally, 4) The death rate among readmitted patients was substantially higher than among non-readmitted patients (33% vs. 19%). A possible bias in our analysis resulting from exclusion of the costs of the readmission is that those readmitted had less days (i.e., the time they were in the hospital) to experi- ence outpatient costs than those not readmitted. Thus, our approach may underestimate the cost differences between Table 4 Average Cost and Percent of Cases Below the Threshold: Readmitted and Non-Readmitted Patients Threshold $ Average Cost % Cases Average Cost % Cases Non-Readmitted Cases (NRC) Readmitted Cases (RC) Ratio of Average Cost NRC to RC 130,000 120,000 110,000 100,000 90,000 80,000 70,000 60,000 50,000 40,000 30,000 20,000 15,000 23,718 22,637 21,717 20,684 19,395 17,947 16,601 15,006 13,264 11,551 9264 7045 5740 0.93 0.92 0.91 0.90 0.88 0.87 0.85 0.82 0.78 0.74 0.68 0.59 0.53 35,558 33,679 32,179 29,467 27,875 25,453 24,183 21,256 18,022 14,745 12,376 9164 7368 0.86 0.84 0.83 0.80 0.78 0.75 0.73 0.68 0.62 0.55 0.49 0.39 0.32 1.50 1.49 1.48 1.42 1.44 1.42 1.46 1.42 1.36 1.28 1.34 1.30 1.28 Zheng et al. BMC Health Services Research (2019) 19:155 Page 9 of 10 the two groups. To examine this, we reran the analysis after eliminating costs in the first 30 days after discharge for both groups. The change was not in the expected direc- tion, i.e., a larger difference between the 2 groups. In the original analysis, the cost difference was $16,516; in the suggested analysis that looks at cost differences starting from day 30 after the index admission, the cost difference was $13,850. The reason for this is that when we eliminate the first 30 days after discharge from the index admission, we eliminate a period of time during which, in addition to the readmission, the readmitted group has much higher outpatient healthcare utilization. For example, the readmit- ted group had 12.4 more physician visits in the 30-days post discharge than the non-readmitted. Studies have indicated a reduction in readmission rates nationwide following CMS’ introduction of annual reporting of hospital performance in readmissions (Hos- pital Compare program) and CMS’ penalty program for excess readmissions (Hospital Readmissions Reduction Program) [35–37]. A recent study, however, suggests that the impact on readmissions in prior studies is at a minimum only half that previously estimated and at a maximum statistically similar to the declines in two con- trol samples [38, 39]. As our study was based predomin- antly on data prior to these programs, it would be useful to use examine the robustness of our findings using more recent data, something feasible since the CMS method- ology for estimation of risk-adjusted readmissions per- formance has largely remained unchanged. We recognize several limitations of this study. Because our study population was limited to a small sample of Medicare participants with few index hospitalizations from the same hospitals, we were unable to examine hospital- level differences in readmissions. Also, our choice of add- itional variables measuring readmission risk were limited to those available in MCBS data; however, we were able to identify measures covering most of the domains covered in previous studies [3]. On the positive side, because the MCBS sample is a stratified national sample of Medicare enrollees, we expect that the findings are representative of patients across all hospitals. Conclusions In summary, our study suggests that on average readmit- ted patients are “sicker” than non-readmitted patients and that current models of risk adjustment for 30-day re- admission, even when supplemented with self-reported measures of patient health behavior, functional health status, family and social support, prior utilization and so- cioeconomic disadvantage, are unable to adjust for these differences. Therefore, use of such models for profiling hospital performance on 30-day readmission may system- atically underestimate performance of hospitals with high rates of observed readmissions. Finally, our findings do call in to question studies that have estimated the benefits of a reduction in 30-day readmissions as “true” savings to the health care system. Given the likely increased morbid- ity and disease severity of the readmitted group, it seems probable that many of the “prevented” 30-readmissions will be readmitted after the 30-day period. Additional file Additional file 1: Table S1. Comparison of sample characteristics before vs. after propensity score matching. Table S2. Models of Medicare spending 1-year following index discharge (N = 4684). Table S3. Sensitiv- ity of estimates to model functional form: Overall Medicare spending 1- year following index discharge (N = 4684). Table S4. Sensitivity of esti- mates to exclusions: Overall Medicare spending 1-year following index discharge. Exclusions B1. Keeping Eligible Admissions in the Second Year. Exclusions B2. Other Exclusions. (DOCX 36 kb) Abbreviations CMS: Centers of Medicare & Medicaid Services; DRG: Diagnostic Related Groups; GLM: Generalized Linear Model; MCBS: Medicare Current Beneficiary Survey; OLS: Ordinary Least Squares Acknowledgements Not applicable. Funding Not applicable. Availability of data and materials The data that support the findings of this study are available from Centers for Medicare & Medicaid Services. Medicare Current Beneficiary Survey (MCBS). https://www.cms.gov/Research-Statistics-Data-and-Systems/Research/ MCBS/. Authors’ contributions Obtained data: AH. Made substantial contributions to conception and design or analysis and interpretation of data: SZ, AH, MS. Gave final approval to current version of the paper: SZ, AH, MS. Agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved: SZ, AH, MS. Ethics approval and consent to participate The Institutional Review Board of Boston University Medical Campus approved this study. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Author details 1University of Victoria Gustavson School of Business, 3800 Finnerty Rd, Victoria, BC V8P 5C2, Canada. 2Boston University School of Medicine, 801 Massachusetts Ave Crosstown Center, Boston, MA 02118, USA. 3Operations and Technology Management Department, Boston University Questrom School of Business, 595 Commonwealth Avenue, Boston, MA 02215, USA. Zheng et al. BMC Health Services Research (2019) 19:155 Page 10 of 10 cohort StudyHospital profiling for stroke readmission. Ann Intern Med. 2014; 161(11):775–84. 22. Philbin EF, Dec GW, Jenkins PL, DiSalvo TG. Socioeconomic status as an independent risk factor for hospital readmission for heart failure. Am J Cardiol. 2001;87(12):1367–71. 23. Pilote L, Tu JV, Humphries K, Behouli H, Belisle P, Austin PC, Joseph L. Socioeconomic status, access to health care, and outcomes after acute myocardial infarction in Canada's universal health care system. Med Care. 2007:638–46. 24. Rathore SS, Foody JM, Wang Y, et al. Race, quality of care, and outcomes of elderly patients hospitalized with heart failure. Jama. 2003;289(19):2517–24. Jencks SF et al. Rehospitalizations among Patients in the Medicare Fee-for- Service Program. N Engl J Med. 2009:360(14): 1418–28. 25. 26. Hosmer WD, Lemeshow S. Interpretation of the fitted logistic regression model. In: Applied Lostistic Regression. Edn. New York: Sons Inc; 2000. 27. Centers for Medicare & Medicaid Services (2016). Medicare Current Beneficiary Survey (MCBS). https://www.cms.gov/Research-Statistics-Data- and-Systems/Research/MCBS/. 28. Goodman DC, Fisher ES, Chang C. After hospitalization: a Dartmouth Atlast report on post-acute care for Medicare beneficiaries. Hanover, NH: The Dartmouth Institute for Health Policy and Clinical Practice; 2011. 29. Bernheim SM, Lin Z, Grady J, Bhat KR, Wang H, Wang Y, Abedin Z, Desai MM, Li S-X, Vellanky S, et al. 2011 measures maintenance technical report: acute myocardial infarction, heart failure, and pneumonia 30-day risk-standardized readmission measures. New Haven, CT: Yale New Haven Health Services Corporation & Center for Outcomes Research & Evaluation; 2011. 30. Bureau of Labor Statistics. Consumer Price index - all urban consumers. Washington DC: Bureau of Labor Statistics; 2017. 31. Charlson M, Pompei P, Ales K, MacKenzie C. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–83. The Dartmouth Institute for Health Policy & Clinical Practice: Dartmouth Atlas of Healthcare. In. Hanover, NH: http://www.dartmouthatlas.org/; 2016. 32. 33. Austin PC. An introduction to propensity score methods for reducing the 34. effects of confounding in observational studies. Multivariate Behav Res. 2011;46(3):399–424. Fisher ES, Wennberg JE, Stukel TA, Sharp SM. Hospital readmission rates for cohorts of Medicare beneficiaries in Boston and New Haven. N Engl J Med. 1994;331(15):989–95. 35. Zuckerman RB, Sheingold SH, Orav EJ, Ruhter J, Epstein AM. Readmissions, observation, and the hospital readmissions reduction program. N Engl J Med. 2016;374(16):1543–51. 36. Carey K, Lin MY. Readmissions to New York hospitals fell for three target conditions from 2008 to 2012, consistent with Medicare goals. Health Aff (Millwood). 2015;34(6):978–85. 37. Desai NR, Ross JS, Kwon J, et al. Association between hospital penalty status under the hospital readmission reduction program and readmission rates for target and nontarget conditions. JAMA. 2016;316(24):2647–56. Ibrahim AM, Dimick JB, Sinha SS, Hollingsworth JM, Nuliyalu U, Ryan AM. Association of coded severity with readmission reduction after the hospital readmissions reduction program. JAMA Intern Med. 2018;178(2):290–2. 38. 39. Ody C, Msall L, Dafny LS, Grabowski DC, Cutler DM. Decreases in readmissions credited to Medicare’s program to reduce hospital readmissions have been overstated. Health Aff. 2019;38(1):36–43. Received: 13 October 2018 Accepted: 1 March 2019 References 1. 2. 3. 4. 5. 6. 7. 8. 9. Anderson GF, Steinberg EP. Hospital readmissions in the Medicare population. N Engl J Med. 1984;311(21):1349–53. Ross JS, Mulvey GK, Stauffer B, Patlolla V, Bernheim SM, Keenan PS, Krumholz HM. Statistical models and patient predictors of readmission for heart failure: a systematic review. Arch Intern Med. 2008;168(13):1371–86. Kansagara D, Englander H, Salanitro A, Kagen D, Theobald C, Freeman M, Kripalani S. Risk prediction models for hospital readmission. JAMA. 2011; 306(15):1688–98. Calvillo-King L, Arnold D, Eubank KJ, Lo M, Yunyongying P, Stieglitz H, Halm EA. Impact of social factors on risk of readmission or mortality in pneumonia and heart failure: systematic review. J Gen Intern Med. 2013; 28(2):269–82. Joynt KE, Jha AK. Thirty-day readmissions — truth and consequences. N Engl J Med. 2012;366(15):1366–9. Press MJ, Scanlon DP, Navathe AS, Zhu J, Chen W, Mittler JN, Volpp KG. The importance of clinical severity in the measurement of hospital readmission rates for Medicare beneficiaries, 1997-2007. Med Care Res Rev. 2013;70(6): 653–65. Barnett ML, Hsu J, McWilliams JM. Patient characteristics and differences in hospital readmission rates. JAMA Intern Med. 2015;175(11):1803–12. van Walraven C, Bennett C, Jennings A, Austin PC, Forster AJ. Proportion of hospital readmissions deemed avoidable: a systematic review. Cmaj. 2011; 183(7):E391–402. Hansen LO, Young RS, Hinami K, Leung A, Williams MV. Interventions to reduce 30-day rehospitalization: a systematic review. Ann Intern Med. 2011; 155(8):520–8. 11. 10. Austin JM, Jha AK, Romano PS, Singer SJ, Vogus TJ, Wachter RM, Pronovost PJ. National hospital ratings systems share few common scores and may generate confusion instead of clarity. Health Aff (Millwood). 2015;34(3):423–30. Krumholz HM, Keenan PS, Brush JE Jr, Bufalino VJ, Chernew ME, Epstein AJ, Heidenreich PA, Ho V, Masoudi FA, Matchar DB, et al. Standards for measures used for public reporting of efficiency in health care: a scientific statement from the American Heart Association interdisciplinary council on quality of care and outcomes research and the American College of Cardiology Foundation. Circulation. 2008;118(18):1885–93. 12. Boccuti C, Casillas G: Aiming for Fewer Hospital U-turns: The Medicare Hospital Readmission Reduction Program. In. https://www.kff.org/medicare/ issue-brief/aiming-for-fewer-hospital-u-turns-the-medicare-hospital- readmission-reduction-program/ Kaiser Family Foundation; 2017. Kocher RP, Adashi EY. Hospital readmissions and the affordable care act. JAMA. 2011;306(16):1794–5. 13. 14. Amarasingham R, Moore BJ, Tabak YP, Drazner MH, Clark CA, Zhang S, Reed WG, Swanson TS, Ma Y, Halm EA. An automated model to identify heart failure patients at risk for 30-day readmission or death using electronic medical record data. Med Care. 2010;48(11):981–8. 15. Coleman EA, Min SJ, Chomiak A, Kramer AM. Posthospital care 16. transitions: patterns, complications, and risk identification. Health Serv Res. 2004;39(5):1449–65. Krumholz HM, et al. Hospital 30-Day Acute Myocardial Infarction Readmission Measure: Methodology. Centers for Medicare & Medicaid Services. 2008. www.qualitynet.org. 17. Martsolf GR, Barrett ML, Weiss AJ, Washington R, Steiner CA, Mehrotra A, Coffey RM: Impact of race/ethnicity and socioeconomic status on risk- adjusted readmission rates: implications for the hospital readmissions reduction program. INQUIRY: The Journal of Health Care Organization, Provision, and Financing 2016, 53:0046958016667596. 18. Coffey RM, Misra A, Barrett M, Andrews RM, Mutter R, Moy E. Congestive heart failure: who is likely to be readmitted? Med Care Res Rev. 2012;69(5):602–16. 19. Dailey EA, Cizik A, Kasten J, Chapman JR, Lee MJ. Risk factors for readmission of orthopaedic surgical patients. JBJS. 2013;95(11):1012–9. 20. Desai MM, Stauffer BD, Feringa HH, Schreiner GC. Statistical models and patient predictors of readmission for acute myocardial infarction: a systematic review. Circ Cardiovasc Qual Outcomes. 2009;2(5):500–7. Keyhani S, Myers LJ, Cheng E, Hebert P, Williams LS, Bravata DM. Effect of clinical and social risk factors on hospital profiling for stroke ReadmissionA 21.
10.1186_s12889-023-15780-y	Saunders et al. BMC Public Health (2023) 23:1176 https://doi.org/10.1186/s12889-023-15780-y BMC Public Health ‘It’s been a lifelong thing for me’: parents’ experiences of facilitating a healthy lifestyle for their children with severe obesity Liz A. Saunders1,2,3* Elizabeth A. Davis2,5 , Ben Jackson1,2 , Lyndsey Price5 , Lisa Y. Gibson2,4 and Timothy Budden1,2 , Justine Doust3,5 , James A. Dimmock1,2,6 , Abstract Objective For parents and guardians, assisting children/adolescents with severe obesity to lose weight is often a key objective but a complex and difficult challenge. Our aim in this study was to explore parents’ (and guardians’) perspectives on the challenges they have faced in assisting their children/adolescents with severe obesity to lead a healthy lifestyle. Methods Thirteen parents/guardians were interviewed from a pool of families who had been referred but did not engage between 2016 and 2018 (N = 103), with the Perth Children’s Hospital Healthy Weight Service, a clinical obesity program for children/adolescents (parent age M = 43.2 years, children age M = 10.3 years). Using semi-structured interviews and thematic analysis, we identified 3 broad themes. Results Parental weight-related factors reflected parents’ own lifelong obesity narrative and its effect on their own and their families’ ability to live a healthy lifestyle. Perceived inevitability of obesity in their child reflected parents’ feelings that the obesity weight status of their children/adolescent was a persistent and overwhelming problem that felt ‘out of control’. Lastly, parents reported challenges getting medical help stemming from co-morbid medical diagnosis in their child/adolescent, and difficulties with medical professionals. Conclusion This study demonstrates that parents face challenges in supporting healthy lifestyle for children/ adolescents with severe obesity due to parents own internal weight biases and their negative experiences within the healthcare system when seeking help. Keywords Parental levels of obesity, Severe obesity, Fixed mindset, Obesity narrative, Weight stigma, Isolated and unsupported parents *Correspondence: Liz A. Saunders liz.saunders@research.uwa.edu.au 1School of Human Sciences (Exercise and Sport Science), The University of Western Australia, Perth, Australia 2Telethon Kids Institute, The University of Western Australia, Subiaco, WA, Australia 3Paediatric Consultation Liaison Program, Child and Adolescent Mental Health Service, Perth Children’s Hospital, Western, Australia 4School of Medical and Health Sciences, Edith Cowan University, Perth, Western, Australia 5Healthy Weight Service, Department of Endocrinology and Diabetes, Perth Children’s Hospital, Western, Australia 6Department of Psychology, College of Healthcare Sciences, James Cook University, Townsville, Australia © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessPage 2 of 12 Introduction Severe obesity in children/adolescents is defined as as a body mass index (BMI) ≥ 120% of the 95th percentile for age and sex, or an absolute BMI ≥ 35 kg/m2[1]. Obesity in childhood consequences are many and predominantly negative. Obesity in childhood has clear associations with adverse physical health outcomes [2–4], and there is compelling evidence that obesity in childhood is predic- tive of obesity in adulthood [2, 3, 5]. Social consequences of obesity include bullying [5–7], discrimination [8], per- ceived external negative perceptions [9, 10], and social stigma [11]. Psychological consequences can include low self-esteem and disturbances in body image that often persist into adulthood [5, 12]. Long-term outcomes for children/adolescents with severe obesity are especially poor, with some studies highlighting that severe obesity in childhood increases the risk of premature adult mor- tality, independent of current adult BMI [13]. However, in relation to treating severe obesity, the notion of calories in, calories out is an oversimplified solution to a complex disease [10] and does not acknowledge that eating well and engaging in physical activity is a recommendation for all people regardless of their health or illness status. What is not yet clear, however, is the range of barriers and challenges that parents experience when attempting to support a healthy lifestyle for their child/adolescent with severe obesity. Children/adolescents with severe obesity and their families are a critically important but often underrepresented population in paediatric health research as these families are often hard to access [14], especially when families are not engaged in treatment programs [15], and thus are not as well studied as ‘non- clinical’ populations (i.e., those without severe obesity). Parents (and guardians) can be especially influential in affecting the weight status of children/adolescents. As well as exerting uncontrollable influence through genetic, biological and epigenetic factors [2, 16], parents can indirectly influence children/adolescents by acting as role models, and can directly facilitate change by influ- encing family health behaviours and choices [17]. There is evidence in the obesity literature related to children/ adolescents that obesogenic contextual lifestyle behav- iours (i.e., unhealthy diets, physical inactivity, sedentary behaviours) of parents may lead to children/adolescents engaging in these same behaviours [18]. However, it’s unclear what effect these factors have on—and how they are managed —in families with a child/adolescent with severe obesity. Obesity in childhood treatment most commonly takes the form of lifestyle modification programs and it is generally acknowledged that a parent-led, family- based approach involving both parent and child/adoles- cent is the most effective for managing or reducing the child/adolescent’s level of obesity [19]. Although child/ adolescent weight loss is not always the primary focus of interventions (i.e., programs often focus on healthy behaviours and relationships with exercise, food choices, and body positivity), at the level of severe obesity, reduc- ing a child/adolescent’s level of obesity is an important health goal. Optimal program structure combines par- ent-led family behaviour modification (such as parenting skills) and educational topics, such as healthy nutrition, appropriate exercise guidelines, and a formal program of exercise for the child/adolescent [20, 21]. For children/ adolescents with severe obesity, treatment is typically delivered in tertiary hospital-based settings, i.e. a clini- cal program, involving multidisciplinary specialists to address various problematic and maladaptive aspects of the obesity and to manage medical comorbidities [22]. However, although these lifestyle modification family- based programs are the cornerstone of clinical treat- ment for severe obesity in children/adolescents, it seems that many of these programs do not elicit immediate or long-term weight loss outcomes [22, 23]. To date, much of the research on reasons for unsuccessful clinical pro- gram outcomes has focussed on parents reported reasons for attrition from clinical programs [24, 25], the experi- ence within a clinical program itself [26], or evaluating the outcomes of pilot or new types of programs [27]. Although there is well-developed literature document- ing the difficulties that parents face supporting children/ adolescents with lower levels of overweight or obesity [14, 28, 29], relatively little research has been undertaken on the broad challenges that parents face in support- ing weight loss and lifestyle behaviours among children/ adolescents with severe obesity. In addition, much of the research on severe obesity in children/adolescents takes a retrospective approach, i.e., focused on families that have actively sought treatment for their child/ado- lescent’s severe obesity. As such, among these families, less is known regarding the everyday challenges faced by those parents who have not or choose not to engage with a clinical tertiary program. Guided by the gap in the lit- erature outlined above, the purpose of the current study was to solicit insight into the experiences of this hard-to- access group using a qualitative approach. Specifically, we sought to better understand the experiences and chal- lenges of parents who were referred but did not engage in a tertiary clinical intervention program, in facilitating a healthy lifestyle for their child/adolescent with severe obesity. Method Philosophical perspective This study was guided by an interpretivist paradigm, underpinned by a relativist ontology (i.e. the notion that there are multiple ‘realities’), and a subjectivist episte- mology (i.e. one’s life experiences shapes their perception Saunders et al. BMC Public Health (2023) 23:1176 of the world) [30]. Within this paradigm, it is understood that knowledge is co-created between participants and the researcher(s), and researchers focus on finding mean- ing within their own and (where possible) participants’ historical, temporal, cultural, and subjective circum- stances. A reflexive approach was also adopted through- out the study. Reflexivity requires an acknowledgment that researchers’ perspectives may influence the research process and influence their research findings [31]. With this in mind, the author team had research experience in the area of obesity [32, 33] as well as programming exper- tise in clinical and community settings, and some were also parents. The lead author had a longstanding rela- tionship with the clinical obesity intervention program from which participants were recruited—this relation- ship allowed access to a hard-to-reach clinical population and meant that the lead author held knowledge about the structure of programs, the staff within the program, and the broad experiences of participants referred to the program, and a desire to help these families share their experiences. Participants and procedure Ethical approval to conduct this study was obtained from the Child and Adolescent Health Services Human Research Committee (CAHS HRC) and the University of Western Australia Human Research Ethics Commit- tee. The Healthy Weight Service (HWS), based in Perth, Western Australia, is a 12-month, multidisciplinary, psycho-educational, family lifestyle intervention pro- gram. Participation in the HWS is voluntary and costs are covered by the Australian Government Medicare Scheme. Eligibility criteria for the HWS include a BMI- z score > 97th percentile; diagnosed pre-diabetes (i.e., an impaired fasting glucose and impaired glucose tolerance), or BMI-z score > 95th percentile and two co-morbidities (i.e. high blood pressure, depression, or obesity-related Table 1 Family Demographic Factors Factor Child/Adolescent Referral BMI-z Child/Adolescent BMI-z at interview Child/Adolescent age at referral (years) Child/Adolescent age at interview (years) PCG BMI n Mean (SD) Range 13 11 13 13 11 2.61 (0.43) 1.85– 3.19 2.48 (0.47) 1.36– 3.06 9.33 (3.48) 4.6– 15.3 10.34 (3.57) 6.4– 17.1 23.5– 52.4 36.33 (10.04) 43.2 (4.63) 36–50 1–10 5 (3.11) PCG Age (years) SEIFA Ranking WA Note. N = 13; BMI range (kg/m2)-Normal weight (18.5–24.9), Overweight (25-29.9), Obesity Class I (30-34.9), Obese class II (35-39.9), Obesity class III SEVERE (≥ 40); PCG = Patient Care Giver; Socio-Economic Indexes for Areas (SEIFA) ranking range from 1 = most disadvantaged to 10 = Least disadvantaged 5 13 Page 3 of 12 joint or bone problems). Participants were recruited using criterion-based and maximum variation sampling methods [34]. The criteria for inclusion were: A) par- ents (both Mother and Father or primary caregiver) of any family that had been referred to the HWS from 2016 to 2018, B) parents of families living in the Perth met- ropolitan area, C) parents of children/adolescents aged under 18 years, and D) parents from families who had not undertaken a HWS program prior to their interview date (i.e., had no experience of this clinical obesity pro- gram for children/adolescents). This sampling approach intended to capture varying socioeconomic backgrounds due to the wide inclusion criteria and because the recruitment program represented the only clinical obe- sity speciality treatment centre for children/adolescents with severe obesity in Western Australia. Eligible parents were mailed an invitation letter outlin- ing the purpose and nature of the study. The first author subsequently contacted all parents or primary caregiver by phone (with the exception of those who returned a written response declining to participate) to gauge inter- est in participating. Parents/primary caregivers of eligible families elected to participate in the study. Participants provided written informed consent prior to their partici- pation in their interview. Participants also provided con- sent for the lead author to access anthropometric data relevant to the child/adolescent (i.e., referral weight and height) as per Ethics approval. All contact and outcomes were recorded (e.g., interview times and dates). In total, 103 families were contacted between Octo- ber 2017 and May 2018 either by return mail or through phone contact with the first author. A final sample of 13 families were interviewed. The sample included eleven parent/caregiver-only interviews with one parent/care- giver from each family, and one interview with parents from two separate families. Participant demographics are listed in Table 1. Parents BMI’s ranged from 23.5 to 52.4 (M = 36.4, SD = 10.20), with two parents in the ‘normal weight’ range, one ‘overweight’, and 10 classified as ‘obese’ according to their BMI. Parents were aged 36 to 50 years (M = 43.2, SD = 4.63) and had children/adolescents aged 6.8 years to 17.1 years at the time of interview (M = 10.34, SD = 3.49). Eight children/adolescents were male and five were female. Two parents interviewed were male and the remaining 11 female. Seven families were intact (parents of the child/adolescent were still in a relationship), two families were separated from the other parent but with new partners, three families were separated from the other parent and the parent was single, and one person interviewed was a grandmother with primary custody of the child/adolescent. Interviews lasted from 42 min to 1 h and 57 min (M = 105 min, SD = 62.85). Saunders et al. BMC Public Health (2023) 23:1176 Page 4 of 12 Data collection A semi-structured interview guide was developed by the authors, all of whom are experienced in conducting semi- structured interviews and familiar with relevant psychol- ogy and obesity literature. All questions were structured in an open-ended format, with follow-up probing ques- tions used for elaboration and clarification to glean as much information from participants as possible [34]. All interviews were conducted by the first author, who was responsible for audio-recording and taking written notes on each discussion [34]. All families who participated in this study were unknown to the lead author. At the con- clusion of the interviews, participants self-reported their own and their child’s current height and weight. Data collection was guided by the principle of prag- matic saturation. Data saturation is typically considered the point where information obtained from participants becomes repetitive, and thus conducting further inter- views is unlikely to yield new information [34]. This work does not (and cannot) claim to represent the views of all parents referred to a clinical obesity intervention pro- gram for children/adolescents with severe obesity; it’s entirely plausible that further themes may have emerged if additional interviews had been conducted. However, a pragmatic perspective to saturation was adopted—data collection was ceased at the point at which, following dis- cussion with all co-authors, a sufficiently detailed account of parents’ experiences had been obtained. Data analysis Audio recordings were reviewed by the first author and transcribed verbatim. Data were analysed using a reflexive thematic analysis process that was assisted by computer analysis data guidelines [34]. Analyses were conducted using the following steps: familiarisation with data through immersion by the lead researcher listening to audio recordings of interviews and reading transcripts; generating initial codes involving identifying and label- ling common themes using NVivo qualitative data analy- sis Software (QSR International Pty Ltd); ordering data within transcripts into meaning units and categorising these meaning units into themes; and assigning emergent code/theme names. Themes were constantly reviewed, consisting of revising extracts and considering whether clear patterns had formed within each theme. Origi- nal transcripts were cross-checked to ensure all themes were represented, using a constant comparison method to ensure meaning units were reflective of identified themes. Themes were then named and defined allowing the identification of sub-themes (authors LS, J.A.D and BJ). An initial thematic framework was identified includ- ing both themes of a semantic (i.e., expressly described) and latent (i.e., meaning was ‘underneath’ what was expressly stated) nature. Analysis involved a series of ‘critical friends’ meetings (i.e., with either two or three co-authors familiar with thematic analysis techniques) to discuss the initial framework, meaning units, themes, and interpretation (authors LS, TB, J.A.D and BJ). This process encouraged reflection and exploration of alter- native meanings and interpretations [35]. Through this process, themes were renamed, reordered, some themes were condensed, and some completely discarded as they did not fit the scope of the project or research question (i.e., the challenges facing parents with severe levels of obesity in assisting their child/adolescent with severe obesity to live a healthy lifestyle). Lastly, the report was written, allowing for additional interpretation of data and consolidation of final themes. Results The aim of this study was to better understand the expe- riences and challenges of parents/caregivers who were not or had chosen not to engage in a clinical tertiary obe- sity intervention program for children/adolescents with severe obesity, in facilitating a healthy lifestyle for their child/adolescent with severe obesity. Three major themes were identified—parents’ own lifelong obesity narrative, parents’ perceived inevitability of child/adolescent lev- els of obesity and parents’ perceived challenges in get- ting medical help for child/adolescent reduction in levels of obesity. Themes, sub-themes, and exemplar meaning units are displayed in Tables 2 and 3, and 4 and described in detail below. Parent lifelong obesity narrative In describing challenges associated with helping their child/adolescent live a healthy lifestyle, many parents described their own history of lifelong issues or chal- lenges with weight and shape. This history included obe- sity in childhood and bullying, parents’ levels of obesity being ‘out of control’, children/adolescents being aware of their parents’ struggle with their own weight, perceived cyclical failed attempts at weight loss, feelings of sad- ness and self-defeating evaluation, and acknowledging the importance of, but avoiding engagement in, health- related behaviours. The parents who reported previous attempts at weight loss all had levels of overweight or obesity at the time of interview. The negative narrative of obesity was long-standing and entrenched for many parents. In describing challenges in supporting their child/adolescent’s healthy lifestyle, par- ents spoke of their own experiences of childhood and their parents’ negative weight-related behaviours, and how this impacted on their own health behaviour beliefs for their own children, such as Mel (P4): I think it made us obsessed about food because there was just no flavour and no texture, and it was just Saunders et al. BMC Public Health (2023) 23:1176 Table 2 Themes, sub-themes and exemplar meaning units reflecting Parental own weight-related factors Theme Parent life- long obesity narrative Sub-theme Parent fam- ily history of weight- related issues influencing in the present Exemplar meaning unit (P7) I grew up with knowing a lot of this stuff. My Mum and my Step-Dad were mega health freaks sort of thing. So, I know all this, but throughout my life there have been things that have brought on depression. That is where I struggle. I know all the healthy things and exercise with growing up with it. (P4) Growing up [Mum] was always doing things. She was doing the rebound exercising, the trampoline in the ‘70s and then at the same time she’d be like, we need to clean our systems out and she’d do this chicory soup thing which was literally lettuce in boiling water in a bowl with a bit of olive oil on the top and salt and that was dinner with a bit of brown bread. We all had to do it….we hated it. (P12) I’m not going to try anymore. I get a bit defeated and yeah - I don’t want to. (P2) I’ve been putting it on hold too going “You know, I can’t be bothered, I haven’t got time”, but I’ve got 30 kilos, or nearly 40 kilos to lose myself to be back to my healthy, what I used to be. I think I’m using – it’s not an excuse, but I keep saying “Next year when I finish I’m going to have more time”. But I feel like I’ve just put my life on hold now for three years. (P4) I haven’t really stuck to anything. So I don’t even know if I can say that I’ve tried. (P12) The soup diet and was doing Jenny Craig for a couple of months when I was younger, Herbalife at one stage when I was about 20, but that wasn’t very long you know. I’m a shock- er you know, I give it a go for a month or a week and that’s it you know. (P4) I have actually never succeeded on a diet. The only diet that worked for me was when his dad and I split up and he hooked up with a ballerina after me and I lost my appetite. (P2) I was so desperate to make that change to lose that weight again, and go “You know what, I’m sick of being… that was after babies, yeah. Not after my last one, that was after [child name], yeah. And I haven’t been back since. I basically just got disheartened, then I went back to my studies full time, so that stopped everything, Weight out of control, not knowing how to get started Parent failed weight-loss attempts impact quite miserable. Well, I think that we all—we’re emotional eaters and I can’t plan… so if something good happens then we celebrate with a treat. If something bad happens we commiserate with a treat and [sister name] and I, both of us have talked about Page 5 of 12 Sub-theme Want to help and fault of genetics Table 3 Themes, sub-themes and exemplar meaning units reflecting Parent perceived inevitability of child with obesity Theme Parent per- ceived inevi- tability of child with obesity Exemplar meaning unit (P6) At the end of the day I will do anything for him. I do want him to be healthier and I could see his weight was gaining. (P12) You know that Danny De Vito and Ar- nold Schwarzenegger movie Twins, well [child name] is Danny De Vito, he got all the shit. (P9) I think it’s a lot of genetics as well…We all do (have an issue with any weight or shape), that’s why I think it’s a lot of genetics as well. (P3) I didn’t want him to be self-conscious about it or ashamed or any of that sort of stuff because it’s going to be part of his life and I don’t want him sort of body-shaming and all that sort of stuff. (P5) [Child name] has obviously got the big issue. We just tell him we don’t want him to be like us…Battle of the bulge. We don’t want him to grow that way. We actually said to him, we don’t want any cause for bullying, we don’t want that extra pressure for him. (P3) Theres lots of things that we miss out on and there’s lots of restrictions in life and I think it’s always having to be the fun police. (P12) They want people to be individual but everybody has still got to stick to the norm, you know, you’ve still got to be, it just shits me sometimes. Acknowl- edgement of the gravity of problem and worry about child future Frustration about their child’s level of obesity this. Parents commonly reported various negative weight- related experiences such as obesity in childhood and a history of weight-related teasing: “[My husband] was bul- lied for his weight when he was younger. But like us, his Mum took him to the doctor, and they just said, you’ll grow, it’s fine. It’s not so fine” (Robyn, P5). Parents also described their feelings about their own current weight, their attitude towards making health related behaviour change, and often the effect this had on their child/adolescent. Parents reported their own weight as having ‘got out of hand’ or not knowing how to get started with weight loss. As Sharon (P2) noted, “And then when they put you on the scales and you’re like “Have I put on 10 more kilos since Christmas? Oh my God. It’s like it’s not happening then”. Parents spoke about know- ing their children had awareness of their (the parent’s) own struggle with their level of obesity. For instance, Robyn (P5) described, “I go to the gym six times a week, and the boys know that”, highlighting her children’s awareness of their parents’ behaviour, and their knowl- edge that she is perpetually aware of her weight status. Parents often spoke about feelings of sadness and self- critical evaluation of their current weight status, such as Karen (P7); “I’m slowly getting more and more over- weight and older and increasing my chances of diabetes and stuff. Knowing that I have been fit and healthier in Saunders et al. BMC Public Health (2023) 23:1176 Sub-theme Attribution of level of obe- sity to medical diagnoses Table 4 Themes, sub-themes and exemplar meaning units reflecting Parent perceived challenges in getting medical help for child weight loss Theme Child comorbid medical diagnosis makes it hard to achieve any child weight loss Exemplar meaning unit (P6) He’s got low muscle tone, he doesn’t really retain a lot. (P4) We struggle because [Child name], with his autism, he’s come a long way with his food now but when he was younger he was very picky and fussy about food, to the point that if he was made to eat something he didn’t like he would literally throw up at the table. (P3) I think because we have been doing everything we can [to reduce child’s obesity], and given [child’s brain condition diagnosis] condition, it’s just something I can’t seem to get on top of. (P5) The GP just said no, it’s okay. We kept going back to different GPs. (P2) We want to help her, because she’s been seeing the paediatrician for two years, but nothing changed. (P8) The doctor wouldn’t give me one [referral]. He said “You don’t need it. What do you need it for?”. And I said “Look at this child”. And so that’s when he said ‘No, we’ll just test the thyroid’. (P3) I will try anything for the better and this is where I got a bit frustrated with [Child Psychologist]. (P4) I would guarantee that there’s a lot of kids on the spectrum that have weight issues because they have sen- sory issues and issues with persisting unless they’re interested. (P5) We went to Doctor [Specialist number 2] because we hadn’t heard from anyone at [weight loss program name], so we were sick of waiting. We went to Doctor [specialist number 1] first, then we went to Doctor [Specialist number 2]. (P5) We went and saw the dietician and he [child] actually got upset at that appointment, but any information that I’ve gotten has been either really obvi- ous information that everybody should know or it’s completely unreasonable, like instead of having fizzy drink buy soda water, make carrot juice and put soda water in carrot juice. (P8) I find them quite dismissive, they just don’t want to listen to what you’ve got to say. (P13) The dietitian’s advice was not very helpful. Well, just, you know, it was common sense. Page 6 of 12 the past and remembering those times”, with several par- ents actively labelling themselves as lazy such as Chrissy (P9), “We’re all lazy”. Some parents, like Alexis (P12), described knowing their weight was a problem but were not actively interested in doing anything to change this “Everything is just emotional. That’s it and sometimes it’s just too hard”. Many parents also described examples of knowing they should, but avoiding engaging in, health- related behaviours with their children and family, like Sharon (P2): With [child name] trying to say “Mum, let’s go walk, let’s go walk”, and I feel like I’m letting her down, because she’s trying so hard to make changes, and going “I want to walk, I want to walk, but I’m like, yeah, “Next week darling, I really can’t focus right now, I’m sorry, leave me alone, I’m on the computer”. Parents described their own history of multiple and repeated attempts at weight loss—attending the gym, restrictive fad diets or dieting, weight loss supplements or pills, weight loss programs, health professional advice, and two parents had undergone or were planning bariat- ric surgery. Many parents spoke of a perceived failure of weight loss as their reason for stopping any health-related behaviour change: I went to the gym for a year I was toning, so that was the good thing about it, but I wasn’t losing anything, and I wanted to go and lose weight. So, I lost a kilo in six months, and I had three days a week a personal trainer, I was going to aerobics classes, I was trying so desperately. Sharon (P2). This narrative of perceived failure of weight loss was described in a cyclical nature—that is, parents tried a health-related behaviour change, perceived failure, stopped the newly adopted behaviour, and returned to their previous unhealthy lifestyle. This cycle was often described as negatively influencing parents’ own cur- rent motivation, where parents described the capacity to change as being too late or too hard, like Alexis (P12): I was walking like an hour a day for five days a week and didn’t lose anything. It just deflates me and now it’s just even worse now because I’m not doing any- thing, that’s my fault. I’ve got my own issues. Parents described examples of acknowledging, but not acting on, their own and their child/adolescent’s level of obesity in instances requiring them to also engage in health behaviours, like Sharon (P2): “They’re [all the kids] always punishing me, because I didn’t do it. They’re always saying “You promised me and we haven’t gone Repeated efforts to get help from Primary care medical profes- sionals - GPS Difficulties with Medi- cal profes- sionals in seeking support and advice for child weight loss Frustration with the negative and unhelpful attitudes of specialists Loss of confi- dence in spe- cialists’ advice Saunders et al. BMC Public Health (2023) 23:1176 Page 7 of 12 walking”. Nearly all the parents with obesity described instances of unhealthy family eating behaviours, often explained through a lens of rationalising these unhealthy options, such as Karen (P7); “We used to have cooked meals like a few years ago. Especially when our fridge died, and I couldn’t afford to get another fridge, that is when we started with the frozen meals. Convenience and money”. Parents also described how their lifetime of obesity had a negative effect on their attempts at health- related behaviour change their child’s current levels of obesity, like Karen (P7), He [child/adolescent] knows about the healthier options because he’s had them in the past, just my motivation and energy. Yes, and just the confidence because I haven’t really learnt the skills throughout my life to do it and stuff as well. In summary, it appeared that in this theme parents described a negative and pervasive lifelong obesity nar- rative which had a detrimental effect on their child, chil- dren, and/or family, and that parents felt was unlikely to change. Perceived inevitability of child with obesity In discussing challenges experienced when supporting their child/adolescent to engage in a healthy lifestyle, parents felt a level of obesity was ‘inevitable’ for their child/adolescent, and that it presented an overwhelming long-term challenge no matter how hard parents tried to change it. At the time of interview, all but one child/ adolescent had BMI-z scores (standardized score for gen- der and age) at or above the 99th percentile. All parents reported previous failed attempts at reducing their child’s level of obesity. Parents described wanting to help but that their motivation had been undermined by a sense of inevitability and hopelessness, feeling ineffectual due to past failures of reducing their child/adolescents level of obesity, ‘bad genes’, feelings of desperation due to past failures at reduction in obesity level, the gravity of their child’s severe obesity, feelings that their child/adolescents level of obesity was out of control, worry about their child’s long-term future, and child awareness of the grav- ity of the severity of obesity, frustration at the effort with no success, unwillingness to restrict child/adolescent diet, and defense of inactivity. Although parents expressed a desire to help their child/ adolescent, it seemed that parents’ motivation had been eroded or undermined by a sense of inevitability, hope- lessness, and feelings of desperation for success in reduc- ing their child\adolescents level of obesity, such as Carly (P3); “I’d do anything to learn more or help more”. Many parents attributed both their child’s severe obesity and often their own to “bad genes”, such as Chrissy (P9): “She’s built like her dad so she’s short and stumpy. She’s got a big belly. She’s got, yeah, unfortunately she’s got the bad genes”. Other parents described these ‘bad genes’ as limiting their capacity to change the child’s trajectory of severe obesity and describing it as being out of their con- trol. Mel (P4), for example, noted: “I’m not sporty. [Dad’s] not sporty. We’re musos. It’s not even in his genes and he hasn’t been - unfortunately we couldn’t encourage him to do something that we don’t know how to do”. All parents acknowledged the gravity of the severe level of obesity for their child and worried about their child’s long-term future, and described feeling sad for their child/adolescent due to their current severe obesity, such as Delia (P8): “I feel sad for her, because I think “Oh God, you’re only six, what are you going to be like when you’re 15, 20, if we don’t get on top of this?”. Many parents also perceived that their child was aware of the gravity of their severe obesity, such as Sharon (P2): “We want to help her, because she’s been seeing the paediatrician for two years, but nothing changed and at the last visit her weight went right up, and I saw her face change, and I thought ‘She’s not happy’. Parents expressed strong feelings of frustration about their child’s level of obesity, which was expressed in sev- eral ways. Parents described the negative impact placed on themselves by the constant awareness and time focused on the severe obesity, such as Carly (P3): “I think it’s also very taxing on my husband and I”. This feeling of frustration also resulted in parent inaction which was counterproductive to goals of reducing their child/ado- lescent’s level of obesity. For instance, parents described an unwillingness to place dietary restrictions on their child as they wanted their child to just be ‘normal child’, such as Robyn (P5): “I just want him to be a normal kid. He’s going to get enough structure in his life, so I just think – I don’t want to have to say, oh, no mate”. Lastly, parents also described this frustration in a defensive manner promoting active rejection of any health-related change, even when faced with serious consequences for their family, such as Alexis (P12): We’re all overweight,. And school is tough. My hus- band gets a bit edgy, not anti, but just…Yeah, like when Department of Children Services – they are going to take him off us [for child obesity level]- we were trying to do as best as we can, you know. Challenges getting medical help for reducing child/ adolescent level of obesity All parents described seeking help in various forms from medical professionals to help their child/adolescents level of obesity. Parents described difficulties with child comorbid medical diagnoses which made it difficult Saunders et al. BMC Public Health (2023) 23:1176 Page 8 of 12 to achieve any reduction in child levels of obesity, and problems in seeking support and guidance from medical professionals. Child comorbid medical diagnoses make it hard to reduce level of obesity Nearly all parents discussed a chronic medical condi- tion for their child/adolescent as an additional layer of complexity alongside their child’s severe obesity. Chil- dren/adolescents health conditions/diagnoses included, respiratory conditions (asthma, allergic rhinitis, sleep apnoea, ear nose and throat issues), neurological abnor- malities (absent posterior bright spot), muscle and joint conditions (low muscle tone, join pain), mental health diagnosis (autism, attention-deficit/hyperactivity disor- der, anxiety, depression), and metabolic conditions (fatty liver disease, high blood pressure). Many parents directly attributed their child’s severe obesity to their child’s comorbid medical diagnoses, and often minimised the role of unhealthy foods and behaviours, such as Sharon (P2): I think it’s her asthma medications in the past, I think it’s a lot to do with her health that’s been unhealthy, her asthma, her allergic rhinitis, it’s con- stant, her hay fever that’s constant, probably her sleep apnoea. So, I don’t think it’s totally just been her diet, I think it’s, because she was active and fit, and she does her school stuff, sport and things. And she’s not one to say “No, I don’t want to go walking”. Parents described wanting to support their child/ado- lescent in living a healthy lifestyle but being unsure how best to engage in health-related behaviour change that would be appropriate for their child’s medical diagnoses, such as Doug (P1): Because with his asthma and everything – I was not an asthmatic kid so I didn’t have a problem. [Mum name] was not an asthmatic kid. Both of us, when we were young we were active and with this we’re trying to work out what we can and what we can’t do and what he can eat and try to help him. Parents reflected on the negative effects of comorbid conditions on their children’s severe obesity and feel- ings of hopelessness about future reductions of this level, even when engaging in making health related behaviour change, such as Carly (P3): He’s got an absent posterior bright spot which is the bit that controls your hypothalamus. So it’s not his fault. The messages aren’t quite – they’re misfiring to tell him that he’s either full or had enough. It slows his metabolism down so even though we do such a restricted diet and we do a lot of exercise, it’s just he’s always going to be carrying a little bit extra. Difficulties with medical professionals in seeking support for reductions in child level of obesity Parents all described seeking health-related behaviour advice from differing medical doctors (often termed “General Practitioners”, or “GPs”, in Commonwealth countries such as Australia) and specialists to address their child/adolescent’s severe obesity, and advice specific for their child/adolescent’s comorbid medical diagnoses. Parents described repeated efforts to get help, but experi- enced minimisation of their children’s health problems by professionals, professionals blocking access to specialist services, feelings of frustration with negative and unhelp- ful attitudes of medical professionals, perceptions that specialist advice did not help, feeling judged by special- ists who assumed parents were unwilling to follow their advice, and loss of confidence in specialist advice. All families described repeatedly seeking help from primary care medical professionals (GPs) when seeking support specific to their child/adolescent’s severe obesity, such as Robyn (P5) “[Child names] been through quite a bit already, with regards to weight, so…Well, he’s been to – so, lots of GP visits, because we’re concerned about it”. Parents often described their worries were minimised by these professionals, like Delia (P8): Over the years I’ve brought it up so many times with different GPs about her weight, and they all just kind of fob it off, like “Oh, she’s not that bad”. I’m like “Well, it’s not normal for a child to be this size”. Parents also described being actively blocked by GPs in seeking additional help from other specialist services: “That’s why I’m thinking “What’s wrong with this kid?”. That’s why I said I wanted to go and see an endocrinolo- gist, but this GP has just said no, just do the blood tests”. Delia (P8). Many families were or had engaged with medical spe- cialists, such as psychologists, paediatricians or dieti- cians, to address their child/adolescent’s severe obesity. Parents often described feeling frustrated with the nega- tive and unhelpful attitudes of these professionals, like Carly (P3), He [child] was probably 18 months old or maybe it was a bit earlier when I got the diagnosis. I thought there was something wrong – he [Paediatrician] said to me, “There’s no name for it. It is what it is. It’s just a birth defect and there’s nothing you can really do, maybe a bit of physio.” So, he [the Paediatrician] was Saunders et al. BMC Public Health (2023) 23:1176 Page 9 of 12 a bit of a douche. Parents described that medical specialists were unable to provide guidance that resulted in any change in health behaviours or reduction in the level of obesity in their child, as suggestions had been tried but had failed, such as Jane (P13): They gave me a food diary, how much portion I should give to [Child name] and just the cycle. But I wasn’t able to get back to the dietician this year. It’s not helpful because what he eats is still the same. So there’s no variety or whatever and it’s just morn- ing tea, lunch or whatever. It’s still the same. There’s nothing…There’s no change at all. Some parents described feeling that specialists assumed parents were unwilling to follow their advice, leading to frustration. Such judgment was often expressed as lead- ing to a loss of confidence in advice from medical profes- sionals, such as Mel (P4): “When you see a nutritionist or a dietician they don’t seem to actually understand about autism, that it’s not just being a brat, it’s an actual - it’s not going to go away”. Discussion The purpose of this study was to better understand the experiences and challenges of parents who were not or had chosen not to engage in a clinical tertiary obe- sity intervention program for children/adolescents in facilitating a healthy lifestyle for their child with severe obesity. Taken together, our findings advance our under- standing of the familial challenges associated with sup- porting healthy lifestyles for children/adolescents with severe obesity by highlighting several broad difficulties. More specifically, these results highlight that the way parents view “weight loss”—or their perceived inability to achieve it—for themselves or their child/adolescent was a key factor limiting their capacity to support health behaviours. In turn, these negative beliefs appeared to reinforce parents’ beliefs about the inevitability of contin- ued severe obesity for their child/adolescent. In addition to these challenges, parents also described difficulties getting support from medical professionals, suggesting that this group may be isolated and missing key allies or supporters of change. Some of the barriers to supporting a child/adolescent with severe obesity living a healthy lifestyle described by the parents with obesity in this study are echoed in the obesity and weight management literature in adults —factors such as previous failed attempts [36], feel- ings that weight loss was out of their control [37], moti- vation problems [38] and body image and self-esteem issues [38]. However, we consider our findings important insofar as they demonstrate that parents with obesity may view health-related behaviour change—for them- selves and also their child/adolescent—through their own lifelong-obesity narrative lens. This narrative, as well as other parental experiences reported in this study (e.g., inevitability), may be due partly to the notion that par- ents have learned or adopted a fixed mindset (i.e. that they believe an attribute to be unchangeable) [39] about their ability to achieve any weight loss, which is under- pinned by a perceived lifetime of cyclical failed attempts of achieving weight loss. Parents’ perception that their child/adolescents obesity was beyond parents control has been discussed previously [6], although our findings are novel in that they highlight this perception among par- ents of children with severe obesity. Parents in our study also commonly felt frustrated that they had failed to promote weight loss in their child/ adolescent, a feeling of failure that has been reported in other studies [14, 26]. It is possible that parents’ feelings of frustration at repeated failed attempts of child/adoles- cent weight loss and internalised weight biases [40], may activate their own lifelong obesity narrative, and may also explain why parents with obesity in this study perceived obesity in their child/adolescent as hopeless and inevi- table. This inevitability belief has been found in studies demonstrating that adults with obesity more strongly endorse the notion that obesity is inherited—a belief that has been shown to be associated with lower physi- cal activity levels and poorer food choices [41]. Thus, our results highlight an important implicit barrier that appears to negatively frame parents’ beliefs about their everyday ability to support their child/adolescent with severe obesity in living a healthy lifestyle. Therefore, in seeking to better understand and address possible mind- set-related barriers, it may be valuable to examine fac- tors associated with changing mindsets—from a fixed to a more growth-focused perspective—in a health context [42]. The results of this study also highlight several problem- atic barriers associated with seeking medical support. First, our results highlight differences to other studies in that parents of children/adolescents with severe obesity often avoid seeking medical help due to being blamed for the level of obesity in their child [26] - this did not appear to be the majority experience for parents in our study. Participants in the present study did express a desire to seek help, but also reported being blocked from seeking further specialist treatment by primary medical care doc- tors (GPs). Authors of other studies have also reported similar issues with healthcare professionals, where par- ents specifically sought referrals to specialists for their child/adolescent with severe obesity as the desired out- come from primary care doctors due to the belief that these doctors did not have the skills to manage treatment Saunders et al. BMC Public Health (2023) 23:1176 Page 10 of 12 of the obesity level [26]. What is also problematic is that even once parents received a referral to a specialist medi- cal practitioner, these specialists were perceived, in some cases, to be judgemental of parents, and the help and guidance they received was often ineffective in changing their child/adolescents’ level of obesity. This judgement from health professionals may stem from professionals’ weight bias or stigma against individuals with obesity [40]. Weight stigma refers to the idea that individuals devalue people because of their weight [40]. Other stud- ies have supported the notion of implicit weight biases held by health professionals working within paediatric weight treatment programs [9]. In addition, other stud- ies have shown that health care specialists may express bias towards parents with obesity of children/adolescents with obesity, rather than the child/adolescent themselves, and blame parents for poor lifestyle adherence [26, 43] or infer parenting failure due to the parent’s own level of obesity [9, 43]. Our results highlight that parents are experiencing significant difficulties in seeking support from medical professionals, creating an additional barrier in supporting their child/adolescent with severe obesity. Clinical implications Our results prompt several important clinical implica- tions. Firstly, clinical intervention teams for severe child/ adolescent obesity are encouraged to consider and inves- tigate the history of parental levels of obesity and health beliefs prior to a child/adolescent and family engaging in a program. This should not preclude parents of families that have longstanding obesity histories; rather, parents may benefit from support around their own—and their child/adolescent’s—obesity narratives and mindsets prior to engaging in a treatment program. This suggestion is in line with literature in treatment of obesity in adults, where psychological barriers such as maladaptive eating patterns and motivation, emotional regulation problems and maladaptive/unhelpful thought patterns are best addressed prior to focussing on treatment for weight loss [44]. Secondly, it appears that one of the factors that complicates help seeking is that seeking help itself was made difficult by medical professionals. Parents per- ceived difficulties in accessing primary care and support from medical professionals, and when they did receive professional support, they often felt negatively judged. These parental perceptions align with literature in pri- mary medical care settings where GP’s identified deficits in their own skills to treat and support parents of a child/ adolescent with severe obesity [45]. These parental per- ceptions also suggest that medical professionals working with severe obesity should be encouraged to consider and address their own weight biases, and where identi- fied, seek appropriate professional education or train- ing. These results highlight the fact that severe levels of obesity are more complicated, and comorbidities further complicate a difficult situation. Therefore, parents may require more targeted specialised help, and specialists may need upskilling and education to be able to address and support these parents to deliver and encourage fami- lies to engage in meaningful long-term changes. Addi- tionally, our results highlight implications for the design and delivery of programs, as they provide insight into the complex experiences and histories that are important to consider alongside program structure and delivery issues. Strengths and opportunities for future research This study differs from existing reports in that the paren- tal opinions are reflective of differing socioeconomic backgrounds (i.e. participants SEIFA rankings ranged from 1 to 10), and importantly, while many studies have been limited to either very young children (2–5 years), young children (6–12 years), or adolescents (13 years plus), the parents in this study had children of varying ages. Further, this population of families—character- ised by children/adolescents who have severe obesity but are not engaged in treatment—are often considered hard to access and are, as a result, not as well under- stood in terms of experiences and challenges compared to the more ‘general’ populations with (less severe) lev- els of overweight or obesity. Additionally, as people with obesity are known to be subject to stigma due to their weight [10, 46], it may be that parents with varying lev- els of obesity are not always willing to participate in research for fear of judgement from researchers [10, 14, 15] or medical professionals [26]—sourcing opinions form this cohort is important in designing future inter- ventions. This study also directly ties the obesity status of parents, and distorted cognitive beliefs about weight loss, as a specific barrier that could be addressed in the design of future interventions for child/adolescents with severe obesity. There are several aspects of this study that could be explored in future research. Firstly, participants resided in the metropolitan area of Perth and therefore its unknown if families in outer areas, or regionally, have experienced similar or additional difficulties. The rise of the use of online platforms may enable future research to reach more families. Secondly, the role of parents inter- nalised weight biases and experiences of weight stigma which is often reinforced by interactions with health- care and society, should be investigated on a larger scale. Lastly, whilst the lead researcher invited families from a large recruitment pool (N = 103), only 13 families con- sented to participate. Further, both parents were invited to attend interviews, only one parent attended per fam- ily, and most commonly mothers. Future research could investigate in a larger sample, whether fathers or whole family perspectives (both fathers and mothers or pri- mary caregivers) may present different perspectives on Saunders et al. BMC Public Health (2023) 23:1176 challenges in supporting their child/adolescent with severe obesity. Conclusion For some parents, negative underlying fixed cognitive views of weight loss may prohibit active and commit- ted support of their child/adolescent’s journey of weight loss. Adding to these cognitive orientations, parents in this study reported difficulties, negative judgments, and poor advice from medical professionals in terms of their child/adolescent’s level of obesity. Within the literature there is a negative and stigmatizing narrative that par- ents ignore the gravity of their child/adolescent’s level of obesity or ignore the advice from specialists. Importantly, this study offers an alternative perspective—that medi- cal professionals may be contributing to barriers and parents’ feelings of internalised hopelessness about their child/adolescent’s level of obesity. Consideration of this perspective is crucial, and clinically relevant to child and adolescent treatment programs for severe obesity. Acknowledgements The authors would like to thank the Perth Children’s Hospital, Healthy Weight Service and its team members, who allowed the lead author to work alongside them in their important work, but also for the unwavering support from all team members in this project. Additionally, the authors would like to thank the Perth Children’s Hospital, Child and Adolescent Mental Health Acute Services, Paediatric consultation Liaison Program who provided in-kind support for this project. Authors’ contributions All authors were involved in the conception and planning of this study. Author LS designed, recruited, conducted all interviews, analysed and interpreted participant data and prepared the manuscript. Authors LS, LG, JD, JAD and BJ contributed to the interview guide. Authors TB, JAD and BJ were involved in analysis and contributed to the editing of the manuscript. All authors read and approved the final manuscript. Funding The lead author was supported by the Australian Research Training Program, funded through The University of Western Australia. The subsequent authors report there are no funding interests to declare. Data Availability Due to the nature of this research, participants of this study did not agree for their data to be shared publicly (as participants individual privacy could be compromised), so supporting data is not available. Please contact liz. saunders@research.uwa.edu.au for further details. Declarations Competing interests The authors declare no competing interests. Ethics approval and consent to participate All methods were carried out in accordance with relevant guidelines and regulations Ethical approval to conduct this study was obtained from the Perth Children’s Hospital, Child and Adolescent Health Services Human Research Committee (CAHS HRC – PRN: RGS0000002542) and The University of Western Australia Human Research Ethics Committee (PRN: RA/4/1/8478). Participants provided written informed consent prior to their participation in their study and for the lead author to access anthropometric data relevant to the child (i.e., referral weight and height). Page 11 of 12 Consent for publication Participants provided written informed consent prior to their participation in their interview. Participants also provided consent for the lead author to access anthropometric data relevant to the child (i.e., referral weight and height). Received: 12 September 2022 / Accepted: 28 April 2023 References 1. Hales CM, Freedman DS, Akinbami L, Wei R, Ogden CL. Evaluation of alterna- tive body Mass Index (BMI) Metrics to Monitor Weight Status in Children and Adolescents with extremely high BMI using CDC BMI-for-age growth charts. Vital Health Stat. 2022;1(197):1–42. Kelsey MM, Zaepfel A, Bjornstad P, Nadeau KJ. Age-related consequences of childhood obesity. Gerontology. 2014;60(3):222–8. Juonala M, Magnussen CG, Berenson GS, Venn A, Burns TL, Sabin MA, et al. Childhood adiposity, adult adiposity, and Cardiovascular Risk factors. N Engl J Med. 2011;365(20):1876–85. 3. 2. 4. Horesh A, Tsur AM, Bardugo A, Twig G. Adolescent and childhood obesity and excess morbidity and mortality in Young Adulthood—a systematic review. Curr Obes Rep. 2021;10(3):301–10. 5. Gibson LY, Allen KL, Davis E, Blair E, Zubrick SR, Byrne SM. The psychosocial burden of childhood overweight and obesity: evidence for persisting difficul- ties in boys and girls. Eur J Pediatrics. 2017;176(7):925–33. 6. Davidson K, Vidgen H. Why do parents enrol in a childhood obesity manage- ment program?: a qualitative study with parents of overweight and obese children. BMC Public Health. 2017;17(1):159. Roberts KJ, Polfuss ML, Marston EC, Davis RL. Experiences of weight stigma in adolescents with severe obesity and their families. J Adv Nurs. 2021;77(10):4184–94. Puhl RM, Lessard LM. Weight stigma in Youth: prevalence, Consequences, and considerations for clinical practice. Curr Obes Rep. 2020;9(4):402–11. 7. 8. 9. Halvorson EE, Curley T, Wright M, Skelton JA. Weight Bias in Pediatric Inpatient Care. Acad Pediatr. 2019;19(7):780–6. 10. Rubino F, Puhl RM, Cummings DE, Eckel RH, Ryan DH, Mechanick JI, et al. Joint international consensus statement for ending stigma of obesity. Nat Med. 2020;26(4):485–97. 11. Puder JJ, Munsch S. Psychological correlates of childhood obesity. Int J Obes. 12. 2010;34(S2):37–43. Jacobson D, Melnyk BM. Psychosocial correlates of healthy beliefs, choices, and behaviors in overweight and obese school-age children: a primary care healthy choices intervention pilot study. J Pediatr Nurs. 2011;26(5):456–64. 13. Franks PWP, Hanson RLMDMPH, Knowler WCMDD, Sievers MLMD, Bennett PHMBF, Looker HCMBBS. Childhood obesity, other Cardiovascular risk factors, and premature death. N Engl J Med. 2010;362(6):485–93. 14. Douglas F, Clark J, Craig L, Campbell J, McNeill G. It’s a balance of just getting things right”: mothers’ views about pre-school childhood obesity and obesity prevention in Scotland. BMC Public Health. 2014;14:1009. 15. Gibson LY, Byrne SM, Zubrick SR. Childhood obesity: what we have learnt from our failure to engage mothers in our intervention. Aust N Z J Public Health. 2013;37(4):389. 16. Loos RJF, Yeo GSH. The genetics of obesity: from discovery to biology. Nat Rev Genet. 2022;23(2):120–33. 17. Mitchell GL, Farrow C, Haycraft E, Meyer C. Parental influences on children’s eating behaviour and characteristics of successful parent-focussed interven- tions. Appetite. 2013;60(0):85–94. 18. Schmelzer L, Krishnagiri S. Exploring the complexities of promoting health in families in an obesogenic environment: Occupation, participation and health. OTJR. 2014;34(2):61–71. 19. Sung-Chan P, Sung YW, Zhao X, Brownson RC. Family-based models for childhood-obesity intervention: a systematic review of randomized con- trolled trials. Obes Rev. 2013;14(4):265–78. 20. Steele RG, Aylward BS, Jensen CD, Cushing CC, Davis AM, Bovaird JA. Com- parison of a family-based group intervention for youths with obesity to a brief individual family intervention: a practical clinical trial of positively fit. J Pediatr Psychol. 2012;37(1):53–63. Saunders et al. BMC Public Health (2023) 23:1176 Page 12 of 12 21. Jansen E, Mulkens S, Jansen A. Tackling childhood overweight: treating parents exclusively is effective. Int J Obes. 2011;35(4):501–9. 22. Ryder JR, Fox CK, Kelly AS. Treatment options for severe obesity in the Pediatric Population: current Limitations and Future Opportunities. Obesity. 2018;26(6):951–60. 23. Hagman E, Danielsson P, Lindberg L, Marcus C, Committee tBS. Paedi- atric obesity treatment during 14 years in Sweden: Lessons from the swedish childhood obesity treatment Register—BORIS. Pediatr Obes. 2020;15(7):e12626. 24. Faus AL, Leibowitz KL. Predictors of Attrition From a Behavior Change–Based Pediatric Lifestyle Changes Program for Overweight/Obese Youth. Clin Pedi- atr. 2014. 25. Dhaliwal J, Perez AJ, Holt NL, Gokiert R, Chanoine J-P, Morrison KM, et al. Why do parents discontinue health services for managing paediatric obesity? A multi-centre, qualitative study. Obes Res Clin Pract. 2017;11(3):335–43. 26. Turner KM, Salisbury C, Shield JPH. Parents’ views and experiences of child- hood obesity management in primary care: a qualitative study. Fam Pract. 2011;29(4):476–81. 27. Veloso S, Matos M, Palmeira A, Martins S, Fonseca H, Diniz J. Weight Manage- ment Program based on self-determination theory: comparing Parents-Child Data. Psicologia Saúde e Doenças. 2015;16(1):86–99. 28. Watson PM, Dugdill L, Pickering K, Hargreaves J, Staniford LJ, Owen S et al. Distinguishing factors that influence attendance and behaviour change in family-based treatment of childhood obesity: A qualitative study. Br J Health Psychol. 2020;n/a(n/a). 29. Putter K, Jackson B, Thornton A, Willis C, Goh K, Beauchamp M et al. Percep- tions of a family-based lifestyle intervention for children with overweight and obesity: a qualitative study on sustainability, self-regulation, and program optimization. BMC Public Health. 2022;22. 35. Smith B, McGannon KR. Developing rigor in qualitative research: problems and opportunities within sport and exercise psychology. International Review of Sport and Exercise Psychology. 2018;11(1):101–21. 36. Burke LE, Swigart V, Warziski Turk M, Derro N, Ewing LJ. Experiences of Self- Monitoring: successes and struggles during treatment for weight loss. Qual Health Res. 2009;19(6):815–28. 37. Greener J, Douglas F, van Teijlingen E. More of the same? Conflicting perspec- tives of obesity causation and intervention amongst overweight people, health professionals and policy makers. Soc Sci Med. 2010;70(7):1042–9. 38. Herriot AM, Thomas DE, Hart KH, Warren J, Truby H. A qualitative investigation of individuals’ experiences and expectations before and after complet- ing a trial of commercial weight loss programmes. J Hum Nutr Dietetics. 2008;21(1):72–80. 39. Dweck CS, Leggett EL. A Social-Cognitive Approach to Motivation and Personality. Psychol Rev. 1988;95(2):256–73. 40. Puhl RM, Himmelstein MS, Pearl RL. Weight stigma as a psychosocial con- tributor to obesity. Am Psychol. 2020;75(2):274–89. 41. Wang C, Coups EJ. Causal beliefs about obesity and associated health behav- iors: results from a population-based survey. Int J Behav Nutr Phys activity. 2010;7(1):19–p. 42. Orvidas K, Burnette JL, Schleider JL, Skelton JA, Moses M, Dunsmore JC. Healthy body, healthy mind: a mindset intervention for obese youth. J Genet Psychol. 2020;181(6):443–57. 43. Silver MP, Cronin SM. Health Care Providers’ perspectives on Family Compli- ance and Behavior Change in a childhood obesity program. Health Educ Behav. 2019;46(4):582–91. 44. Garip G, Yardley L. A synthesis of qualitative research on overweight and obese people’s views and experiences of weight management. Clin Obes. 2011;1(2–3):110–26. 30. Denzin NK, Lincoln YS. Strategies of qualitative inquiry. 4th ed. ed. Thousand 45. King LA, Loss JHM, Wilkenfeld RL, Pagnini DL, Booth ML, Booth SL. Australian Oaks: SAGE Publications; 2013. 31. Finlay L, Gough B. Reflexivity: a practical guide for researchers in health and social sciences. Malden, MA: Blackwell Science; 2003. 32. Budden T, Dimmock JA, Smith B, Beauchamp MR, Rosenberg M, Jackson B. Overweight and obese men’s experiences in a sport-based weight loss intervention for men. Psychol Sport Exerc. 2020;50:101750. 33. Budden T, Dimmock JA, Beauchamp MR, Rosenberg M, Fitzpatrick I, Jackson B. MAN v FAT Soccer: feasibility trial and preliminary efficacy of a sport-based weight loss intervention for overweight and obese men. Journal of Sport and Exercise Psychology. In press. 34. Sparkes AC, Smith B. Qualitative research methods in Sport, Exercise and Health: from process to product. London, UNITED KINGDOM: Routledge; 2013. GPs’ perceptions about child and adolescent overweight and obesity: the weight of Opinion study. Br J Gen Pract. 2007;57(535):124–9. 46. Puhl RM, Heuer CA. The stigma of obesity: a review and update. Obesity. 2009;17(5):941–64. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Saunders et al. BMC Public Health (2023) 23:1176
10.1186_s12909-023-04345-7	Cardalda‑Serantes et al. BMC Medical Education (2023) 23:365 https://doi.org/10.1186/s12909‑023‑04345‑7 BMC Medical Education RESEARCH Open Access Would anti‑choking devices be correctly and quickly managed by health science students? A manikin crossover trial Borja Cardalda‑Serantes1, Aida Carballo‑Fazanes2,3,4* and Antonio Rodríguez‑Núñez2,3,4,5,7 , Emilio Rodríguez‑Ruiz2,4,5, Cristian Abelairas‑Gómez2,3,4,6 Abstract Background The brand‑new anti‑choking devices (LifeVac® and DeCHOKER®) have been recently developed to treat Foreign Body Airway Obstruction (FBAO). However, the scientific evidence around these devices that are available to the public is limited. Therefore, this study aimed to assess the ability to use the LifeVac® and DeCHOKER® devices in an adult FBAO simulated scenario, by untrained health science students. Methods Forty‑three health science students were asked to solve an FBAO event in three simulated scenarios: 1) using the LifeVac®, 2) using the DeCHOKER®, and 3) following the recommendations of the current FBAO protocol. A simulation‑based assessment was used to analyze the correct compliance rate in the three scenarios based on the correct execution of the required steps, and the time it took to complete each one. Results Participants achieved correct compliance rates between 80–100%, similar in both devices (p = 0.192). Overall test times were significantly shorter with LifeVac® than DeCHOKER® device (36.6 sec. [31.9–44.4] vs. 50.4 s [36.7–66.9], p < 0.001). Regarding the recommended protocol, a 50% correct compliance rate was obtained in those with prior training vs. 31.3% without training, (p = 0.002). Conclusions Untrained health science students are able to quickly and adequately use the brand‑new anti‑choking devices but have more difficulties in applying the current recommended FBAO protocol. Keywords Airway clearance, FBAO, LifeVac®, DeCHOKER®, Nursing and medical students, Simulation 6 Faculty of Education Sciences, Universidade de Santiago de Compostela, Santiago de Compostela, Spain 7 Pediatric Critical, Intermediate and Palliative Care Section, Pediatric Department. Hospital, Clínico Universitario de Santiago de Compostela, Santiago de Compostela, Spain *Correspondence: Aida Carballo‑Fazanes aida.carballo.fazanes@usc.es 1 Anesthesiology and Intensive Care Medicine Department, University Clinic Hospital of Santiago de Compostela (CHUS), Galician Public Health System (SERGAS), Santiago de Compostela, Spain 2 CLINURSID Research Group, University of Santiago de Compostela, Santiago de Compostela, Spain 3 Primary Care Interventions to Prevent Maternal and Child Chronic Diseases of Perinatal and Developmental Origin (RICORS), Instituto de Salud Carlos III, RD21/0012/0025 Madrid, Spain 4 Simulation, Life Support, and Intensive Care Research Unit (SICRUS), Health Research Institute of Santiago de Compostela (IDIS), Santiago de Compostela, Spain 5 Intensive Care Medicine Department, University Clinic Hospital of Santiago de Compostela (CHUS), Galician Public Health System (SERGAS), Santiago de Compostela, Spain © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Cardalda‑Serantes et al. BMC Medical Education (2023) 23:365 Page 2 of 8 Background Foreign body airway obstruction (FBAO) is a medical emergency that represents the fourth leading cause of potentially preventable and treatable accidental death both at home and in the community [1, 2]. Also, it has been reported as a leading cause of death in 1 to 3 aged kids, ahead of traffic accidents [3, 4]. Kids younger than 3 years old together with people over 65 years old and patients with musculoskeletal and neurologic conditions represent the main population at risk [3, 5]. In case of a FBAO event, early intervention by bystand- ers in out-of-hospital setting has been associated with a better neurological prognosis for the victim [2–6]. In this sense, scientific societies have urged to provide first aid training to parents, education professionals, kids, and elderly caregivers [5, 7]. In addition, previous stud- ies suggest the need for adequate training in this field for laypeople, health science students, and healthcare profes- sionals. The latter are also responsible of broadcast infor- mation about activities and other preventive measures to both their patients and their caregivers [7], therefore, seems essential that their training should be adequate. ineffective and the victim The current recommended protocol to treat FBAO combines back blows and abdominal thrusts for resus- citation of a choking victim, progressing to cardio- pulmonary resuscitation (CPR) if those manoeuvres are loses consciousness [1, 8]. Recently, anti-choking devices (LifeVac® and DeCHOKER®) have been developed to treat FBAO as a second step in the face of the ineffectiveness of standard manoeuvres [9–11]. Besides, these devices are widely available for general population use [12]. Currently, these externally applied, portable, non-powered, suction-gen- erating devices are only registered as Class 1 FDA ‘suc- tion apparatus’ [9, 13, 14]. There is limited high-quality scientific evidence about these devices to support or dis- approve them [10, 13]. Considering the challenge to carry out high-quality research in this field [15], with the hypothesis that peo- ple without training are able to use the brand-new anti- choking devices, this study aimed to assess the ability of health sciences students (Medicine and Nursing) to man- age LifeVac®, DeCHOKER® and the recommended pro- tocol in a simulated adult FBAO scenario. Methods Aim, design and setting of the study This manikin crossover trial study carried out at the Uni- versity of Santiago de Compostela aimed to assess the ability to use the brand-new anti-choking devices (Life- Vac® and DeCHOKER®) in an adult FBAO simulated scenario, by untrained health science students. Participants A convenience sample of 43 health science students (nursing and medical students in any year of their degree) from the University of Santiago de Compostela without prior training in anti-choking devices took part in this study. Before the tests, all the participants signed an informed consent, explaining the study’s aims, agreeing to give up their data for research purposes (always treated anonymously), and informing them that they could leave the study at any time. The study was conducted under the amended Declaration of Helsinki. The Research Ethics Committee of Santiago-Lugo did not consider it neces- sary to review the research protocol since it is a simula- tion study. Procedure A manikin randomized crossover study was performed. Three FBAO events were simulated to each partici- pant, who without prior training, was encouraged to resolve them with 1) LifeVac® device (LifeVac® test), 2) DeCHOKER® device (DeCHOKER® test), and 3) fol- lowing the recommendations of the current protocol for action. The start of these scenarios was randomized for the participants using a random generator. For assessing the anti-choking devices, an adult mani- kin (Little Anne QCPR™; Laerdal) was used as a simu- lated FBAO victim. Participants had to try to resolve it only with the help of the manufacturer’s leaflet instruc- tions, which were provided on paper as they are in real- ity accompanying the corresponding anti-choking device. However, for the test which evaluates the recommended protocol, a real victim who simulated a FBAO event (first mild and finally severe obstruction) was used. Par- ticipants had no instructions beyond their knowledge to solve it. No training was performed before each test, nor was any information provided to them during the tests; letting them act as if they were alone in the FBAO scenario. Data, related to the performance or non-performance and correct or incorrect executions of each step of the above-mentioned tests, were collected in a specific checklist by one researcher while another recorded the time spent on each step and the overall time test. Materials Two anti-choking devices (LifeVac® and DeCHOKER®) and a manikin (Little Anne QCPR™; Laerdal) were used in our study. LifeVac® device consists of a mask with a patented valve to create a seal [9, 13, 14]. The second component, connected to the first one through a one-way valve, con- sists of the plunger. This plunger, when compressed, will Cardalda‑Serantes et al. BMC Medical Education (2023) 23:365 Page 3 of 8 cause a unidirectional suction phenomenon that will try to dislodge the foreign body from the airway, preventing it from moving deeper into the airway [16] It includes 3 types of interchangeable masks: a small paediatric mask (children between 1–4 years old weighing more than 10 kg), a large paediatric mask (children over 4 years old), and one for adults [5, 17, 18]. DeCHOKER® device consists of a plunger-type sys- tem, responsible for generating the negative pressure and unidirectional suction current necessary to dislodge the foreign body (solid or liquid) and clear the airway. Unlike LifeVac®, it also has an oropharyngeal component, which simulates an oropharyngeal airway [13]. It is available in three different sizes: infants (between 1-5 years old), chil- dren (between 5-12 years old) and adults (from 12 years, including wheelchair patients and pregnant) [19]. Little Anne QCPR™ (Laerdal) mannequin was used as a simulated FBAO victim for the resolution of the two tests which evaluate both anti-choking devices. Variables Characteristics of the participants (age, sex and under- graduate degree) were recorded. In addition, data on their knowledge and subjective assessment of their abil- ity to act in the event of a FBAO situation were collected, as well as whether they had witnessed and/or acted in a FBAO event on some occasion and when. The primary variables of this study were: the proper execution of each of the steps required in the handling of the anti-choking devices and in the recommended pro- tocol (categorical variables) and the time (in seconds) it took to resolve the scenarios (continuous variable). The correct compliance rate (%) was calculated according to the next equation (Ʃ steps correctly performed × 100)/ number of steps assessed). LifeVac® correct compliance rate was calculated given correct or incorrect execution of the items of this sequence: 1) insert the mask on the device’s bellows, 2) place the mask correctly covering the victim’s nose and mouth, 3) fix the mask to the victim’s airway, 4) push in the handle/bellows, 5) pull the handle upwards, 6) keep the mask fixed to the victim’s airway throughout the procedure. DeCHOKER® correct compliance rate was calculated by evaluating correct or incorrect execution of the items of this sequence: 1) place the device correctly, 2) fix the mask to the victim’s airway, 3) pull the plunger out with force, 4) keep the mask fixed to the victim’s airway in place throughout the procedure. The current recommended protocol of action cor- rect compliance rate was calculated taking into account the correct or incorrect execution of the items of this sequence: 1) encourages coughing, 2) performs back blows, 3) correctly performs back blows, 4) performs abdominal thrusts, 5) correctly performs abdominal thrusts, 6) continues 5 back blows × 5 abdominal thrusts, 7) correctly continues 5 × 5, 8) indicates initiation of CPR manoeuvres in case of unconsciousness (1). These vari- ables were compared between participants who had prior training in FBAO recommended protocol and partici- pants who had no previous training. Lastly, at the end of each test, participants were ques- tioned on a subjective variable, the election of one of the two anti-choking devices (LifeVac® and DeCHOKER®). Statistical analysis First, a descriptive analysis was performed. Categorical variables were expressed with frequencies and percent- ages. Continuous variables were expressed with median and interquartile range (IQR), according to their adjust- ment to a non-normal distribution (Shapiro–Wilk test). When comparing categorical variables, Chi-square sta- tistic was performed, or Fisher’s Exact Test when the number of cells with expected values ≤ 5 was over 20%. Comparisons between LifeVac® and DeCHOKER® con- tinuous variables were performed using the Wilcoxon signed-rank test and between participants with and without training in the recommended protocol with the Mann Whitney U test. Analysis was performed using the SPSS statistical software (IBM corp., v. 25.0 for Mac), and for all analyses, a p-value of less than 0.05 was statistically significant. Results Overall, 43 health science students from the University of Santiago de Compostela participated in the study: 24 nursing (55.8%) and 19 medical students (44.2%). Char- acteristics of the participants are presented in Table 1. Thirty-one (72.1%) participants referred to had previous training on the current recommended protocol. Before the study 28 (65.1%) participants considered themselves capable of resolving an FBAO event. However, only 5/43 participants had witnessed and 3/5 of them had acted on a FBAO event (Table 1). A descriptive analysis of the participants’ performance with LifeVac® and DeCHOKER® anti-choking devices is shown in Table 2. Even though the median estimated correct compliance rate with both devices is 100%, only 62.8% of participants performed all steps correctly with the LifeVac® device vs. 81.4% with the DeCHOKER® device (p = 0.125). Although there were no significant differences, “to keep the mask fixed to the victim’s air- way throughout the procedure” was the most failed step in both tests (74.5% with LifeVac® vs. 86.0% with DeCHOKER®; p = 0.164). Cardalda‑Serantes et al. BMC Medical Education (2023) 23:365 Page 4 of 8 Table 1 Characteristics of the participants Variables Age (years) Sex Male Female Degree Nursing Medicine Prior training in FBAO Yes No Participants n = 43 21.0 (21.0 – 23.0) 27 (62.8) 16 (37.2) 24 (55.8) 19 (44.2) 31 (72.1) 12 (27.9) Years since training 3.0 (1.0 – 3.0) If you witness a FBAO, would you be able to solve it? Yes No Have you ever witnessed a FBAO? Yes No Years since having witnessed the FBAO Have you intervened when the FBAO? (n = 5) Yes No 28 (65.1) 15 (34.9) 5 (11.6) 38 (88.4) 8.0 (3.3 – 10.5) 3 (60.0) 2 (40.0) kg kilogram; m meters; FBAO Foreign Body Airway Obstruction Continuous variables are expressed with median (interquartile range) Categorical variables are expressed with absolute frequency (relative frequency) Table 3 shows the comparative analysis of test times and estimated correct compliance rate between both anti-choking devices. Overall test times were signifi- cantly shorter when using LifeVac® compared with DeCHOKER® (36.6 s. [31.9 – 44.4] vs. 50.4 s. [36.7 – 66.9], p < 0.001). Participants achieved high and similar correct compliance rates with both devices. Regarding the current recommended protocol per- formance (Table 4), none of the untrained participants performed all the steps, so the overall test time is signifi- cantly shorter compared with trained participants (42.7 vs. 56.5 sec., p = 0.002). The correct compliance rate is significantly lower in those participants without prior training (31.3% vs. 50%, p = 0.002). Concerning the steps of the current recommended protocol, less than 50% of the participants encouraged the victim to cough. Although 71.1% of the participants performed interscapular clapping and 95.3% performed abdominal thrusts, only 51.6% and 17.1% were performed correctly, respectively. Almost all participants who failed to perform the back blows and the abdominal thrusts did so by administering an incorrect number of them. Although 55.8% of the participants continued with the 5 black blows × 5 abdominal thrust sequence, it was only correctly performed in 16.7% of the tests, in most cases by executing abdominal thrusts only (Table 4). Finally, participants were asked about their opinion on which anti-choking device they would choose after hav- ing used both of them, 55.8% of the participants chose the LifeVac® device. Discussion In our study, we have tried to evaluate, in a simulated FBAO scenario, the handling of brand-new anti-choking devices by health science students, as well as observe how they would solve the event following the recommended protocol. In general, the participants have shown greater ease in performing the skills required in the use of anti- choking devices than in handling the currently FBAO recommended protocol. Although these devices are not yet recommended by resuscitation guidelines [1], they are available to every- one in public places such as airports, shopping centers or schools. Among the scarce and low-quality scientific evidence available to date, there are only two studies that compare both anti-choking devices [20, 21]. The remain- ing studies, despite reporting a high success rate of air- way clearance, only independently evaluate either the LifeVac® device [8–11, 14, 22–24] or the DeCHOKER® device [25]. Besides, only a few studies are case series [8, 9, 23, 25], albeit with a very small sample (n ≤ 29). The rest [11–14, 22, 24], are simulation studies on manikins, except for Juliano et al. [14] who conducted a study on cadavers. In our study, we did not evaluate efficacy but the correct execution technique and times in the differ- ent tests. We consider it necessary to evaluate first the knowledge and practical skills in the resolution of the FBAO event both with the devices and with the recom- mended protocol, before assessing their efficacy in terms of successful foreign body removal. In general, we observed a correct execution of all steps recommended by the manufacturers in both devices and no differences between them. The most failed step was, also in Carballo-Fazanes et al. [20] study, “to keep the mask fixed to the victim’s airway throughout the procedure” although more than 70% of the participants achieved this with both devices. This is reflected in a high estimated correct compliance rate with both devices, being slightly higher in the case of the DeCHOKER®. Regarding the overall time of each scenario was signifi- cantly longer with DeCHOKER® than with LifeVac®, as well as Carballo-Fazanes et al. [20] study. These dif- ferences, based on our findings and the participants’ comments, may be related to the relative clarity of the LifeVac® instructions for use. In this regard, our results are in line with those obtained by Patterson et al. [21], they observed a significantly higher success rate of FBAO Cardalda‑Serantes et al. BMC Medical Education (2023) 23:365 Page 5 of 8 Table 2 Descriptive analysis of the participants’ performance with LifeVac® and DeCHOKER® devices during an adult victim FBAO Variables LifeVac® DeCHOKER® Place the mask correctly covering the victim’s nose and mouth Yes No Fix the mask to the victim’s airway Yes No Push in handle Yes No Pull handle (LifeVac®) // Pull the plunger out with force (DeCHOKER®) Yes No Keep the mask fixed to the victim’s airway throughout the procedure Yes No Perform all the steps correctly Yes No Correct compliance rate Continuous variables are expressed with median (interquartile range) Categorical variables are expressed with absolute frequency (relative frequency) FBAO Foreign Body Airway Obstruction p-values calculated by Chi‑square test or Fisher’s Exact Test as appropriate a Wilcoxon test p-value 1.000 0.007 ‑ ‑ 0.164 0.125 38 (88.4) 5 (11.6) 39 (90.7) 4 (9.3) 42 (97.7) 1 (2.3) 43 (100.0) 0 (0.0) 32 (74.4) 11 (25.6) 27 (62.8) 16 (37.2) 42 (97.7) 1 (2.3) 41 (95.3) 2 (4.7) ‑ ‑ 43 (100.0) 0 (0.0) 34 (86.0) 6 (14.0) 35 (81.4) 8 (18.6) 100 (80.0 – 100.0) 100 (100.0 – 100.0) 0.192a Table 3 Comparison of procedure time and compliance rate between LifeVac® and DeCHOKER® Variables Time to device fitting on the victim Total time Correct compliance rate Data are expressed as median (interquartile range) p‑values calculated by Wilcoxon test LifeVac® 29.3 (25.7 – 37.5) 36.6 (31.9 – 44.4) 100 (80.0 – 100.0) DeCHOKER® 36.8 (26.2 – 49.2) 50.4 (36.7 – 66. 9) 100 (100.0 –100.0) p-value ‑ < 0.001 0.192 removal in less than 60 sec. with LifeVac® than with DeCHOKER® (82.2% and 44.4% respectively). By contrast, the current recommended protocol of action for FBAO treatment turned out to be less well known by health science students. The most unknown items of the protocol for our participants were to encourage the victim to cough and the number of back blows and abdominal thrusts. Related to this, we obtained a correct compliance rate of 50%, this rate drops to 31.3% in the case of participants with no prior training. We hypothesise that this could be related to the fact that, despite that some of the participants reported previous training and they are health science students, the current recommended protocol has more steps than anti-choking devices procedures. In addi- tion, anti-choking device tests (which are prepared for laypersons, according to their manufacturers) were car- ried out following the manufacturer’s leaflet instruc- tions, while recommended protocol test was performed without instructions, following their knowledge to solve it. In this sense, Patterson et al. [21] observed a success rate of FBAO removal of 66.7% in less than 60 sec. with the abdominal thrust procedure. This could also mean that performing the recommended protocol properly is more difficult than using the devices. However, in their study, they compared the efficacy and usefulness of both devices (LifeVac® and DeCHOKER®) with only the Cardalda‑Serantes et al. BMC Medical Education (2023) 23:365 Page 6 of 8 Table 4 Descriptive analysis of the participants’ performance of current recommended steps to treat an adult victim with FBAO Overall (n = 43) Prior training (n = 31) No prior training (n = 12) χ2 p-value Variables Encourage to cough Yes No Give 5 back blows Yes No Give black blows correctly Yes No Give 5 abdominal thrusts Yes No Give abdominal thrusts correctly Yes No Yes No Continue to 5 back blows and 5 abdominal thrusts correctly Yes No Start BLS for unconscious victim Yes No Perform all the steps Yes No Correct compliance ratea Time to back blows (sec) 21 (48.8) 22 (51.2) 31 (71.1) 12 (27.9) n = 31 16 (51.6) 15 (48.4) 41 (95.3) 2 (4.7) n = 41 7 (17.1) 34 (82.9) 24 (55.8) 19 (44.2) n = 24 4 (16.7) 20 (83.3) 25 (58.1) 18 (41.9) 7 (16.3) 36 (83.7) Continue to 5 back blows and 5 abdominal thrusts 17 (54.8) 14 (45.2) 27 (87.1) 4 (12.9) n = 27 14 (51.9) 13 (48.1) 30 (96.8) 1 (3.2) n = 30 7 (23.3) 23 (76.7) 20 (64.5) 11 (35.5) n = 20 4 (20.0) 16 (80.0) 20 (64.5) 11 (35.5) 7 (22.6) 24 (77.4) 50.0 (37.5 – 62.5) 50.0 (50.0–62.5) 13.2 (10.3 – 19.0) 13.5 (10.5–19.2) Time to abdominal thrust (sec) 22.5 (14.2 – 29.1) 26.1 (19.2–31.0) Total time (sec) 54.4 (42.9 – 67.6) 56.5 (51.9–71.0) Continuous variables are expressed with median (interquartile range) Categorical variables are expressed with absolute frequency (relative frequency) 4 (33.3) 8 (66.7) 4 (33.3) 8 (66.7) n = 4 2 (50.0) 2 (50.0) 11 (91.7) 1 (8.3) n = 11 0 (0.0) 11 (100.0) 4 (33.3) 8 (66.7) n = 4 0 4 (100.0) 5 (41.7) 7 (58.3) 1.601 0.206 12.429 < 0.001 0.005 0.945 0.509 0.476 3.095 0.079 3,411 0.065 0.960 0.327 1.856 0.173 3.237 0.072 0 (0.0) 12 (100.0) 31.3 (12.5–50.0) 9.4 (6.3–13.2) 14.1 (9.5–19.8) 42.7 (40.7–51.2) ‑ ‑ ‑ ‑ U = 76,000; p = 0.002b U = 26,000; p = 0.107b U = 58,000; p = 0.001b U = 77,000; p = 0.002b FBAO Foreign Body Airway Obstruction, BLS Basic Life Support, sec seconds, U Mann–Whitney U test value p‑values calculated by Chi‑square test a Score calculated according to correct or incorrect execution of the eight steps of the sequence b Mann‑Whitney U test abdominal thrusts step of the recommended protocol. Also, more participants reported prior training in BLS (94.4%) than in our study (72.1%). However, we cannot assume that students are well trained in this field. This study has not tested whether pre-test training would contribute to better results, since we wanted to analyze the first impression and performance with the devices without prior knowledge. However, we hypoth- esise that this might be the same thing as with other basic life support content, such as automated external defibrillators [26] or adult and child CPR [27–29], where brief pre-test training showed better results. We asked participants to give their subjective opin- ion of the devices once they had completed the tests, more than 50% preferred LifeVac®. Although they found DeCHOKER® to be safer, LifeVac® resulted in more intuitive, and practical and had clearer instructions. In previous studies, LifeVac® was also considered superior in terms of ease of use, safety, and confidence by partici- pants [21]. Cardalda‑Serantes et al. BMC Medical Education (2023) 23:365 Page 7 of 8 Simulation-based assessment was applied in this study considering its benefits in the emergency medicine set- ting, as it allows the opportunity to practice clinical skills in a risk-free environment, acquiring special relevance in health science students as it helps to gain self-confidence and willingness in their performance in real clinical set- tings [30–32]. Our study has some limitations to be considered. First, the relatively small and convenience sample and the imbalance between the number of participants with and without prior training in FBAO protocol. COVID-19 pandemic restrictions have made it difficult to recruit a bigger sample, so our findings may be difficult to general- ize. Second, the use of a manikin model in a simulated scenario may not directly translate results to real FBAO events. In this sense, participants may act differently in real-life situations since simulations, especially those involving rapid intervention, often have a Hawthorne effect involving changes in their behaviour because they know they are being observed. On the other hand, a CPR manikin was used (FBAO non-specific) since there is currently no manikin prepared to use the anti-choking devices. The three scenarios (LifeVac® test, DeCHOKER® test, and recommended protocol test) were not com- pared because the correct execution of the recommended protocol technique could not be evaluated with the com- mercially available manikin. Therefore, recommended protocol test was performed on a human victim. Finally, no washout period has been established between the sce- narios due to the existing differences among them; how- ever, the start of the tests has been randomized so that, in case of a possible learning bias, they are equally affected. Conclusions Untrained Medicine and Nursing students are able to quickly and adequately use the brand-new anti-choking devices (LifeVac® and DeCHOKER®) but fail to apply the current recommended FBAO protocol. This must be considered to define the role of such devices in the case of FBAO and to train future healthcare professionals to adequately manage FBAO events. Abbreviations BLS CPR FBAO IQR Basic Life Support Cardiopulmonary Resuscitation Foreign Body Airway Obstruction Interquartile Range Acknowledgements We are grateful for the contribution of the participants to our study. Authors’ contributions BC‑S, AC‑F and ER‑R organised and were involved in data collection. AC‑F conducted the statistical analysis. CA‑G and AR‑N conceived the study and supervised the process. All authors contributed to writing the original draft of the manuscript and reviewed, edited and accepted the final version. Funding The authors have no financial disclosures to report. Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The need for ethical approval was waived by the Research Ethics Committee of Santiago‑Lugo because it did not involve the use of participants’ health data, the collection of biological samples, or intervention of participants. All participants, who were over 18 years of age, signed an informed consent form before they participated in the study. The informed consent included informa‑ tion on the aims of the study, the possibility of leaving the study at any time they wished, and the transfer of data anonymously for research purposes. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Received: 1 September 2022 Accepted: 10 May 2023 References 1. Olasveengen TM, Semeraro F, Ristagno G, Castren M, Handley A, Kuzovlev A, et al. European Resuscitation Council Guidelines 2021: Basic Life Sup‑ port. Resuscitation. 2021;161:98–114. https:// doi. org/ 10. 1016/j. resus citat ion. 2021. 02. 009. 2. Norii T, Igarashi Y, Sung‑Ho K, Nagata S, Tagami T,Yoshino Y, et al. Protocol for a nationwide prospective, observational cohort study of foreign‑ body airway obstruction in Japan: the MOCHI registry. BMJ Open. 2020;10(7):e039689. https:// doi. org/ 10. 1136/ bmjop en‑ 2020‑ 039689. 3. Duckett SA, Waggoner BC, Bartman M, Roten RA. Choking. In: StatPearls. 4. Treasure Island (FL): StatPearls Publishing 2020. https:// www. ncbi. nlm. nih. gov/ books/ NBK49 9941/. Accessed 2 Feb 2022. Instituto Nacional de Estadística (INE). Defunciones según la Causa de Muerte [Deaths by Cause of Death]. Madrid, Instituto Nacional de Estadís‑ tica (INE) (Spain), 2018. https:// www. ine. es/ jaxiT3/ Datos. htm?t= 10803# !tabs‑ tabla. Accessed 2 Feb 2022. 5. Dodson H, Cook J. Foreign Body Airway Obstruction (FBAO). In: StatPearls . Treasure Island (FL): StatPearls Publishing 2020. https:// www. ncbi. nlm. nih. gov/ books/ NBK55 3186/. Accessed 2 Feb 2022. 6. Patel PB, Shapiro NL. Portable, non‑powered, suction‑generating device for management of life‑threatening aerodigestive tract foreign bodies: Novel prototype and literature review. Int J Pediatr Otorhinolaryngol. 2019;118:31–5. https:// doi. org/ 10. 1016/j. ijporl. 2018. 12. 014. Fano C, Lorenzoni G, Azzolina D, Giuliani A, French M, Campagna S, et al. Perception of Choking Injury Risk Among Healthcare Students. J Community Health. 2019;44(5):974–81. https:// doi. org/ 10. 1007/ s10900‑ 019‑ 00662‑5. 7. 8. McKinley MJ, Deede J, Markowitz B. Use of a Novel Portable Non‑pow‑ 9. ered Suction Device in Patients With Oropharyngeal Dysphagia During a Choking Emergency. Front Med (Lausanne). 2022;8:742734. https:// doi. org/ 10. 3389/ fmed. 2021. 742734. Saperstein DM, Pugliesi PR, Ulteig C, Schreiber N. Successful use of a novel device called the LifeVac to resuscitate choking victims ‑ world‑ wide results. Int J Clin Skills. 2018;12:216–9. https:// doi. org/ 10. 4172/ Clini cal‑ Skills. 10001 36. 10. Dunne CL, Peden AE, Queiroga AC, Gomez Gonzalez C, Valesco B, Szpil‑ man D. A systematic review on the effectiveness of anti‑choking suction devices and identification of research gaps. Resuscitation. 2020;153:219– 26. https:// doi. org/ 10. 1016/j. resus citat ion. 2020. 02. 021. Cardalda‑Serantes et al. BMC Medical Education (2023) 23:365 Page 8 of 8 simulated and actual pediatric cardiac arrest. Resuscitation Plus 6:100117. https:// doi. org/ 10. 1016/j. resplu. 2021. 100117. 30. Griswold S, Fralliccardi A, Boulet J, Moadel T, Franzen D, Auerbach M, et al. Simulation‑based education to ensure provider competency within the health care system. Acad Emerg Med. 2018;25(2):168–76. https:// doi. org/ 10. 1111/ acem. 13322. 31. Philippon AL, Truchot J, De Suremain N, Renaud MC, Petit A, Baron GL, Freund Y. Medical students’ perception of simulation‑based assessment in emergency and paediatric medicine: a focus group study. BMC Med Educ. 2021;21(1):586. https:// doi. org/ 10. 1186/ s12909‑ 021‑ 02957‑5. 32. Holmboe E, Rizzolo MA, Sachdeva AK, Rosenberg M, Ziv A. Simulation‑ based assessment and the regulation of healthcare professionals. Simul Healthc. 2011;6(Suppl):S58‑62. https:// doi. org/ 10. 1097/ SIH. 0b013 e3182 283bd7. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub‑ lished maps and institutional affiliations. 11. Lih‑Brody L, Lih A, Brody E, Singer M. LifeVac: a novel apparatus to resusci‑ tate a choking victim. Am J Gastroenterol. 2015;110:S695. 12. Rodriguez‑Ruiz E, Abelairas‑Gómez C, Barcala‑Furelos R, Rodríguez‑Núñez A. Foreign body airway obstruction and anti‑choking suction devices Time to step forward. Resuscitation. 2020;157:133–4. https:// doi. org/ 10. 1016/j. resus citat ion. 2020. 09. 038. 13. Van de Voorde P, de Lucas N. A “foreign body” in the “foreign body airway obstruction” algorithm. Resuscitation. 2020;153:258–9. https:// doi. org/ 10. 1016/j. resus citat ion. 2020. 05. 030. 14. Juliano M, Domingo R, Mooney MS, Trupiano A. Assessment of the Life‑ Vac, an anti‑choking device, on a human cadaver with complete airway obstruction. Am J Emerg Med. 2016;34(8):1673–4. https:// doi. org/ 10. 1016/j. ajem. 2016. 03. 047. 15. Couper K, Hassan AA, Ohri V, Patterson E, Tang HT, Bingham R, et al. Removal of foreign body airway obstruction: A systematic review of interventions. Resuscitation. 2020;156:174–81. https:// doi. org/ 10. 1016/j. resus citat ion. 2020. 09. 007. 16. Consejo Español de Resucitación Cardiopulmonar (CERCP). Comuni‑ cado del CERCP acerca de los nuevos dispositivos publicitados para la desobstrucción de la vía aérea [CERCP statement on the new devices advertised for airway clearance]. Madrid, Consejo Español de Resuci‑ tación Cardiopulmonar (Spain), 2020. https:// www. cercp. org/ area‑ ciuda dana/ area‑ ciuda dana‑2/ 630‑ notic ia‑ 20190 426‑1‑ ciuda danos. Accessed 10 Mar 2022. 17. Ojeda Rodriguez JA, Ladd M, Brandis D. Abdominal Thrust Maneuver (Heimlich). 2021. In: StatPearls. Treasure Island (FL): StatPearls Publishing 2020. https:// www. ncbi. nlm. nih. gov/ books/ NBK53 1467/. Accessed 10 Mar 2022. 18. LifeVac® leaflet: Dispositivo de emergencia en Asfixia por Atragantami‑ ento [Emergency Device in Choking]. 2019. https:// lifev ac. es/ wp‑ conte nt/ uploa ds/ 2019/ 05/ Folle to‑ LifeV ac‑ 2019. pdf. Accessed 10 Mar 2022. 19. Dechoker LLC. Dechoker – Because every second counts. https:// www. decho ker. com/. Accessed 10 March 2022. 20. Carballo‑Fazanes A, Abelairas‑Gómez C, Rodríguez‑Ruiz E, Barcala‑ Furelos R, Rodríguez‑Núñez A. Anti‑choking suction devices use A pilot simulated study with parents and kindergarten teachers. Resuscitation. 2022;177:5–6. https:// doi. org/ 10. 1016/j. resus citat ion. 2022. 06. 005. 21. Patterson E, Tang HT, Ji C, Perkins GD, Couper K. The efficacy and usability of suction‑based airway obstruction: a manikin randomised crossover trial. Resuscitation Plus. 2021;5:100067. https:// doi. org/ 10. 1016/j. resplu. 2020. 100067. 22. Lih‑Brody L, Singer M, Brody E. Lifevac: a novel apparatus for the resuscita‑ tion of the adolescent choking victim. Ann Emerg Med. 2017;70:S149–50. 23. Lih‑Brody L. The World Congress of Gastroenterology. ACG 2017. Suc‑ cessful resuscitation of choking victims using a Lifevac, a non‑powered portable suction device: real world experience: 2017 President Poster Award: 1762. Orlando: Am J Gastroenterol; 2017. https:// lifev ac. es/ wp‑ conte nt/ uploa ds/ 2019/ 05/ 03‑ The‑ World‑ Congr ess‑ of‑ Gastr oente rolog y‑. pdf. Accessed 10 Mar 2022. 24. Timerman S, Giannetti N, Costa A, Fachioli T, Kalil R. Device for the resusci‑ tation of the choking victim. Resuscitation. 2020;155:S22. https:// doi. org/ 10. 1016/j. resus citat ion. 2020. 08. 069. 25. Bhanderi BG, Hil SP. Evaluation of Dechoker, an airway clearance device (ACD) used in adult choking emergencies within the adult care home sector: a mixed methods case study. Front Public Health 8:541885. https:// doi. org/ 10. 3389/ fpubh. 2020. 541885. 26. Yeung J, Okamoto D, Soar J, Perkins GD. AED training and its impact on skill acquisition, retention and performance‑a systematic review of alternative training methods. Resuscitation. 2011;82(6):657–64. https:// doi. org/ 10. 1016/j. resus citat ion. 2011. 02. 035. 27. Anderson R, Sebaldt A, Lin Y, Cheng A. Optimal training frequency for acquisition and retention of high‑quality CPR skills: A randomized trial. Resuscitation. 2019;135:153–61. https:// doi. org/ 10. 1016/j. resus citat ion. 2018. 10. 033. 28. González‑Salvado v, Rodríguez‑Ruiz E, Abelairas‑Gómez C, Ruano‑Raviña A, Peña‑Gil C, González‑Juanatey JR, Rodríguez‑Núñez A. Training adult laypeople in basic life support. A systematic review. Rev Esp Cardiol 2020;73:53‑68. https:// doi. org/ 10. 1016/j. rec. 2018. 11. 013. 29. Donoghue A, Heard D, Griffin R, Abbadessa ML, Gaines S, Je S, et al. Lon‑ gitudinal effect of high frequency training on CPR performance during • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12909-021-02802-9	Guraya et al. BMC Medical Education (2021) 21:381 https://doi.org/10.1186/s12909-021-02802-9 R E S E A R C H A R T I C L E Open Access Preserving professional identities, behaviors, and values in digital professionalism using social networking sites; a systematic review Shaista Salman Guraya1,2* and Muhamad Saiful Bahri Yusoff2 , Salman Yousuf Guraya3 Abstract Background: Despite a rapid rise of use of social media in medical disciplines, uncertainty prevails among healthcare professionals for providing medical content on social media. There are also growing concerns about unprofessional behaviors and blurring of professional identities that are undermining digital professionalism. This review tapped the literature to determine the impact of social media on medical professionalism and how can professional identities and values be maintained in digital era. Methods: We searched the databases of PubMed, ProQuest, ScienceDirect, Web of Science, and EBSCO host using (professionalism AND (professionalism OR (professional identity) OR (professional behaviors) OR (professional values) OR (professional ethics))) AND ((social media) AND ((social media) OR (social networking sites) OR Twitter OR Facebook)) AND (health professionals). The research questions were based on sample (health professionals), phenomenon of interest (digital professionalism), design, evaluation and research type. We screened initial yield of titles using pre-determined inclusion and exclusion criteria and selected a group of articles for qualitative analysis. We used the Biblioshiny® software package for the generation of popular concepts as clustered keywords. Results: Our search yielded 44 articles with four leading themes; marked rise in the use of social media by healthcare professionals and students, negative impact of social media on digital professionalism, blurring of medical professional values, behaviors, and identity in the digital era, and limited evidence for teaching and assessing digital professionalism. A high occurrence of violation of patient privacy, professional integrity and cyberbullying were identified. Our search revealed a paucity of existing guidelines and policies for digital professionalism that can safeguard healthcare professionals, students and patients. * Correspondence: ssalman@rcsi-mub.com 1Royal College of Surgeons Ireland, RCSI - MUB, Busaiteen, Bahrain 2Department of Medical Education, School of Medical Sciences, University Sains Malaysia, Health campus, Kelantan, Kota Bahru, Malaysia Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Guraya et al. BMC Medical Education (2021) 21:381 Page 2 of 12 Conclusions: Our systematic review reports a significant rise of unprofessional behaviors in social media among healthcare professionals. We could not identify the desired professional behaviors and values essential for digital identity formation. The boundaries between personal and professional practices are mystified in digital professionalism. These findings call for potential educational ramifications to resurrect professional virtues, behaviors and identities of healthcare professionals and students. Keywords: Professionalism, Digital professionalism, Professional identity, Professional behaviors, Professional values, Professional ethics, Social media, Social networking sites, Health professionals Background Social media is based on a collection of digital platforms whose content is created, edited and shared by its clients themselves [1]. The expeditious development of social media has transformed the way healthcare professionals and students interact with each other [2]. Facebook, Twitter, LinkedIn, YouTube, Instagram, Wikis, Blogs, Podcasts and WeChat are the most popular social media worldwide [3]. Medical professionalism is a multi- dimensional construct that refers to a set of skills and the professionals are expected to competencies that practice [4] The crossroads of medical professionalism and the use of social media has created a new facet of digital e- professionalism, that reflects the manifestation of trad- itional professional attitudes and behaviors through so- cial media [5]. Digital professionalism refers to the professionals’ use of digital media and the mechanisms in which the profession is evolved by this use [6]. interchangeable with professionalism, The concept of digital professionalism in e-health em- braces the core values that can steer teaching, learning and practice domains in medical disciplines through online platforms. A safe application of digital profession- alism includes professional competence, reputation, and responsibility [7]. Digital media provides enormous interconnectivity that has expanded our range of oppor- tunities for sharing information. However, this unprece- dented opportunity has created interdependency on social media, devices, and users with loss of natural pauses for self-reflection in our livelihood. The ubiquity and easy access of digital media permits free communi- cations that has the potential to thrust its contents into the medical practice. The fluid and complex nature of medical professional virtues, behaviours and identities are more vulnerable in the current era of digital professionalism [8]. Professional virtues and behaviours illustrate the processes of how the professionals enact their role, while professional identity involves an oath for adhering to the values and ethics of medical profession associated with the profession such as being trustworthy, competent, and safe medical practi- tioner. Medical professional identity requires the prac- ticing physician to act as a professional at individual, interpersonal and societal levels [9]. On the hand, digital professional identity pertains to a wide range of distinct personal and professional acts that are manifested in the digital space [10]. Unfortunately, literature has reported erosions of professional identities and behaviours in the current era of digital professionalism [11]. Inappropriate social media behavior has also shown detrimental effects on medical and health sciences students’ approach to- wards humanism, empathy and altruism [5]. Unauthorized postings of patient health information, pictures, patient- doctor communication blogs, and images with clear pa- tient identification are commonly witnessed unprofes- sional behaviors. This practice has blurred personal and professional boundaries in the medical sphere. In digital professionalism, medical educators and policymakers are skeptical about preserving patient confidentiality and priv- acy on social media [12]. There are growing concerns about the absence of a structured program for digital professionalism in the medical and health sciences [13, 14]. In addition, there is literature that can help understand the a paucity of safe-guarding medical professionals’ mechanisms for identities and values in the digital world [15]. This sys- tematic review aimed to review the available body of knowledge that can help identify key concepts and threats to professional identity in the era of digital professionalism. Methods Research questions Our research questions were based on Sample, Phenomenon of Interest, Design, Evaluation and Re- search type (SPIDER) [16] as shown in Table 1. We framed the following specific questions for our systematic review; I. What are the desired values and behaviors of digital professionalism that are needed for maintaining digital professional identity? II. What is the impact of social media on medical professionalism? III. How can values and behaviors of digital professionalism be made measurable and reproducible in teaching and assessment? Guraya et al. BMC Medical Education (2021) 21:381 Page 3 of 12 Table 1 Selection criteria for the studies in this systematic review using SPIDER (n = 44) Variables Inclusion criteria Sample Phenomenon of Interest Design Evaluation Research type - Health professionals -Medical and nursing undergraduate and postgraduate students and/or residents -Physicians and fellows -Medical educators or school administrators -Medical school websites -Respondents from medicine, pharmacy, dentistry, and nursing Digital professionalism, Professional behaviors, attitudes and identity in digital world Questionnaire/Survey Interview Focus group Observational studies Ethnography Content analysis Views Experiences Opinions/Attitudes/Perceptions/Beliefs/Ideas Knowledge/Understanding Behaviours -Qualitative -Quantitative -Mixed method S Sample; PI Phenomenon of Interest; D Design; E Evaluation; R Research type Search strategy In May 2020, we searched the databases of PubMed, ProQuest, ScienceDirect, Web of Science, and EBSCO host using keywords (professionalism AND (profession- alism OR (professional identity) OR (professional behav- iors) OR (professional values) OR (professional ethics))) AND ((social media) AND ((social media) OR (social networking sites) OR Twitter OR Facebook)) AND (health professionals) terms and text words for the English language articles published during 1st January 2015 till 30th April, 2020. The search focused on titles about definitions, analyses and relationships of health professionals about professional identity, virtues, behav- iors, medical professionalism and digital professionalism. PubMed was the mainstay to systematically develop a search string, which was later extrapolated to other data- bases. All selected keywords were searched in the fields “Abstract” and “Article Title” (alternatively “Topic”) and in MeSH/Subject Headings/Thesaurus where available. Language, document type, and publication year restric- tions were instead included in the exclusion criteria for the screening process. We defined healthcare profes- sionals in undergraduate, graduate, and continuing edu- cation, postgraduates and practicing physicians/nurses, deans, directors, and faculty. For this study we defined healthcare professionals as individuals who may be in- volved in healthcare delivery (for example: physicians, nurses, dentists, physiotherapists, and pharmacists). A full search log, including detailed search strings for all included information sources, results and notes is avail- able in Appendix I. Data collection, eligibility criteria and the selection of articles We used the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines was used for data mining and selection of the studies for this systematic review [17]. The original research articles that conducted qualitative, quantitative and mixed methods studies about definitions of digital professionalism, e-professionalism in the digital age, guidelines for the usage of social media, and the degree and extent of usage of social media by health professionals for educational, professional and personal purposes were included. The participants of the selected studies were medical and allied health sciences students, physicians, fac- ulty and program directors. We excluded systematic re- views, meta-analysis, editorials, and commentaries from our search. The studies about professionalism in non-medical fields were also excluded. SS reviewed the titles and abstracts of the studies re- trieved during initial search and grouped relevant arti- cles for possible inclusion. Then we reviewed full text of the selected articles for their further matching with the inclusion criteria. To mitigate research bias, the entire search process was reviewed by SS, SYG and MSBY. We resolved research disagreements and disputes through discussions until we reached a consensus. Data extraction and synthesis This step included review of information from the arti- cles, publication year, author, country of study, single study level, health professionals’ center/multicenter, Guraya et al. BMC Medical Education (2021) 21:381 Page 4 of 12 discipline, ethical approval, methodology, study purpose, results and Medical Education Research Study Quality Instrument (MERSQI) [18] score (Appendix II). The data was organized in charts for the descriptive analysis of the quantity and quality of the selected studies. We performed thematic analysis using emerging con- cepts and theories from the selected studies, which gen- erated different concepts. The leading themes and concepts were further analyzed in discussion to reach consensus for future implementations. We coded the findings of the selected articles and constructed a coding tree. Later, all researchers critically analyzed preliminary themes, which refined the coding process and helped in adding more strings such as assessment and policy about digital professionalism. We also used biblioshiny® from R Statistical Package to carry out bibliometric analysis [19]. Using the hierarchical clustering strategy, we la- belled each keyword as a cluster item, and then merged clusters with maximum similarity into a large new clus- ter. Finally, the multiple cluster analysis was graphically generated for review. Quality assessment We used the MERSQI tool for the evaluation of studies of quantitative educational research. The MERSQI checklist has 10 items in six domains: study design, sampling, type of data, validity evidence, data analysis, and type of out- comes with a maximum score of three in each domain. A study can have a maximum MERSQI score of 18 (highest quality). SS individually scored each study and in case of score discrepancies, SYG re-assessed the scoring and the results were cross verified among researchers. Quality assurance All (SS, SYG and MSBY) objectively researchers reviewed the workflow for the selection of studies. In case of discrepancies, the researchers reached consensus by comparing the studies with inclusion criteria and key words. The discrepancies, inconsistencies and controver- sies were resolved with consensus until all the concerns were resolved. Results Initial search retrieved 4,055 titles, and after eliminating duplicates and retaining only English language publica- tions, we included 1,319 for further abstract analysis. During the exclusion phase, 1,277 titles were excluded as they could not meet the inclusion criteria. Lastly, 126 full text articles were excluded from the remaining 170 publications. After full paper review, we included 44 arti- cles in our systematic review for deeper analysis. The en- tire process using PRISMA guidelines is illustrated in Fig. 1. The yearly publication pattern of the selected 44 arti- cles about professional identity, behaviors and virtues in the digital world is shown in Fig. 2. A maximum number of 10 articles were published in 2016. From a different perspective, the graphical representa- tion of countries of origin of the selected 44 studies is displayed in Fig. 3. Most studies were based in the USA (15/34 %), while other studies were based in Canada (8/18 %), UK (4/9 %), China (3/7 %), UAE (3/7 %), and New Zealand (1/3 %). Most commonly used methodologies were cross-sectional surveys (27/61 %) and analysis of the publicly available Internet con- tent such as Facebook profiles, Twitter streams, or blogs (8/ 18 %). Other methods used in the selected studies included focus group discussions, mixed-methods by semi-structured interviews and survey. Of note, of all the survey-based stud- ies, about half of these studies had response rates of 50 % or greater, while three studies either did not explicitly report a response rate. For the 11 studies that analyzed the publicly available Internet content, four (36 %) did not mention any methods to increase study rigor of data extraction and ana- lysis expected of content analyses. in this systematic Interestingly, 32 (73 %) studies had clear ethical state- ments either with institutional board approval, exemption, or undertaking that ethical approval was not necessary. Table 2 shows descriptive analysis of the data about med- ical disciplines and study levels from the selected 44 stud- ies study populations, 23 (52 %) involved postgraduates and/or resi- dents, physicians and fellows, and practising nurses, 18 (41 %) included undergraduate students (medical, dental or nursing) while 7 (6 %) conducted studies involved dean, directors, and faculty. Approximately 50 % of survey-based studies had response rates of 50 % or more, and postgrad- uates & practicing physicians/nurses were the most com- mon group of studied participants. In terms of review. A graphical relationship among the selected keywords for our systematic review using the bibliometric analysis is illustrated in Fig. 4. The plane distance between key- words reflects the degree of similarity and commonal- ities among them. The keywords approaching centre of the figure indicate that they have received high attention in the recent years. Probity received maximum attention, while cybercivilty received least attention and similarity with other keywords. The analysis of MERSQI showed that 32 quantitative studies had average score of 12, while other 12 qualita- tive studies did not qualify for quality check. A max- imum number of 14 studies had a primary research objective of exploring the beliefs and attitudes of the participants towards usage of social media use and pro- fessional behaviours. Other leading research objectives of the selected studies in our systematic review are out- lined in Table 3. Guraya et al. BMC Medical Education (2021) 21:381 Page 5 of 12 Fig. 1 The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) flow diagram for the selection of studies in this systematic review Fig. 2 The yearly publication pattern of articles about professional identity, behaviors and virtues in the digital world during 2015–2020 (n = 44). This search was conducted in May 2020, which explains lower number of articles in 2020 Guraya et al. BMC Medical Education (2021) 21:381 Page 6 of 12 Fig. 3 The country-wise pattern of articles published about professional identity, behaviors and virtues in the digital world during 2015–2020 (n = 44) Our systematic review generated four main themes; I. Usage of social media by healthcare professionals and students [1, 3, 11, 20–42]. II. The impact of social media on medical professionalism [20, 25–29, 31, 33–35, 37, 40, 41, 43–50]. III. Blurring of professional values, behaviors, and identity in the digital era [5, 11, 20, 22, 26, 28, 29, 31, 32, 35, 37, 42, 43, 46, 48, 50–56]. IV. Limited evidence for teaching and assessing professionalism in the digital era [5, 11, 20, 27, 29, 31, 32, 34, 42, 48, 49, 54, 55]. By and large, the usage of social media by health pro- fessionals has escalated during the last decade [1, 3, 11, 20–42, 45–49], there is a negative impact of social media usage on medical professionalism as reflected by erosion of professional integrity [20, 25, 29, 33–35], an upsurge of awareness about professional identity but rise in un- professional behaviors in the digital era [11, 20, 22, 29, 32, 35, 47, 48, 51, 52, 57] and some evidence of en- hanced acquisition of knowledge about digital profes- sionalism by in curricula [5, 20, 27, 29, 32, 34, 42, 48, 49, 54, 55]. incorporating structured modules Table 2 Descriptive analysis of the data about medical disciplines and study levels from the selected studies in this systematic review (n-44) Features Analysis Disciplines Medicine Nursing Dentistry Pharmacy Physiotherapy Study level Postgraduates & Practicing physicians/nurses Undergraduate Dean, directors, and faculty No. 27 9 5 2 1 23 18 7 Discussion This systematic review reports a rapid rise in the usage of social media by healthcare professionals and students with a negative impact of social media as reflected by substan- tial unprofessional behaviors leading to blurred profes- sional identities. There is a compelling evidence that the awareness of social media by healthcare professionals and students is getting better, nevertheless, there is a recipro- cal increase in the prevalence of unprofessional behaviors in the digital era. We could find limited and unsatisfactory data about the appropriate acquisition of knowledge and structured curriculum for teaching and assessment of digital professionalism. Unfortunately, this review could not identify the desired values and behaviors of digital professionalism that are needed for maintaining digital professional identity. Of all traits of medical professional- ism, probity found highest attention in the studies selected in our systematic review. Probity in medical disciplines is an ever needed professional characteristic that enriches the faculty-student and physician-patient relationships with elements of honesty and trust [58]. The four leading themes of this systematic review are elaborated in the following parts of discussion. Theme I: Usage of social media by health professionals and students The use of social media is among the most innovative but, unfortunately, the most destructive necessary evil of the current era. Currently, more than 40 % of the health care consumers use social media for their healthcare needs worldwide [59]. In medical education, social media is being increasingly used for learning and teaching, research, hos- pital care quality, and for assessment of online behaviour of healthcare professionals [60]. Only in the USA, nearly 65 % of the adult population use social media for different rea- sons and this usage has sharply risen in the last decade [61]. Understandably this usage is ubiquitous among young adults (90 %) and notable among older adults (77 %), this Guraya et al. BMC Medical Education (2021) 21:381 Page 7 of 12 Fig. 4 Bibliometric analysis illustrating the cluster and multiple interconnections of frequently used keywords difference being reflected by being digital native and digital immigrants, respectively. In medical education, 94 % of medical students, 79 % of medical residents, and 42 % of practicing physicians use social media [62]. This exponential growth in social media usage provides health information, facilitate live chat platforms for patient-to-patient and patient-to-health professional, data collection on patient perspectives, health promotion and education, and offer telemedicine for online consultations and treatments [43, 63]. Use of physician-bloggers has also risen that foster sharing of health information and marketing campaigns. Cognizant with this rise in usage of social media, medical educators, physicians, and students are util- izing contents of social media regardless of its accur- acy and authenticity. Table 3 Leading research purposes of the selected studies in this systematic review (n-44) Study purposes Explore beliefs and attitudes regarding social media use and professional behaviours Quantify and evaluate professional digital media use Describe professional and personal information and activities, perceptions of online professional behavior and opinions on guidelines in this area. Examine the effects of an educational intervention to assess students’ change in social media use practices Determine educational use of Social media Identify and characterize the types of unprofessional and concomitant personal and institutional risks. Describe appropriate patient-physician relationship on social media. Describe the characteristics of professional/unprofessional online posts or tweets Describe perceptions of confidentiality, accountability, and e-professionalism Describe relationship between anonymity and professionalism n = 44 14 7 5 4 3 3 3 2 2 1 Guraya et al. BMC Medical Education (2021) 21:381 Page 8 of 12 Theme II: The impact of social media on medical professionalism Research has provided compelling evidence that social media has bipolar effect on professionalism [3] and this has led to erosion of professionalism integrity [56, 64, 65]. In a multi-site survey-based study by Garg et al., most of the identified unprofessional behaviors grouped as high-risk-to-professionalism events (HRTPE) were re- ported by residents [50]. The investigators have detailed that HRTPE included posting identifiable patients’ demographics, a clinical or radiological image, and in- appropriate pictures of intoxicated colleagues or unpro- fessional remarks. The study has eluded that such events pose substantial threats to the healthcare professionals and their associated institution. Laliberté at al., have cau- tioned that due to blurred boundaries between profes- the Facebook sional and unprofessional friendship can potentially lead to mixing of professional and personal lives [28]. The occurrence of such phe- nomena is more vulnerable in hospital departments that provide intense and lengthy sessions such as rehabilita- tion centers. territories, There is an apparent dissonance between the med- ical students’ understanding of e-professionalism while using social media and being aware of its impact of losing professional identity [31]. Social media is con- sidered Powerful, Public and Permanent and the im- pact of risk of these three Ps potentially carries disseminating misleading and inaccurate information particularly if influenced by confliction and biased in- terests [35]. Research has diligently proven that habit- ual use of social networking sites adversely affects behavioural relations [66]. In addition, there are grow- ing concerns about negative impact of social media such as extroversion, loneliness, eccentric personality characteristics and social dissonance. Conversely, literature has reported some benefits of use of social media by health professionals; self-directed learning by staying current, listening to patients’ opin- ions and needs, and patient education can potentially lead to better patient care [27, 33, 37, 41, 43, 44]. The use of social media offers valid opportunities to enhance engagement, effective feedback, professional collabor- ation and competence [40]. Chretien et al., have reported that, using social media, the medical students seemingly benefit from listening to patient perspectives and tend to embrace deeper cultural knowledge [26]. At the same time, using virtual patient communities on various inter- faces of social media can enrich the students’ under- standing of the patient perspectives [27]. Interestingly, non-hand held devices (desktop, laptop) have been shown to have a better impact on development of pro- fessional values and behaviors than hand-held devices (mobile phone, iPad, tablets) [56]. Theme III: Blurring of professional values, behaviors, and identity in the digital era Our review has identified a wealth of unprofessional be- haviors that the researchers have welded with social media in the digital world. These include, but not lim- ited to, indecorous description roles of pharmacists, breaches in the code of patient privacy, and offensive promotion of pharmaceutical products [20]. Profanity, sexually explicit conducts, derogatory remarks, patient demeaning, references of racism and ethnics, are some other unprofessional behaviors that have been reported in social media [51]. Physicians mostly publish pictures or other information about their patients on social media and approximately only 5 % of them obtain formal permission from their patients prior to posting [22]. Interestingly, a study has reported that 89 % physicians believed better quality of care for their patients who are connected to them through Facebook versus other pa- tients [47]. In a survey-based investigation Marnocha et al., the authors have described that out of 293 nursing students, 77 % had encountered at least one event of un- professional content posted by fellow students [48]. Be- sides, recurring types of unprofessional remarks were posted about patients, peers, their work (58 %), profanity (37 %), patient privacy environment (31 %), prejudicial language (29 %) and cyberbullying (11 %). This study has signaled a rising prevalence of un- professional online behaviours by nursing students and have emphasized the crucial role of policies and formal training of digital professionalism among nursing profes- sionals and educators. the most In a report by Lefebvre et al., as many as 80 % of digitally natives were not concerned about patients’ online privacy or data protection [29]. The same re- port has revealed that a highest number of nurses in 36–45 years age group believed that making a patient friend or from patient’s family on social media is ac- ceptable. Expression of humor online is considered to be more perilous than face-to-face conversations with patients. Fraping, deliberate posting of inappropriate material on Facebook, after hacking into someone else’s account is a new emerging cyber-crime [11]. This bring up another aspect of security of cyber net- working that can safe-guard public’s privacy and self- esteem [32]. A study has shown extremely poor awareness about privacy regulations of social media among surgical trainees and established surgeons [35]. Furthermore, Facebook owners can access all data of their clients that they have uploaded for personal or corporate use. Lastly, a past history of posting unpro- fessional content on Facebook strongly predicts the occurrence of same event in future as well [52]. In other words, future professional behavior is predicted by past behaviors. Guraya et al. BMC Medical Education (2021) 21:381 Page 9 of 12 Social media notifications are a constant source for distraction and stress. Even on silent mode, haptic alerts can still arrive and cause distraction [29]. All types of notifications such as visual, auditory, or haptic lead to leads to poor work efficiency and memory. Additionally, media updates dismantle emotional states and amplifies stress levels [28]. Currently, this negative impact of so- cial media on healthcare professionals’ health and well- being is essentially ignored. The advantages of using social media with positive professional behaviors include sharing patient empathy, effective patient management, online publication of rec- ommended dosage and side-effects of drugs, and rebuttal of misleading health information [20, 26]. A study has increase in medical students’ under- shown a gradual standing towards considering a change in their online professional behavior [5]. This change can be attributed to increasing awareness about the deleterious effects of social media. The growing knowledge about ethical and moral use of social media can enhance its positive im- pact in the medical profession [50, 54]. Maintaining professional identity on social media is a daunting task as the boundary between professionalism and unprofessional conduct is delicate and invisible. Ed- ucators find it hard to define the extent to which the on- line identity should be allowed to reflect the concerned professional [49]. An interesting term of a “dual-citizen model” has been coined that can be applied by creating different online profiles. In a study, the participants were able to generate three leading themes that the authors argued to incorporate into existing curricula; “negotiat- ing identities”, “maintaining distance” and “recognizing and minimizing risks” [55]. Negotiating identities as stu- dents were placed in learning climate without any role in patient care; maintaining distance by separating two crucial but unique roles; and recognizing and minimiz- ing risks by being vigilant to new roles where profession- alism might be compromised by social media. This approach can make students aware of their transitional status during their studies that will potentially mitigate risks from consequences of possible transgressions. Theme IV: Teaching and assessing professionalism in the digital era Educators have advocated the incorporation of student- centered domains for social media in teaching and as- sessment [20]. Since the landscape of social media cli- mate is continuously changing, there is a need for reciprocal curricular interventions to harmonize educa- tional impact in academic institutions. The ensuing adult and active learning would aim at developing professional identities of healthcare professionals and students [5]. A great majority of medical educators and health policy- makers have argued that the use of social media in medical disciplines should be taught early in medical education, and this module should include; professional standards for the use of social media, integration of so- cial media into clinical practice, professional networking [49], and research [27]. Furthermore, essential remedial measures should be inculcated into this module that can explain concerns about social media and professionalism [29, 48]. These interventional processes would require multidisciplinary and cross-sectoral input from patients, academic and physician leaders, social media experts, and interprofessional stakeholders [34]. Finally, institu- tional policies about online privacy, maintaining digital professionalism, protecting medical information of pa- tients, and the sanctions for breaching the policies should be developed and implemented. Research has shown that prior familiarity with social media policies was positively correlated with improved academic per- formance during online professionalism [55]. This sheds light on the educational awareness of healthcare profes- sionals and students about online professionalism. Social media guidance, particularly about the ele- ments of patient confidentiality and privacy is crucial for its appropriate implementation. This may highlight the need to educate all stakeholders for essential self- disclosure for using social media [31]. Taking an oath or signing a declaration by healthcare professionals and students for adhering to the policies and regula- tions guidance about social media is a valid but difficult-to-implement option [11]. In summary, few institutions have incorporated a structured module of e-professionalism in their curricula despite a staggering rise in the use of social media for networking and education. Currently, academic institu- tional responses to cyber irregularities have typically taken the form of disciplinary actions with punitive in- tent. This has unintentionally created a hidden curricu- lum of digital unprofessionalism. This study calls for assistance and guidance in training the digitally en- hanced learning in preparation for their future digitally driven clinical practice. Due to the multi-dimensional construct of professionalism, it is hard to assess all do- mains in the medical field. To add to its complexity, the assessment of digital professionalism is still in its in- fancy. This review could not identify a reliable and standard assessment tool. However, few studies have in- dicated initial success by comparing the posts, tweets, content, privacy settings, and personal disclosures using pre-post-intervention [34] and a longitudinal follow-up [52] with time and seniority. Study limitations The term “digital professionalism” is a relatively recent term, and researchers have been writing about medical students’ and professionals’ behaviour in social media Guraya et al. BMC Medical Education (2021) 21:381 Page 10 of 12 for more than 15 years. Despite these efforts, the quality of research has been found to be of low quality. Sec- ondly, the study was not limited to one study level and has included all healthcare practitioners across the con- tinuum. Furthermore, the studies included in this review were heterogeneous and hence we were not able to in- vestigate the structural differences of the 44 studies due to variations in the type of information provided. Conclusions Our systematic review reports an escalating rise of use of social media among health care professionals and stu- dents. This study signals unprofessional behaviors on so- cial media among healthcare professionals. Current body of literature reports a high prevalence of breach of pa- tient confidentiality due to an absence of existing social media policies. Nevertheless, there is a corresponding but less strong surge of awareness about the adverse ef- fects of social media. The dignified and noble medical professionals’ identity is blurred and hazy in this digital era. Since the climate of social media is rapidly trans- forming, there is a need for corresponding curricular modifications that can balance legal and effective use of social media in medical education. This intervention can potentially resurrect professional virtues, behaviors and identities of healthcare professionals and students. This study highlights the need to adapt a unified policy for usage of social media among health professionals and students that can mitigate the risk of cyberbullying, pa- tient confidentiality and professional integrity. Abbreviations SPIDER: Sample, Phenomenon of Interest, Design, Evaluation and Research type; PRISMA: Preferred Reporting Items for Systematic Reviews and Meta- analyses; MERSQI: Medical Education Research Study Quality Instrument; HRTPE: High-risk-to-professionalism events Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12909-021-02802-9. Additional file 1. Additional file 2. Acknowledgements SS thanks Dr Bindhu Nair, AHIP Deputy and Research Support Librarian at RCSI-MUB for her expertise and assistance in reviewing the search strategy and study protocol. Declaration of interest This systematic review is performed as part of a PhD program in Health Professions Education at Universiti Sains Malaysia. All authors report no declarations of interest. Authors' contributions SS made a substantial contribution to the conception of the work, created the search strategy, conducted the literature search, hand searched and screened the titles and abstract, extracted the data collected, analysed and interpreted the data, drafted the initial manuscript. Other team members SYG and MSBY critically evaluated the search strategy and revised the manuscript for important intellectual content. All members (SS, SYG and MSBY) approved the final version, and agreed to be accountable for all aspects of the work. Funding The publication cost of this article will be funded by the RCSI-MUB if accepted by the journal. Availability of data and materials All data generated or analysed during this study are included in this submitted article as Appendix I and II. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1Royal College of Surgeons Ireland, RCSI - MUB, Busaiteen, Bahrain. 2Department of Medical Education, School of Medical Sciences, University Sains Malaysia, Health campus, Kelantan, Kota Bahru, Malaysia. 3Clinical Sciences Department, College of Medicine, University of Sharjah, Sharjah, United Arab Emirates. Received: 20 July 2020 Accepted: 24 June 2021 References 1. 4. 2. 3. Collins K, Shiffman D, Rock J. How Are Scientists Using Social Media in the Workplace? PLoS One. 2016;11(10):e0162680. Smailhodzic E, Hooijsma W, Boonstra A, Langley DJ. Social media use in healthcare: A systematic review of effects on patients and on their relationship with healthcare professionals. BMC Health Serv Res. 2016;16(1): 1–14. Guraya SY, Almaramhy H, Al-Qahtani MF, Guraya SS, Bouhaimed M, Bilal B. Measuring the extent and nature of use of Social Networking Sites in Medical Education (SNSME) by university students: results of a multi-center study. Medical education online. 2018;23(1):1505400. Guraya SY, Sulaiman N, Guraya SS, Yusoff MSB, Roslan NS, Al Fahim M, et al. Understanding the climate of medical professionalism among university students; A multi-center study. Innovations in Education and Teaching International. 2020:1–10. Gomes AW, Butera G, Chretien KC, Kind T. The development and impact of a social media and professionalism course for medical students. Teach Learn Med. 2017;29(3):296–303. Gabbard GO. Digital professionalism. Acad Psychiatry. 2019;43(3):259–63. Ellaway RH, Coral J, Topps D, Topps M. Exploring digital professionalism. Med Teach. 2015;37(9):844–9. Chretien KC, Tuck MG. Online professionalism: A synthetic review. Int Rev Psychiatry. 2015;27(2):106–17. Cruess RL, Cruess SR, Steinert Y. Amending Miller’s pyramid to include professional identity formation. Acad Med. 2016;91(2):180–5. 10. Ruan B, Yilmaz Y, Lu D, Lee M, Chan TM. Defining the Digital Self: A 6. 7. 5. 8. 9. Qualitative Study to Explore the Digital Component of Professional Identity in the Health Professions. J Med Internet Res. 2020;22(9):e21416. 11. Curtis F, Gillen J. “I don’t see myself as a 40-year-old on Facebook”: medical students’ dilemmas in developing professionalism with social media. Journal of Further Higher Education. 2019;43(2):251–62. 12. Panahi S, Watson J, Partridge H. Social media and physicians: exploring the benefits and challenges. Health informatics journal. 2016;22(2):99–112. 13. Ricciardi W. Assessing the impact of digital transformation of health services: Opinion by the Expert Panel on Effective Ways of Investing in Health (EXPH). Eur J Pub Health. 2019;29(Supplement_4):ckz185. 769. Guraya et al. BMC Medical Education (2021) 21:381 Page 11 of 12 14. Regmi K, Jones L. A systematic review of the factors–enablers and barriers– affecting e-learning in health sciences education. BMC medical education. 2020;20:1–18. Filkins BL, Kim JY, Roberts B, Armstrong W, Miller MA, Hultner ML, et al. Privacy and security in the era of digital health: what should translational researchers know and do about it? American journal of translational research. 2016;8(3):1560. 15. 16. Cooke A, Smith D, Booth A. Beyond PICO. the SPIDER tool for qualitative evidence synthesis. Qual Health Res. 2012;22(10):1435–43. 17. Guraya S. Prognostic significance of circulating microRNA-21 expression in esophageal, pancreatic and colorectal cancers; a systematic review and meta-analysis. Int J Surg. 2018;60:41–7. 18. Reed DA, Cook DA, Beckman TJ, Levine RB, Kern DE, Wright SM. Association between funding and quality of published medical education research. Jama. 2007;298(9):1002–9. 19. Patil S. Global Library & Information Science Research seen through Prism of Biblioshiny. Studies in Indian Place Names. 2020;40(49):157–70. 20. Benetoli A, Chen TF, Schaefer M, Chaar B, Aslani P. Pharmacists’ perceptions of professionalism on social networking sites. Research in Social Administrative Pharmacy. 2017;13(3):575–88. 21. Benetoli A, Chen TF, Schaefer M, Chaar BB, Aslani P. Professional use of social media by pharmacists: a qualitative study. J Med Internet Res. 2016; 18(9):e258. 22. De Gagne JC, Hall K, Conklin JL, Yamane SS, Roth NW, Chang J, et al. Uncovering cyberincivility among nurses and nursing students on Twitter: A data mining study. International journal of nursing studies. 2019;89:24–31. 23. Dobson E, Patel P, Neville P. Perceptions of e-professionalism among dental students: a UK dental school study. British dental journal. 2019;226(1):73–8. 24. Guraya SY, Al-Qahtani MF, Bilal B, Guraya SS, Almaramhy H. Comparing the extent and pattern of use of social networking sites by medical and non medical university students: a multi-center study. Psychology research behavior management. 2019;12:575. 25. Call T, Hillock R. Professionalism, social media, and the Orthopaedic Surgeon: What do you have on the Internet? Technol Health Care. 2017; 25(3):531–9. 27. 26. Chretien KC, Tuck MG, Simon M, Singh LO, Kind T. A digital ethnography of medical students who use Twitter for professional development. J Gen Intern Med. 2015;30(11):1673–80. Klee D, Covey C, Zhong L. Social media beliefs and usage among family medicine residents and practicing family physicians. Fam Med. 2015;47(3): 222–6. Laliberté M, Beaulieu-Poulin C, Campeau Larrivée A, Charbonneau M, Samson É. Ehrmann Feldman D. Current uses (and potential misuses) of Facebook: an online survey in physiotherapy. Physiotherapy Canada. 2016;68(1):5–12. Lefebvre C, McKinney K, Glass C, Cline D, Franasiak R, Husain I, et al. Social media usage among nurses: perceptions and practices. JONA: The Journal of Nursing Administration. 2020;50(3):135–41. 28. 29. 30. Avcı K, Çelikden SG, Eren S, Aydenizöz D. Assessment of medical students’ attitudes on social media use in medicine: a cross-sectional study. BMC medical education. 2015;15(1):1–6. 31. Duke VJ, Anstey A, Carter S, Gosse N, Hutchens KM, Marsh JA. Social media in nurse education: Utilization and E-professionalism. Nurse Educ Today. 2017;57:8–13. Kenny P, Johnson IG. Social media use, attitudes, behaviours and perceptions of online professionalism amongst dental students. British dental journal. 2016;221(10):651–5. 32. 33. Pereira I, Cunningham AM, Moreau K, Sherbino J, Jalali A. Thou shalt not tweet unprofessionally: an appreciative inquiry into the professional use of social media. Postgraduate medical 91(1080):561–4. journal. 2015; 35. 34. Walton JM, White J, Ross S. What’s on YOUR Facebook profile? Evaluation of an educational intervention to promote appropriate use of privacy settings by medical students on social networking sites. Medical education online. 2015;20(1):28708. Silva DAF, Colleoni R. Patient’s Privacy Violation on Social Media in the Surgical Area. The American Surgeon. 2018;84(12):1900–5. Jensen KK, Gögenur I. Nationwide cross-sectional study of Danish surgeons’ professional use of social media. Danish medical journal. 2018;65(9):A5495. Long X, Qi L, Ou Z, Zu X, Cao Z, Zeng X, et al. Evolving use of social media among Chinese urologists: Opportunity or challenge? PloS one. 2017;12(7): e0181895. 36. 37. 38. McKnight R, Franko O. HIPAA compliance with mobile devices among 39. ACGME programs. Journal of medical systems. 2016;40(5):129. Salem J, Borgmann H, Baunacke M, Boehm K, Hanske J, MacNeily A, et al. Widespread use of internet, applications, and social media in the professional life of urology residents. Canadian Urological Association Journal. 2017;11(9):E355. 40. Wagner JP, Cochran AL, Jones C, Gusani NJ, Varghese TK Jr, Attai DJ. Professional use of social media among surgeons: results of a multi- institutional study. Journal of surgical education. 2018;75(3):804–10. 41. Wang Z, Wang S, Zhang Y, Jiang X. Social media usage and online professionalism among registered nurses: A cross-sectional survey. International journal of nursing studies. 2019;98:19–26. 42. Barnable A, Cunning G, Parcon M. Nursing students’ perceptions of confidentiality, accountability, and e-professionalism in relation to facebook. Nurse Educator. 2018;43(1):28–31. 43. Campbell L, Evans Y, Pumper M, Moreno MA. Social media use by physicians: a qualitative study of the new frontier of medicine. BMC Med Inf Decis Mak. 2016;16(1):1–11. 44. Phillips HW, Chen J-S, Wilson B, Udawatta M, Prashant G, Nagasawa D, et al. Social media use for professional purposes in the neurosurgical community: a multi-institutional study. World neurosurgery. 2019;129:e367-e74. 45. Neville P. Social media and professionalism: a retrospective content analysis of Fitness to Practise cases heard by the GDC concerning social media complaints. British dental journal. 2017;223(5):353. 46. Bacaksiz FE, Eskici GT, Seren AKH. “From my Facebook profile”: What do nursing students share on Timeline, Photos, Friends, and About sections? Nurse education today. 2020;86:104326. 47. Desai DG, Ndukwu JO, Mitchell JP. Social Media in Health Care: How Close Is Too Close? The health care manager. 2015;34(3):225 – 33. 48. Marnocha S, Marnocha MR, Pilliow T. Unprofessional content posted online among nursing students. Nurse Educ. 2015;40(3):119–23. 49. Henning MA, Hawken S, MacDonald J, McKimm J, Brown M, Moriarty H, et al. Exploring educational interventions to facilitate health professional students’ professionally safe online presence. Med Teach. 2017;39(9): 959–66. 50. Garg M, Pearson DA, Bond MC, Runyon M, Pillow MT, Hopson L, et al. Survey of individual and institutional risk associated with the use of social media. Western Journal of Emergency Medicine. 2016;17(3):344. 51. Parmar N, Dong L, Eisingerich AB. Connecting with your dentist on 52. facebook: patients’ and dentists’ attitudes towards social media usage in dentistry. J Med Internet Res. 2018;20(6):e10109. Koo K, Bowman MS, Ficko Z, Gormley EA. Older and wiser? Changes in unprofessional content on urologists’ social media after transition from residency to practice. BJU Int. 2018;122(2):337–43. 53. Dimitri D, Gubert A, Miller AB, Thoma B, Chan T. A quantitative study on 54. 55. anonymity and professionalism within an online free open access medical education community. Cureus. 2016;8(9). Lefebvre C, Mesner J, Stopyra J, O’Neill J, Husain I, Geer C, et al. Social media in professional medicine: new resident perceptions and practices. J Med Internet Res. 2016;18(6):e119. Khandelwal A, Nugus P, Elkoushy MA, Cruess RL, Cruess SR, Smilovitch M, et al. How we made professionalism relevant to twenty-first century residents. Med Teach. 2015;37(6):538–42. 56. Mariano MCO, Maniego JCM, Manila HLMD, Mapanoo RCC, Maquiran KMA, Macindo JRB, et al. Social media use profile, social skills, and nurse-patient interaction among registered nurses in tertiary hospitals: a structural equation model analysis. International journal of nursing studies. 2018;80:76–82. 57. Chretien K, Tuck M, Simon M, Singh L, Kind T, Chretien KC, et al. A Digital Ethnography of Medical Students who Use Twitter for Professional Development. JGIM: Journal of General Internal Medicine. 2015;30(11):1673-80. 58. Guraya SY. Comparing recommended sanctions for lapses of academic integrity as measured by Dundee Polyprofessionalism Inventory I: Academic integrity from a Saudi and a UK medical school. Journal of the Chinese Medical Association. 2018;81(9):787–95. Surani Z, Hirani R, Elias A, Quisenberry L, Varon J, Surani S, et al. Social media usage among health care providers. BMC Res Notes. 2017;10(1):654. 59. 60. Riddell J, Brown A, Kovic I, Jauregui J. Who Are the Most Influential Emergency Physicians on Twitter? West J Emerg Med. 2017;18(2): 281–7. Guraya et al. BMC Medical Education (2021) 21:381 Page 12 of 12 61. Perrin A. Social media usage. Pew research center. 2015:52–68. 62. Shah V, Kotsenas AL. Social media tips to enhance medical education. Academic Radiology. 2017;24(6):747–52. 63. Masters KLT, Johannink J, Al-Abri R, Herrmann-Werner A. Surgeons’ Interactions With and Attitudes Toward E-Patients: Questionnaire Study in Germany and Oman. Journal of medical 2020;22 ((3)). Internet research 64. Guraya SY, Norman RI, Roff S. Exploring the climates of undergraduate professionalism in a Saudi and a UK medical school. Med Teach. 2016;38(6): 630–2. 65. Guraya SY, Norman RI, Khoshhal KI, Guraya SS, Forgione A. Publish or Perish mantra in the medical field: A systematic review of the reasons, consequences and remedies. Pakistan journal of medical sciences. 2016; 32(6):1562. 66. Akram W, Kumar R. A study on positive and negative effects of social media on society. International Journal of Computer Sciences Engineering. 2017; 5(1):351–4. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12889-023-15713-9	Widnall et al. BMC Public Health (2023) 23:745 https://doi.org/10.1186/s12889-023-15713-9 BMC Public Health Implementing a regional School Health Research Network in England to improve adolescent health and well-being, a qualitative process evaluation Emily Widnall1, Lorna Hatch1, Patricia N Albers1, Georgina Hopkins1, Judi Kidger1, Frank de Vocht1, Eileen Kaner2, Esther MF van Sluijs3, Hannah Fairbrother4, Russell Jago1,5 and Rona Campbell1 Abstract Background There is an increased need for prevention and early intervention surrounding young people’s health and well-being. Schools offer a pivotal setting for this with evidence suggesting that focusing on health within schools improves educational attainment. One promising approach is the creation of School Health Research Networks which exist in Wales and Scotland, but are yet to be developed and evaluated in England. Methods This qualitative process evaluation aimed to identify the main barriers and facilitators to implementing a pilot School Health Research Network in the South West of England (SW-SHRN). Semi-structured interviews were conducted with school staff, local authority members, and other key stakeholders. Interview data were analysed using the 7-stage framework analysis approach. Results Four main themes were identified from the data: (1) ‘Key barriers to SW-SHRN’ (competing priorities of academic attainment and well-being, schools feeling overwhelmed with surveys and lack of school time and resource); (2) ‘Key facilitators to SW-SHRN: providing evidence-based support to schools’ (improved knowledge to facilitate change, feedback reports and benchmarking and data to inform interventions); (3) ‘Effective dissemination of findings’ (interpretation and implementation, embedding findings with existing evidence and policy, preferences for an online platform as well personalised communication and the importance of involving young people and families); and (4) ‘Longer-term facilitators: ensuring sustainability’ (keeping schools engaged, the use of repeat surveys to evaluate impact, informing school inspection frameworks and expanding reach of the network). Conclusion This study identifies several barriers to be addressed and facilitators to be enhanced in order to achieve successful implementation of School Health Research Networks in England which include providing a unique offering to schools that is not too burdensome, supporting schools to take meaningful action with their data and to work closely with existing organisations, services and providers to become meaningfully embedded in the system. Keywords Mental health, Well-being, Adolescents, Schools, School Health Research Network Correspondence: Emily Widnall emily.widnall@bristol.ac.uk Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessPage 2 of 13 Background Adolescence offers a key opportunity for early interven- tion with preventive approaches to promote health and well-being across the life course [1, 2]. There is a clear and well-evidenced link between young people’s physi- cal health, emotional health and well-being, and their cognitive development and learning [3–5]. Schools offer a pivotal setting for this with evidence suggesting that focusing on health within schools improves educational attainment [4–7]. International guidance has focused on adopting a whole school approach to young people’s health and well-being for several years, namely the World Health Organization’s (WHO) Health Promoting Schools (HPS) Framework [8] which has been re-advocated in recent years with WHO calling for making every school a health promoting setting [9]. Whole school approaches involve all parts of the school working together and sharing a commitment, ethos and culture towards health and well- being. The HPS Framework comprises of health educa- tion being addressed in the school curriculum, health and well-being promotion through changes to the school environment and schools engaging with families and communities to help strengthen these health messages. Public Health England published guidance on the 8 prin- ciples to promoting a whole school approach to mental health and well-being more specifically, which include; enabling student voice to influence decisions, working with parents and carers and identifying need and moni- toring impact of interventions [3]. Literature on embed- ding whole-school approaches to health and well-being discusses developing supportive policy (e.g. anti-bul- lying), the potential for schools to re-shape their iden- tity through prioritising values such as care, respect and empathy, as well as schools creating a culture that enables young people to feel confident talking about how they feel [10, 11]. Review-level evidence suggests that a whole- school approach is effective in encouraging healthy behaviours in young people including physical activity, healthy eating, and in prevention of tobacco use and bul- lying [12]. Despite growing recognition of school-based health improvement, there remain a number of barriers to improving health and well-being in this context, includ- ing financial constraints, schools focussing on educa- tional outcomes and school performance and limited understanding about effective health interventions [13]. One established method for overcoming these barriers has been the creation of School Health Research Net- works (SHRNs). SHRNs use a whole system approach to facilitate health improvement in schools in that it brings together stakeholders and communities to develop a shared understanding of how best to improve school- aged children’s health and well-being [14], a collaborative model that goes beyond typically commissioned school surveys. System-based approaches look at the interrela- tionships between components of a system (e.g. a school) and the broader system as a whole (e.g. wider educational and government systems) [15]. Although established SHRNs exist with the UK (SHRN, Wales; https://www. shrn.org.uk/ and SHINE Scotland; https://shine.sphsu. gla.ac.uk/) as well as internationally (COMPASS, Can- ada; https://uwaterloo.ca/compass-system/), a SHRN has yet to be implemented in England. These networks help schools work with researchers to generate and use good quality evidence regarding health improvement [16]. Each country has their own unique context and while we can learn from experiences of SHRNs in other coun- tries, we cannot simply replicate what these networks have done and expect it to work in the same way. We therefore require country-specific research to understand the unique barriers and facilitators to developing and sustaining SHRNs. In comparison to Wales and Scotland, England has a diverse school system with a variety of school types including Grammar schools that select stu- dents based on academic achievement, Academy schools that are state-funded but independent from local author- ities and therefore decide on their own curriculums, and Free schools which are similar to academies but run by charities. Only a very small proportion of schools in Eng- land are still maintained by local government (11%). Academy schools, have autonomy over their national curriculum as well as how they support and teach about mental health and well-being [17].A recent qualitative study revealed a wide amount of variability amongst academy trust leaders in how they perceive the role of academies in promoting health and well-being amongst students [13]. This study also revealed differences in whether multi-academy trusts (those responsible for more than one school) adopt a centralised strategy to health promotion, or allow individual schools autonomy. Existing structures in England means that there are dif- ferent decision making approaches for health and well- being in different schools and therefore a SHRN needs to be sufficiently flexible to fit in with these varying struc- tures, and this research will help us understand how best to do this. One existing study in England testing a similar model to a SHRN is the BeeWell study (https://gmbeewell. org/), an annual well-being survey of secondary school pupils across Greater Manchester. Although BeeWell have adopted a regional approach in England, Greater Manchester is a city-region with a combined authority (a group of two or more local government councils that collaborate/take collective action). SW-SHRN is more ambitious in that it is seeking to create a network across a larger geographic area, made up of 15 separate local government administrative areas. Therefore, we want Widnall et al. BMC Public Health (2023) 23:745 Page 3 of 13 to understand the barriers and enablers to doing this at scale. Our pilot study created a network of 18 schools from 6 local authorities in the South West of England. This paper reports on a qualitative process evaluation of implement- ing this pilot network to determine the barriers and facilitators to inform the expansion and continuation of the existing pilot network. A working logic model of SW- SHRN can be found within the study protocol paper [18]. We aimed to answer the following four research questions: i. What are the key issues that impact the successful delivery and running of the SW-SHRN? ii. What key information is required by schools to maximise the impact of the SW-SHRN? iii. What data does the SW-SHRN need to provide to be successful and informative? iv. What is required for the SW-SHRN to be sustainable long term? (sustaining school recruitment, retention and sustaining partnerships to best support schools to improve student health and well-being) Methods Design and participants This process evaluation forms part of a larger pilot study of the SW-SHRN in which Year 8 (age 12–13) and Year 10 (age 14–15) secondary school students (n = 5,211) participated in an online health and well-being survey in school time (within one school lesson)[18]. The survey topics included mental health and well-being, physical activity and eating behaviour, sexual health, risky behav- iours (smoking and alcohol use), body image, sleep, peer support, cyberbullying, social media use and the school connectedness. Parental opt-out informed consent is obtained prior to the student survey as well as students providing informed consent at the beginning of the sur- vey. Full methodological details can be found in the pilot study protocol paper [18]. Schools (n = 18) and local authorities (n = 6) receive tailored feedback reports on the student data and researchers worked closely with schools in order to suggest key areas in which to make changes and to facilitate sharing of best practice between schools across the South West of England. This process evaluation was based on a series of semi- structured interviews with school staff, local authority members, and wider key stakeholders. The key school contact at each participating school (n = 18) was invited to participate in a feedback interview. This was the mem- ber of staff involved in organising and delivering the SW- SHRN student survey in school and involved in receiving feedback reports and working with the team to make changes. Local authority staff from participating and non-participating schools in South West England were also invited to participate in an interview, these staff had school specific roles and some staff supported recruit- ing schools to the network. Other key organisations and individual stakeholders were identified by the research team at the study outset; these consisted of staff within charities, government departments, universities, acad- emy trusts, and local councils whose remit was to work with schools. Local authorities and wider stakeholders were approached by a member of the research team via email with an information sheet and consent form and invited to participate in an interview. For stakeholders who had no prior knowledge of the network, an overview of SW-SHRN was provided in advance of the interview. Participants all received a full information sheet and con- sent form to sign in advance of the interview taking place, where written consent was not received before the inter- view took place, verbal consent was taken (and recorded) before the interview began. Ethical approval for the study was granted by University of Bristol’s Faculty of Health Sciences Research Ethics Committee (Ref. 110,922). Data collection and analysis EW, a female mixed methods public health researcher with experience in conducting qualitative interviews and mental health research in schools conducted the research interviews. All interviews took place either over the phone or via an online video conferencing platform (e.g. Microsoft Teams). Interviews followed a topic guide (additional file 1 & 2). The local authority and stakeholder topic guide included questions on stakeholder views on the network, their perceived barriers and facilitators, what outputs they would like to see from the network and how to make the network sustainable and scalable. The school staff topic guide included questions on school recruitment methods, experiences of participation, feedback on administering the student survey, views on tailored school reports, how they would use the data pro- vided with their school and what would encourage them to continue being part of the network. Interview data were analysed by EW and LH using Gale and colleagues 7-stage framework analysis approach [19]. NVivo version 12 software (QSR International) was used to aid data management [20]. Audio record- ings were transcribed verbatim, reviewed, and checked for accuracy by EW prior to analysis (stage 1). All tran- scripts were initially read by EW to gain familiarity with the interview data, EW recorded any initial contextual notes or early interpretative thoughts. (stage 2). EW and LH then independently read and annotated six randomly selected transcripts; two school contact interviews, two local authority member interviews and two wider key stakeholder interviews to generate an initial list of codes and create a draft framework (stage 3). EW and LH then met to discuss and compare these initial codes and agree on a final set of codes to apply to the remaining Widnall et al. BMC Public Health (2023) 23:745 Table 1 Summary of Key Stakeholder Interviews by Organisation and Role Type Interview Role Type KS1 KS2 KS3 KS4 KS5 KS6 KS7 KS8 KS9 KS10 LA1 LA2 LA3 LA4 LA5 SC1 SC2 SC3 SC4 SC5 SC6 SC7 SC8 SC9 SC10 SC11 Organisation Type Charity Government department Government department Government department Government department Government department University Academy Trust NHS Government department Local authority Local authority Local authority Local authority Local authority Participating school Participating school Participating school Participating school Participating school Participating school Participating school Participating school Participating school Participating school Participating school Mental health lead Mental health, national Public health, national Public health, regional Research lead, national Public health, national Clinical Psychologist/Academic Governor Mental Health Support Team Mental health, regional Advanced Public Health Practi- tioner, Health & Well-being Health Improvement Special- ist: Children & Young People Children & Families Commissioning Lead for Health and Well-being (Education and Learning) Services for Children and Young People Children and Families Deputy Head Teacher Pastoral Support Worker Deputy Head Teacher Head of Personal Develop- ment Curriculum Deputy Head Teacher, Student Welfare & Behaviour Music Teacher, Lead for Looked After Children Mental Health & Well-being Coordinator PSHE Lead Assistant Headteacher Deputy of PE and Health, PSHE Lead Deputy Head Teacher ‘KS’ = key stakeholder, ‘LA’ = Local Authority ‘SC’ = School Contact, PSHE = Personal, social, health and economic education Page 4 of 13 interview transcripts. A draft analytical framework was then produced(stage 4). Although there were some dis- tinct differences between school contact interviews com- pared to wider stakeholders, there was sufficient overlap to allow all transcripts to be coded using the same ana- lytical framework. Our analytic framework was then applied to all remaining transcripts which were single- coded by either EW or LH(stage 5), with further regular discussions to expand or refine the framework as needed. Charting then took place which Gale and colleagues describe as ‘summarizing the data by category’ (p.5)[19]. EW and LH charted the data into the framework matrix by creating summaries and identifying key quotes to rep- resent each category (stage 6). EW and LH met regularly to interpret the data, identifying central characteristics and comparing data categories between and within cases to generate a set of themes and subthemes (stage 7). The final set of themes and subthemes identified were then discussed, revised, and agreed by all co-authors. Codes were both deductive (generated from our topic guide and research questions) and inductive (generated from interview data). The Framework Method was cho- sen due to its ability to incorporate both inductive and deductive codes as well as the strengths of the charting/ matrix process embedded within this approach which ensured that researchers were able to pay close atten- tion to describing the data of each organisation type (school, authority, government department etc.) before comparing similarities and differences across organisa- tions. Charting also allows the views of each research participant to remain connected to other aspects of their account within the matrix which avoids losing the con- text of individual viewpoints [19]. The researchers conducting and analysing the inter- views were working on a project that was focussed on the creation of a school health research network and it is therefore possible that were unconscious biases towards the promotion of the network in the interpretation of the data. Results A total of 26 semi-structured interviews were conducted with key school contacts (n = 11 from 11 individual schools); local authorities (n = 5) and wider key stake- holders (n = 10). Table 1 summarises the interviews by organisation and role type. The four key themes identified from the data were (1) Key barriers to SW-SHRN; (2) Key facilitators to SW- SHRN: providing evidence-based support to schools; (3) Effective dissemination of findings; and (4) Longer-term facilitators: ensuring sustainability. Theme 1 and 2 relate to research question 1, identifying key issues that impact the successful delivery and running of SW-SHRN. Theme 3 relates to research questions 2 and 3 by identifying key Widnall et al. BMC Public Health (2023) 23:745 Page 5 of 13 information required by schools to maximise the impact of the network and identifying what data SW-SHRN needs to provide to be successful. Theme 4 relates to research question 4 and identifies what is required for SW-SHRN to be sustainable long-term. Figure 1 pro- vides an illustrative overview of the four key themes and subthemes. Theme 1: key barriers to SW-SHRN Stakeholders suggested a number of potential barriers to the successful roll out and growth of SW-SHRN which were divided into five subthemes: (1) academic attain- ment vs. health and well-being; (2) schools overwhelmed with surveys; (3) competing with commissioned surveys; (4) scarcity of school time and resource and (5) reduced role of local authorities. This theme discusses these five key barriers to SW-SHRN, as well as detailing suggested facilitators considered by stakeholders to reduce the impact of these barriers. “Ultimately, schools get incentivised for academic achievements and for attendance. Therefore…when we are asking them to allow time and to prioritise other things… how do we argue the case for why this is beneficial? The evidence tells us that young people with depression have less school attendance, and do less well in terms of educa- tional achievement…but how do we get schools to buy into that?” (KS 7). Stakeholders also believed that communicating the link between health and attainment became particularly important when thinking about how to engage harder to reach schools. “I mean, if you are really getting into the harder to engage schools, then there might be more really strong links on why this is important for their academic out- comes, so being able to demonstrate that. That might be better than just why this is good for your kids’ health and well-being outcomes.” (KS 2). Schools feel overwhelmed with surveys Academic attainment vs. health and well-being Schools differed in their levels of priority for student health and well-being, with academic achievement and attendance remaining the central priority in schools. Stakeholders addressed the importance of continued communication about the strong links between health and attainment to schools. Schools and stakeholders discussed the large volume of surveys currently being offered to schools, particularly because of the COVID-19 pandemic. Surveys discussed were not exclusively research surveys but also included surveys from councils, companies and charities. This has led to schools feeling overwhelmed with survey offers, and having to weigh up which ones to take part in and Fig. 1 Barriers and facilitators to implementing a School Health Research Network in the South West of England Widnall et al. BMC Public Health (2023) 23:745 Page 6 of 13 sometimes actively trying to reduce the number of sur- veys going out to students. “Schools get offered a lot of stuff. People will constantly, “Can we do this with you? There is this initiative.” Some of which are government backed, some of which are uni- versity backed, some of which come from elsewhere. They are… very, very busy and so miss stuff at times, even if it is good stuff, and even if they want to do it they are not able to do it.” (KS 8). “I think my view is, yes, there are a lot of surveys going on…the one area I’ve looked at is, can we reduce the amount of surveys that are going out?” (SC 10). Competing with commissioned surveys Schools and wider stakeholders expressed several ben- efits of working with universities to conduct health and well-being surveys including expertise, reputation, and quality. “Knowledge that they [schools] will be getting really topi- cal information, learning from quite renowned, maybe, experts in the field. It’s obviously accredited with an edu- cation establishment, such as the university. Just that robustness of it, that it’s evidence-based, but also, in terms of academia, it’s also recognised by the bodies that schools would know about. I think that would be helpful, just in terms of gauging support.” (LA 1). Despite these benefits, several local authorities (LAs) and schools across the South West have existing rela- tionships with commissioned survey providers which presented as a very clear barrier to involvement in SW- SHRN. Strengths of commissioned surveys included sup- plying raw data to schools, having run the same survey for several years and therefore allowing year-on-year comparison, as well as commissioned surveys being open to all year groups and running in both primary and sec- ondary schools. Because of these established relation- ships with commissioned survey providers, stakeholders described SW-SHRN needing to go above and beyond what these existing surveys were providing. “You’ll kind of come up against the other suppliers of school surveys in the region, no doubt…I suppose if they’re funding them at the minute, you would have to pitch them the advantage of moving from a supplier that they have been used to using.” (KS 6). “You’d be coming as the new person versus the people who they currently have a relationship with, who they are already working with and they probably are satisfied to some extent…cost saving would be a good argument.” (LA 3). LAs also described the importance of having input to survey content and a level of ownership over the data which many had with their commissioned survey pro- viders. To compete with other survey providers, it would be important for SW-SHRN to offer local authorities the ability to shape the content of the survey and for them to be able to interact with and analyse the data for their own purposes. “Local authorities really like having control over things… I think to kind of lose that control and not have that…I don’t know if ownership is the right word, but not have that flexibility, not have that entire control over that process or the content I think is something that might be a bit difficult for a local authority.” (LA 3). “We always make some recommendations about local questions, that we’d like to see…It’s always a negotiation, and we never get all the questions that we want, but usu- ally, we get some of it…So, the possibility of adding a small number of local questions would be good.” (LA 5). Scarcity of school time and resource A key issue for schools was a lack of time and resource outside of the planned curriculum. Staff mentioned dif- ficulties organising the survey around planned lessons for two entire year groups and the need for a dedicated member of staff, as well as administrative or IT support, to assist with this. “So logistically organising something, I mean in a school it’s always difficult…but just logistically pulling two whole year groups out of lessons to do the survey it takes time. You’ve got to organise that. And I think having a member of staff that takes responsibility for it is the only way that that’s actually going to happen.” (SC 13). The opt-out consent process that was managed by the research team benefited schools and reduced burden. “I must say though the way that you handled the opt out thing was a great help to this school. I think if we were having to deal with consent, it would have been a night- mare….Actually, en masse, most parents are happy for their children to take part in something like that. If you’d said to me you need to make sure that you get parental consent and student consent for all 280 year 8s, and 260 year 10s, then I would have been pulling my hair out.”(SC 9). Another benefit was schools having access to Univer- sity iPads to support data collection and to ease the pres- sure on booking computer rooms for the survey. “I think having the iPads as well was a massive, massive bonus because we’re such a big school, and the amount of classes that use the computer spaces it was just going to be nearly impossible to get a day where we’d get most of the year groups done.” (SC 9). Theme 2: key facilitators to SW-SHRN: providing evidence- based support for schools A key feature that attracted schools to join SW-SHRN was access to evidence-based support for schools. This theme was broken down into 4 subthemes: (1) Improved knowledge of children and young people’s health and Widnall et al. BMC Public Health (2023) 23:745 Page 7 of 13 well-being; (2) Feedback reports and benchmarking; (3) Data to inform interventions & monitor impact and (4) Interpretation and implementation. Improved knowledge to facilitate change Stakeholders referred to the importance of measuring young people’s health and well-being at scale to improve their overall knowledge and understanding of this popu- lation in order to create meaningful and targeted change in areas of need. “For us it [survey] has the potential to be really helpful, because we just do not have any other way of soliciting the views of such a wide group of the population, because we have not done a questionnaire, like so many other areas. So to be able to get such a large amount of data is very helpful.” (LA 2). “There is massive potential for school improvement and the use of the network to grow- measurement practices to grow data use, evidence-informed decision-making prac- tices. Which, in turn, should improve children’s outcomes. There is no reason why it wouldn’t, if you are understand- ing needs well and finding evidence-based ways to respond to those needs.” (KS 2). School staff also discussed how the data could iden- tify groups at need and indicate health topics that may require more focus, as well as hopes that the data may facilitate more open conversations about health with students. “This survey has helped identify particular groups of students and areas which we could now work on rather than just shooting in the dark at what we could do and offer the students.” (SC 9). “I hope it will open up more conversations with our stu- dents, and us knowing what areas they need support on, and then being able to target these areas without saying to them, “What do you need support with?” Instead, actually saying to them, “We’re going to do this topic,” and knowing that this topic has been highlighted through the study that that has affected them.” (SC 15). Feedback reports and benchmarking Another key benefit of SW-SHRN to schools was the use of individualised school feedback reports and info- graphics as well as the use of benchmarking data to allow schools to compare their results with other participating schools across the South West. One teacher also reflected on the data challenging their assumptions. “I really liked the graphic, that was probably the most powerful part of it. The breakdown of the questions, that was really fascinating, in having the bar graph for the pupil premium versus the non-pupil premium and free school meals, boy and girl ratio, that was really, really powerful. There were some things where we probably assumed things about, say, a group of boys in year ten and actually it’s come back the opposite of what we assumed.” (SC 9). “I thought the report was brilliant, it was really clear. I really liked the benchmarking that you did and the break- down of boys and girls in different groups. That was really useful. That’s not something we’ve had from the [commis- sioned survey provider] before. So I was able to identify Year 10 boys that are pupil premium students, we’ve got a real problem with this. Being able to identify that specifi- cally is really, really useful.” (SC 13). Although benchmarking was considered a benefit by all schools, local authorities highlighted possible sensitivi- ties when schools fall below average when benchmarked on certain health areas. “I thought it was really cool to think about doing really big comparisons across much larger sets of data. But then again…schools are really sensitive to that. So, schools are really happy when they’re doing better than other places, but obviously they’re not so happy when they find out they’re not doing that as well.” (LA 3). Data to inform interventions Using SW-SHRN data to inform potential interventions was vital to schools and local authorities. Academy staff also discussed the usefulness of having data to inform intervention suggestions to their academy trust and senior leadership team. “We want to know where the kids are at, where their needs lie, and what needs we are meeting, and then what needs we need to work on. In getting interventions to go through our trust, and to go through our SLT (senior lead- ership team), if there’s data provided to say, “We need to do this intervention, because this data has shown us that that is what we need to work on,” it’s so much more mean- ingful to be able to approach an intervention with the senior members of staff ” (SC 15). “I think it’s using [the data] as we move forward, it’s those conversations about how do we benefit the well- being of the students…what changes can we make, or what sorts of things do we need to look at bringing in?” (SC10). One local authority also discussed involvement of the University to either develop an intervention or suggest existing interventions. “I do wonder whether the university can see that there are common issues, which are cropping up across several areas. And whether or not it would not be possible for the university to either develop an intervention itself, which is then bought in by different areas, or whether the univer- sity…could suggest interventions which already exist.” (LA 2). Theme 3: effective dissemination of findings Several suggestions were provided in terms of how to effectively share network findings to make them Widnall et al. BMC Public Health (2023) 23:745 Page 8 of 13 accessible and meaningful to schools, students, parents/ caregivers, and wider stakeholders. This consisted of supporting schools to make use of their individual data reports as well as how best to share findings more broadly across the network and with a wider community of stake- holders. This theme was divided into four subthemes: 3.1 Interpretation and implementation; 3.2Embedding the findings with existing evidence and policy; 3.3) Central- ised online platform and direct personalised communi- cation; and 3.4 Sharing findings with young people and families. Interpretation and implementation The need to support schools with interpreting network findings as well as supporting with recommendations for interventions to implement. This included helping to signpost schools to evidence-based interventions, resources and organisations, as well as supporting the school to better focus the curriculum to cover focus areas addressed in the feedback reports. “It would just be thinking about what support they [schools] have afterwards. Because it is often quite hard to interpret that kind of data, if that’s not what your job is. And then to decide what the next steps are and to drive real change. I think there is that support training…that support to translate data. So, improving school level skills in using the data and interpreting the data. And poten- tially that signposting and some sort of gateway through to the ‘what works’ evidence as well.” (KS 1). Teachers also suggested researcher-provided tailored resources for each health area covered in the survey to allow schools to directly act on the findings. “I mean really if I was going to attain my dream it would be to have linked to the report if your school is below aver- age in this, here are some resources to address it for all of the different areas that you survey on.” (SC 13). Embedding the findings with existing evidence and policy Stakeholders discussed the importance of integrating network findings with the existing evidence-base and existing policy and practice. Therefore, rather than just being sent SW-SHRN data alone, schools and stakehold- ers were keen for researchers to put the data into con- text for them as well as circulating suggested resources. Schools did not want a one-off interaction, but were keen to keep up to date with the latest evidence and policy, highlighting the importance of the network being ‘live’ and regularly updated. “Like any network…providing a kind of noticeboard, really, about what the latest developments are in policy and strategy; the findings about what the evidence base is. If the network is a resourced one where it can host, and it becomes trusted and it has got a website, webpages or a newsletter, and we can post things in there about develop- ing evidence, research.” (KS 4). Schools suggested that the network could act as a sign- posting service between national policy, emerging initia- tives and schools as well as helping schools understand current local statistics to provide local context to their report data, as well as providing suggested resources. “Within mental health, and that broadness of the PSHE and the network, there are a lot of changes coming out nationally, and things that are being pushed on in terms of awareness, so the opportunity to regularly engage with those, so having article postings.” (SC 1). “I suppose resources; if you could have some current sta- tistics for the area, so on percentages, perhaps, even from a social norms perspective, so: “Actually, not as many stu- dents or kids as you think are drinking or smoking.” You know? And has mental health in the South West declined? Has it increased? What resources are out there?” (SC 7). Centralised online platform and direct personalised communication Stakeholders offered a range of methods and platforms for sharing data, the primary preference was having an interactive online hub to store updates, network data, ongoing network events, as well as providing an overview of the wider evidence and policy context as discussed above. “I think having it as a bit of a hub, almost… And then I think just having ‘planned events’…to say ‘this is what we have got planned moving forwards’…I don’t know, data protection-wise, but having somewhere to go on and look at maybe comparing ourselves against other schools. Just having a normal website and then, for the schools that are part of the network, having an access-only part as well. Just combining the two.” (SC 10). Schools were also keen to receive evidence summaries, headline findings, and regular updates about the network through email bulletins or newsletters. “Being in a school you can become very insular and focus on your own things, but actually having a quick bul- letin of ten things that have happened to help with well- being in other schools just gives you that quick, ‘Right, actually, that’s a good idea, we could try that maybe in our school’. I think anything like a bulletin that’s clear… that’s got headline facts or figures, or good practice…that’s probably the most valuable thing that we could receive.” (SC 9). “I know that our school does and I know a lot of schools have these rolling screens, so television screens that can have rolling data…simple things like, “Did you know less than – I don’t know – 5% have tried smoking or more than 95% of students in the South West have never done this?“ One headline at a time… that gives students a chance to actually take in the information.” (SC 7). Widnall et al. BMC Public Health (2023) 23:745 Page 9 of 13 Stakeholders also suggested the use of policy briefings and success stories or case studies of schools implement- ing change as a result of SW-SHRN findings. and stakeholders also noted the importance of ongoing involvement of young people in shaping the network and its resources. Sharing findings with young people and families Stakeholders suggested a number of people to involve when disseminating findings, but particularly empha- sised the importance of sharing findings with young people and using a ‘you said, we did’ approach to allow young people to see the results of their survey responses in action. “Children and young people get asked to do loads of sur- veys… there needs to be a version of “you said” and “what we’re going to do or what we can do to help support what you have said”. The school gets theirs [data], but then sometimes we think about that third part of the triangle, which is the young people who have done it.“ (LA 4). As well as sharing the findings with young people, stakeholders also highlighted the importance of sharing findings with parents and school governors in an acces- sible way as well as using the findings to shape school policy, curriculum and development plans. “Put stuff in the parent newsletter over the next year or two. Just drip feed a few facts about what times children go to bed, or what they’re eating, or if they’ve had break- fast. Things that parents might be interested in.“ And obvi- ously, the good news, share it with governors. Try and get something in the school development plan, share it with the PSHE coordinator. Just get it out there.” (LA 5). Although one local authority staff member suggested sharing ‘good news’ with governors, careful consideration is needed on how to meaningfully share the challenges schools may be facing, for example when they are below the benchmark in a particular health area. Theme 4: longer-term facilitators: ensuring sustainability Participants also discussed a number of longer-term facilitators which focussed on the sustainability of the network and potential ideas to expand the grown of the network in the future. This comprised of four subthemes: (1) Keeping schools engaged; (2) Repeat surveys and eval- uating intervention impact; (3) Informing school inspec- tion frameworks and (4) Expanding reach and enhancing accessibility. Keeping schools engaged Schools and stakeholders highlighted the importance of maintaining contact over time and continuing to provide updates and share latest findings to keep people engaged during the gap between biennial surveys. This links to suggestions for an online hub for schools to keep up to date with network activity and latest evidence. Schools also mentioned the issue of staff turnover and the need for regular communication to maintain links. Schools “It’s keeping them in the loop regularly, just to remind them about this network…, it’s difficult, isn’t it, because it’s like a balance between keeping them in the loop and keeping them up-to-date…but also not asking them to do anything because you don’t want to burden them.” (KS 5). “I think just having an ongoing conversation, and issuing reports, and picking out findings from the previous survey keeps it alive. It reminds people of what the survey was, the value of it, what it can do, how you can respond to it.“ (LA 5). It was also acknowledged that schools would like to be recognised for their involvement in the network which was another means of keeping schools engaged long-term, for example if they were working towards an accreditation. “I suspect they would quite like to be named because they’ll be seen as kind of trailblazers for working on this, and with the push, like we say, about the curriculum changes and that kind of thing, being shown as one of the front-runners of linking into this kind of network could well be quite a kind of status thing for the schools.” (KS10). Repeat surveys and evaluating intervention impact Something that school staff and local authorities noted as lacking is the ability to evaluate the impact of health and well-being interventions. Stakeholders therefore discussed the value of monitoring change over time and supported the suggestion of repeat biennial surveys to monitor change when new policies or interventions are implemented. One school detailed how they were always looking to ‘monitor, track, improve and reflect’ (SC15) as a type of audit and feedback approach. However, repeat surveys would need to be carefully planned given the existing barriers discussed regarding lack of time and burden on schools. “How can you use the data to inform a strategic approach to health and well-being and monitor it. I guess that’s the thing, monitor your changes. So, you change something, have you had the effect you wanted to have? This is where they were at, this is what they implemented, this is the benefit.” (KS2). “I think the value of the questionnaire that is being done, the value very much lies in repeating it, doesn’t it? Because it is not much use to schools if they do loads of work to address an issue it is not particularly helpful if they cannot find out whether they have made a difference or not.” (LA2). Informing school inspection frameworks Stakeholders provided their thoughts on health and well-being becoming a bigger focus in future Ofsted Widnall et al. BMC Public Health (2023) 23:745 Page 10 of 13 (The Office for Standards in Education; the government department responsible for inspecting education institu- tions in England) frameworks and how this may impact sustainability of the network. Overall stakeholders agreed that if it were to become a more central focus this would lead to increased buy-in to the network. Stakeholders also mentioned the potential for Ofsted to use SW-SHRN data to target particular areas for inspection. However, some stakeholders were hesitant on schools being scruti- nised on health and well-being outcomes by Ofsted due a wide range of factors external to the school impacting on this. Overall, stakeholders saw promise in schools being able to demonstrate awareness of need and targeted action to Ofsted inspectors as a result of SW-SHRN data. “The survey would help identify those needs [child health needs], and you could say that the schools’ PSHE programme is informed by the data they’ve collected about their needs and behaviour. It would be really good to be able to present that to Ofsted, saying we’re aware of the needs of our children, and we’ve responded as a school to the data which we’ve collected, just like they would for data about academic subjects or anything else.” (LA 5). “I think it could go both ways, couldn’t it? It depends quite how they mandate it, whether they mandate that it must be a particular measure or a particular time…But on the other hand, if schools have to evidence that they are already doing something, actually that might really help buy in to this. I hope Ofsted would not go down a line of saying, “It has to be this measure at this time,” because I do not think there is a perfect measure out there”. (KS 7) Expanding reach and enhancing accessibility In terms of the network being sustainable long-term, a common query or suggestion amongst stakeholders was whether we could extend the survey to include primary schools, as well as offering schools the option to run the survey with all secondary school year groups. The value of reaching older children (16 +), as well as children who are home schooled or attending alternative provision academies was also mentioned. Surveying children as early as possible e.g., in nursery and/or primary was seen as optimal as the data would support early preventative work. “There are many thousands more primary schools than there are secondary, so you’re able to reach a much larger audience. But most importantly, it is preventative work. So, I would be looking at nurseries as well. If you’re not surveying kids until they’re in Year 8, that is quite late. So, I would be interested in trying to understand more, as early as possible. It will presumably give you much more data and greater benchmarking and ultimately more power.” (KS 1). “I think in terms of provision for children who are… home educated, just being aware that we have a virtual school and the…Alternative Provision Academies, for children who have been expelled or excluded. I think, for us, it’s just making sure that we tailor any resources and needs to those more niche audiences. I think schools are always looking at how they are as inclusive as possible.” (LA 1). As well as future expansion, schools discussed the importance of the survey being inclusive and accessible for students with special educational needs or lower reading abilities in mainstream secondary schools. “Having the option to have it read to them probably would help…I know it is hard if you’re using validated surveys, some of the language is a little bit inaccessible. So I think it’s got to be inclusive and accessible.” (SC 13). Discussion The aim of this study was to identify the key barriers and facilitators involved in setting up a regional SHRN in the South West of England and to identify opportuni- ties for refinement of the network to enhance its sustain- ability. We identified four key themes (1) Key barriers to SW-SHRN; (2) Key facilitators to SW-SHRN: provid- ing evidence-based support to schools; (3) Effective dis- semination of findings; and (4) Longer-term facilitators: ensuring sustainability. Barriers incorporated pressures on school time, dif- ferent levels of prioritisation on student health and well-being in comparison to academic attainment, and competing with existing commissioned health and well-being surveys. These barriers are consistent with a recent systematic review of sustaining school-based mental health and well-being interventions [21]. The review found that competing priorities and responsibili- ties often led to intervention delivery challenges and also highlighted the need for school interventions to be easy to use or implement and well-organised. These two find- ings are in line with our study results relating to compet- ing school priorities and discussion of time pressures and reducing burden on schools. Although the links between health improvement and educational attainment are well-evidenced within the academic literature, it seems particularly important to clearly communicate this link to school staff, local authorities and academy trusts, particularly with ref- erence to our findings regarding competing priorities between health and well-being and academic achieve- ment and reassuring schools that focussing on health and well-being is not diverting resource away from the core curriculum and attainment. Previous research from the Welsh SHRN demonstrates emerging evidence of bet- ter educational outcomes in schools with more extensive health improvement policies and practices [22] which is another important factor when communicating the ben- efits to schools of participating in a SHRN. Widnall et al. BMC Public Health (2023) 23:745 Page 11 of 13 To address these barriers, SW-SHRN aims to provide collaborative opportunities for schools to share best practice between one another and across different local authorities in an effort to create an active learning net- work. By building an active learning network that mul- tiple partners benefit from (similar to the Welsh and Scottish model) we hope to make the research/survey burden worthwhile for schools and go above and beyond existing survey provider offerings. As SW-SHRN grows and more schools participate, we hope the network can offer a more standardised approach to health and well-being surveys across the region and in turn reduce the number of survey requests that secondary schools receive. There are also unknowns on how commercial survey companies deal with ethical requirements, data security and ownership of data, therefore a university- led SHRN hopes to provide schools a robust and secure method of collecting student data. It will be important to take an inclusive approach in terms of promoting the network and recruiting to the network to ensure all the relevant education infrastruc- tures are incorporated to maximise the growth of the network given the diverse school system in England. Pre- vious literature has evidenced that collaboration with the education sector is critical when developing health-pro- moting schools programmes [23]. A key facilitator to SW-SHRN is the ability to provide schools with evidence-based information to enhance their understanding of mental health and well-being in school populations as well as identifying health needs and challenges, for example subgroups of students requiring more support or intervention. What seemed to set SW-SHRN apart from existing school surveys was the individualised feedback reports. Within these reports, schools valued the use of benchmarking data to allow them to see where they sit in the context of all partici- pating schools in the region as well as the break down of data by gender, year group and socio-economic status. In turn, these detailed reports aim to allow schools to more effectively target health areas both within the curriculum as well as through targeted resources and interventions. There was also a need to make network outcomes and impacts clear to schools, local authorities, and wider stakeholders, which echoes findings from intervention developer perspectives of evidence-based interventions in schools [24]. A common suggestion to make outcomes and impacts visible was to provide an online platform containing network data which incorporates evidence summaries and policy impacts which would be available to schools and all key stakeholders. A key finding was the need to embed SW-SHRN findings in the wider evidence base and put the findings in context of existing knowl- edge of young people’s health and well-being, as well as linking findings to existing policies and practice [24]. This also aligned with supporting schools to interpret their data and implement meaningful change, schools felt they required support from the research team to translate sur- vey data into action. Schools and stakeholders reflected on how to ensure sustainability of SW-SHRN. Sustainability within the con- text of the network refers to how to sustain the growth of the network (number of schools, academy trusts and local authorities involved), sustaining active involve- ment from participating schools (e.g. engaging in repeat surveys and acting on findings) and sustaining meaning- ful collaborations between stakeholders. Stakeholders discussed the importance of the network maintaining a wider systems perspective, continued conversations with key stakeholders and embedding network findings within wider national policy. One important aspect relating to sustainability is the role future Ofsted frameworks could play in sustain- ing SW-SHRN if health and well-being were to form a larger part of future frameworks. Stakeholders saw value in making use of SW-SHRN data to inform student need and modifying provision accordingly (e.g. PSHE cur- riculum), which could then be presented to Ofsted to showcase meaningful health and well-being activity. An important area of future research could focus on how best to mandate routine monitoring of health and well- being provision in schools. Findings revealed the potential SW-SHRN data holds to support and inform both regional and national policy and planning, and the implications this may have on who may support funding the network in the future. Working at a systems level has been effective for the Welsh School Health Research Network, their network has been effec- tively embedded into the system and plays a key role in national and regional planning [14]. The suggestions from participants regarding joined up working, influencing questions to drive policy, and understanding challenging areas through comparison with other schools demonstrate the need for connecting multiple systems and structures and a requirement for the network to monitor and intervene at multiple levels (e.g. school level, local authority level, government level). Together, these suggestions reflect the need for the net- work to take a systems-based approach. A possible area of future expansion for the network noted by several stakeholders could be the inclusion of primary schools, as well as 16+, and alternative provision settings, to allow SW-SHRN to provide a more complete picture of health and well-being across all school settings and in all age groups. Primary schools were of particular interest as an area for expansion, both to allow for ear- lier intervention, to allow for longitudinal tracking of health and attainment outcomes and also due to many multi-academy trusts comprising of both primary and Widnall et al. BMC Public Health (2023) 23:745 Page 12 of 13 secondary schools and therefore wanting a network that was accessible for all of their schools. However, expand- ing to primary schools would need careful consideration, particularly in terms of how to sustain such a large net- work if expanded given the barriers identified so far. Sustainability of public health interventions in schools remains relatively underexplored in comparison to health care and a recent review highlights particular difficulties with retaining senior leadership contacts given frequent staff turnover in schools [25]. Additionally, it will be important to refine definitions of sustainability relating to SW-SHRN as the network continues to develop [26]. One incentive to join the network could lie in its multi- ple forms of research participation, particularly for those schools who are less active in research. SW-SHRN offers involvement in a population health survey, 1:1 feedback on a tailored school report, qualitative interviews and focus groups with young people, as well as the school environment survey that may help schools reflect on their current health and well-being policies. A possibility for the network as it develops could be offering schools flexibility on which aspects of research they participate in. An important overall finding from this study was the general unified opinions or advice given from key stake- holders, suggesting agreement and consensus around the importance of routine collection of health and well-being outcomes in young people. However, there were varied opinions and priorities across individual schools, par- ticularly how schools would make use of SW-SHRN data and how much support schools felt they needed from the University in making meaningful changes as a result of the data they received from the network. This reinforces the individual nature and unique set-up of each school or academy and the need to offer a flexible and tailored research agenda to meet individual school needs. SW- SHRN in the future, for example, could offer different levels of school involvement depending on individual school preferences. Findings from this evaluation will be used to develop, adapt and enhance the expansion of School Health Research Networks in England, with particular focus towards creating meaningful change in schools and sup- porting schools to effectively make use of the data gen- erated from these networks. SW-SHRN will continue to routinely seek feedback from participating schools, local authorities and academy trusts to continue refining the model and prioritise areas of future expansion. Strengths and limitations This is the first regional School Health Research Network to be set-up in England. This study benefits from seeking perspectives from a wide variety of school staff, six dif- ferent local authorities across the South West, as well as advice from a wide range of relevant stakeholders includ- ing government departments, charities, researchers, and existing providers of health and well-being initiatives for young people. However, some limitations must also be acknowledged. Although this pilot study tests a regional School Health Research Network in the South West of England, school staff and local authority interviews only covered seven of the 15 local authorities in the region, therefore the findings may not translate to the whole region and it will be important for future SW-SHRN recruitment to target these remaining eight local authori- ties to gain their perspectives. Another limitation is that only one individual per school and local authority were interviewed which means we were not able to explore how far there were diverse opinions within schools or local authorities. Future work could benefit from the use of focus groups to allow discussion between mem- bers of staff and perhaps include combinations of school staff, local authority staff, and wider key stakeholders to encourage conversation around differing viewpoints. Conclusion To ensure effective implementation and sustained growth, School Health Research Networks in England need to provide clear benefits to schools and ensure par- ticipation is not overly burdensome. Schools should be provided with detailed data reports to improve knowl- edge, facilitate change and inform interventions, and should be supported in interpreting report findings in order to take meaningful data-driven action. The network should develop in partnership and close communication with existing organisations and service providers to max- imise relevance, avoid repetition and become meaning- fully embedded in existing policy and practice. Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12889-023-15713-9. Supplementary Material 1 Acknowledgements We would like to thank all participating schools and all school staff, local authorities and wider stakeholders who participated in an interview. EK also acknowledges NIHR Senior Investigator support. For the purpose of Open Access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising. Author Contribution RC and RJ conceived the study and wrote the grant application that secured study funding. All authors were involved in the design and set-up of SW-SHRN. EW conducted the qualitative interviews and EW and LH led on data analysis. EW drafted the initial manuscript, all authors contributed to the development of the manuscript and approved the final version. Funding This study is funded by the National Institute for Health and Care Research (NIHR) School for Public Health Research (SPHR), Grant Reference Number Widnall et al. BMC Public Health (2023) 23:745 Page 13 of 13 9. World Health Organization. Making every school a health-promoting school: implementation guidance. 2021. 10. Glazzard J. A whole-school approach to supporting children and young people’s mental health. J Public Mental Health. 2019;18(4):256–65. 11. Stirling S, Emery H. A whole school framework for emotional well-being and mental health. London: National Children’s Bureau; 2016. 12. Langford R, Bonell C, Jones H, Pouliou T, Murphy S, Waters E, et al. The World Health Organization’s Health promoting schools framework: a Cochrane systematic review and meta-analysis. BMC Public Health. 2015;15:130. Jessiman PE, Campbell R, Jago R, Van Sluijs EMF, Newbury-Birch D. A qualita- tive study of health promotion in academy schools in England. BMC Public Health. 2019;19(1):1186. 13. 14. Murphy S, Littlecott H, Hewitt G, MacDonald S, Roberts J, Bishop J, et al. A Transdisciplinary Complex Adaptive Systems (T-CAS) Approach to develop- ing a National School-Based Culture of Prevention for Health Improve- ment: the School Health Research Network (SHRN) in Wales. Prev Sci. 2021;22(1):50–61. 15. Trochim WM, Cabrera DA, Milstein B, Gallagher RS, Leischow SJ. Practical challenges of systems thinking and modeling in public health. Am J Public Health. 2006;96(3):538–46. 16. Gobat N, Littlecott H, Williams A, McEwan K, Stanton H, Robling M, et al. Developing a whole-school mental health and wellbeing intervention through pragmatic formative process evaluation: a case-study of innovative local practice within the School Health Research network. BMC Public Health. 2021;21(1):154. 17. Bhattacharya B. Academy schools in England. Child Educ. 2013;89(2):94–8. 18. Sharp CA, Widnall E, Albers PN, Willis K, Capner C, Kidger J, et al. Creation of a Pilot School Health Research Network in an English Education infrastructure to improve Adolescent Health and Well-Being: a study protocol. Int J Environ Res Public Health. 2022;19(20):13711. 19. Gale NK, Heath G, Cameron E, Rashid S, Redwood S. Using the framework method for the analysis of qualitative data in multi-disciplinary health research. BMC Med Res Methodol. 2013;13(1):1–8. 20. QSR International Pty Ltd. NVivo (version 12). [Software]; 2018. 21. Moore A, Stapley E, Hayes D, Town R, Deighton J. Barriers and facilitators to sustaining School-Based Mental Health and Wellbeing Interventions: a systematic review. Int J Environ Res Public Health. 2022;19(6):3587. 22. Littlecott HJ, Long S, Hawkins J, Murphy S, Hewitt G, Eccles G, et al. Health improvement and educational attainment in secondary schools: comple- mentary or competing priorities? Exploratory analyses from the school health research network in Wales. Health Educ Behav. 2018;45(4):635–44. 23. Rowling L, Jeffreys V. Capturing complexity: integrating health and education research to inform health-promoting schools policy and practice. Health Educ Res. 2006;21(5):705–18. 24. Forman SG, Olin SS, Hoagwood KE, Crowe M, Saka N. Evidence-based interventions in schools: Developers’ views of implementation barriers and facilitators. School Mental Health. 2009;1(1):26–36. 25. Herlitz L, MacIntyre H, Osborn T, Bonell C. The sustainability of public health interventions in schools: a systematic review. Implement Sci. 2020;15(1):4. 26. Wiltsey Stirman S, Kimberly J, Cook N, Calloway A, Castro F, Charns M. The sustainability of new programs and innovations: a review of the empiri- cal literature and recommendations for future research. Implement Sci. 2012;7(1):17. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. PD-SPH-2015. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. EvS was supported by the Medical Research Council [MC_UU_00006/5]. Data Availability The datasets used and analysed during the current study are available from the University of Bristol data archive, https://data.bris.ac.uk/data/. Declarations Ethics approval and consent to participate Ethical approval for the study was granted by University of Bristol’s Faculty of Health Sciences Research Ethics Committee (Ref. 110922). Written or verbal informed consent was obtained prior to all qualitative interviews commencing. As part of the wider study, parental opt-out informed consent is obtained for the student survey. All methods were carried out in accordance with relevant guidelines and regulations. Consent for publication Not Applicable. Competing interests None to declare. Author details 1Population Health Sciences, Bristol Medical School, University of Bristol, Canynge Hall, Bristol BS8 2PL, England 2Faculty of Medical Sciences, Newcastle University, Newcastle upon Tyne, England 3MRC Epidemiology Unit, University of Cambridge, Cambridge, England 4Health Sciences School, University of Sheffield, Sheffield, England 5Centre for Exercise Nutrition & Health Sciences, School for Policy Studies, University of Bristol, Bristol, England Received: 16 November 2022 / Accepted: 19 April 2023 References 1. Viner RM, Ross D, Hardy R, Kuh D, Power C, Johnson A, et al. Life course epide- miology: recognising the importance of adolescence. BMJ Publishing Group Ltd; 2015. Lane C, Brundage CL, Kreinin T. Why we must invest in early adolescence: early intervention, lasting impact. J Adolesc Health. 2017;61(4):10–S1. Lavis P, Robson C. Promoting children and young people’s emotional health and wellbeing: a whole school and college approach. Public Health England; 2015. Bonell C, Humphrey N, Fletcher A, Moore L, Anderson R, Campbell R. Why schools should promote students’ health and wellbeing. British Medical Journal Publishing Group; 2014. Basch CE. Healthier students are better learners: a missing link in school reforms to close the achievement gap. J Sch Health. 2011;81(10):593–8. Bonell C, Farah J, Harden A, Wells H, Parry W, Fletcher A et al. Systematic review of the effects of schools and school environment interventions on health: evidence mapping and synthesis. Public Health Research. 2013;1(1). Brooks F. The link between pupil health and wellbeing and attainment. Public Health England Wellington House. 2014:133–55. 2. 3. 4. 5. 6. 7. 8. WHO. Promoting health through schools: report of a WHO expert committee on comprehensive school health education and promotion. World Health Organization; 1997. Widnall et al. BMC Public Health (2023) 23:745
10.1155_2019_3172647	Hindawi Mediators of Inﬂammation Volume 2019, Article ID 3172647, 16 pages https://doi.org/10.1155/2019/3172647 Research Article CD36-Mediated Lipid Accumulation and Activation of NLRP3 Inflammasome Lead to Podocyte Injury in Obesity-Related Glomerulopathy Jing Zhao, Hong-liang Rui , Min Yang, Li-jun Sun, Hong-rui Dong, and Hong Cheng Division of Nephrology, Beijing Anzhen Hospital, Capital Medical University, Beijing 100029, China Correspondence should be addressed to Hong Cheng; Drchengh@163.com Received 7 December 2018; Revised 24 February 2019; Accepted 4 March 2019; Published 11 April 2019 Guest Editor: Divya P. Kumar Copyright © 2019 Jing Zhao et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Podocyte injury critically contributes to the pathogenesis of obesity-related glomerulopathy (ORG). Recently, lipid accumulation and inﬂammatory responses have been found to be involved in podocyte injury. This study is to explore their role and relationship in podocyte injury of ORG. In animal experiments, the ORG mice developed proteinuria, podocyte injury, and hypertriglyceridemia, accompanied with deregulated lipid metabolism, renal ectopic lipid deposition, activation of NOD-like receptor protein 3 (NLRP3) inﬂammasome, and secretion of IL-1β of the kidney. The expression of adipose diﬀerentiation- related protein (ADRP), CD36, sterol regulatory element-binding protein 1 (SREBP-1), and peroxisome proliferator-activated receptor α (PPARα) in renal tissue were increased. In in vitro cell experiments, after cultured podocytes were stimulated with leptin, similar to ORG mice, we found aggravated podocyte injury, formatted lipid droplet, increased expression of ADRP and CD36, activated NLRP3 inﬂammasome, and released IL-1β. In addition, after blocking CD36 with inhibitor sulfo-N- succinimidyl oleate (SSO) or CD36 siRNA, activation of NLRP3 inﬂammasome and release of IL-1β are downregulated, and podocyte injury was alleviated. However, after blocking NLRP3 with MCC950, although podocyte injury was alleviated and release of IL-1β was decreased, there was no change in the expression of CD36, ADRP, and intracellular lipid droplets. Taken together, our study suggests that CD36-mediated lipid accumulation and activation of NLRP3 inﬂammasome may be one of the potential pathogeneses of ORG podocyte injury. 1. Introduction Obesity is one of the major public health concerns with prev- alence rates rapidly rising worldwide. In China, according to the 2010 China chronic disease monitoring program, the prevalence of obesity and central obesity of Chinese adults is 12.0% and 40.7%, respectively [1]. Obesity may directly lead to kidney injury, known as obesity-related glomerulopa- thy (ORG), and the incidence of ORG has increased concur- rently with obesity [2, 3]. A retrospective study by D’Agati et al. showed that among cases that underwent renal biopsy, the percentage of ORG increased 13.5 times from 1986 to 2015 [3]. We also analyzed the cases that underwent renal biopsy in our hospital from 2010 to 2015 and calculated that the incidence of ORG with or without other kidney diseases accounted for 9.85% [4]. Podocyte injury critically contributes to the pathogenesis of ORG. Glomerular hyper- trophy accompanied with podocyte hypertrophy and podo- cyte process fusion are the main pathological features of ORG, secondary focal segmental glomerulosclerosis may have occurred on this pathological basis, and clinical mani- festations are proteinuria and progressive renal dysfunction; some patients may eventually develop end-stage renal disease (ESRD) [2, 3]. CD36, also known as a fatty acid transporter protein, is a single-chain transmembrane surface glycoprotein and belongs to the class B scavenger receptor family [5, 6]. CD36 is a multifunctional receptor that binds to two types of ligands: one is lipid-related ligands, including long-chain free fatty acids (FFA), oxidative low-density lipoprotein (ox-LDL), and oxidized phospholipids, and the other is protein-related ligands, such as advanced oxidized protein 2 Target Nephrin Podocin Desmin ADRP CD36 SREBP-1 PPARα NLRP3 Pro-caspase1 IL-1β GAPDH Table 1: Primer sequences for PCR analysis in animal and cellular experiments. Mediators of Inﬂammation Primer sequence (5′-3′) Forward GTCTGGGGACCCCTCTATGA Reverse CAGGTCTTCTCCAAGGCTGT Forward CAGAAGGGGAAAAGGCTGCT Reverse GATGCTCCCTTGTGCTCTGT Forward GTTTCAGACTTGACTCAGGCAG Reverse TCTCGCAGGTGTAGGACTGG Forward CTGGTGAGTGGCCTGTGT TA Reverse AAGCACACGCCTTGAGAG AA Forward ATGGGCTGTGATCGGAACTG Reverse AGCCAGGACTGCACCAATAAC Forward GCGTGGTTTCCA ACATGACC Reverse TAGTGCCTCCTTTGCCACTG Forward CTGCAGAGCAACCATCCAGAT Reverse GCCGAAGGTCCACCATTTT Forward TCTGCACCCGGACTGTAAAC Reverse CATTGTTGCCCAGGTTCAGC Forward ACAAGGCACGGGACCTATG Reverse TCCCAGTCAGTCCTGGAAATG Forward CGCAGCAGCACATCAACAAG Reverse GTGCTCATGTCCTCATCCTG Forward TGTGAACGGATTTGGCCGTA Reverse GATGGGCTTCCCGTTGATGA Length (bp) 209 200 106 199 60 188 70 131 237 118 202 products (AOPPs), thrombin-sensitive protein-1 (TSP1), and amyloid protein [5]. In the kidney, CD36 is mainly expressed in podocytes, mesangial cells, and tubular epithe- lial cells. It binds to long-chain FFA ligands in podocytes and mediates lipid uptake, apoptosis, and release of reactive oxygen species (ROS) [5, 7–9]. Current studies have shown that CD36 may play an important role in kidney injury associated with metabolic diseases. The expression CD36 was upregulated in renal tissue of cases with diabetic nephropathy [7]. Sulfo-N- succimidyl oleate (SSO), an inhibitor of CD36, could alleviate albuminuria of high-fat diet-fed mice [10]. In in vitro exper- iments, after podocytes were stimulated by palmitate, a type of saturated FFA, the expression of CD36 was increased and a lipid droplet was formatted; the apoptosis of podocytes was also increased [7]. These studies suggest that CD36, which is related to lipid uptake, may play a role in podocyte injury of ORG. In addition, according to literature and our previous studies, inﬂammatory responses and activation of NOD-like receptor protein 3 (NLRP3) inﬂammasome may be involved in CD36-mediated lipid accumulation and podo- cyte injury [11, 12]. NLRP3 inﬂammasome is composed of NOD-like recep- tor protein 3 (NLRP3), apoptosis-related speck protein (ASC), and cysteine aspartate-1 precursor (pro-caspase-1) [13, 14]. Activation of NLRP3 results in the activation of cas- pase-1, which cleaves the proinﬂammatory cytokines IL-1β and IL-18 to their active forms. Mature IL-1β and IL-18 are resulting in a sterilized released outside of the cell, inﬂammatory response [14]. Numerous studies have shown that the activated NLRP3 inﬂammasome was involved in the pathogenesis of metabolic diseases, such as type 2 dia- betes mellitus, atherosclerosis, and obesity [15, 16]. Our previous studies have observed the activation of NLRP3 inﬂammasome and podocyte injury in the ORG mouse model and cultured podocyte stimulated by leptin, and blocking one of the upstream receptors of NLRP3, puri- nergic ligand-gated ion channel 7 receptor (P2X7R), could ameliorate leptin-induced podocyte injury and inﬂamma- tory response [12]. CD36-mediated lipid accumulation may be associated with activation of the NLRP3 inﬂammasome. Such phenom- enon was ﬁrst reported in the study of arteriosclerosis [17]. OxLDL-induced IL-1β secretion promotes foam cell forma- tion, which was mainly via CD36-mediated release of ROS production and activation of the NLRP3 inﬂammasome [18]. In terms of renal disease, Yang et al. reported that in a nephrotic syndrome animal model, increased expression of CD36 could mediate the apoptosis of podocyte through acti- vating the NLRP3 inﬂammasome [11]. Is CD36-mediated lipid accumulation involved in podo- cyte injury of ORG? Is it associated with NLRP3 inﬂamma- some activation? Our study showed that both in ORG mouse models and in leptin-stimulated cultured podocytes, formatted lipid droplets, increased expression of ADRP and CD36, activated NLRP3 inﬂammasome, and released IL-1β were found. In addition, blocking of CD36 reduced podocyte injury and activation of the NLRP3 inﬂammasome, while Mediators of Inﬂammation 3 blocking NLRP3 could alleviate podocyte injury, but did not decrease the expression of CD36 and adipose diﬀerentiation- related protein (ADRP). Taken together, our study suggests that CD36 mediated lipid accumulation and NLRP3 inﬂam- masome activation, which may be one of the potential pathogeneses of ORG podocyte injury. 2. Materials and Methods 2.1. Animals and Grouping. Twenty 6-week-old male C57BL/6J mice (SPF Biotechnology, China) were housed in an animal room of speciﬁc-pathogen-free cleanliness grade with 50-60% humidity at temperature 20-26°C. Mice were randomly divided into 2 groups: control group (n = 10), which were fed a common diet ad libitum that contained fat accounting for 10% kcal (Beijing Huafukang Biological Technology Co. Ltd., Beijing, China), and ORG model group (n = 10), which were fed a high-fat diet that contained fat accounting for 60% kcal (Research Diet, USA) as described previously [4]. All mice were sacriﬁced after anesthesia with pentobarbital at the end of the 12th week. One fourth of renal tissue was ﬁxed in 4% neutral formaldehyde solution for light microscopy; one fourth of renal tissue was ﬁxed in 2.5% glu- taraldehyde solution for electron microscopy; and the renal cortex of the remaining part was rapidly preserved in liquid nitrogen for real-time quantitative polymerase chain reaction (PCR) analysis and Western blot assay. All animal care and experimental protocols complied with the US National Institutes of Health Guide for the Care and Use of Laboratory Animals (publication no. 85-23, 1996) and were approved by the Institutional Animal Care and Use Committee of Capital Medical University (Beijing, China). 2.2. Biological Parameters. Body weight was measured at baseline and every 4 weeks. Kidney weight was measured after mice were sacriﬁced. Nocturnal 12 h urine protein was collected and measured at the 0 and 12th weeks by using the Bradford protein assay kit (Beyotime Biotechnology, China) according to the user’s instruction. Following sacri- ﬁce, the blood samples were collected for the measurement of serum creatinine, serum triglyceride, serum cholesterol, and blood glucose levels, as well as urine creatinine levels. The measurement was carried out by using the Olympus AU5400 Chemistry Analyzer (Olympus, Japan). The calcula- tion methods of Lee’s index, visceral fat index, and creatinine clearance rate were described as previously [19]. 2.3. Pathological Examination. The mouse renal cortical tis- sues were ﬁxed, dehydrated, embedded, sectioned (3 μm), and stained with periodic acid-Schiﬀ reagent as described [19] for light microscopy. Twenty images of glomerular max- imal proﬁles with a vascular pole and/or urinary pole were taken under a high-power microscope (×400, Olympus, Japan) and were analyzed by Nikon NIS-Elements BR image analysis software (Nikon, Japan). The length (μm) of two longest perpendicular diameters in every glomerular capil- lary tuft without Bowman’s space was measured, and then the mean value was calculated. Table 2: Primary and secondary antibodies for Western blot assays. Primary antibody Rabbit anti-nephrin pAb (Abcam) Rabbit anti-podocin pAb (Sigma) Rabbit anti-desmin pAb (Abcam) Rabbit anti-ADRP pAb (Abcam) Rabbit anti-CD36 pAb (Abcam) Rabbit anti-SREBP-1 pAb (Santa Cruz) Rabbit anti-PPARα pAb (Santa Cruz) Rabbit anti-NLRP3 pAb (Novus) Rabbit anti-caspase1 P10 pAb (Santa Cruz) Rabbit anti-IL-1β pAb (Abcam) Mouse anti-β-actin mAb (Sigma) Secondary antibody IRDye 800 conjugated goat anti- rabbit IgG antibody (LI-COR) Ditto Ditto Ditto Ditto Ditto Ditto Ditto Ditto Ditto IRDye 700 conjugated goat anti-mouse IgG antibody (LI-COR) Table 3: Biological parameters in the diﬀerent groups at the 12th week (x ± s). Group Body weight (g) Kidney weight (mg) Lee’s index (g/cm) Visceral fat index Urinary protein (mg/d) Serum creatinine (μmol/L) Creatinine clearance rate (mL/min) Serum triglyceride (mmol/L) serum cholesterol (mmol/L) Blood glucose (mmol/L) ∗P < 0 05 vs. control group. Control 28 9 ± 1 8 0 34 ± 0 03 16 86 ± 0 38 0 02 ± 0 01 Model ∗ 31 9 ± 1 9 ∗ 0 37 ± 0 02 ∗ 17 64 ± 0 56 ∗ 0 05 ± 0 02 801 5 ± 178 9 ∗ 1001 7 ± 230 1 11 3 ± 0 8 0 4 ± 0 1 0 5 ± 0 1 2 2 ± 0 4 9 5 ± 0 6 10 4 ± 1 1 ∗ 0 7 ± 0 2 ∗ 0 9 ± 0 4 ∗ 3 6 ± 0 3 9 8 ± 1 3 The ultra-thin section renal cortical tissues were stained with uranium acetate-lead citrate for electron microscopy. Brieﬂy, for each specimen, ten photographs (×20 000 magni- ﬁcation) covering diﬀerent regions in the glomerular cross section were taken separately. The length (μm) of the periph- eral GBM was measured, and the number of slit pores over- lying this GBM length was counted by Nikon NIS-Elements BR image analysis software (Nikon, Japan). The average width of the foot process was calculated as described [12]. 2.4. Podocyte Culture and Grouping. The conditionally immortalized mouse podocyte cell line was kindly provided by Professor Maria Pia Rastaldi (S. Carlo Hospital, University of Milan). For the culture of podocytes, we followed the 4 Mediators of Inﬂammation ⁎ ⁎ ⁎ 2 l e v e l 1 n o i s s e r p x e A N R m e v i t a l e R 0 Control Model Nephrin Podocin Desmin Control Model 100 kD 42 kD 55 kD 42 kD ⁎ ⁎ ⁎ Nephrin Podocin Desmin 훽-Actin 2 n o i s s e r p x e n i e t o r p e v i t a l e R l e v e l 1 0 (A) (B) (C) ⁎ 60 40 20 0 r e t e m a i d r a l u r e m o l g e g a r e v A ) M 휇 ( (D) ⁎ 0.4 0.3 0.2 0.1 0 s s e c o r p t o o f ) M 휇 ( e h t f o h t d i w e g a r e v A Control Model Control Model Control Model (a) Nephrin Podocin Desmin (b) Figure 1: Changes in glomerular diameter and podocyte of renal tissue in the ORG model. (a) Histology of renal tissues of diﬀerent groups. A and B Light microscope of PAS staining (×400). The average glomerular diameter was measured and compared. C and D, Electron microscope of glomeruli (×20000). The average width of the foot process was measured and compared. Values are represented as mean ± SD (n = 10). (b) The relative mRNA and protein expression levels of nephrin, podocin, and desmin of the renal cortex were measured by real-time quantitative PCR and Western blot assay. The relative protein expression level was expressed as the target protein/β-actin ratio. Values are represented as mean ± SD. ∗P < 0 05 vs. control group, #P < 0 05 vs. ORG model group. methods of Wang et al. [20]. Podocytes were incubated in RPMI-1640 medium (Thermo Fisher Scientiﬁc, USA) con- taining 10% inactivated fetal bovine serum (FBS, Thermo Fisher Scientiﬁc, USA) and 10 μ/mL interferon-γ (IFN-γ, Cell Signaling Technology, USA) at 33°C in humidiﬁed air with 5% CO2. When the cells reached 80-90% conﬂu- ence, transferred to RPMI-1640 medium containing 10% inactivated FBS without IFN-γ and incu- bated at 37°C in humidiﬁed air with 5% CO2 for 10-14 they were days to allow diﬀerentiation. Well-diﬀerentiated podocytes were used for experiments, and diﬀerent stimulants were added to the cells for diﬀerent times according to experi- mental requirements. 2.5. Reverse Transcription and Real-Time Quantitative PCR. Total RNA was extracted from the mouse renal cortex tissue or cultured podocytes using TRIzol reagent (Thermo Fisher Scientiﬁc, USA) following the manufacturer’s instruction. Mediators of Inﬂammation n o i s s e r p x e n i e t o r p e v i t a l e R P R D A f o l e v e l 2.5 2 1.5 1 0.5 0 ADRP β-Actin n o i s s e r p x e n i e t o r p e v i t a l e R P R D A f o l e v e l 2 1.5 1 0.5 0 Control Model (A) (B) (a) ADRP Nephrin Merge Control Model 5 ⁎ Control Model Control Model 48 kD 42 kD ⁎ Control Model (b) (c) Figure 2: Changes of lipid accumulation of renal tissue in ORG model. (a) Representative Oil Red O staining images of renal tissue in diﬀerent groups (magniﬁcation ×1000). (b) The relative mRNA and protein expression levels of ADRP of the renal cortex were measured by real-time quantitative PCR and Western blot assay. The relative protein expression level was expressed as the target protein/β-actin ratio. Values are represented as mean ± SD. ∗P < 0 05 vs. control group, #P < 0 05 vs. ORG model group. (c) Double immunoﬂuorescence staining of ADRP and nephrin of the ORG model. The localization of ADRP (red spots), nephrin (green spots), and merged image (yellow spots) in the frozen section of renal tissue of the ORG model (×400) is shown as indicated. 2 μg total RNA from each sample was reverse-transcribed to cDNA with EasyScript First-Strand cDNA Synthesis Super- Mix (TransGen Biotech). The gene-speciﬁc primers (SBS Genetech, China) are listed in Table 1. Real-time PCR was performed using SYBR Green RT-PCR Master Mix (TransGen Biotech, China) according to the manufacturer’s instruction. GAPDH was set as the internal control gene in the animal and cellular experiments. The relative quantity of mRNA expression was calculated according to the formula 2− targetgeneCt−GAPDHCt × 103, in which Ct was the threshold cycle number. All assays were repeated at least in triplicate independently. 6 Mediators of Inﬂammation ⁎ ⁎ ⁎ Control Model CD36 SREBP-1 PPAR훼 훽-Actin l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R 3 2 1 0 88 kD 125 kD 55 kD 42 kD ⁎ ⁎ ⁎ Control Model Control Model CD36 SREBP1 PPAR훼 CD36 (a) Nephrin CD36 SREBP1 PPAR훼 Merge l e v e l n o i s s e r p x e A N R m e v i t a l e R 3 2 1 0 Control Model (b) Figure 3: Expression of molecules related to lipid metabolism in renal tissue of ORG mice. (a) The relative mRNA and protein expression levels of CD36, SREBP1, and PPARα of the renal cortex were measured by real-time quantitative PCR and Western blot assay. The relative protein expression level was expressed as the target protein/β-actin ratio. Values are represented as mean ± SD. ∗P < 0 05 vs. control group, #P < 0 05 vs. ORG model group. (b) Double immunoﬂuorescence staining of CD36 and nephrin of the ORG model. The localization of CD36 (red spots), nephrin (green spots), and merged image (yellow spots) in the frozen section of renal tissue of the ORG model (×400) is shown as indicated. 2.6. Western Blot Assay. Total protein lysates were extracted from the mouse renal cortex tissue or cultured podocytes using RIPA lysis buﬀer (ComWin Biotech, China). Protein samples were sonicated ﬁve times for 1 s each, centrifuged at 12 000 rpm for 10 min at 4°C, and then boiled for 5 min. Protein samples were separated by 10-12% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes (General Electric Co). After being blocked with 5% skim milk in phosphate-buﬀered saline with 0.1% Tween 20 for 1 h, the membranes were incubated with primary antibody at 4°C overnight and then incubated with secondary antibody at room temperature for 1 h. Details regarding primary and sec- ondary antibodies are listed in Table 2. The blotted proteins were quantiﬁed using the Odyssey Infrared Imaging System (LI-COR Biosciences). β-Actin was set as an internal control. The relative expression level of each target protein was dis- played as a ratio of target protein/β-actin protein. All the assays were performed at least in triplicate independently. 2.7. Small Interfering RNA (siRNA) of CD36. Well-diﬀerenti- ated podocytes were transiently transfected with 50 pmol mouse CD36 siRNA (Santa Cruz, USA) or control siRNA- A (Santa Cruz, USA) using Lipofectamine 3000 (Life technol- ogies, USA) according to the manufacturer’s instruction. The transfected cells were cultured for 24 h. The eﬃciency of CD36 siRNA after 24 hours of transfection was conﬁrmed by quantitative real-time PCR of CD36 mRNA. Then, cells Mediators of Inﬂammation l e v e l n o i s s e r p x e A N R m e v i t a l e R 3 2 1 0 ⁎ ⁎ ⁎ Control Model NLRP3 Pro-caspase-1 IL-1훽 훽-Actin l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R 3 2 1 0 118 kD 45 kD 17 kD 42 kD 7 ⁎ ⁎ ⁎ Control Model Control Model NLRP3 Pro-caspase-1 IL-1훽 NLRP3 (a) Nephrin NLRP3 Pro-caspase-1 IL-1훽 Merge Control Model (b) Figure 4: Expression of NLRP3 inﬂammasome and IL-1β in renal tissue of ORG mice. (a) The relative mRNA and protein expression levels of NLRP3, pro-caspase-1, and IL-1β of the renal cortex were measured by real-time quantitative PCR and Western blot assay. The relative protein expression level was expressed as the target protein/β-actin ratio. Values are represented as mean ± SD. ∗P < 0 05 vs. control group, #P < 0 05 vs. ORG model group. (b) Double immunoﬂuorescence staining of NLRP3 and nephrin of the ORG model. The localization of NLRP3 (red spots), nephrin (green spots), and merged image (yellow spots) in the frozen section of renal tissue of the ORG model (×400) is shown as indicated. were incubated with fresh medium with or without leptin for another 12 h or 24 h according to real-time quantitative PCR analysis or Western blot assay, respectively. 2.8. Oil Red O Staining. The lipid accumulation in mouse renal tissues and cultured podocytes was evaluated by Oil Red O staining. Brieﬂy, the sections were ﬁxed with 4% para- formaldehyde, rinsed with 60% isopropanol, stained with Oil Red O for 20 min, and rinsed with 60% isopropanol. Finally, the samples were counterstained with haematoxylin for 5 min. The results were examined by light microscopy (Nikon, Japan). 2.9. Double Immunoﬂuorescence Staining. For double stain- ing of an indirect immunoﬂuorescence assay of proteins and podocyte marker nephrin, frozen renal tissues of mice were ﬁxed in 4% paraformaldehyde, cut into 5 μm thick sec- tions, permeabilized with 0.1% Triton X-100, and blocked with 2% BSA. After blocking, the sections were incubated overnight at 4°C with a rabbit monoclonal anti-ADRP (1 : 100, Abcam, UK), rabbit monoclonal anti-CD36 antibody (1 : 100, Abcam, UK), or rabbit monoclonal anti-NLRP3 (1 : 100, Novus, USA) and guinea pig polyclonal anti- nephrin (1 : 100, Progen Biotechnik, German), and then washed with PBS for three times. Next, the sections were 8 l e v e l n o i s s e r p x e A N R m e v i t a l e R 2 1 0 Control Leptin Leptin+SSO 88 kD 48 kD 42 kD ⁎ ⁎ ⁎ # Control Leptin Leptin +SSO CD36 ADRP CD36 ADRP 훽-Actin 3 2 1 0 l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R (a) ⁎ ⁎ ⁎ # Control Leptin Leptin +SSO CD36 ADRP l e v e l n o i s s e r p x e A N R m e v i t a l e R 2 1 0 Control Leptin Leptin+SSO ⁎ # # # ⁎ ⁎ Nephrin Podocin Desmin 훽-Actin Control Leptin Leptin +SSO Nephrin Podocin Desmin (c) l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R 2 1 0 100 kD 42 kD 55 kD 42 kD Mediators of Inﬂammation Leptin Control Leptin + SSO (b) ⁎ ⁎ ⁎ # # # Control Leptin Leptin +SSO Nephrin Podocin Desmin Figure 5: Eﬀects of inhibiting CD36 by SSO on leptin-induced podocyte injury and lipid accumulation. (a) and (c) Diﬀerentiated podocytes were incubated in medium, medium containing 15 ng/mL leptin, or medium containing both 15 ng/mL leptin and 50 μM SSO, respectively. The relative mRNA and protein expression levels of CD36 and ADRP (a) and nephrin, podocin, and desmin (c) of podocytes were measured by real-time quantitative PCR and Western blot assay. (b) Representative Oil Red O staining images of podocytes in diﬀerent groups as indicated (magniﬁcation ×400). Values are represented as mean ± SD (n = 3). ∗P < 0 05 vs. control group, #P < 0 05 vs. leptin + SSO group. incubated with rhodamine-labeled goat anti-mouse antibody (ZSBiO, China) and FITC-labeled rabbit anti-guinea pig antibody (Abcam, UK) for 2 h at room temperature as sec- ondary antibodies, respectively. After staining, the tissue sections were observed with a ﬂuorescent microscope (Nikon, Japan). 2.10. Statistical Analysis. All the data of continuous variables were represented as mean ± standard deviation (SD) and analyzed by using SPSS 21.0 statistical software (IBM, USA). Statistical signiﬁcance between groups was analyzed by one-way ANOVA. P < 0 05 was considered to indicate a statistically signiﬁcant diﬀerence. 3. Results 3.1. Biological Parameters and Renal Histological Changes in ORG Animal Models. The average body weight, kidney weight, Lee’s index, visceral fat index, and 12 h urinary pro- tein excretion were signiﬁcantly increased in the ORG model group compared with the control group at the 12th week (P < 0 05), while there was no signiﬁcance in serum Mediators of Inﬂammation 9 Control siRNA CD36 siRNA 88 kD 42 kD ⁎ CD36 훽-Actin l e v e l n o i s s e r p x e n i e t o r p 6 3 D C e v i t a l e R 1 0.5 0 ⁎ Control siRNA CD36 siRNA Control siRNA CD36 siRNA l e v e l n o i s s e r p x e A N R m 6 3 D C e v i t a l e R 1.5 1 0.5 0 Control Leptin+ control siRNA Leptin+ CD36 siRNA 88 kD 48 kD 42 kD ⁎ ⁎ # # (a) CD36 ADRP 훽-Actin l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R 2 1 0 ⁎ ⁎ # # l e v e l n o i s s e r p x e A N R m P R D A e v i t a l e R 2 1 0 Control Leptin +control siRNA Leptin +CD36 siRNA CD36 ADRP Leptin +control siRNA Leptin +CD36 siRNA Control CD36 ADRP (b) Figure 6: Continued. 10 l e v e l n o i s s e r p x e A N R m e v i t a l e R 2 1 0 Control Nephrin Podocin Desmin Nephrin Podocin Desmin 훽-Actin ⁎ ⁎ ⁎ # # # Leptin +control siRNA Leptin +CD36 siRNA Control Leptin+ control siRNA Leptin+ CD36 siRNA Mediators of Inﬂammation l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R 2 1 0 100 kD 42 kD 55 kD 42 kD ⁎ # # # ⁎ ⁎ Leptin +control siRNA Leptin +CD36 siRNA Control Nephrin Podocin Desmin (c) Figure 6: Eﬀects of inhibiting CD36 by small interfering RNA on leptin-induced podocyte injury and lipid accumulation. (a) Diﬀerentiated podocytes were transiently transfected with 50 pmol control siRNA or CD36 siRNA, respectively. After 24 h (for mRNA) or 36 h (for protein), cells were harvested and the relative mRNA and protein expression levels of CD36 were measured by real-time quantitative PCR and Western blot assay. (b) and (c) Diﬀerentiated podocytes were transiently transfected with 50 pmol control siRNA or CD36 siRNA, respectively. After 24 h, cells were incubated in medium or medium containing 15 ng/mL leptin as indicated, respectively. Following 24 h (for mRNA) or 36 h (for protein) stimulation by leptin, the relative mRNA and protein expression levels of CD36 and ADRP (b) or nephrin, podocin, and desmin (c) of podocytes were measured by real-time quantitative PCR and Western blot assay. Values are represented as mean ± SD (n = 3). ∗P < 0 05 vs. control siRNA group (a) or control group (B and C), #P < 0 05 vs. leptin + control siRNA group. creatinine levels among the two groups (P > 0 05). The creat- inine clearance rate (CCr) of mice in the ORG model group signiﬁcantly increased than that in the control group (P < 0 05) (Table 3). Hyperlipidemia was found in ORG model mice. At the end of the 12th week, levels of serum triglyceride and choles- terol in the ORG model group were signiﬁcantly higher than those in the control group (P < 0 05). There was no signiﬁ- cant diﬀerence in blood glucose levels between the two groups (P > 0 05) (Table 3). Renal tissue pathological examination showed that the mean glomerular diameter in the ORG model group was sig- niﬁcantly longer than that in the control group at the end of the 12th week (P < 0 05, Figure 1(a)). Under a transmission electron microscope, there was mild and segmental foot pro- cess eﬀacement in the ORG model group, and the mean foot process width in the ORG model group was signiﬁcantly wider than that in the control group at the end of the 12th week (P < 0 05, Figure 1(a)). 3.2. Podocyte Injury of ORG Animal Models. We next investi- gated the changes of podocyte-associated molecules in renal tissues, including nephrin and podocin. The expression of desmin was also measured, which is a sensitive marker of podocyte injury [21]. Compared with the control group, the mRNA and protein expressions of podocyte-associated mol- ecules in the ORG model group were signiﬁcantly decreased, and expression of desmin was signiﬁcantly increased in the renal cortical tissue of mice at the end of the 12th week (P < 0 05) (Figure 1(b)). 3.3. Lipid Accumulation of Renal Tissues in ORG Animal Models. The results of Oil Red O staining revealed that there was an obvious lipid droplet formation in the glomeruli (Figure 2(a)). We also investigate the expression of adipose diﬀerentiation-related protein (ADRP), which is a major constituent located in the lipid droplet surface. Our results showed that the mRNA and protein expression of ADRP in the ORG model group was signiﬁcantly upregulated (P < 0 05) (Figure 2(b)). Immunoﬂuorescence staining of renal tissue showed that increased expression of ADRP in the glomerulus and its position overlapped with the podocyte marker protein nephrin, which suggested that there are lipid droplets in the podocytes during ORG (Figure 2(c)). 3.4. Expression of CD36 and Other Molecules Related to Lipid Metabolism of Renal Tissue in the ORG Model. We further examined the expression changes of CD36 and other mole- cules associated with lipid metabolism. At the end of the 12th week, the expression of CD36 was signiﬁcantly increased in the ORG model group compared with the control group (P < 0 05). The expression of sterol regulatory element- binding protein 1 (SREBP-1), which regulates genes required for fatty acid synthesis, was signiﬁcantly upregu- lated (P < 0 05). The expression of peroxisome proliferator- activated receptor α (PPARα), the main regulator of FFA oxidation, was also upregulated (P < 0 05) (Figure 3(a)). The result suggests that hyperlipidemia of ORG leads to imbalance of renal lipid metabolism. Immunoﬂuorescence staining of renal tissue showed high expression of CD36 in Mediators of Inﬂammation l e v e l n o i s s e r p x e A N R m e v i t a l e R 3 2 1 0 ⁎ ⁎ ⁎ Control Leptin Leptin+SSO # # # NLRP3 Pro-caspase-1 IL-1훽 훽-Actin l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R 4 3 2 1 0 118 kD 45 kD 17 kD 42 kD ⁎ ⁎ # ⁎ 11 # # Control Leptin Leptin +SSO Control Leptin Leptin +SSO NLRP3 Pro-caspase-1 IL-1훽 l e v e l n o i s s e r p x e A N R m e v i t a l e R 2 1 0 ⁎ ⁎ ⁎ NLRP3 # # # Pro-caspase-1 IL-1훽 훽-Actin Control Leptin +control siRNA Leptin +CD36 siRNA NLRP3 Pro-caspase-1 IL-1훽 (a) Control Leptin+ control siRNA Leptin+ CD36 siRNA l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R 3 2 1 0 118 kD 45 kD 17 kD 42 kD NLRP3 Pro-caspase-1 IL-1훽 ⁎ ⁎ ⁎ # # # Control Leptin +control siRNA Leptin +CD36 siRNA NLRP3 Pro-caspase-1 IL-1훽 (b) Figure 7: Eﬀects of inhibiting CD36 on leptin-induced activation of NLRP3 inﬂammasome and IL-1β secretion of podocyte. (a) Diﬀerentiated podocytes were incubated in medium, medium containing 15 ng/mL leptin, or medium containing both 15 ng/mL leptin and 50 μM SSO, respectively. (b) Diﬀerentiated podocytes were transiently transfected with 50 pmol control siRNA or CD36 siRNA, respectively. After 24 h, cells were incubated in medium or medium containing 15 ng/mL leptin as indicated, respectively. (a) and (b) The relative mRNA and protein expression levels of NLRP3, pro-caspase-1, and IL-1β of podocytes were measured by real-time quantitative PCR and Western blot assay. Values are represented as mean ± SD (n = 3). ∗P < 0 05 vs. control group, #P < 0 05 vs. leptin + SSO group (a) or leptin + CD36 siRNA group (b). both renal tubules and glomeruli (Figure 3(b)). Expression of CD36 in glomerulus was overlapped with nephrin, suggest- ing that CD36-midiated lipid uptake and lipid accumulation increased in podocytes during ORG (Figure 3(b)). 3.5. Expression of NLRP3 Inﬂammasome and IL-1β of Renal Tissue in the ORG Model. In order to detect renal inﬂam- matory response in the ORG model, we measured the components of NLRP3 inﬂammasome expression of (NLRP3 and pro-caspase-1) and the downstream inﬂam- matory factor, IL-1β. The results showed that the mRNA and protein expressions of NLRP3, pro-caspase-1, and IL-1β were signiﬁcantly upregulated in the ORG model group (P < 0 05) (Figure 4(a)). Immunoﬂuorescence stain- ing of renal tissue showed a high expression of NLRP3 in the glomerulus (Figure 4(b)). Overexpressed NLRP3 overlapped with nephrin, suggesting that the overexpressed NLRP3 inﬂammasome exists in podocytes during ORG (Figure 4(b)). Taken together, our results suggest that podocyte injury of ORG may be related with CD36-mediated lipid accumula- tion and activation of the NLRP3 inﬂammasome. We will next conﬁrm the results in cellular experiments. 3.6. Eﬀects of CD36 Inhibitor SSO on Leptin-Induced Podocyte Injury and Lipid Accumulation. To mimic podocyte injury of ORG, cultured podocytes were stimulated by 15 ng/mL leptin, a key adipocytokine that regulates satiety and body fat [4, 12, 19, 22]. After leptin stimulation, the expression of CD36 and ADRP was signiﬁcantly increased, and Oil Red O staining showed intracellular lipid droplet formation in podocyte (Figures 5(a) and 5(b)). The expression of nephrin 12 l e v e l n o i s s e r p x e A N R m 훽 1 - L I e v i t a l e R 2 1 0 l e v e l n o i s s e r p x e A N R m e v i t a l e R 2 1 0 ⁎ # IL-1훽 훽-Actin Nephrin Podocin Desmin 훽-Actin Control Leptin Leptin +MCC950 ⁎ # # ⁎ ⁎ # Control Leptin Leptin +MCC950 Nephrin Podocin Desmin Control Leptin Leptin+ MCC950 l e v e l n o i s s e r p x e n i e t o r p 훽 1 - L I 1 e v i t a l e R 2 1 0 17 kD 42 kD (a) Control Leptin Leptin+ MCC950 l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R 2 1 0 100 kD 42 kD 55 kD 42 kD Mediators of Inﬂammation ⁎ # Control Leptin Leptin +MCC950 ⁎ # # # ⁎ ⁎ Control Leptin Leptin +MCC950 Nephrin Podocin Desmin Figure 8: Eﬀects of inhibiting NLRP3 by MCC950 on leptin-induced podocyte injury and IL-1β secretion. (a) and (b) Diﬀerentiated podocytes were incubated in medium, medium containing 15 ng/mL leptin, or medium containing both 15 ng/mL leptin and 1 μM MCC950, respectively. The relative mRNA and protein expression levels of nephrin, podocin, and desmin (a) and IL-1β (b) of podocytes were measured by real-time quantitative PCR and Western blot assay. Values are represented as mean ± SD (n = 3). ∗P < 0 05 vs. control group, #P < 0 05 vs. leptin + MCC950 group. (b) and podocin is signiﬁcantly downregulated (P < 0 05), while signiﬁcantly upregulated the expression of desmin is (P < 0 05) (Figure 5(c)). While when a speciﬁc inhibitor of CD36, SSO, was added, the eﬀects of leptin on podocyte were signiﬁcantly inhibited. After SSO was added, although there was no signiﬁcant change in CD36 expression in the leptin+SSO group com- pared with the leptin group, the expression of ADRP was downregulated in podocyte (P < 0 05, Figure 5(a)). In addition, Oil Red O staining showed the decreased intra- cellular lipid droplet formation (Figure 5(b)). Compared with the leptin group, the expression of nephrin and podocin in the leptin+SSO group was signiﬁcantly upregu- lated, and desmin expression was signiﬁcantly downregu- lated (P < 0 05) (Figure 5(c)). These results suggest that inhibition of CD36 by SSO could ameliorate lipid accumula- tion and podocyte injury. 3.7. Eﬀects of CD36 siRNA on Leptin-Induced Podocyte Injury and Lipid Accumulation. We also used CD36 siRNA to inhibit the expression of CD36 in podocyte. Our results showed that CD36 siRNA could eﬀectively inhibit the expression of CD36 in podocyte (Figure 6(a)). Compared with the leptin + control siRNA group, the expressions of podocin and nephrin in the leptin + CD36 siRNA group were signiﬁcantly upregulated and the expression of desmin was signiﬁcantly downregulated (P < 0 05); also, the expression of CD36 and ADRP was signiﬁcantly downregulated by CD36 siRNA (P < 0 05) (Figure 6(b)). 3.8. Eﬀects of Blocking CD36 on Leptin-Induced NLRP3 Inﬂammasome Activation and IL-1β Secretion. We also found that stimulation of leptin could increase the expression of NLRP3 inﬂammasome and secretion of inﬂammatory cytokine IL-1β (Figures 7(a) and 7(b)). Could inhibition of Mediators of Inﬂammation l e v e l n o i s s e r p x e A N R m e v i t a l e R 2 1 0 ⁎ ⁎ ⁎ ⁎ Control Leptin Leptin+ MCC950 ADRP CD36 훽-Actin l e v e l n o i s s e r p x e n i e t o r p e v i t a l e R 2 1 0 48 kD 88 kD 42 kD 13 ⁎ ⁎ ⁎ ⁎ Control Leptin Leptin +MCC950 ADRP CD36 Control Leptin Leptin +MCC950 ADRP CD36 Control (a) Leptin Leptin+MCC950 Figure 9: Eﬀects of inhibiting NLRP3 by MCC950 on leptin-induced lipid uptake and accumulation of podocyte. (a) and (b) Diﬀerentiated podocytes were incubated in medium, medium containing 15 ng/mL leptin, or medium containing both 15 ng/mL leptin and 1 μM MCC950, respectively. (a) The relative mRNA and protein expression levels of ADRP and CD36 of podocytes were measured by real-time quantitative PCR and Western blot assay. (b) Representative Oil Red O staining images of podocytes in diﬀerent groups as indicated (magniﬁcation ×400). Values are represented as mean ± SD (n = 3). ∗P < 0 05 vs. control group, #P < 0 05 vs. leptin + MCC950 group. (b) CD36 aﬀect such eﬀect? Our results showed that either SSO or CD36 siRNA could signiﬁcantly inhibit leptin-induced podocyte NLRP3 inﬂammasome activation and inﬂamma- tory factor IL-1β secretion. After SSO or CD36 siRNA was given, mRNA and protein expressions of NLRP3, pro-caspase-1, and IL-1β were signiﬁcantly downregulated (P < 0 05) (Figures 7(a) and 7(b)). This result suggests that leptin-induced podocyte NLRP3 inﬂammasome acti- vation and IL-1β secretion might be mediated by CD36. 3.9. Eﬀects of Blocking NLRP3 Inﬂammasome on Leptin- Induced Podocyte Injury and Lipid Accumulation. On the other hand, we used MCC950, a speciﬁc inhibitor of NLRP3, to observe the eﬀects of blocking NLRP3 inﬂammasome on leptin-induced podocyte injury and lipid accumulation. Our results showed that after MCC950 was added, the secretion of IL-1β was reduced (Figure 8(a)) and leptin-induced podocyte injury was also alleviated. The expression of nephrin and podocin was signiﬁcantly upregulated, and the expression of desmin was signiﬁcantly downregulated (P < 0 05) (Figure 8(b)). However, there was no signiﬁcant diﬀerence in the expression of mRNA and protein of CD36 and ADRP (P > 0 05) (Figure 9(a)), and Oil Red O staining also showed that lipid droplets in podocyte did not decrease (Figure 9(b)). Our results show that although inhibition of the NLRP3 inﬂammasome attenuated inﬂammatory response and podocyte injury, it did not alter leptin-induced lipid uptake and lipid accumulation of podocytes. This result suggests that CD36-mediated lipid uptake and lipid accumu- lation play a vital role in podocyte injury of ORG, which might achieve such eﬀects through CD36-mediated NLRP3 inﬂam- masome activation and IL-1β secretion. 4. Discussion The imbalance release of adipocytokine, including patho- genic adipocytokine (such as leptin) and protective adipo- cytokine (such as adiponectin), may be an important mechanism for ORG [23]. There were many studies on the role of adipocytokine in the pathogenesis of ORG [2, 23]. We also successfully established the ORG cell model by leptin-stimulated podocytes [4, 12, 19]. In recent years, the role of abnormal lipid metabolism in the pathogen- esis of ORG has gradually attracted attention. This study is to observe the abnormal lipid metabolism and its pathogenesis in ORG. ORG patients manifested metabolically unhealthy obe- sity, which means a central or visceral body fat distribution [24]. Verani found that obesity-associated focal segmental 14 Mediators of Inﬂammation Leptin FFAs CD36 Podocyte IL-1훽 CD36 ASC Caspase-1 NLRP3 ADRP Lipid droplets Podocyte injury Figure 10: Schematic model depicting a possible mechanism that contributes to CD36-mediated podocyte injury of ORG. In the pathogenesis of ORG, after stimulation of adipocytokine such as leptin, the expression of CD36 increased, which is responsible for the FFA uptake in podocytes. Increased CD36 promotes lipid droplet formation in podocytes and further activates NLRP3 inﬂammasome; the cells release the mature form of inﬂammatory cytokines such as IL-1β, which causes the injury of podocytes. FFAs: free fatty acids. glomerulosclerosis or glomerulomegaly was not associated with the amount of obesity per se, but rather with serum tri- glycerides and renal deposition of lipid [25]. Our study also found that serum triglyceride and cholesterol levels were signiﬁcantly elevated in ORG mice. Abdominal adipose tissue is thought to generate high concentrations of circulating FFAs. Excessive FFAs are trans- ferred and accumulated in the liver and kidney through blood circulation and formed ectopic lipid deposits [24]. de Vries et al. reported that in ORG patients, there are signiﬁ- cant lipid accumulation and lipid droplet formation in mesangial cells and podocytes of the glomerulus [24]. We also observed lipid droplet formation and upregulated ADRP in renal tissue and podocyte of ORG mice, which suggests that there are ectopic lipid deposits and lipid accumulation in ORG. Is lipid accumulation the causative factor of ORG? The classical “lipotoxicity” emphasizes the important role of LDL and ox-LDL, while ORG patients mainly have an ele- vated TG level, which means there may be diﬀerent patho- genic mechanisms of ORG [24, 26, 27]. We hypothesized that CD36 may play an important role in lipid accumulation and pathogenesis of ORG. CD36 is the main receptor for lipid uptake of podocytes. In transgenic mice overexpressing CD36 in the kidney, lipid deposition of renal tissues was signiﬁcantly increased [28]. When cultured podocytes were stimulated with palmitic acid, upregulated expression of CD36, formation of intracellular lipid droplet, and release of ROS were observed [7]. Our study showed that during ORG, the expression of CD36 in renal tissue was upregulated, which was consistent with lipid droplet formation and podocyte injury. When cultured podocytes were stimulated with leptin, expressions of CD36 and ADRP were also upregulated. In addition, after blocking CD36 with inhibitor SSO or CD36 siRNA, the lipid deposition of podocytes was signiﬁcantly reduced, ADRP expression was downregulated, and podocyte injury was signiﬁcantly reduced. This experiment initially conﬁrmed the important role CD36 in podocyte injury of ORG. Renal lipid metabolism includes fatty acid intake, fatty acid synthesis, and oxidation and utilization of fatty acids [3, 29]. In an ORG mouse model, we also observed an upregulated expression of SREBP-1 and PPARα. SREBP- 1 is responsible for fatty acid synthesis, and PPARα reg- ulates lipid oxidation utilization [3]. It has been reported that the expression of PPARα in kidney tissue is down- regulated in the mouse model of high-fat diet and cases of stage IV of diabetic nephropathy [30, 31], which is inconsistent with our results. One of the possible reasons for inconsistency may be that the modeling time of ORG mice is about 12 weeks and the renal lesion seems to maintain in the early stage. So the lipid uptake and the lipid might be increased in oxidation utilization of synchronization. We have previously observed an activation of the NLRP3 inﬂammasome in the ORG model and in leptin-induced podocyte injury [12]. Is there a relationship between CD36- mediated podocyte lipid accumulation and activation of NLRP3 inﬂammasome in ORG? Sheedy et al. reported that upregulation of CD36 promotes NLRP3 activation and IL-1β secretion in macrophages [17]. Liu et al. found that during the formation of foam cells in the pathogenesis of Mediators of Inﬂammation 15 atherosclerosis, lipids cause chronic inﬂammatory response through CD36-mediated release of ROS and activation of the NLRP3 inﬂammasome [18]. Therefore, we hypothesized that CD36-mediated lipid accumulation of podocyte may lead to podocyte injury through activating the NLRP3 inﬂammasome and releasing IL-1β. Our study showed that NLRP3 inﬂammasome activation and IL-1β secretion were increased in podocytes and renal tissue of ORG mice; after blocking CD36, lipid deposition in podocytes was reduced, activation of NLRP3 inﬂammasome and secretion of IL-1β were also inhibited, and podocyte injury was alleviated. This result suggests that CD36- mediated lipid accumulation leads to activation of the NLRP3 inﬂammasome and release of inﬂammatory factors, resulting in podocyte injury of ORG. Previous studies have also shown that activation of the NLRP3 inﬂammasome may promote the expression of CD36, thereby promoting the uptake of lipids by cells. Yang et al. found that in the chronic inﬂammation model induced by injection of casein, the expression of CD36 in renal tissue was upregulated, and the deposition of renal fat in mice was aggravated [32]. In vitro experiments showed that mesangial cells stimulated with TNFα lead to upregulated expression of CD36, increased lipid uptake, and intracellular lipid accumu- lation [32]. Gnanaguru et al. found in macrophage studies inhibition of NLRP3 inﬂammasome downregulates that CD36 expression and reduces lipid uptake by cells [33]. However, our study showed that inhibition of the NLRP3 inﬂammasome could improve podocyte injury, but there was no signiﬁcant change in CD36 and ADRP expression in podocytes, and lipid deposition was also not reduced. This result further suggests that CD36-mediated lipid uptake and lipid accumulation might be the initiating factors that promote inﬂammatory responses by activating NLRP3 inﬂammasome, leading to podocyte injury of ORG. Taken together, we showed that obesity and maladjusted lipid metabolism in ORG lead to renal lipid accumulation and podocyte injury, which are partially mediated by CD36; CD36-mediated lipid accumulation activates NLRP3 inﬂam- masome, releases inﬂammatory factor IL-1β, and induces podocyte injury of ORG; inhibition of CD36 also inhibits NLRP3 activation inﬂammasome and ameliorates podocyte injury (Figure 10). Diﬀerent from increased CD36 and activated NLRP3 inﬂammasome induced by LDL and ox-LDL in atherosclero- sis and foam cell formation [33–35], high-level triglycerides and FFAs conduct CD36-mediated lipid accumulation, NLRP3 inﬂammasome activation, and podocyte injury of ORG. In summary, we believe that CD36 may play a central role that links the pathogenesis of ORG and abnormal lipid metabolism. CD36 may become one of the important targets for ORG treatment in the future. Data Availability The original data of the current study are available in the https://ﬁgshare.com/s/ following 75bd66bf6fb68b4b0106. website: Conflicts of Interest The authors declare that they have no conﬂict of interests. Authors’ Contributions Jing Zhao and Hong-liang Rui contributed equally to this work. Acknowledgments This research was supported by the following 3 grants: National Natural Science Fund Project (81573745), Beijing Municipal Natural Science Foundation (7192050), and Capital Medical Development Research Fund (2018-2-1051). References [1] C. Weiwei, G. Runlin, L. Lisheng et al., “Outline of the report on cardiovascular diseases in China, 2014,” European Heart Journal Supplements, vol. 18, Supplement F, pp. F2–F11, 2016. [2] T. Xu, Z. Sheng, and L. Yao, “Obesity-related glomerulopathy: pathogenesis, pathologic, clinical characteristics and treat- ment,” Frontiers of Medicine, vol. 11, no. 3, pp. 340–348, 2017. [3] V. D. D'Agati, A. Chagnac, A. P. J. de Vries et al., “Obesity- related glomerulopathy: clinical and pathologic characteristics and pathogenesis,” Nature Reviews Nephrology, vol. 12, no. 8, pp. 453–471, 2016. [4] J. J. Zhu, Y. P. Chen, M. Yang et al., “Aldosterone is involved in the pathogenesis of obesity-related glomerulopathy through activation of Wnt/β-catenin signaling in podocytes,” Molecu- lar Medicine Reports, vol. 17, no. 3, pp. 4589–4598, 2018. [5] X. Yang, D. M. Okamura, X. Lu et al., “CD36 in chronic kidney disease: novel insights and therapeutic opportunities,” Nature Reviews Nephrology, vol. 13, no. 12, pp. 769–781, 2017. [6] H. Yokoi and M. Yanagita, “Targeting the fatty acid transport protein CD36, a class B scavenger receptor, in the treatment of renal disease,” Kidney International, vol. 89, no. 4, pp. 740– 742, 2016. [7] W. Hua, H. Z. Huang, L. T. Tan et al., “CD36 mediated fatty acid-induced podocyte apoptosis via oxidative stress,” PLoS One, vol. 10, no. 5, article e0127507, 2015. [8] W. Cui, H. Maimaitiyiming, Q. Zhou, H. Norman, C. Zhou, and S. Wang, “Interaction of thrombospondin1 and CD36 contributes to obesity-associated podocytopathy,” Biochimica et Biophysica Acta (BBA) - Molecular Basis of Disease, vol. 1852, no. 7, pp. 1323–1333, 2015. [9] S. Xu, S. M. Nam, J. H. Kim et al., “Palmitate induces ER cal- cium depletion and apoptosis in mouse podocytes subsequent to mitochondrial oxidative stress,” Cell Death & Disease, vol. 6, no. 11, article e1976, 2015. [10] L. C. Li, J. L. Yang, W. C. Lee et al., “Palmitate aggravates proteinuria-induced cell death and inﬂammation via CD36- inﬂammasome axis in the proximal tubular cells of obese mice,” American Journal of Physiology-Renal Physiology, vol. 315, no. 6, pp. F1720–F1731, 2018. [11] X. Yang, Y. Wu, Q. Li et al., “CD36 promotes podocyte apo- ptosis by activating the pyrin domain-containing-3 (NLRP3) inﬂammasome in primary nephrotic syndrome,” Medical Science Monitor, vol. 24, pp. 6832–6839, 2018. 16 Mediators of Inﬂammation [26] J. Sieber and A. W. Jehle, “Free fatty acids and their metabo- lism aﬀect function and survival of podocytes,” Frontiers in Endocrinology, vol. 5, p. 186, 2014. [27] X. Z. Ruan, Z. Varghese, and J. F. Moorhead, “An update on the lipid nephrotoxicity hypothesis,” Nature Reviews Nephrol- ogy, vol. 5, no. 12, pp. 713–721, 2009. [28] H. M. Kang, S. H. Ahn, P. Choi et al., “Defective fatty acid oxi- dation in renal tubular epithelial cells has a key role in kidney ﬁbrosis development,” Nature Medicine, vol. 21, no. 1, pp. 37– 46, 2015. [29] A. Fornoni, S. Merscher, and J. B. Kopp, “Lipid biology of the podocyte—new perspectives oﬀer new opportunities,” Nature Reviews Nephrology, vol. 10, no. 7, pp. 379–388, 2014. [30] F. V. Francisqueti, A. J. T. Ferron, F. K. Hasimoto et al., “Gamma oryzanol treats obesity-induced kidney injuries by modulating the adiponectin receptor 2/PPAR-α axis,” Oxida- tive Medicine and Cellular Longevity, vol. 2018, Article ID 1278392, 9 pages, 2018. [31] M. Herman-Edelstein, P. Scherzer, A. Tobar, M. Levi, and U. Gafter, “Altered renal lipid metabolism and renal lipid accumulation in human diabetic nephropathy,” Journal of Lipid Research, vol. 55, no. 3, pp. 561–572, 2014. [32] P. Yang, Y. Xiao, X. Luo et al., “Inﬂammatory stress promotes the development of obesity-related chronic kidney disease via CD36 in mice,” Journal of Lipid Research, vol. 58, no. 7, pp. 1417–1427, 2017. [33] G. Gnanaguru, A. R. Choi, D. Amarnani, and P. A. D'Amore, “Oxidized lipoprotein uptake through the CD36 receptor acti- vates the NLRP3 inﬂammasome in human retinal pigment epithelial cells,” Investigative Ophthalmology & Visual Science, vol. 57, no. 11, pp. 4704–4712, 2016. [34] L. Chen, Q. Yao, S. Xu, H. Wang, and P. Qu, “Inhibition of the NLRP3 inﬂammasome attenuates foam cell formation of THP-1 macrophages by suppressing ox-LDL uptake and pro- moting cholesterol eﬄux,” Biochemical and Biophysical Research Communications, vol. 495, no. 1, pp. 382–387, 2018. [35] J. P. Rhoads, J. R. Lukens, A. J. Wilhelm et al., “Oxidized low- density lipoprotein immune complex priming of the nlrp3 inﬂammasome involves TLR and FcγR cooperation and is dependent on CARD9,” The Journal of Immunology, vol. 198, no. 5, pp. 2105–2114, 2017. [12] X. X. Hou, H. R. Dong, L. J. Sun, M. Yang, H. Cheng, and Y. P. Chen, “Purinergic 2X7 receptor is involved in the podocyte damage of obesity-related glomerulopathy via activating nucleotide-binding and oligomerization domain-like receptor protein 3 inﬂammasome,” Chinese Medical Journal, vol. 131, no. 22, pp. 2713–2725, 2018. [13] J. Rheinheimer, B. M. de Souza, N. S. Cardoso, A. C. Bauer, and D. Crispim, “Current role of the NLRP3 inﬂammasome on obesity and insulin resistance: a systematic review,” Metabo- lism, vol. 74, pp. 1–9, 2017. [14] F. Martinon, K. Burns, and J. Tschopp, “The inﬂammasome: a molecular platform triggering activation of inﬂammatory cas- pases and processing of proIL-β,” Molecular Cell, vol. 10, no. 2, pp. 417–426, 2002. [15] N. Esser, S. Legrand-Poels, J. Scheen, and N. Paquot, “Inﬂammation as a link between obesity, metabolic syndrome and type 2 diabetes,” Diabetes Research and Clinical Practice, vol. 105, no. 2, pp. 141–150, 2014. J. Piette, A. [16] B. Ke, W. Shen, X. Fang, and Q. Wu, “The NLPR3 inﬂamma- some and obesity-related kidney disease,” Journal of Cellular and Molecular Medicine, vol. 22, no. 1, pp. 16–24, 2018. [17] F. J. Sheedy, A. Grebe, K. J. Rayner et al., “CD36 coordinates NLRP3 inﬂammasome activation by facilitating intracellular nucleation of soluble ligands into particulate ligands in sterile inﬂammation,” Nature Immunology, vol. 14, no. 8, pp. 812– 820, 2013. [18] W. Liu, Y. Yin, Z. Zhou, M. He, and Y. Dai, “OxLDL-induced IL-1beta secretion promoting foam cells formation was mainly via CD36 mediated ROS production leading to NLRP3 inﬂam- masome activation,” Inﬂammation Research, vol. 63, no. 1, pp. 33–43, 2014. [19] B. L. Liu, Y. P. Chen, H. Cheng et al., “The protective eﬀects of curcumin on obesity-related glomerulopathy are associated with inhibition of wnt/β-catenin signaling activation in podocytes,” Evidence-Based Complementary and Alternative Medicine, vol. 2015, Article ID 827472, 12 pages, 2015. [20] C. Wang, X. X. Hou, H. L. Rui et al., “Artiﬁcially cultivated Ophiocordyceps sinensis alleviates diabetic nephropathy and its podocyte injury via inhibiting P2X7R expression and NLRP3 inﬂammasome activation,” Journal of Diabetes Research, vol. 2018, Article ID 1390418, 16 pages, 2018. [21] J. Y. Chen, D. Y. Jian, C. C. Lien et al., “Adipocytes play an etiological role in the podocytopathy of high-fat diet-fed rats,” Journal of Endocrinology, vol. 231, no. 2, pp. 109–120, 2016. [22] D. K. Papafragkaki and G. Tolis, “Obesity and renal disease: a possible role of leptin,” Hormones, vol. 4, no. 2, pp. 90–95, 2005. [23] C. P. Kovesdy, S. L. Furth, C. Zoccali, and on behalf of the World Kidney Day Steering Committee, “Obesity and kidney disease: hidden consequences of the epidemic,” Blood Puriﬁca- tion, vol. 43, no. 4, pp. 346–354, 2017. [24] A. P. de Vries, P. Ruggenenti, X. Z. Ruan et al., “Fatty kidney: emerging role of ectopic lipid in obesity-related renal disease,” The Lancet Diabetes & Endocrinology, vol. 2, no. 5, pp. 417– 426, 2014. [25] R. R. Verani, “Obesity-associated focal segmental glomerulo- sclerosis: pathological features of the lesion and relationship with cardiomegaly and hyperlipidemia,” American Journal of Kidney Diseases, vol. 20, no. 6, pp. 629–634, 1992.
10.1186_s12889-019-7608-1	Shanyinde et al. BMC Public Health (2019) 19:1291 https://doi.org/10.1186/s12889-019-7608-1 R E S E A R C H A R T I C L E Open Access Is physician assessment of alcohol consumption useful in predicting risk of severe liver disease among people with HIV and HIV/HCV co-infection? Milensu Shanyinde1* Pietro Caramello7, Fiona C. Lampe1, Antonella D’Arminio Monforte8, Alessandro Cozzi-Lepri1 and for the ICONA Foundation Study Group , Enrico Girardi2, Massimo Puoti3, Andrea De Luca4, Laura Sighinolfi5, Uberti Foppa Caterina6, Abstract Background: Alcohol consumption is a known risk factor for liver disease in HIV-infected populations. Therefore, knowledge of alcohol consumption behaviour and risk of disease progression associated with hazardous drinking are important in the overall management of HIV disease. We aimed at assessing the usefulness of routine data collected on alcohol consumption in predicting risk of severe liver disease (SLD) among people living with HIV (PLWHIV) with or without hepatitis C infection seen for routine clinical care in Italy. Methods: We included PLWHIV from two observational cohorts in Italy (ICONA and HepaICONA). Alcohol consumption was assessed by physician interview and categorized according to the National Institute for Food and Nutrition Italian guidelines into four categories: abstainer; moderate; hazardous and unknown. SLD was defined as presence of FIB4 > 3.25 or a clinical diagnosis of liver disease or liver-related death. Cox regression analysis was used to evaluate the association between level of alcohol consumption at baseline and risk of SLD. Results: Among 9542 included PLWHIV the distribution of alcohol consumption categories was: abstainers 3422 (36%), moderate drinkers 2279 (23%), hazardous drinkers 637 (7%) and unknown 3204 (34%). Compared to moderate drinkers, hazardous drinking was associated with higher risk of SLD (adjusted hazard ratio, aHR = 1.45; 95% CI: 1.03–2.03). After additionally controlling for mode of HIV transmission, HCV infection and smoking, the association was attenuated (aHR = 1.32; 95% CI: 0.94–1.85). There was no evidence that the association was stronger when restricting to the HIV/HCV co-infected population. Conclusions: Using a brief physician interview, we found evidence for an association between hazardous alcohol consumption and subsequent risk of SLD among PLWHIV, but this was not independent of HIV mode of transmission, HCV-infection and smoking. More efforts should be made to improve quality and validity of data on alcohol consumption in cohorts of HIV/HCV-infected individuals. Keywords: HIV-infected, HIV/HCV co-infection, Alcohol consumption, Severe liver disease * Correspondence: milensu.shanyinde.14@ucl.ac.uk 1Institute of Global Health, University College London, Royal Free Hospital, London, UK Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Shanyinde et al. BMC Public Health (2019) 19:1291 Page 2 of 13 people (cART), Background Identifying unhealthy levels of alcohol consumption in HIV patients seen for routine clinical care is important because of the possible role alcohol plays in HIV disease progression as well as non HIV-related comorbidities such as liver failure [1]. In the era of combination antiretroviral treatment living with HIV/AIDS (PLWHIV) are now living longer and long term effects of alcohol consumption are likely to affect people’s quality of life and survival [2]. Therefore, knowledge of alcohol con- sumption behaviour and risk of disease progression asso- in the ciated with hazardous drinking are important overall management of HIV disease [3]. Alcohol con- sumption is common in PLWHIV with estimates of current alcohol use reported to be 50% in studies of HIV- positive people [1, 4–8] and hazardous drinking, has re- ported prevalence ranging between 8 and 12%. Hazardous drinking can lead to harmful consequences such as liver disease progression or liver related mortality [7, 9–18]. Assessment of alcohol consumption in HIV cohort studies varies because of the measurement tools imple- mented, mode of assessment and risk groups under in- vestigation [18–25]. Most studies have used methods of alcohol assessment based on brief self-reported ques- frequency of tionnaires drinks consumed [26]. Others studies have used patient interviews, biomarkers or breath tests to assess level of alcohol consumption [27, 28]. These different measure- tools has led to methodological challenges in ment quantifying estimates of alcohol consumption amongst PLWHIV [29]. relating to quantity and/or In this analysis, we use data routinely collected by treating physicians in two cohorts of PLWHIV seen for routine clinical care in Italy, including questions related to both frequency and quantity of alcohol consumed. Our objective is two-fold. Firstly, we aim to categorise drinking behaviour using data routinely collected in our cohorts by mapping the questions on the electronic case report form (CRF) to those used in national drinking guidelines known as the National Institute for Food and Nutrition (NIFN) in Italy. Secondly, to assess the associ- ation between alcohol consumption and risk of severe liver disease (SLD) among PLWHIV with or without HCV infection. Methods Study participants This analysis includes all PLWHIV (with and without HCV co-infection) enrolled up to June 2016 in the ICONA Foundation Study and HepaICONA prospective cohorts who were free from SLD (see definition in para- graph below) at study enrolment. Patients enrolled prior to 1st January 2002 were excluded from this study as al- cohol assessment in the cohorts was not in standard use prior to this date. Both cohorts are observational studies of PLWHIV and details of both cohorts have been pub- lished elsewhere [30, 31]. In brief, the ICONA Founda- tion cohort began to enrol PLWHIV in 1997 as long as they were antiretroviral (ART)-naïve at time of enrol- ment. Patients’ demographics, clinical and laboratory data are recorded using an electronic data collection form. Occurrence of any clinical event, including liver- related events and causes of death (classified using CoDe) are recorded [30]. The HepaICONA cohort began in 2013 and it enrols HIV/HCV co-infected and HCV viremic individuals who are naive to direct acting antivirals drugs (DAA) at study entry. Similar data col- lections processes are implemented as those in place for the ICONA cohort, including the questions related to al- cohol consumptions [31]. All patients have given in- formed consent to participate in the study and ethic committee approval from all participating centres was obtained for both cohorts (Additional file 1: Table S1). Classification of alcohol consumption Alcohol consumption is collected by physician interview at study enrolment and at subsequent clinical visits (at least every 6 months) during follow-up. This analysis only includes assessments carried out at enrolment (baseline), which for the ICONA cohort is prior to ART initiation. Exact questions in the patients’ interview (with possible responses) are as follows; 1) Do you currently drink alco- hol? (Yes/No/Unknown); 2) How frequently do you drink alcohol? (Daily/Non-daily/Unknown); 3) How many units of (Wine/Beer/Spirits) do you consume per day? Frequency and quantity of units of drink consumed was translated into drinking categories by mapping the data to the definitions described in the NIFN guidelines. A unit of alcohol in Italy is defined as containing 14 g of pure alcohol which corresponds to 125 ml of wine, 330 ml of a can of beer and 40 ml of liquor [32]. Hazardous drinking is defined as > 3 units/day for men and for women > 2 units/day. In cases where drinking was reported as ‘non-daily’ WHO guidelines were used which state that > 4 drinks per occasion is considered hazardous drinking [33]. People were classified as mod- erate drinkers if they consumed an amount below the hazardous drinking thresholds. Abstainers were people who reported not drinking alcohol at all. Alcohol consumption at baseline was categorised into these three groups described above and an unknown cat- egory if there was missing data for alcohol consumption as shown in Fig. 1. In some individuals who reported more than one type of drink, the drink with the highest quantity of alcohol was used in the classification process. In instances were information on alcohol was given in other metrics e.g. ml, this was converted to units (e.g. 500 ml of wine a day equated to 500 ml/125 ml = 4 units/ Shanyinde et al. BMC Public Health (2019) 19:1291 Page 3 of 13 Fig. 1 Alcohol classification based on responses from Physician assessment of alcohol consumption day). In few instances were information reported as hav- ing a drink with a meal, this was assumed 2 units/day to take into account of the 2 meals a day. Definitions of covariates Demographics were collected at study enrolment. HIV- related variables (previous AIDS diagnosis, CD4 and HIV-RNA), hepatitis co-infection (HCV) status and smoking status were also collected at study enrolment. HCV status was based on HCVAb+ test or, when the re- sults for antibody test was not available, using HCV- RNA > 615 IU/mL or a positive HCV-RNA qualitative test or evidence of the determination of HCV genotype. Definition of severe liver disease Time to the development of severe liver disease in follow- up was a composite endpoint defined at the time of experiencing one of the following events (first of these oc- curring): (i) a FIB4 > 3.25 (where FIB4 was calculated based on available blood test results using the formula (ageAST)/PLTsqrt (ALT) and assessed at each clinical visit [34];(ii) a clinical diagnosis of liver disease from med- ical records (ascites, decompensated cirrhosis, hepatocel- lular carcinoma, hepatic encephalopathy, oesophageal varices); (iii) liver-related death. Statistical analyses Baseline was defined as date of enrolment between 1st January 2002 and 30th June 2016 in ICONA and be- tween 1st October 2013 and 30th June 2016 for people in HepaICONA. Individuals were followed up until the date of experiencing the composite endpoint or their follow-up was censored at the date of their clinical visit at which they were last seen free from SLD. Summary statistics were used to describe the study participants overall and after stratification by alcohol consumption category at baseline. Formal comparisons of categorical characteristics by groups of alcohol consumption were performed using chi-squared tests. Time to SLD was estimated using the Kaplan Meier method overall and after stratifying by baseline alcohol consumption category. Univariable and multivariable Cox regression models were fitted to estimate hazard ratios (HRs) of the risk of SLD associated with baseline levels of alcohol consumption. In the Cox regression model, only time-fixed covariates measured at baseline were included. Potential confounders measured at baseline were consid- ered in a series of separate multivariable models fitted sequentially as follows: Model #1 controlling for demo- graphic factors (gender, age, nationality, geographical re- gion, calendar year of enrolment); Model #2 (model #1 plus previous AIDS diagnosis, CD4, HIV-RNA, and HBV infection status); Model #3 (model #2 plus mode of HIV transmission, and HCV infection status) and finally model #4 (model #3 plus smoking status). Results are presented as adjusted hazard ratios (aHRs) with 95% confidence in- tervals (95% CI). An interaction term between HCV infec- tion and alcohol consumption was formally tested to determine whether the effect of alcohol consumption on risk of SLD differed by HCV infection status. This could be done only for the ICONA cohort as HepaICONA in- cludes only HIV/HCV co-infected individuals. Two differ- ent methods for handling missing data on alcohol consumption were applied: (i) people with unknown alco- hol consumption were included as a separate ‘unknown’ category of the alcohol variable (main analysis); (ii) an Shanyinde et al. BMC Public Health (2019) 19:1291 Page 4 of 13 analysis in which people with unknown information were re-classified as ‘Abstainer’, ‘Moderate’ or ‘Hazardous’ drinker using multiple imputation (MI). A Fully Condi- tional Specification (FCS) imputation algorithm was im- plemented owing to the discrete nature of the alcohol consumption variable. We assumed 10 imputes with 100 iterations to be sufficient. Variables considered predictors of unreported alcohol consumption included in the MI model were: age, gender, nationality, geographical region, calendar year enrolled, mode of HIV transmission, AIDS diagnosis, CD4, HIV-RNA, HCV infection, smoking status and SLD. Separate univariable and multivariable Cox models were fitted similar to those used for the main ana- lysis. Using Rubin’s combination rules, which is a method that combines estimates from imputed datasets to esti- mate standard errors, confidence intervals, and p-values to give an overall estimate of the imputed datasets. All ana- lyses were performed using SAS (version 9.4, SAS Insti- tute, Cary North Carolina USA). Results Alcohol consumption We included 9542 patients who satisfied the inclusion criteria (8876 from ICONA and 666 from HepaICONA). When mapping our questions to the NIFN guidelines the distribution according to alcohol consumption was; abstainers 3422 (36%; 95%CI [35–37]), moderate users 2279 (23%; 95%CI [23–25]), hazardous drinkers 637 (7%; 95%CI [6, 7]), and unknown alcohol status 3204 (34%; 95%CI [33–35]). The distribution according to alcohol consumption in individuals with available data on alco- hol consumption (n = 6338) was; abstainers (54%; 95%CI [53–55]), moderate users (36%; 95%CI [35–37]) and haz- ardous drinkers (10%; 95%CI [9–11]). After using MI to reclassify people with missing data, overall distribution of alcohol consumption was as follows: abstainers (53%), moderate users (37%) and hazardous drinkers (10%). Patient characteristics Table 1 shows baseline characteristics for (n = 6338), over- all and stratified by alcohol consumption. The majority of individuals were male (77%); median age [IQR] 38 [31–46] years. Compared to moderate drinkers, hazardous drinkers were more likely to be male (p < 0.001), of older age (p < 0.001), injecting drug users (p < 0.001), smokers (p < 0.001) and HIV/HCV co-infected (p < 0.001). Com- pared to individuals with complete data on alcohol con- sumption, those with missing data were more likely to be male (p < 0.001), of older age (p < 0.001), to have acquired HIV through IDU (p < 0.001), to not report smoking sta- tus (p < 0.001), to have no test result for HCV infection (p < 0.001) and to be enrolled in later calendar years (p < 0.001) (Additional file 2: Table S2). Alcohol consumption and risk of severe liver disease Patients were followed-up for a median [IQR] of 25.2 months [6.1–55.6]. A total of 617 (7%) people experienced the composite SLD outcome (n = 506 FIB4 > 3.25, n = 110 clinical diagnosis of liver disease, n = 1 liver-related mor- tality). Fig. 2 shows the Kaplan-Meier estimates of the time to SLD according to baseline alcohol consumption level. The estimated cumulative risk of experiencing SLD by 60 months (95% CI) from baseline in abstainers, mod- erate, hazardous or unknown alcohol consumption were 8.4% (7.1–9.7), 7.9% (6.3–9.5), 10.7% (7.4–14.1) and 11.4% (9.9–12.9), respectively [log-rank p < 0.001]. In univariable Cox regression analyses with moderate drinkers as the comparator group, hazardous drinking and unknown alcohol status were strongly associated with increased risk of SLD (unadjusted HR = 1.61 [95% CI: 1.16–2.26]; p = 0.005 and 1.67 [95% CI: 1.34–2.09]; p < 0.001 respectively. In contrast, there was no evidence for a difference between abstaining and moderate con- sumption (unadjusted HR = 1.09 [95% CI: 0.87–1.38]; p = 0.446) Table 2. After controlling for age, gender, na- tionality, region, calendar year enrolled, HIV related fac- tors and HBV, and still using the moderate consumption as the comparator, adjusted HR (95%CI) for hazardous consumption were drinking and unknown alcohol [aHR = 1.45 (1.03–2.03; p = 0.031) and aHR = 1.37 (1.09– 1.72; p = 0.007)] respectively. However, after additionally adjusting for mode of HIV transmission and HCV infec- tion status, the effect of hazardous drinking was attenu- ated (aHR = 1.30 [95% CI: 0.92–1.82]; p = 0.129) but unknown alcohol consumption remained associated with risk of SLD (aHR = 1.43 [95% CI: 1.13–1.80]; p = 0.003). After further adjustment for smoking status, alcohol consumption was no longer associated with risk of SLD; (aHR = 1.09 [95% CI: 0.86–1.38]; global p = 0.446) Table 2. An interaction term between HCV infection and alco- hol use was not significant, indicating that the associ- ation between level of alcohol consumption and risk of SLD did not differ by HCV status (p = 0.740 Fig. 3). In the univariable Cox regression analyses, after combining MI estimates from separate multivariable models, results were similar to those in the main ana- lysis. Hazardous drinking was associated with the risk of SLD (unadjusted HR = 1.70 [95% CI: 1.24–2.34]; p = 0.002 and abstaining was not associated with risk SLD (unadjusted HR = 1.15 [95% CI: 0.90–1.47]; p = 0.261] Table 2. However, as in the main analysis, after adjust- ing for potential confounders including mode of HIV transmission and HCV infection hazardous drinking was no longer associated with risk of SLD (aHR = 1.29 [95% CI: 0.93–1.78]; p = 0.120). Further adjustment for smoking status, alcohol consumption was not associ- ated with risk of SLD (aHR = 1.13 [95%CI: 0.88–1.45]; global p = 0.724) Table 2. Shanyinde et al. BMC Public Health (2019) 19:1291 Page 5 of 13 Table 1 Patients’ characteristics stratified by alcohol consumption classification at baseline Baseline characteristics Abstainers (N = 3422) Moderate (N = 2279) Hazardous (N = 637) Total (N = 6338) Gender, n(%) Male Female Age, years 2363 (69.1%) 1059 (30.9%) 1954 (85.7%) 325 (14.3%) 567 (89.0%) 70 (11.0%) 4884 (77.1%) 1454 (22.9%) Median (IQR) 38 (31, 47) 37 (30, 45) 41 (34, 49) 38 (31, 46) Mode of HIV Transmission, n(%) PWID Homosexual contacts Heterosexual contacts Other/Unknown Nationality, n(%) Italian Region, n(%) North Center South AIDS diagnosis, n(%) 367 (10.7%) 1317 (38.5%) 1517 (44.3%) 221 (6.5%) 250 (11.0%) 1124 (49.3%) 757 (33.2%) 148 (6.5%) 114 (17.9%) 222 (34.9%) 269 (42.2%) 32 (5.0%) 731 (11.5%) 2663 (42.0%) 2543 (40.1%) 401 (6.3%) 2572 (75.2%) 1915 (84.0%) 516 (81.0%) 5003 (78.9%) 1577 (46.1%) 1366 (39.9%) 479 (14.0%) 1144 (50.2%) 866 (38.0%) 269 (11.8%) 382 (60.0%) 217 (34.1%) 38 (6.0%) 3103 (49.0%) 2449 (38.6%) 786 (12.4%) Yes 336 (9.8%) 156 (6.8%) 43 (6.8%) 535 (8.4%) CD4 count cells/mm3, n(%) ≤ 300 301–500 ≥ 501 Unknown HIV-RNA viral load, n(%) ≤ 5000 5001–10,000 10,001–100,000 ≥ 100,001 Unknown Smoking, n(%) No Yes Unknown Hepatitis B, n(%) Yes HCV Infection, n(%) Negative Positive Not tested Calendar year, n(%) 2002–2006 2007–2012 2013–2016 Follow-up (months) 1156 (33.8%) 810 (23.7%) 1035 (30.2%) 421 (12.3%) 605 (17.7%) 208 (6.1%) 1222 (35.7%) 1025 (30.0%) 362 (10.6%) 2201 (64.3%) 1092 (31.9%) 129 (3.8%) 580 (25.4%) 593 (26.0%) 861 (37.8%) 245 (10.8%) 389 (17.1%) 172 (7.5%) 922 (40.5%) 567 (24.9%) 229 (10.0%) 924 (40.5%) 1268 (55.6%) 87 (3.8%) 172 (27.0%) 166 (26.1%) 216 (33.9%) 83 (13.0%) 114 (17.9%) 48 (7.5%) 231 (36.3%) 179 (28.1%) 65 (10.2%) 188 (29.5%) 413 (64.8%) 36 (5.7%) 1908 (30.1%) 1569 (24.8%) 2112 (33.3%) 749 (11.8%) 1108 (17.5%) 428 (6.8%) 2375 (37.5%) 1771 (27.9%) 656 (10.4%) 3313 (52.3%) 2773 (43.8%) 252 (4.0%) 107 (3.1%) 59 (2.6%) 25 (3.9%) 191 (3.0%) 2366 (69.1%) 410 (12.0%) 646 (18.9%) 473 (13.8%) 1113 (32.5%) 1836 (53.7%) 1549 (68.0%) 250 (11.0%) 480 (21.1%) 313 (13.7%) 671 (29.4%) 1295 (56.8%) 409 (64.2%) 119 (18.7%) 109 (17.1%) 69 (10.8%) 218 (34.2%) 350 (54.9%) 4324 (68.2%) 779 (12.3%) 1235 (19.5%) 855 (13.5%) 2002 (31.6%) 3481 (54.9%) Shanyinde et al. BMC Public Health (2019) 19:1291 Page 6 of 13 Table 1 Patients’ characteristics stratified by alcohol consumption classification at baseline (Continued) Baseline characteristics Abstainers (N = 3422) Moderate (N = 2279) Hazardous (N = 637) Median (IQR) 26.5 (7.4, 57.1) 23.4 (4.8, 53.5) 25.6 (5.6, 54.8) Total (N = 6338) 25.2 (6.1, 55.6) Discussion One of the objectives of this analysis was to classify drink- ing behaviour using data on alcohol consumption rou- tinely collected through physician assessment in our cohorts of PLWHIV seen for routine care in Italy. In this analysis involving 9542 PLWHIV, using the NIFN guide- lines of whom 6338 (66%) had data on alcohol consump- tion. In our study population, 10% of individuals were classified as hazardous drinkers, 36% as moderate drinkers and 54% as abstainers. The overall estimate of the current prevalence of alcohol consumption was 46%. Other HIV studies in which alcohol consumption was measured with similar questionnaires have reported prevalence of current use ranging between [27–95%] [18, 21–25]. Our estimate of hazardous use of 10% was similar to reported estimates in other studies ranging from [8–12%] [5, 16, 18, 23, 35]. Other studies using different assessment tools reported higher estimates of alcohol consumption. For example, Samet J et al assessed alcohol consumption in HIV posi- tive individuals in USA by means of patient’s interviews using a series of questions on quantity and frequency and also carried out breath alcohol level test [1]. These authors reported 27% with moderate drinking behaviour and 16% with risky drinking. In addition the use of different meas- urement tools is a further reason to explain lower preva- lence of hazardous drinking in our study could be due to collection of alcohol consumption via face to face rather than anonymously. Thirty four percent of individuals in this analysis had missing data on alcohol consumption. This large propor- tion of missing data highlights the challenges of collecting complete data on alcohol consumption in observational studies of PLWHIV. It is also unclear whether unreported alcohol consumption may have arisen from questions not being asked by the physician or the participant being un- willing to give information for other unknown reasons. The prevalence of people with missing information was generally consistent with that seen in other HIV studies showing estimates of under-reporting ranging between [7–41%] [8, 25, 36, 37]. Possible reasons for under- reporting of alcohol use include social desirability and fear of the impact on antiretroviral therapy initiation [27, 35, 36]. Of note, it is part of the Italian culture to assume that a drink with a meal is normal consumption which might also explain the under-reporting. Some studies have assessed the extent of under-reporting by comparing self-reported alcohol consumption with blood tests or biomarkers, or interviews carried out by professionals and found a lack of agreement between these measures [27, 36, 38]. Physician interview like ours are likely to measure alcohol consumption even less accurately than self-administered questionnaires [29]. Fig. 2 Kaplan Meier estimate of the risk of severe liver disease stratified by alcohol consumption classification Shanyinde et al. BMC Public Health (2019) 19:1291 Page 7 of 13 e u a v l 6 4 4 0 . 8 9 4 0 . 7 0 1 0 . , 6 8 0 ( . , 4 9 0 ( . 9 0 1 . ) 8 3 1 . 0 0 1 . 2 3 1 . ) 5 8 1 . 7 1 2 0 . 7 1 2 0 . 1 0 0 < . 1 0 0 < . 7 7 8 0 . 7 7 8 0 . 3 5 3 0 . 1 0 0 < . 6 9 5 0 . 0 4 2 0 . 1 0 0 < . 1 0 0 < . 8 0 0 0 . 8 0 0 0 . 1 0 0 < . 1 9 4 0 . , 2 9 0 ( . , 4 3 1 ( . , 4 7 0 ( . , 9 7 0 ( . , 9 8 0 ( . , 6 4 0 ( . , 9 4 0 ( . , 9 0 1 ( . , 9 6 0 ( . ) 6 4 1 . 5 1 1 . ) 3 4 1 . 6 4 1 . ) 8 5 1 . 8 9 0 . ) 8 2 1 . 0 0 1 . 5 9 0 . ) 4 1 1 . 9 1 1 . ) 0 6 1 . 0 0 1 . 9 5 0 . ) 5 7 0 . 4 6 0 . ) 2 8 0 . 0 4 1 . ) 0 8 1 . 0 0 1 . 1 9 0 . ) 9 1 1 . 8 0 4 0 . , 6 8 . . 0 ( 2 1 1 . 3 0 0 0 . , 3 1 1 ( . l a b o g l - p e u a v l - p ] I C % 5 9 [ H R 4 l e d o M - p l a b o g l e u a v l e u a v l - p ] I C % 5 9 [ H R 3 l e d o M l a b o g l - p e u a v l - p ] I C % 5 9 [ H R 2 l e d o M l a b o g l - p e u a v l - p ] I C % 5 9 [ H R 1 l e d o M l a b o g l - p e u a v l e u a v l e u a v l n o i t a t u p m i l e p i t l u m h t i w / t u o h t i w e s a e s i d r e v i l e r e v e s r o f l s i s y a n a n o i s s e r g e r x o C l e b a i r a v i t l u m d n a l e b a i r a v n U 2 i e l b a T 3 1 2 0 . 3 1 2 0 . 1 0 0 < . 1 0 0 < . 9 2 8 0 . 9 2 8 0 . , 2 9 0 ( . , 6 3 1 ( . , 4 7 0 ( . ) 6 8 1 . 3 4 1 . ) 0 8 1 . 5 1 1 . ) 3 4 1 . 7 4 1 . ) 9 5 1 . 7 9 0 . ) 7 2 1 . 9 2 1 0 . , 5 9 0 ( . 0 3 1 . 9 0 0 0 . 3 1 4 0 . , 7 8 0 ( . 0 1 1 . ) 0 4 1 . 0 0 1 . 7 0 0 0 . 0 4 6 0 . , 4 8 0 ( . 9 5 2 0 . 9 5 2 0 . 1 3 0 0 . 7 0 0 0 . , 3 0 1 ( . , 9 0 1 ( . , 2 9 0 ( . 6 0 1 . ) 4 3 1 . 0 0 1 . 5 4 1 . ) 3 0 2 . 7 3 1 . ) 2 7 1 . 3 1 1 . ) 9 3 1 . 1 0 0 < . 6 0 5 0 . 8 2 0 0 . 1 0 0 < . 6 8 2 0 . 6 8 2 0 . , 5 8 0 ( . , 4 0 1 ( . , 4 2 1 ( . , 1 9 0 ( . 8 0 1 . ) 7 3 1 . 0 0 1 . 5 4 1 . ) 4 0 2 . 6 5 1 . ) 5 9 1 . 2 1 1 . ) 8 3 1 . 4 8 1 0 . 4 8 1 0 . , 2 9 0 ( . ) 9 5 1 . 0 2 1 . ) 6 5 1 . 1 8 0 0 . 1 8 0 0 . , 7 9 0 ( . ) 7 6 1 . 6 2 1 . ) 5 6 1 . 1 0 0 < . 1 0 0 < . , 7 3 1 ( . 8 4 1 . 1 0 0 < . 1 0 0 < . , 5 4 1 ( . 6 5 1 . 1 0 0 < . 1 0 0 < . , 0 5 1 ( . 1 0 0 < . 1 0 0 < . , 6 3 1 ( . 5 7 1 . n a i l a t I - n o N s v n a i l a t I 7 1 1 0 . 7 1 1 0 . , 6 9 0 ( . 1 0 0 < . 6 4 4 0 . , 7 8 0 ( . 9 0 1 . i r e n a t s b A 5 0 0 0 . 1 0 0 < . , 6 1 1 ( . 1 6 1 . ) 6 2 2 . , 4 3 1 ( . 7 6 1 . ) 9 0 2 . ) 8 3 1 . 0 0 1 . e t a r e d o M s u o d r a z a H n w o n k n U e u a v l - p d e t s u d a n U j ] I C % 5 9 [ H R n o i t p m u s n o c l o h o c A l 1 9 3 0 . 0 0 1 . 2 4 2 0 . 0 0 1 . 1 4 3 0 . 0 0 1 . 6 9 0 0 . 3 5 5 0 . 1 9 2 0 . , 9 7 0 ( . 4 9 0 . ) 3 1 1 . , 7 8 0 ( . 6 1 1 . ) 8 5 1 . 7 5 1 0 . 4 8 5 0 . , 4 7 0 ( . , 1 8 0 ( . 8 8 0 . ) 5 0 1 . 9 0 1 . ) 6 4 1 . 6 2 2 0 . 1 0 6 0 . , 5 7 0 ( . 0 9 0 . ) 7 0 1 . , 1 8 0 ( . 8 0 1 . ) 4 4 1 . 1 0 0 < . 0 0 1 . 1 0 0 < . 0 0 1 . 1 0 0 < . 0 0 1 . 1 0 0 < . 8 0 0 0 . 8 0 0 0 . 5 0 0 0 . , 4 5 0 ( . , 9 0 1 ( . ) 9 7 0 . 9 6 0 . ) 9 8 0 . 0 4 1 . ) 9 7 1 . 1 2 0 0 . 1 2 0 0 . 1 0 0 0 . 1 0 0 < . 0 0 1 . 1 0 0 < . 7 8 4 0 . , 8 6 0 ( . 1 9 0 . ) 9 1 1 . 0 8 2 0 . , 3 5 0 ( . ) 5 6 0 . 7 6 0 . ) 6 8 0 . , 5 0 1 ( . 4 3 1 . ) 3 7 1 . , 5 6 0 ( . 6 8 0 . ) 3 1 1 . 0 0 1 . 1 0 0 < . , 9 4 0 ( . 3 6 0 . 1 0 0 < . , 0 4 0 ( . 1 5 0 . 1 0 0 < . 9 6 2 0 . , 1 4 0 ( . , 0 7 0 ( . 2 5 0 . ) 6 6 0 . 8 8 0 . ) 1 1 1 . – – – 1 0 0 < . 1 0 0 < . 1 0 0 < . 5 0 0 0 . 1 3 0 0 . 9 3 7 0 . 1 0 0 < . 1 1 5 0 . 7 1 1 . ) 3 4 1 . 0 6 1 . ) 1 7 1 . l e a m e F s v l e a M r e d n e G r e d o l s r a e y 0 1 r e p s r a e y , e g A y t i l a n o i t a N , 9 6 0 ( . 3 8 0 . ) 8 9 0 . , 2 7 0 ( . 5 9 0 . ) 7 2 1 . , 5 4 0 ( . 7 5 0 . ) 1 7 0 . , 6 8 0 ( . 8 0 1 . ) 5 3 1 . 0 0 1 . , 0 3 1 ( . 4 6 1 . ) 7 0 2 . ) 5 2 2 . 0 0 1 . , 4 5 0 ( . 0 0 1 . 9 6 0 . ) 0 9 0 . d e l l o r n e h t r o N r e t n e C h t u o S 6 0 0 2 – 2 0 0 2 2 1 0 2 – 7 0 0 2 r a e y r a d n e a C l 6 1 0 2 – 2 1 0 2 i n o g e R l i a c h p a r g o e G s i s o n g a D S D A I i o N , s v s e Y 0 0 5 – 1 0 3 0 0 3 ≤ t n u o c 4 D C Shanyinde et al. BMC Public Health (2019) 19:1291 Page 8 of 13 l a b o g l - p e u a v l 0 1 0 0 . e u a v l - p 8 0 0 0 . 1 0 0 < . 4 3 0 0 . 1 0 0 0 . 1 0 0 0 . 2 5 6 0 . 1 0 0 < . 1 0 0 < . 1 0 0 < . 1 0 0 < . 1 0 0 0 . 1 0 0 < . 1 0 0 < . 4 0 0 0 . 1 0 0 < . 7 0 0 0 . 5 0 4 0 . 6 0 0 0 . ] I C % 5 9 [ H R 4 l e d o M - p l a b o g l e u a v l e u a v l - p ] I C % 5 9 [ H R 3 l e d o M l a b o g l - p , 8 0 1 ( . , 2 7 1 ( . , 3 0 1 ( . , 1 2 1 ( . , 2 2 1 ( . , 0 8 0 ( . , 1 4 1 ( . , 3 3 0 ( . , 2 3 0 ( . , 5 4 0 ( . , 4 2 2 ( . , 8 0 1 ( . , 3 7 0 ( . , 2 1 1 ( . 9 3 1 . ) 0 8 1 . 3 4 2 . ) 4 4 3 . 0 0 1 . 5 5 1 . ) 3 3 2 . 9 5 1 . ) 0 1 2 . 4 6 1 . ) 2 2 2 . 6 0 1 . ) 0 4 1 . 3 0 2 . ) 3 9 2 . 0 0 1 . 4 4 0 . ) 8 5 0 . 3 4 0 . ) 7 5 0 . 2 6 0 . ) 6 8 0 . 0 0 1 . 4 9 2 . ) 6 8 3 . 4 3 1 . ) 7 6 1 . 0 0 1 . 1 9 0 . ) 3 1 1 . 6 4 1 . 2 1 0 0 . 0 0 1 . 1 0 0 < . 1 4 0 0 . 1 0 0 0 . 1 0 0 0 . 2 5 7 0 . , 2 0 1 ( . , 9 1 1 ( . 3 5 1 . ) 0 3 2 . 7 5 1 . ) 7 0 2 . , 1 2 1 ( . 3 6 1 . ) 9 1 2 . , 9 7 0 ( . 5 0 1 . ) 8 3 1 . 0 1 0 0 . 1 0 0 < . , 7 0 1 ( . , 6 8 1 ( . 9 3 1 . ) 8 7 1 . 2 6 2 . ) 0 7 3 . e u a v l e u a v l - p 3 1 1 0 . 1 0 0 < . 0 7 6 0 . 0 1 4 0 . 6 6 5 0 . 1 0 0 < . 1 0 0 < . 1 0 0 < . 1 0 0 < . 1 0 0 < . 8 0 0 0 . 1 0 0 < . 5 0 0 0 . , 4 3 0 ( . 5 4 0 . ) 9 5 0 . , 3 3 0 ( . 4 4 0 . ) 8 5 0 . , 7 4 0 ( . 4 6 0 . ) 9 8 0 . , 2 2 2 ( . 1 9 2 . ) 2 8 3 . , 0 1 1 ( . 6 3 1 . ) 9 6 1 . 0 0 1 . ) 4 8 2 . 0 0 1 . – – – 1 0 0 < . 1 0 0 < . , 7 3 1 ( . 7 9 1 . 1 0 0 < . 1 0 0 < . ] I C % 5 9 [ H R 2 l e d o M l , 5 9 0 ( . 2 2 1 . ) 6 5 1 . , 0 6 3 ( . 3 9 4 . ) 5 7 6 . , 3 7 0 ( . 9 0 1 . ) 2 6 1 . , 6 8 0 ( . 2 1 1 . ) 5 4 1 . , 2 8 0 ( . 8 0 1 . ) 3 4 1 . , 4 4 0 ( . 7 5 0 . ) 6 7 0 . 0 0 1 . , 4 3 1 ( . 2 9 1 . ) 7 7 2 . – – – – – – – – – – a b o g l - p e u a v l - p ] I C % 5 9 [ H R 1 l e d o M l a b o g l - p e u a v l e u a v l – – – – – – – – – – – – – – – – – – e u a v l - p 4 2 7 0 . 1 0 0 < . d e t s u d a n U j ] I C % 5 9 [ H R , 7 7 0 ( . 6 9 0 . ) 0 2 1 . , 2 7 2 ( . 6 3 3 . ) 6 1 4 . n w o n k n U 0 0 5 > 1 0 0 < . 0 0 1 . 8 1 0 0 . , 4 4 0 ( . 4 6 0 . , 0 0 0 0 1 – 0 0 0 5 0 0 0 5 ≤ 1 0 0 < . , 0 5 0 ( . ) 3 9 0 . 3 6 0 . ) 9 7 0 . , 0 0 0 0 0 1 – 0 0 0 0 1 , 1 0 0 < . 1 0 0 < . , 3 3 1 ( . 1 9 1 . ) 4 7 2 . 5 0 0 0 . , 2 1 1 ( . ) 0 9 0 . 3 4 1 . ) 3 8 1 . n w o n k n U o N , s v s e Y n o i t c e f n i V B H 5 0 0 0 . , 6 5 0 ( . 1 7 0 . , 0 0 0 0 0 1 > n o i s s i m s n a r T I V H f o e d o M 1 0 0 < . 1 0 0 < . , 3 1 0 ( . 5 1 0 . ) 9 1 0 . 0 0 1 . 1 0 0 < . , 5 1 0 ( . 8 1 0 . 1 0 0 < . 1 0 0 < . 1 0 0 < . 1 0 0 < . , 4 2 0 ( . ) 2 2 0 . 2 3 0 . ) 3 4 0 . 0 0 1 . , 9 8 5 ( . 0 1 7 . ) 5 5 8 . , 7 3 1 ( . 7 6 1 . ) 5 0 2 . 1 0 0 < . 0 0 1 . 1 0 0 0 . , 3 1 1 ( . 9 3 1 . 1 0 0 < . , 9 8 1 ( . ) 0 7 1 . 0 3 2 . l a u x e s o r e t e H s t c a t n o c r e h t O n o i t c e f n i V C H o N s e Y n w o n k n U s u t a t s g n i k o m S n w o n k n U o N s e Y I D W P M S M L m i / s e p o c , d a o l l a r i V ) d e u n i t n o C ( n o i t a t u p m i l e p i t l u m h t i w / t u o h t i w e s a e s i d r e v i l e r e v e s r o f l s i s y a n a n o i s s e r g e r x o C l e b a i r a v i t l u m d n a l e b a i r a v n U 2 i e l b a T Shanyinde et al. BMC Public Health (2019) 19:1291 Page 9 of 13 e u a v l e u a v l e u a v l e u a v l l a b o g l - p e u a v l - p ] I C % 5 9 [ H R 4 l e d o M - p l a b o g l e u a v l e u a v l - p ] I C % 5 9 [ H R 3 l e d o M l a b o g l - p e u a v l - p ] I C % 5 9 [ H R 2 l e d o M l a b o g l - p e u a v l - p ] I C % 5 9 [ H R 1 l e d o M l a b o g l - p e u a v l - p d e t s u d a n U j ] I C % 5 9 [ H R ) d e u n i t n o C ( n o i t a t u p m i l e p i t l u m h t i w / t u o h t i w e s a e s i d r e v i l e r e v e s r o f l s i s y a n a n o i s s e r g e r x o C l e b a i r a v i t l u m d n a l e b a i r a v n U 2 i e l b a T 4 2 7 0 . 3 5 3 0 . 4 1 1 0 . , 8 8 0 ( . , 4 9 0 ( . ) 1 9 1 . 3 1 1 . ) 5 4 1 . 0 0 1 . 0 3 1 . ) 1 8 1 . 5 1 3 0 . , 9 8 0 ( . 3 1 1 . 4 9 3 0 . , 7 8 0 ( . 0 2 1 0 . , 3 9 0 ( . ) 4 4 1 . 0 0 1 . 9 2 1 . ) 8 7 1 . 7 3 0 0 . , 3 0 1 ( . 1 1 1 . ) 2 4 1 . 0 0 1 . 3 4 1 . ) 9 9 1 . 9 3 3 0 . , 8 8 0 ( . 3 1 1 . 1 6 2 0 . 0 2 0 0 . , 7 0 1 ( . 8 4 1 . ) 5 0 2 . ) 5 4 1 . 0 0 1 . 2 0 0 0 . , 0 9 0 ( . , 4 2 1 ( . ) 0 8 2 . 5 1 1 . ) 7 4 1 . 0 0 1 . 0 7 1 . ) 4 3 2 . n o i t a t u p m I l e p i t l u M h t i W n o i t p m u s n o c l o h o c A l i r e n a t s b A e t a r e d o M s u o d r a z a H s u t a t s i g n k o m S + 3 l e d o M . 4 ; n o i t c e f n i V C H d n a n o i s s i m s n a r t I V H f o e d o M + 2 l e d o M . 3 ; V B H + s r o t c a f l d e t a e r - V H + 1 I l e d o M . 2 ; d e l l o r n e r a e y r a d n e a c l , i n o g e r l a c i h p a r g o e g , y t i c i n h t e , r e d n e g , e g A . 1 Shanyinde et al. BMC Public Health (2019) 19:1291 Page 10 of 13 Fig. 3 Cox regression adjusted HRs stratified by HCV status for risk of severe liver disease Different classifications of alcohol consumption were used across studies making it difficult to make valid comparisons. It has to be noted that most of other pub- lished studies were enriched with people in specific risk groups who are more likely to have alcohol problems [8, 36, 39]. In contrast, our estimates are from a hetero- geneous cohort with 12% PWID, 42% MSMs and 40% who acquired HIV through heterosexual contacts. In- deed, we did find that PWID were more likely to have missing data for alcohol consumption which in turn was associated with higher risk of SLD. This analysis also set out to investigate whether our measure of alcohol consumption was useful to predict the risk of SLD. Seven percent of our study population experi- enced SLD over follow-up. Although the risk of SLD ap- peared marginally lower for moderate drinkers compared with abstainers, this was not statistically significant. A lower risk in moderate drinkers compared to abstainers has been previously documented and a possible explan- ation could be due to the fact that patients who are cur- rently abstaining may include individuals who were never drinkers as well as those who previously drank and had to abstain due to medical reasons or other reasons [40]. In our study it was not possibly to separate these groups. We found an association of hazardous drinking with increased risk of SLD that was largely independent of baseline demographic and HIV related factors and HBV. However, after further adjusting for mode of HIV transmission, HCV infection and smoking status the association was largely attenuated. Lim JK et al, in 2111 PLWHIV also found a moderate increased risk of advanced fibrosis in people with hazardous alcohol consumption (aOR = 1.26 (95%CI: 0.87–1.82) compared to non-hazardous drinkers) after adjusting for a number of potential confounders in- cluding HCV infection [11]. In another study including 308 PLWHIV, in which heavy alcohol consumption was defined as > 2 drinks/day or ≥ 5 drinks per occasion and > 1 drink per day or ≥ 4 drinks per occasion for men and women respectively, reported that 10% of the study popu- lation developed of liver fibrosis. Consistent with our re- sults, the authors found a moderate difference in risk according to alcohol consumption and no significant asso- ciation between heavy alcohol consumption and risk of advanced liver fibrosis (aOR = 1.14 [95%CI: 0.47–2.77] compared to non-heavy alcohol consumption; p = 0.77) [41]. In contrast, Chaudhry et al. 2009 did find an associ- ation between alcohol consumption and liver fibrosis mea- sured using the APRI score after adjusting for potential confounders including HCV infection[aRR = 2.30 (95%CI: 1.26–4.17)]. Of interest, also in their analysis there was no evidence that the association between alcohol consump- tion and the risk of SLD varied by HCV infection status. Our study has several other limitations that should be addressed. First, we used a time-fixed covariate at enrol- ment to classify individuals’ drinking behaviour for the study period, although it is possible that drinking habits changed over follow-up potentially leading to a dilution of the association. This was done mainly to simplify the analysis as mechanisms of time-dependent confounding in this context are largely unexplored and potentially dif- ficult to address by means of a standard Cox regression analysis. Secondly, as typical in the observational setting, we have to assume that results cannot be explained by residual or unmeasured confounding. In addition, be- cause of the large proportion of people with missing data, selection bias cannot be entirely ruled out. People Shanyinde et al. BMC Public Health (2019) 19:1291 Page 11 of 13 lacking information on alcohol were indeed different from those with complete data for a number of factors known to be associated with the risk of SLD (Additional file 2: Table S2). However, use of multiple imputation gave similar results to the main analysis and the amount of missing data observed in out cohorts is consistent with other HIV cohorts collecting data on alcohol use. Thirdly, data collected on mode of HIV transmission in ICONA and HepaICONA is not able to distinguish be- tween ex-PWID and current PWID, leading to potential residual confounding due to misclassification. Finally, it’s possible that the physicians may not ask the questions on alcohol use in standardised fashion in accordance with the format on the eCRF. Conclusion In conclusion, we evaluated the value of information on alcohol consumption obtained by brief physician interview of PLWHIV to predict their future risk of occurrence of SLD. We found an association between alcohol consump- tion and risk of SLD which was however partly explained by differences in HCV status, HIV mode of transmision and smoking. There was no evidence that the association was stronger when restricting the analysis to the HCV- infected population. It is conceivable that the weak associ- ation found is due to misclassification of the exposure, so efforts should be made in order to collect more accurate information on alcohol consumption in cohorts of HIV/ HCV infected individuals. In particular data collection on historical alcohol consumption including items which could allow to distinguish between people who had cur- rently stopped drinking from those who never drank would be useful for future studies. Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s12889-019-7608-1. Additional file 1. Table S1. List of Ethic Committees that Approved Icona Study. Additional file 2. Table S2. Patients’ characteristics stratified by reported and non-reported alcohol consumption. Abbreviations ART: Antiretroviral treatment; CRF: Case report form; DAA: Direct acting antivirals; HCV: Hepatitis C virus; HIV: Human immunodeficiency virus; NIFN: National institute for food and nutrition; PLWHIV: People living with HIV; PWID: People who inject drugs; SLD: Severe liver disease; WHO: World health organisation Acknowledgements ICONA Foundation Study Group and HepaICONA study group BOARD OF DIRECTORS A d’Arminio Monforte (Vice-President), M Andreoni, G Angarano, A Antinori, F Castelli, R Cauda, G Di Perri, M Galli, R Iardino, G Ippolito, A Lazzarin, CF Perno, F von Schloesser, P Viale. SCIENTIFIC SECRETARY A d’Arminio Monforte, A Antinori, A Castagna, F Ceccherini-Silberstein, A Cozzi-Lepri, E Girardi, S Lo Caputo, C Mussini, M Puoti. STEERING COMMITTEE M Andreoni, A Ammassari, A Antinori, C Balotta, A Bandera, P Bonfanti, S Bonora, M Borderi, A Calcagno, L Calza, MR Capobianchi, A Castagna, F Ceccherini-Silberstein, A Cingolani, P Cinque, A Cozzi-Lepri, A d’Arminio Monforte, A De Luca, A Di Biagio, E Girardi, N Gia- notti, A Gori, G Guaraldi, G Lapadula, M Lichtner, S Lo Caputo, G Madeddu, F Maggiolo, G Marchetti, S Marcotullio, L Monno, C Mussini, S Nozza, M Puoti, E Quiros Roldan, R Rossotti, S Rusconi, MM Santoro, A Saracino, M Zaccarelli. STATISTICAL AND MONITORING TEAM A Cozzi-Lepri, I Fanti, L Galli, P Loren- zini, A Rodano, M Shanyinde, A Tavelli. BIOLOGICAL BANK INMI F Carletti, S Carrara, A Di Caro, S Graziano, F Petrone, G Prota, S Quartu, S Truffa. PARTICIPATING PHYSICIANS AND CENTERS A Giacometti, A Costantini, C Valeriani (Ancona); G Angarano, L Monno, C Santoro (Bari); F Maggiolo, C Suardi (Bergamo); P Viale, V Donati, G Verucchi (Bologna); F Castelli, E Quiros Roldan, C Minardi (Brescia); T Quirino, C Abeli (Busto Arsizio); PE Manconi, P Piano (Cagliari); B Cacopardo, B Celesia (Catania); J Vecchiet, K Falasca (Chieti); L Sighinolfi, D Segala (Ferrara); F Mazzotta, F Vichi (Firenze); G Cassola, C Viscoli, A Alessandrini, N Bobbio, G Mazzarello (Genova); C Mastroianni, V Belvisi (Latina); P Bonfanti, I Caramma (Lecco); A Chiodera, P Milini (Macerata); M Galli, A Lazzarin, G Rizzardini, M Puoti, A d’Arminio Monforte, AL Ridolfo, R Piolini, A Castagna, S Salpietro, L Carenzi, MC Moioli, C Tincati, G Marchetti (Milano); C Mussini, C Puzzolante (Modena); A Gori, G Lapadula (Monza); N Abrescia, A Chirianni, G Borgia, R Orlando, F Di Martino, L Maddaloni, I Gentile, G Bonadies (Napoli); A Cascio, C Colomba (Palermo); F Baldelli, E Schiaroli (Perugia); G Parruti, T Ursini (Pescara); G Magnani, MA Ursitti (Reggio Emilia); R Cauda, M Andreoni, A Antinori, V Vullo, A Cristaudo, A Cingolani, G Baldin, S Cicalini, L Gallo, E Nicastri, R Acinapura, M Capozzi, R Libertone, S Savinelli, A Latini, G Iaiani, L Fontanelli Sulekova (Roma); M Cecchetto, F Viviani (Rovigo); MS Mura, G Madeddu (Sassari); A De Luca, B Rossetti (Siena); D Francisci, C Di Giuli (Terni); P Caramello, G Di Perri, GC Orofino, S Bonora, M Sciandra (Torino); M Bassetti, A Londero (Udine); G Pellizzer, V Manfrin (Vicenza). Authors’ contributions MS conceived and designed the analysis, performed the analysis, interpreted the output and wrote the manuscript. EG, MP, ADL and ADM commented on the draft of the manuscript. LS and UFC and PC collected the data. ACL and FL conceived and designed the analysis and provided critical revisions of the manuscript. All authors read and approved the final version. Funding The authors received no specific funding for this work. Availability of data and materials The datasets generated and/or analysed during the current study are not publicly available due [reasons below] but are available from the corresponding author on reasonable request. As a group we are very open to collaboration and in involving other researchers in our work. However, we strongly feel that we cannot make a full dataset publicly available for several reasons: 1. We are extremely concerned about confidentiality – since these patients may be identified by combinations of person-specific characteristics within the database, and the database includes sensitive data such as HIV status, HCV status, as well as data on intravenous drug use as well as sexual preferences. 2. Further, being a multicentre study, the study has gone through a long process of being approved by the individual clinics and national Ethical Committees. Such approvals do not include granting public access to the data being collected. This would mean that we would have to go back for renewed evaluation by all clinics as well as by national Ethical Committees in all sites. Ethics approval and consent to participate All participants have given written informed consent signed by the individual to participate the in the study and ethic committee approval from all participating centres was obtained for both cohorts. IRBs at each participating clinical site have approved the Icona protocol [Additional file 1: Table S1]. Consent for publication Not applicable. Shanyinde et al. BMC Public Health (2019) 19:1291 Page 12 of 13 Competing interests The authors declare that they have no competing interests. Author details 1Institute of Global Health, University College London, Royal Free Hospital, London, UK. 2Lazzaro Spallanzani National Institute for Infectious Diseases, Rome, Italy. 3Niguarda ca Grande Hospital, Milan, Italy. 4University Hospital of Siena, Siena, Italy. 5Division of Infectious Diseases, Hospital of Ferrara, Ferrara, Italy. 6Infectious disease university San Raffaele, San Raffaele, Italy. 7Infectious and Tropical Diseases Unit I, Department of Infectious Diseases, Amedeo di Savoia Hospital, Torino, Italy. 8Clinic of Infectious Diseases, San Paulo Hospital, Milan, Italy. Received: 29 October 2018 Accepted: 10 September 2019 References 1. Samet JH, Phillips SJ, Horton NJ, Traphagen ET, Freedberg KA. Detecting alcohol problems in HIV-infected patients: use of the CAGE questionnaire. AIDS Res Hum Retrovir. 2004;20(2):151–5. Justice A, Sullivan L, Fiellin D. HIV/AIDS, comorbidity, and alcohol: can we make a difference? Alcohol Res Health. 2010;33(3):258–66. Conigliaro J, Gordon AJ, McGinnis KA, Rabeneck L, Justice AC. How harmful is hazardous alcohol use and abuse in HIV infection: do health care providers know who is at risk? J Acquir Immune Defic Syndr (1999). 2003; 33(4):521–5. Lefevre F, O'Leary B, Moran M, Mossar M, Yarnold PR, Martin GJ, et al. Alcohol consumption among HIV-infected patients. J Gen Intern Med. 1995; 10(8):458–60. Galvan FH, Bing EG, Fleishman JA, London AS, Caetano R, Burnam MA, et al. The prevalence of alcohol consumption and heavy drinking among people with HIV in the United States: results from the HIV cost and services utilization study. J Stud Alcohol. 2002;63(2):179–86. Bonacini M. Alcohol use among patients with HIV infection. Ann Hepatol. 2011;10(4):502–7. 2. 3. 4. 5. 6. 7. Muga R, Sanvisens A, Fuster D, Tor J, Martinez E, Perez-Hoyos S, et al. Unhealthy alcohol use, HIV infection and risk of liver fibrosis in drug users with hepatitis C. PLoS One. 2012;7(10):e46810. 8. Wandera B, Tumwesigye NM, Nankabirwa JI, Kambugu AD, Parkes-Ratanshi R, Mafigiri DK, et al. Alcohol consumption among HIV-infected persons in a large urban HIV clinic in Kampala Uganda: a constellation of harmful behaviors. PLoS One. 2015;10(5):e0126236. Benhamou Y, Bochet M, Di Martino V, Charlotte F, Azria F, Coutellier A, et al. Liver fibrosis progression in human immunodeficiency virus and hepatitis C virus coinfected patients. The Multivirc group. Hepatology (Baltimore, Md). 1999;30(4):1054–8. 9. 10. Cooper CL, Cameron DW. Effect of alcohol use and highly active 11. antiretroviral therapy on plasma levels of hepatitis C virus (HCV) in patients coinfected with HIV and HCV. Clin Infect Dis. 2005;41(Suppl 1):S105–9. Lim JK, Tate JP, Fultz SL, Goulet JL, Conigliaro J, Bryant KJ, et al. Relationship between alcohol use categories and noninvasive markers of advanced hepatic fibrosis in HIV-infected, chronic hepatitis C virus-infected, and uninfected patients. Clin Infect Dis. 2014;58(10):1449–58. 12. Marcellin P, Pequignot F, Delarocque-Astagneau E, Zarski JP, Ganne N, Hillon P, et al. Mortality related to chronic hepatitis B and chronic hepatitis C in France: evidence for the role of HIV coinfection and alcohol consumption. J Hepatol. 2008;48(2):200–7. Salmon-Ceron D, Lewden C, Morlat P, Bevilacqua S, Jougla E, Bonnet F, et al. Liver disease as a major cause of death among HIV infected patients: role of hepatitis C and B viruses and alcohol. J Hepatol. 2005;42(6):799–805. Tsui JI, Cheng DM, Libman H, Bridden C, Saitz R, Samet JH. Risky alcohol use and serum aminotransferase levels in HIV-infected adults with and without hepatitis C. J Stud Alcohol Drugs. 2013;74(2):266–70. 13. 14. 15. Zakhari S, Li TK. Determinants of alcohol use and abuse: impact of quantity and frequency patterns on liver disease. Hepatol (Baltimore, Md). 2007;46(6):2032–9. 16. Chaudhry AA, Sulkowski MS, Chander G, Moore RD. Hazardous drinking is associated with an elevated aspartate aminotransferase to platelet ratio index in an urban HIV-infected clinical cohort. HIV Med. 2009;10(3):133–42. Saunders JB, Aasland OG, Amundsen A, Grant M. Alcohol consumption and related problems among primary health care patients: WHO collaborative 17. project on early detection of persons with harmful alcohol consumption--I. Addiction. 1993;88(3):349–62. 18. Chander G, Lau B, Moore RD. Hazardous alcohol use: a risk factor for non- adherence and lack of suppression in HIV infection. J Acquir Immune Defic Syndr (1999). 2006;43(4):411–7. 19. Marcellin F, Lions C, Winnock M, Salmon D, Durant J, Spire B, et al. Self- reported alcohol abuse in HIV–HCV co-infected patients: a better predictor of HIV virological rebound than physician's perceptions (HEPAVIH ARNS CO13 cohort). Addiction. 2013;108(7):1250–8. 20. Westerberg VS. Alcohol measuring scales may influence conclusions about the role of alcohol in human immunodeficiency virus (HIV) risk and progression to acquired immunodeficiency syndrome (AIDS). Am J Epidemiol. 1992;135(7):719–25. 22. 21. Crum RM, Galai N, Cohn S, Celentano DD, Vlahov D. Alcohol use and T- lymphocyte subsets among injection drug users with HIV-1 infection: a prospective analysis. Alcohol Clin Exp Res. 1996;20(2):364–71. Samet JH, Horton NJ, Traphagen ET, Lyon SM, Freedberg KA. Alcohol consumption and HIV disease progression: are they related? Alcohol Clin Exp Res. 2003;27(5):862–7. Samet JH, Cheng DM, Libman H, Nunes DP, Alperen JK, Saitz R. Alcohol consumption and HIV disease progression. J Acquir Immune Defic Syndr (1999). 2007;46(2):194–9. 23. 24. Wu ES, Metzger DS, Lynch KG, Douglas SD. Association between alcohol use and HIV viral load. J Acquir Immune Defic Syndr (1999). 2011;56(5):e129–30. 25. Conen A, Fehr J, Glass TR, Furrer H, Weber R, Vernazza P, et al. Self- reported alcohol consumption and its association with adherence and outcome of antiretroviral therapy in the Swiss HIV cohort study. Antivir Ther. 2009;14(3):349–57. 26. Holmes JM, Gerhardstein KR, Griffin PT. Brief screening for alcohol use disorders in HIV primary care. HIV Clin. 2011;23(4):8–13. 27. Asiimwe SB, Fatch R, Emenyonu NI, Muyindike WR, Kekibiina A, Santos G-M, et al. Comparison of traditional and novel self-report measures to an alcohol biomarker for quantifying alcohol consumption among HIV-infected adults in sub-Saharan Africa. Alcohol Clin Exp Res. 2015;39(8):1518–27. 28. Hahn JA, Samet JH. Alcohol and HIV disease progression: weighing the evidence. Curr HIV/AIDS Rep. 2010;7(4):226–33. 29. Dawson DA. Methodological issues in measuring alcohol use. Alcohol Res Health. 2003;27(1):18–29. 30. d'Arminio Monforte A, Lepri AC, Rezza G, Pezzotti P, Antinori A, Phillips AN, et al. Insights into the reasons for discontinuation of the first highly active antiretroviral therapy (HAART) regimen in a cohort of antiretroviral naive patients. I.CO.N.A. study group. Italian Cohort Antiretroviral-Naive Patients AIDS (London, England). 2000;14(5):499–507. 31. HepaICONA. Hepatitis ICONA Cohort 2016 [Available from: http://www. fondazioneicona.org/_new/pages/HepaICONA/. 32. NIFN. National Institute for food and nutrition : Bevande alcoliche se si in quantita controllata report no 7 2012 [Available from: http://nut. entecra.it/656/Bevande_alcoliche__se_si__solo_in_quantit_agrave__ controllata.html. Accessed 16 May 2016. 33. WHO. Alcohol consumption: levels and patterns. 2014. 34. McPherson S, Stewart SF, Henderson E, Burt AD, Day CP. Simple non- invasive fibrosis scoring systems can reliably exclude advanced fibrosis in patients with non-alcoholic fatty liver disease. Gut. 2010;59(9):1265–9. 35. Hormes JM, Gerhardstein KR, Griffin PT. Under-reporting of alcohol and substance use versus other psychiatric symptoms in individuals living with HIV. AIDS Care. 2012;24(4):420–3. 36. Roux P, Cohen J, Lascoux-Combe C, Sogni P, Winnock M, Salmon-Ceron D, et al. Determinants of the underreporting of alcohol consumption by HIV/ HCV co-infected patients during face-to-face medical interviews: the role of the physician. Drug Alcohol Depend. 2011;116(1–3):228–32. Santos GM. Self-reported alcohol abstinence associated with ART initiation among HIV-infected persons in rural Uganda, vol. 134; 2014. 37. 38. Hahn JA, Bwana MB, Javors MA, Martin JN, Emenyonu NI, Bangsberg DR. Biomarker testing to estimate under-reported heavy alcohol consumption by persons with HIV initiating ART in Uganda. AIDS Behav. 2010;14(6):1265–8. Tran BX, Nguyen N, Ohinmaa A, Duong AT, Nguyen LT, Van Hoang M, et al. Prevalence and correlates of alcohol use disorders during antiretroviral treatment in injection-driven HIV epidemics in Vietnam. Drug Alcohol Depend. 2013;127(1–3):39–44. Stockwell T, Zhao J, Panwar S, Roemer A, Naimi T, Chikritzhs T. Do “moderate” drinkers have reduced mortality risk? A systematic review and 39. 40. Shanyinde et al. BMC Public Health (2019) 19:1291 Page 13 of 13 41. meta-analysis of alcohol consumption and all-cause mortality. J Stud Alcohol Drugs. 2016;77(2):185–98. Fuster D, Tsui JI, Cheng DM, Quinn EK, Bridden C, Nunes D, et al. Impact of lifetime alcohol use on liver fibrosis in a population of HIV-infected patients with and without hepatitis C coinfection. Alcohol Clin Exp Res. 2013;37(9): 1527–35. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12909-023-04421-y	Elnaga et al. BMC Medical Education (2023) 23:433 https://doi.org/10.1186/s12909-023-04421-y BMC Medical Education RESEARCH Open Access Virtual versus paper-based PBL in a pulmonology course for medical undergraduates Heba H. Abo Elnaga1, Manal Basyouni Ahmed2,3, Marwa Saad Fathi3,4 and Sanaa Eissa2,3 Abstract Background Problem-based learning (PBL) remains a valid and effective tool for small-group medical education. Using Virtual patients (VP) case simulation in PBL is a recognizable educational method that has successfully prepared students to focus learning on core information that uses realistic patient-based cases relating to everyday clinical scenarios. Using other modalities as the virtual patient in PBL instead of the paper-based methods remains debatable. This study aimed to evaluate the effectiveness of using VP case simulation mannequin in PBL versus the PBL in paper- based cases in improving the cognitive skills by comparing the grades of a multiple-choice question test and assess its ability to reach students’ satisfaction using questionnaire with Likert survey instrument. Methods The study was conducted on 459 fourth-year medical students studying in the pulmonology module of the internal medicine course, Faculty of Medicine, October 6 University. All students were divided into 16 PBL classes and randomly divided into groups A and B by simple manual randomization. The groups were parallel with a con- trolled cross-over study between paper-based and virtual patient PBL. Results The pre-test showed no significant difference between both, while post-test scores were significantly higher in both VP PBL cases 1 discussing COPD (6.25 ± 0.875) and case 2 discussing pneumonia (6.56 ± 1.396) compared to paper-based PBL (5.29 ± 1.166, 5.57 ± SD1.388, respectively) at p < 0.1 When students in Group A experienced PBL using VP in case 2 after paper-based PBL in case 1, their post-test score improved significantly. (from 5.26 to 6.56, p < .01). Meanwhile, there was a significant regression in the post-test score of the students in Group B when they experienced the paper-based PBL session in case 2 after using PBL using VP in case 1, (from 6.26 to 5.57, p < .01). Most of the students recommended using VP in PBL as they found VP was more engaging and inducing concentration in gathering the information needed to characterize the patient’s problem than in a classroom- paper-based cases ses- sion. They also enjoyed the teaching of the instructor and found it a suitable learning style for them. Conclusion Implementing virtual patients in PBL increased knowledge acquisition and understanding in medical students and was more motivating for students than paper based PBL to gather the needed information. Keywords Problem-based learning, Virtual patient, Medical education *Correspondence: Heba H. Abo Elnaga hebahelmy73.med@o6u.edu.eg Sanaa Eissa drsanaa_mohamed@med.asu.edu.eg Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Elnaga et al. BMC Medical Education (2023) 23:433 Page 2 of 10 Background The importance of the educational environment in medi- cal schools has attracted the attention of educational institutes. Students’ study significantly impacts their course satisfaction, sense of well-being, aspirations, and knowledge acquisition [1–8]. Around fifty years ago, col- laborative learning through case-based or PBL scenarios has been a brilliant way to gain and develop workplace knowledge associated with specific competencies [9]. The process of PBL with numerous variants of the techniques craft the students to detect their knowledge’s limitations; they detect learning gaps essential to answer questions deficient in their reserved knowledge [3]. A problem is generally provided to students on paper or in digital format, and it is then studied and discussed in small groups across two or three sessions. This mode of delivery, however, lacks the realism of a patient encounter; it does not develop or assess students’ behav- ioural talents, reasoning, decision making, or communi- cation skills [10]. Virtual patients are interactive computer simulations of real-life clinical scenarios used for training, education, or assessment in the health professions [11]. This includes various systems that employ diverse technologies and satisfy varied learning requirements [12]. The student is placed in the role of a health care professional, mak- ing judgments concerning the type and order of clinical information gathered, differential diagnosis, and patient management and follow-up [8]. It is predicted that virtual patients will largely address clinical reasoning learning demands [13, 14]. However, the impact of using virtual patients on other educational outcomes has not been thoroughly investigated [12, 15]. Some literature reported that the educational use of virtual patients as one of the techniques for applying PBL provided a successful learn- ing tool where virtual patients provide learners with simulated healthcare experiences while also providing methods for information collecting and clinical decision- making in the case scenario [16]. Another study confirmed that medical students need to be exposed to simulation education experiences on a reg- ular basis in order to maintain psychological stability and provide competent medical care in a clinical setting [17]. Several studies have discovered that simulation can be transformed into a powerful strategy for appropriately training health professionals to effectively address today’s changing world’s challenges [18]. Furthermore, a virtual patient platform in conjunction with a diagnostic reasoning framework could be used for education, diagnostic assessment, and improved correct diagnosis [19]. When comparing PBL using paper-based, and virtual patients, they concluded that employing virtual patients can more effectively increase abilities and at least as effec- tively improve knowledge. Clinical reasoning, procedural skills, and a combination of procedural and team skills improved; proof of effectiveness from various income countries indicates virtual patients’ global applicability. Further research should be conducted to investigate the the potential positive effects of VPL on clinical reasoning that might be gained by introducing such learning sup- port [20, 21]. Many countries are beginning to develop simulation and recent innovations in medical education for both undergraduates and postgraduates, with the hope of sup- porting and improving patient care delivery [22]. The use of VP mannequins is expected to progressively replace the present less genuine paper-based versions of PBL, presuming that future generations are looking for- ward to more creative learning methods [16]. Using other modalities as the virtual patient in PBL instead of the paper-based methods remains debatable. This study aimed to evaluate the the potential positive effects of using VP case simulation mannequin in PBL versus the teaching modalities of PBL in paper-based cases in improving the cognitive skills by comparing the grades of a multiple-choice question test and assess its ability to reach students’ satisfaction using questionnaire with Likert survey instrument. Methods Study design This is a single-center, randomized, parallel-group with a controlled cross-over study design conducted on 459 fourth-year medical students, 279 males (60.8%) and 180 females (39.2%), in the pulmonology module of inter- nal medicine course, Faculty of Medicine, October 6 University. The research was conducted as part of a pre-existing problem-based learning curriculum. Students enrolled in the course voluntarily participated. All study activities, including the completion of multiple-choice self-assess- ment questions to evaluate learning, were completed within the framework of the course, in accordance with the guidelines and legal regulations of the faculty of med- icine, October 6 University without extracurricular activ- ities required. As is customary for all courses at our medical school, all fourth-year medical students were divided into 16 PBL classes at the start of the module. In our study, 8 of which were randomly assigned to group A (233 students) and 8 of which were assigned to group B (226 students) by sim- ple manual randomization method [23]. The PBL curriculum required all groups to attend a two-hour meeting each week and another 2 h session for discussion and performing the posttest. During the first Elnaga et al. BMC Medical Education (2023) 23:433 Page 3 of 10 week of the module, In a paper-based PBL format, Group A faced structured teaching objectives related to man- aging chronic obstructive pulmonary disease (COPD). Meanwhile, Group B conducted the same COPD case using the virtual patient simulator by interviewing a mannequin that delivered information that matched the written script [24], Fig. 1. The students were crossed over to the opposite modal- ity in the second week. During this week, Group A encountered structured teaching points related to the management of a case of pneumonia by the virtual patient, interviewing the mannequin simulator, while Group B completed the same case in a paper-based PBL with identical teaching points. The study’s virtual patient mannequin, ALEX-PCS, is a high-fidelity manikin patient communication simula- tor that incorporates the most recent computer hard- ware technology, similar to the SimMan. This wireless high-fidelity manikin has been programmed to provide an extremely realistic full-body patient presentation. It also achieves the highest level of realism and provides a diverse set of high stakes learning scenarios. Cases were developed from real clinical records col- lected locally. Their content was matched for topic and difficulty level and set up to be as similar as possible in terms of complexity and case type. All Students partici- pated in other similar activities throughout the pulmo- nology curriculum as lectures, small group teaching, and skill lab. All PBL faculty facilitators involved in teaching either the paper-based PBL or using the virtual patient mannequinare underwent training workshops and were given particular instructions to reinforce the teaching points related to managing the assigned PBL cases in both groups. The study was approved by the ethical committee of the Faculty of Medicine of both October 6 University and Ain Shams University (ethical approval number: MS 769/ 2021). Informed, written consent was obtained from all participants of the study. All methods were performed in accordance with the legal regulations of the faculty of medicine, October 6 University without extracurricular activities. Assessment methods Each student completed an online multiple-choice ques- tion and submitted it successfully as a pre-experience and post-experience test specific to the topic area. The tests covered the pulmonology case PBL session either by using virtual patient case simulation PBL or paper-based teaching modalities of PBL both pre-experience and post-experience tests assess the learning of the teaching points and students’ application of knowledge.Both pre- experience and post-experience tests were drawn from the question bank of the faculty of medicine, October 6 University which was previously used in different exams. they were tested before for their validity and reliability by the expert members of faculty of medicine according to the results of facility and discrimination indices and the presence of functional distractors of each question calcu- lated by the software program. Student performance on pre- and post-experience assessment tests had no bearing on their final course grade. Both pre-tests and post-tests were scored using a web-based course management system using an answer Fig. 1 Flow chart of participant selection Elnaga et al. BMC Medical Education (2023) 23:433 Page 4 of 10 key written prior to administration, and each student received an individual score [25]. At the end of the module, all students were asked to complete a questionnaire using Likert survey instru- ment with response options ranging from 0 = strongly disagree to 4 = strongly agree) to reflect their experi- ence and satisfaction with the virtual patient case sim- ulation in Problem-based learning (PBL) compared to paper-based teaching modalities of PBL of a written case scenario [26]. The structured self-administered questionnaire was translated into Arabic to make it simple and understand- able for all participants. We contacted the students via the Microsoft Teams application. We invited them to participate in the study via an electronic link with a ques- tionnaire after explaining the purpose of this research. The questionnaires were designed following thorough research of related literature. They were authorized by Ain Shams University’s specialist staff members in pub- lic health and community medicine, who reviewed their validity and reliability [27, 28]. The questionnaire included the students’ socio-demo- graphic characteristics involving age, sex, and nationality. Moreover, the consent, which feeds agreeing of the stu- dent to participate in the study and to answer the ques- tionnaire with the researcher’s contacts, was comprised. Inclusion criteria This randomized, controlled cross-over study included all undergraduate fourth-year medical students enrolled in the Internal Medicine course in the 2021–2022 academic year of the pulmonology course at the faculty of medi- cine, October 6 University. The pulmonology module is an integral component of the internal medicine course’s curriculum, lasting four weeks and consisting of an aver- age of eight lectures per week, small group learning and discussion sessions (centered on solving clinical cases), and others for demonstration and teaching clinical exam- ination of real patients. All the students enrolled in the study had equivalent knowledge and skills in pre-clinical sciences, and their IT skills were comparable. Therefore, no assessable differ- ences could be observed between the included students at the study’s baseline. Exclusion criteria • Students who were not registered in the chest dis- eases module. • Students who failed previously in the module and are repeating it, • Students who did not regularly attend the teaching sessions of PBL or simulation by virtual patient. Data analysis The average score for each of the pre-test and the post- test each week were calculated and compared between the two groups of the paper-based PBL and Virtual patient group for both the COPD and pneumonia cases. Comparisons between the two groups were made using independent t-tests, and comparisons within the same group pre and post-tests were made using paired t-tests. Analysis was performed using statistical software (SPSS version 23; SPSS, Inc., Chicago, IL). Data are presented as mean ± SD for quantitative data and frequency (%) for categorical data; a p-value of < 0.05 was considered statis- tically significant. Results The study included 459 students studying in the Res- piratory module of internal medicine course in grade 4, Faculty of Medicine, October 6 University. The partici- pants were 279 males and 180 females, with a mean age of 20.81 ± 2.6 years (Table 1). The participants studied two different case scenarios. During the study on each case, a pre-test was done for all students, and then the students were divided into two groups; Group A, where paper-based learning (PBL) was implemented, and Group B, where virtual-patient learn- ing (VP) was implemented after which a cross-over rear- rangement was done for the second case scenario. A post-test was done after studying each case. In both cases, after conducting PBL by either paper- based or virtual patient approach, the post-tests students’ scores were significantly higher than their pre-tests’ scores; p < 0.0001 (Table 2). There was no significant difference between groups A and B regarding the total pre-test scores (p > 0.05). At the same time, the post-test total scores are significantly increased in the virtual PBL Table 1 Characteristic data of participating students Age (years) Gender Female Male Nationality Egypt Jordan Nigeria Yemen Botswana Sweden Lebanon Total Mean ± SD 20.81 ± 2.601 Frequency (n.) 180 279 Percent (%) 39.2 60.8 Frequency (n.) Percent (%) 421 14 10 8 2 2 2 459 91.7 3.1 2.2 1.7 0.4 0.4 0.4 100.0 Elnaga et al. BMC Medical Education (2023) 23:433 Page 5 of 10 Table 2 Comparison between Pre and Posttest mean scores of the students in group A and the students in group B in both case1 and case 2 Table 4 Comparison between mean post-test results of students in groups A and B when they experienced the paper- based PBL versus PBL by VP in both cases Group A Group B Mean ± SD Significance Mean ± SD Significance Case 1 Pre-test 4.30 ± 1.450 Post-test 5.26 ± 1.154 3.55 ± 1.338 Post-test 6.56 ± 1.396 T = -8.774 P < 0.0001 T = -21.253 P < 0.0001 4.39 ± 1.402 T = -17.437 P < 0.0001 6.26 ± 0.872 3.50 ± 1.351 T = -22.181 P < 0.0001 5.57 ± 1.388 Case 2 Pre-test Data are presented as Mean ± SD; p-value < 0.05 is statistically significant Group A: who experienced paper-based PBL in case 1 and PBL using VP in case 2 Group B: who experienced PBL using VP in case 1 and paper-based PBL in case 2 group (group B in case 1 and group A in case 2); p < 0.01 (Table 3). Comparing the scores of post-tests between case 1 and case 2. There was a significantly higher score in case 2 in Group A (from 5.26 to 6.56, p < 0.01). Meanwhile, the score was significantly decreased in Group B (from 6.26 to 5.57, p < 0.01) (Table 4). Survey. A survey questionnaire reflecting the students’ experi- ence and reach of their satisfaction with the (PBL) using VP simulation compared to paper-based teaching modal- ities of PBL of a written case scenario. The student’s responses to the survey were distributed on a Likert scale score of 0–4. There are significant differences in the responses to various questions (p < 0.01) (Fig. 2). Regarding question 1, 142 (30.9%) students strongly agreed, and 208 (45.32%) students agreed that the teaching method of the case by the virtual patient was more helpful and effective than a classroom- paper- based case session. Table 3 Comparison between group A and group B mean total scores after Pre and Post-test in both Case 1 and case 2 Group A Mean ± SD Group B Mean ± SD Significance Case 1 Pre-test 4.30 ± 1.450 4.39 ± 1.402 Post-test 5.26 ± 1.154 6.26 ± 0.872 Case 2 Pre-test 3.55 ± 1.338 3.50 ± 1.351 Post-test 6.56 ± 1.396 5.57 ± 1.388 T = -0.650 P = 0.516 T = -9.106 P < 0.01 T = 0.321 P = 0.748 T = 6.849 P < 0.01 Data are presented as Mean ± SD; p-value < 0.05 is statistically significant Group A: who experienced paper-based PBL in case 1 and PBL using VP in case 2 Group B: who experienced PBL using VP in case 1 and paper-based PBL in case 2 Post-test paper- based PBL Mean ± SD Group A Group B 5.26 ± 1.154 5.57 ± 1.388 Post-test PBL-VP p Mean ± SD 6.56 ± 1.396 6.26 ± .872 < 0.0001 < 0.0001 Data are presented as Mean ± SD; p-value < 0.05 is statistically significant Group A: who experienced paper-based PBL in case 1 and PBL using VP in case 2 Group B: who experienced PBL using VP in case 1 and paper-based PBL in case 2 While in question 2, 134 (29.2%) students strongly agreed, and 212 (46.19%) students agreed that the teaching method of the case by the virtual patient pro- vided them with learning materials and activities to promote seeking knowledge more than a classroom- paper-based cases. By asking whether the students enjoyed teaching the case by the virtual patient more than a classroom- paper- based cases session in question 3, 143 (31.2%) students strongly agreed, and 207 (45.1%) students agreed. Concerning question 4,137 (29%) students strongly agreed, and 199 (43.35%) students that the teaching method of the case by the virtual patient was motivat- ing and helped them to learn more than a classroom- paper-based cases session. Considering question 5, 136 (29.6%) students strongly agreed, and 198 (43.13%) students agreed that the teaching method of the case by the virtual patient was engaging and seemed to reinforce the students effec- tively to introduce new knowledge they needed, to characterize the patient’s problem more than a class- room—paper-based cases session. (6), 141 Concerning question (30.7%) students strongly agreed, and 200 (43.57%) students agreed that the way their instructor(s) taught using the virtual patient suited their learning style more than a class- room- paper-based cases session. Regarding the recommendation of the students to use VP in PBL in upcoming teaching sessions (question 7), 150 (32.7%) students strongly agreed, and 185 (40.31%) agreed, while only 33 (7.2%) students disagreed and 13 (2.8%) strongly disagreed. Finally, question 8 answer adopted a different theme where 140 (30.50%) students rated the quality of the teaching session with the virtual patient as "Good," 126 (27.5%) as very good, and 96 (20.9%) as excellent, in com- parison to the classroom- paper-based cases session. A discussion was made by a focal group of faculty members, who are the practitioners of both PBL teaching Elnaga et al. BMC Medical Education (2023) 23:433 Page 6 of 10 Fig. 2 The responses to various questions differ significantly.(p < 0.01). Likert scale of questions 1 to 7: 0 = strongly disagree, 1 = Disagree, 2 = Neutral, 3 = Agree, 4 = strongly agree. Likert scale of question 8 is 0 = poor, 1 = Neutral, 2 = Good, 3 = Very good, 4 = Excellent methods.Overall, they provided positive and encourag- ing responses that virtual PBL method for teaching pul- monary cases was an enjoyable experience. However, we could not make conclusions about this issue due to the small number of staff participating in this study. Discussion PBL is abundantly used in advanced medical education; in 2005, over 70% of medical schools reported employ- ing PBL in some small group teaching for medical stu- dents in the pre-clinical years [29]. However, despite the importance of PBL as a pedagogical method of learning, in improving students’ abilities such as clinical reasoning, problem-solving and critical thinking [30–32]. The data about the efficacy and outcomes of using other modali- ties as the virtual patient compared with the paper-based methods remains inadequate [33, 34], especially in Egypt. After studying each of the 2 cases (COPD and Pneumo- nia), the chances of the medical students answering the post-test questions correctly after completing the PBL session via virtual patient recorded significantly better scores than their colleagues who completed PBL by the paper-based. These results also reveal that virtual patient PBL increased post-test performance independent of the week the students received the intervention, This is congruent with the findings of other studies, which found that incorporating digital learning objects in PBL increased cognitive, metacognitive, affective, and total learning processes or outcomes [35] This was evident by comparing the post-test scores between case 1 and case 2 in each group. There was a significantly higher score in case 2 in Group A in which they studied PBL by the vir- tual patient Meanwhile, the score of post-tests between case 1 and case 2 was significantly decreased in Group B. Ultimately There was no discernible difference between the pre-test scores of the 2 cases of the two groups of students who participated in the study, which means that both groups of students had an equal level of basic knowledge before they started to study any of the 2 cases. Given the significant improvement in postexperience test scores after using VP in PBL compared with post- experience test scores after classroom- paper-based PBL sessions, we conclude that the method of teaching by PBL is effective in improving the students’ competen- cies. Meanwhile, using the virtual-patient learning in PBL showed excellence over the paper-based methods in enhancing the students’ focus on core information and knowledge relevant to patient case scenarios. According to previous studies, dedicated software can help learn- ers produce explanations, structure exercises, and make them more manageable [36]. We designed this study to compare the application of PBL by both modalities, the paper-based and by the vir- tual patient with cross-over method, to achieve complete fairness for all students to have the same opportunity to experience interactive small group activities of PBL in both methods that revolve around working through a patient case. It was apparent that the questions in both weeks were matched in their level of difficulty. Elnaga et al. BMC Medical Education (2023) 23:433 Page 7 of 10 The results of this study are consistent with previous findings regarding both paper-based PBL and virtual- patient PBL in clinical-level medical students [37]. The current study verified that the primary advantage of PBL using VP enhanced the students learning, com- prehension and recalling of core information relevant to real patient scenarios. This was evidenced by Students’ test scores which significantly improved when undertak- ing VP PBL compared to the paper-based PBL. Previous research indicated that incorporating technology into education provides students with an engaging learning experience, allowing them to stay more engaged in the material without being distracted [38, 39]. When the stu- dents in Group B encountered structured teaching points related to the management of a chronic obstructive pul- monary disease (COPD) case 1 by the virtual patient meeting a mannequin, they achieved significantly higher scores in the post-test of their scores than when they crossed over and encountered structured teaching points related to the management of a case of pneumonia by the paper-based PBL format (case 2). These findings were also reported in group A whose scores on the post-test after encountering structured teaching points related to the management of a case of (COPD) case 1 by the paper-based PBL format improved to be significantly higher than their score in the post- test when they crossed over and encountered structured teaching points related to the management of a case of pneumonia (case 2) by the virtual patient. In addition, there were non-significant statistical dif- ferences between the scores of the pre-test o the 2 cases in the same group, which denotes the equal difficulty of both cases. This study supports the concept advanced by others that such visual, aural, and tactile cues engage learners beyond the solely cognitive features of standard PBL [40]. Another study proved that students prefer the idea of having a dialogue with a patient in order to better under- stand that patient”s problem [15]. As a result, VPs, especially those using a mannequin, can be effectively integrated into clinical education by coordinating their use with other learning activities (e.g., didactics, clinical experiences), by making room in the course by eliminating some lectures and textbook assign- ments, and by taking a voluntary rather than obligatory approach. It was evident that employing a virtual patient in PBL offers the learner an enjoyable environment to make decisions and understand the ramifications of those actions [41]. Such interaction may strengthen learning concepts beyond a vocal discussion of a textual exam- ple, such as VP interactive activities encouraging deeper learning, accentuating understanding and application of knowledge over memory and recall. It was evident in a study that found a general favourable effect from the use of various educational technology in PBL. Making disci- plinary thinking and techniques clear; giving a platform to stimulate articulation, cooperation, and reflection; and lowering perceived cognitive load were all positive conse- quences for student learning.[42]. This was obvious, after analysis of the questionnaire, about the level of satisfaction regarding the use of the virtual patient in PBL at the end of the module where all students had experienced both the written and virtual patient I PBL sessions. Most students preferred learning by VP PBL to paper- based PBL and agreed that VPs were more efficacious concerning learner satisfaction and learning outcomes. Limited previous research on the utility of VPs within medical education found that VPs were equally effective compared to other simulation methods [43–45]. The majority of the students either strongly agreed, or agreed that the virtual patient provided them with learn- ing materials and activities which encouraged exploring new knowledge than a classroom- paper-based cases ses- sion. These findings are consistent with McLean’s who concluded that students consider computer system as the most helpful form of communication and resource deliv- ery for PBL in medicine [46]. This study identified student preferences and agreed with Huwendiek and his colleagues on the usefulness of sequencing and matching VPs with other activities and evaluations. An example of an active learning activity is having students monitor major discoveries presented in a VP mannequin. Creating a summary statement from a VP’s history is an example of positive activity. VPs pro- vide the benefit of a uniform case presentation [47]. Our results showed that most of the students enjoyed the instructor’s approach of presenting the case by the virtual patient more than a classroom- paper-based cases session, as approved in another study in which educa- tional technology assist students and their facilitators in making disciplinary thought clear [36]. The interactive nature of the e-learning model engages the participant, shifting them to an active learning expe- rience. This removes the didactic passivity of a teaching- centered approach, as previously proved by Ruiz and his coworkers [48, 49]. Concerning the motivation for learning, Most of the students strongly agreed, and students agreed that the teaching method of the case by the virtual patient was motivating and helped them to learn more than a classroom-paper-based cases session. This was sup- ported by the findings of Dornan et al., who contended that information system can create motivation to the students [50]. Elnaga et al. BMC Medical Education (2023) 23:433 Page 8 of 10 There was a positive effect of PBL by VPs on engaging the students in a simulation of full-body mannequins with a real-life case scenario that inspired students to gain experience in practice and facilitate high-fidelity learning. Through active engagement and team coop- eration, simulation supplies crucial parts of a reflective practitioner’s future education. This was clear dur- ing measuring student satisfaction where the greatest number, found that the teaching method of the cases by the virtual patient enabled them to characterize the patient’s problem more than the paper-based PBL. This goes in accordance with Gesundheit et al. (2009), who found that medical students were highly satisfied with using VPs, which could be a variable in their engage- ment in learning activities [42, 51]. The results of our study revealed that many students strongly agreed, and agreed that the method their instructor(s) taught the virtual patient was more con- ducive to their learning style than a classroom-paper- based case session. This student’s opinion encourages teachers participating in medical education to seek a modification of the paper-based model of learning and education. This was also recommended by Hmelo-Sil- ver, and his colleagues [36]. However, a review of literature by a fourth-year medi- cal student stated that students did not wish to see paper-based PBL instructor-led instruction but needed enhancement by innovative teaching techniques [52]. This was obvious in the rate of the quality of the teaching session with the virtual patient in compari- son to the classroom- paper-based cases session, where most of students rated it an excellent session,, very good, and "Good." On the other hand, most t students enjoyed how their instructor presented the PBL session using the VP they recommended and favored it as a learning technique in all aspects of the curriculum. This is consistent with the findings of Olaussen and his colleagues study [53]. Moreover, a considerable number of students rec- ommended using virtual patient in upcoming teaching sessions rather than a classroom- paper-based cases session. These findings indicated positive student feed- back on using VP in PBL, which met their expectations during the session. The students had a better area for active engagement by using their whole body and all five senses associated with their intellectual, psycho- logical, and interactional skills. Another study found that VPs as gamification in medical education could successfully inflate skills and effectively improve knowledge, clinical reasoning, pro- cedural skills, and a combination of procedural and team skills [54]. Limitations Limited number of cases conducted ( only two cases) in only one module ( the pulmonary module) in one institu- tion in Egypt,. Also, the completion of the case was dependent on internet connectivity, which could cause the mannequin’s response to be delayed during the session and distract the students’ concentration. Conclusion The findings of this single-center study demonstrated that using virtual patients mannequins was more effec- tive than paper-based PBL for knowledge acquisition and student’satisfaction in fourth-year medical students dur- ing PBL of COPD and pneumonia cases. Abbreviations PBL VP Problem-based learning Virtual patients Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12909- 023- 04421-y. Additional file 1. Acknowledgements Not applicable. Authors’ contributions HHA: Study conception, literature search, study design, collection and analysis of the data, practical work, writing, shared in the revision of the article. MBA: Writing, shared in the revision of the article. MSF: Writing, shared in the revi- sion of the article. SE: Supervision on the practical work, writing, shared in the revision of the article All authors read and approved the final manuscript. Funding Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooperation with The Egyptian Knowledge Bank (EKB). Availability of data and materials The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The study was approved by the ethical committee of the Faculty of Medicine of both October 6 University and Ain Shams University (ethical approval number: MS 769/ 2021). Informed, written consent was obtained from all participants of the study. All methods were approved by the department’s undergraduate medical education committee and performed in accordance with the legal regulations of the faculty of medicine, October 6 university without extracurricular activities required. Consent for publication Not applicable. Elnaga et al. BMC Medical Education (2023) 23:433 Page 9 of 10 Competing interests The authors declare that they have no competing interests. Author details 1 Department of Pulmonary, Faculty of Medicine, October 6 University, 28C, Opera City Compound, Sheikh Zayed Giza, Egypt. 2 Department of Medical Bio- chemistry and Molecular Biology, Faculty of Medicine, Ain Shams University, Cairo, Egypt. 3 Medical Education Department, Faculty of Medicine, Ain Shams University, Cairo, Egypt. 4 Department of Medical Microbiology and Immunol- ogy, Faculty of Medicine, Ain Shams University, Cairo, Egypt. Received: 28 August 2022 Accepted: 1 June 2023 References 1. Plucker JA. The relationship between school climate conditions and 2. student aspirations. J Educ Res. 2010;91:240–6. Sarfraz Khan J. Determination of medical education environment in Pun- jab private and public medical colleges affiliated with University. J Ayub Med Coll. 2009;21(4):162–70. 3. Pimparyon P, Roff S, McAleer S, Poonchai B, Pemba S. Educational envi- ronment, student approaches to learning and academic achievement in a Thai nursing school. Med Teach. 2000;22:359–64. https:// doi. org/ 10. 1080/ 01421 59004 09456. 4. Genn JM. AMEE medical education guide no. 23 (part 1): Curriculum, 5. environment, climate, quality and change in medical education - A unify- ing perspective. Med Teach. 2001;23:337–44. https:// doi. org/ 10. 1080/ 01421 59012 00633 30. Lizzio A, Wilson K, Simons R. University students’ perceptions of the learn- ing environment and academic outcomes: Implications for theory and practice. Stud High Educ. 2002;27:27–52. https:// doi. org/ 10. 1080/ 03075 07012 00993 59. 6. Audin K, Davy J, Barkham M. University quality of life and learning (UNIQoLL): An approach to student well-being, satisfaction and insti- tutional change. J Furth High Educ. 2003;27:365–82. https:// doi. org/ 10. 1080/ 03098 77032 00012 8073. 7. Roff S. Students’ perceptions of educational environment: a comparison of academic achievers and under-achievers at kasturba medical college. India. 2004. https:// doi. org/ 10. 1080/ 13576 28040 00024 45. 8. Abraham R, Ramnarayan K, Vinod P, Torke S. Students’ perceptions of learning environment in an Indian medical school. BMC Med Educ. 2008;8:1–5. https:// doi. org/ 10. 1186/ 1472- 6920-8- 20/ TABLES/4. 9. Mansur DI, Kayastha SR, Makaju R, Dongol M. Problem based learning in medical education. Kathmandu Univ Med J. 2012;10:78–82. 10. World Health Organization. The world health report : 2006 : working together for health. 2006. https:// apps. who. int/ iris/ handle/ 10665/ 43432. 11. Ellaway R, Candler C, Greene P, Smothers V. An architectural model for MedBiquitous virtual patients. MedBiquitous. 2006;6:1–15. 12. Kononowicz AA, Zary N, Edelbring S, Corral J, Hege I. Virtual patients - what are we talking about? a framework to classify the meanings of the term in healthcare education. BMC Med Educ. 2015;15:1–7. https:// doi. org/ 10. 1186/ S12909- 015- 0296-3/ FIGUR ES/3. 13. Cook DA, Triola MM. Virtual patients: a critical literature review and pro- posed next steps. Med Educ. 2009;43:303–11. https:// doi. org/ 10. 1111/J. 1365- 2923. 2008. 03286.X. 14. Posel N, Mcgee JB, Fleiszer DM. Twelve tips to support the develop- ment of clinical reasoning skills using virtual patient cases. Med Teach. 2015;37:813–8. https:// doi. org/ 10. 3109/ 01421 59X. 2014. 993951. 15. Berman NB, Durning SJ, Fischer MR, Huwendiek S, Triola MM. The role for virtual patients in the future of medical education. Acad Med. 2016;91:1217–22. 16. Hamdy H, Al-Moslih A, Tavarnesi G, Laus A. Virtual patients in problem- based learning. Med Educ. 2017;51:557–8. https:// doi. org/ 10. 1111/ MEDU. 13293. 17. Yu JH, Chang HJ, Kim SS, Park JE, Chung WY, Lee SK, et al. Effects of high-fidelity simulation education on medical students’ anxiety and confidence. PLoS One. 2021;16:e0251078. 18. Herrera-Aliaga E, Estrada LD. Trends and innovations of simula- tion for twenty first century medical education. Front Public Heal. 2022;10:619769. 19. Dekhtyar M, Park YS, Kalinyak J, Chudgar SM, Fedoriw KB, Johnson KJ, et al. Use of a structured approach and virtual simulation practice to improve diagnostic reasoning. Diagnosis. 2022;9:69–76. https:// doi. org/ 10. 1515/ DX- 2020- 0160/ MACHI NEREA DABLE CITAT ION/ RIS. 20. Norman G. Research in clinical reasoning: past history and current trends. Med Educ. 2005;39:418–27. https:// doi. org/ 10. 1111/J. 1365- 2929. 2005. 02127.X. 21. Berman N, Fall LH, Smith S, Levine DA, Maloney CG, Potts M, et al. Inte- gration strategies for using virtual patients in clinical clerkships. Acad Med. 2009;84:942–9. https:// doi. org/ 10. 1097/ ACM. 0B013 E3181 A8C668. 22. Piot MA, Dechartres A, Attoe C, Jollant F, Lemogne C, Layat Burn C, Rethans JJ, Michelet D, Cross S, Billon G, Guerrier G, Tesniere A, Falissard B. Simulation in psychiatry for medical doctors: a systematic review and meta-analysis. Med Educ. 2020;54:696–708. 23. Bhardwaj P. Types of sampling in research. J Pract Cardiovasc Sci. 2019;5:157. https:// doi. org/ 10. 4103/ JPCS. JPCS_ 62_ 19. 24. Trullàs JC, Blay C, Sarri E, Pujol R. Effectiveness of problem-based learning methodology in undergraduate medical education: a scoping review. BMC Med Educ. 2022;22:1–12. 25. Eissa S, Sallam RM, Moustafa AS, Hammouda AMA. Large-scale applica- tion of case-based learning for teaching medical biochemistry: a chal- lenging experience with positive impacts. Innov Educ. 2020;2:1–19. 26. Dwivedi R, Pandey NN. Analysis of evaluation data collected on likert type items: humanities-courses. J Data Anal Inf Process. 2021;09:102–21. 27. Al-Dahir S, Bryant K, Kennedy KB, Robinson DS. Online virtual-patient cases versus traditional problem-based learning in advanced phar- macy practice experiences. Am J Pharm Educ. 2014;78(4):76. 28. Ghareeb Mohamed A, Khamis Mohamed L. Tanta scientific nursing journal perceived nursing students’ satisfaction and self-confidence towards the elements of clinical simulation design and educational practice during the outbreak of COVID-19 pandemic. Tanta Sci Nurs J. 2020;19(2):68–98. 29. Kinkade S. A snapshot of the status of problem-based learning in U. S. medical schools, 2003-04. Acad Med. 2005;80:300–1. https:// doi. org/ 10. 1097/ 00001 888- 20050 3000- 00021. 30. Neville AJ, Norman GR. PBL in the undergraduate MD Program at McMaster University: three iterations in three decades. Acad Med. 2007;82:370–4. https:// doi. org/ 10. 1097/ ACM. 0B013 E3180 33385D . 31. Neville AJ, Learning P-B. Medical education forty years on. a review of its effects on knowledge and clinical performance. Med Princ Pr. 2009;18:1–9. 32. Norman G. Medical education: past, present and future. Perspect Med Educ. 2012;1:6–14. https:// doi. org/ 10. 1007/ S40037- 012- 0002-7/ FIGUR ES/2. 33. Noverati N, Naro GR, Fischer RJ, Thompson BM. Using video and virtual patients in problem-based learning: a scoping review. Med Sci Educ. 2020;30(4):1685–91. 34. Plackett R, Kassianos AP, Mylan S, Kambouri M, Raine R, Sheringham J. The effectiveness of using virtual patient educational tools to improve medical students’ clinical reasoning skills: a systematic review. BMC Med Educ. 2022;22. https:// doi. org/ 10. 1186/ s12909- 022- 03410-x. 35. Chi DL, Pickrell JE, Riedy CA. Student learning outcomes associated with video vs. paper cases in a public health dentistry course. J Dent Educ. 2014;78:24–30. 36. Hmelo-Silver CE, Duncan RG, Chinn CA. Scaffolding and achieve- ment in problem-based and inquiry learning: a response to Kirschner, Sweller, and Clark (2006). Educ Psychol. 2007;42:99–107. 37. Car LT, Myint Kyaw B, Dunleavy G, Smart NA, Semwal M, Rotgans JI, et al. Digital problem-based learning in health professions: systematic review and meta-analysis by the digital health education collaboration. J Med Internet Res. 2019;21:1–12. 38. Akbaba-Altun S. Complexity of integrating computer technologies into education in Turkey. J Educ Technol Soc. 2006;9:176–87. 39. Perraton H. Choosing technologies for education. J Educ Media. 40. 2000;25:31–8. Issenberg SB, McGaghie WC, Petrusa ER, Gordon DL, Scalese RJ. Features and uses of high-fidelity medical simulations that lead to effective Elnaga et al. BMC Medical Education (2023) 23:433 Page 10 of 10 41. learning: A BEME systematic review. Medical Teacher. 2005;27:10–28. https:// doi. org/ 10. 1080/ 01421 59050 00469 24. Issenberg SB, McGaghie WC, Hart IR, Mayer JW, Felner JM, Petrusa ER, et al. Simulation technology for health care professional skills training and assessment. JAMA. 1999;282:861–6. https:// doi. org/ 10. 1001/ JAMA. 282.9. 861. 42. Jin J, Bridges SM. Educational technologies in problem-based learning in health sciences education: a systematic review. J Med Internet Res. 2014;16: e251. 43. Consorti F, Mancuso R, Nocioni M, Piccolo A. Efficacy of virtual patients in medical education: a meta-analysis of randomized studies. Comput Educ. 2012;59:1001–8. 44. Triola M, Feldman H, Kalet AL, Zabar S, Kachur EK, Gillespie C, et al. A randomized trial of teaching clinical skills using virtual and live standard- ized patients. J Gen Intern Med. 2006;21:424–9. https:// doi. org/ 10. 1111/J. 1525- 1497. 2006. 00421.X. 45. Cook DA, Erwin PJ, Triola MM. Computerized virtual patients in health professions education: a systematic review and meta-analysis. Acad Med. 2010;85:1589–602. https:// doi. org/ 10. 1097/ ACM. 0B013 E3181 EDFE13. 46. McLean M, Murrell K. WebCT: Integrating computer-mediated communi- cation and resource delivery into a new problem-based curriculum. J Vis Commun Med. 2002;25:8–15. https:// doi. org/ 10. 1080/ 01405 11022 01183 65. 47. Huwendiek S, Duncker C, Reichert F, De Leng BA, Dolmans D, Van Der Vleuten CPM, et al. Learner preferences regarding integrating, sequenc- ing and aligning virtual patients with other activities in the undergradu- ate medical curriculum: a focus group study. Med Teach. 2013;35:920–9. 48. Ruiz JG, Mintzer MJ, Leipzig RM. The impact of e-learning in medical education. Academic Medicine. 2006;81:207–12. https:// doi. org/ 10. 1097/ 00001 888- 20060 3000- 00002. 49. Bridges S, Botelho M, Green JL, Chau ACM. Multimodality in problem- based learning (PBL): An interactional ethnography. In: Problem-Based Learning in Clinical Education: The Next Generation. Springer Nether- lands; 2012. p. 99–120. https:// doi. org/ 10. 1007/ 978- 94- 007- 2515-7_7. 50. Dornan T, Hadfield J, Brown M, Boshuizen H, Scherpbier A. How can medical students learn in a self-directed way in the clinical environ- ment? design-based research. Med Educ. 2005;39:356–64. https:// doi. org/ 10. 1111/J. 1365- 2929. 2005. 02112.X. 51. Gesundheit N, Brutlag P, Youngblood P, Gunning WT, Zary N, Fors U. The use of virtual patients to assess the clinical skills and reasoning of medical students: initial insights on student acceptance. Med Teach. 2009;31:739–42. 52. Kean S. The role of virtual patients in medical education: a review of the literature. Wounds UK. 2017;13:34–41. 53. Olaussen C, Heggdal K, Tvedt CR. Elements in scenario-based simula- tion associated with nursing students’ self-confidence and satisfaction: a cross-sectional study. Nurs Open. 2020;7:170–9. https:// doi. org/ 10. 1002/ NOP2. 375. 54. Krishnamurthy K, Selvaraj N, Gupta P, Cyriac B, Dhurairaj P, Abdul- lah A, et al. Benefits of gamification in medical education. Clin Anat. 2022;35:795–807. https:// doi. org/ 10. 1002/ CA. 23916. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12891-023-06492-w	Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 https://doi.org/10.1186/s12891-023-06492-w BMC Musculoskeletal Disorders Arabic version of the intermittent and constant osteoarthritis pain questionnaire (ICOAP-Ar): translation, cross-cultural adaptation, and measurement properties Ahmed Farrag1, Walaa Elsayed2, Doaa Al Saleh3, Ahmed Hefny4 and Afaf Shaheen1,5 Abstract Background Pain is the most incapacitating symptom of knee osteoarthritis (OA), with intermittent and/or continuous nature as described by the patients. Accuracy of pain assessment tools across different cultures is important. This study aimed to translate and culturally adapt the Intermittent and Constant OsteoArthritis Pain (ICOAP) measure into Arabic (ICOAP-Ar) and evaluate its psychometric properties in patients with knee OA. Methods The ICOAP was cross-culturally adapted following the recommended guidelines from English. Knee OA patients from outpatient clinics were recruited to assess the structural (confirmatory factor analysis) and construct validity (Spearman’s correlation coefficient - rho) to assess the relationship between the ICOAP-Ar and the pain and symptoms subscales of the Knee Injury and Osteoarthritis Outcome Score (KOOS), in addition to internal consistency (Cronbach’s alpha and the corrected item-total correlation). A week later, test-retest reliability (intraclass correlation coefficient (ICC)) was evaluated. Following four weeks of physical therapy treatment, the ICOAP-Ar responsiveness was evaluated using the receiver operating characteristic curve. Results Ninety-seven participants were recruited (age = 52.97 ± 9.9). A model with single pain construct showed acceptable fit (Comparative fit index = 0.92). The ICOAP-Ar total and subscales had a strong to moderate negative correlation with the KOOS pain and symptoms domains, respectively. The ICOAP-Ar total and subscales demonstrated satisfactory internal consistency (α = 0.86–0.93). The ICCs were excellent (ICCs = 0.89–0.92) with acceptable corrected item total correlations (rho = 0.53–0.87) for the ICOAP-Ar items. The ICOAP-Ar responsiveness was good with moderate effect size (ES = 0.51–0.65) and large standardized response mean (SRM = 0.86–0.99). A cut-off point of 51.1/100 was determined with moderate accuracy (Area under the curve = 0.81, sensitivity = 85%, specificity = 71%). No floor or ceiling effects were found. Conclusions The ICOAP-Ar exhibited good validity, reliability, and responsiveness after physical therapy treatment for knee OA, which renders it reliable for evaluating knee OA pain in clinical and research settings. Keywords Pain, Osteoarthritis, Knee, ICOAP Correspondence: Ahmed Farrag ahmedfarrag@cu.edu.eg Full list of author information is available at the end of the article © The Author(s) 2023, corrected publication 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessPage 2 of 11 Introduction Joint osteoarthritis (OA) is an active pathological joint condition associated with considerable multifactorial health burden [1, 2]. It is attributed to an unbalanced repair-damage process of variable joint structures [3]. It is commonly associated with structural changes and pro- inflammatory reactions [4, 5]. Amongst a multiplicity of symptoms, pain is the most incapacitating clinical symp- tom joint OA patients always report to clinicians [6]. Sev- eral sets of criteria have been developed to help clinically diagnose joint OA, and joint pain was the main diagnos- tic symptom in common [7–9]. This indicates the critical importance of joint pain assessment for diagnostic and prognostic purposes of joint OA. Lower extremity joints are the most affected by OA. This made the lower extremity joints the most common joints requiring replacement surgeries [10, 11]. Knee OA comprises 85% of the OA global burden [12]. A year prev- alence of knee OA was recently reported to reach 22.9%, corresponding to 654.1 million individuals aged 40 and above [13]. Typically, OA joint pain is intermittent and may occasionally flare-up with increased intensity and frequency and decreased threshold [14, 15]. Research has reported a significant relationship between the severity of knee OA pain and poor sleep quality and reduced quality of life [16, 17]. The availability of a valid tool for measuring knee OA pain is important for patient assessment and follow-up. The importance is greatly emphasized with the conduc- tion of multicenter and multinational studies targeting patients with knee OA. This necessitates standardized measurement tools for producing comparable and mean- ingful data. This is achieved by cross-cultural adapta- tion of a measurement tool across different languages. Focus groups examined the quality and characteristics of pain experienced by patients with hip and knee OA and reported two distinct types of pain; the intermittent and constant OA pain [18]. Therefore, these groups devel- oped the Intermittent and Constant OsteoArthritis Pain (ICOAP) measurement tool to assess the hip and knee OA pain [19]. The ICOAP questionnaire is an 11-item tool that measures hip and knee “constant pain” and “pain that comes and goes” and their impact on quality of life in terms of mood and sleep disturbance. The ICOAP scale showed adequate psychometric properties regard- ing inter-item correlation, content and construct validity, internal consistency and test-retest reliability [19]. The ICOAP questionnaire has been cross-culturally adapted to different languages other than the original English version [20–25]. Recently, it was cross-cultur- ally adapted into the Arabic language [26]. However, the published Arabic version suffered significant limitations [27]. The issues noticed are mainly concerned with the correctness of the translation used in the Arabic version of the ICOAP questionnaire. The authors did not appro- priately translate the ICOAP questionnaire. They used Arabic terms and item structure that do not correctly infer comparable meaning. These inconsistencies would prevent reaching equivalence between the English and Arabic versions of the ICOAP questionnaire. Accord- ingly, the content validity of the Arabic version would be significantly deficient, which renders it inequivalent to the original ICOAP and, consequently, invalid and inap- propriate for assessing OA pain in the Arab population in its current state. Additionally, responsiveness was not reported for the Arabic ICOAP, which is an important psychometric property to assess the tool’s sensitivity to changes occurring in the measured outcomes [28]. Therefore, the purposes for the current study were to appropriately translate and cross-culturally adapt the ICOAP measurement tool into Arabic and assess the psychometric properties of the culturally adapted Ara- bic version (ICOAP-Ar). The examined properties were validity (content, structural and construct validity), reli- ability (test-retest reliability and internal consistency), and responsiveness. Materials and methods Participants Patients with Knee OA were recruited from those referred to the outpatient physical therapy clinic of King Fahd Military Medical Complex, Saudi Arabia, and Ther- apy & Rehab. Center, Egypt. The study was approved by the Institutional Review Board of the King Fahd Military Medical Complex, Saudi Arabia (AFHER-IRB-2020-024). All participants received verbal and written information about the study and signed the consent form. Eligible participants had to meet the knee OA diagnos- tic criteria according to the American College of Rheu- matology (ACR) [7]. These include knee pain and at least three of the following additional symptoms: morning stiffness ≤ 30 min, crepitation, bone margin tenderness, bony enlargement and no palpable warmth. Patients were excluded if they: had rheumatoid arthritis, serious path- ological conditions (inflammatory arthritis and malig- nancy), total or partial arthroplasty of the affected joint, or could not read and understand documents written in Arabic. Procedures This psychometric testing study had two phases. Phase I for the translation and cross-cultural adaptation of the ICOAP measurement tool to create its equivalent Arabic version (ICOAP-Ar). Phase II was conducted to evaluate the psychometric properties of the ICOAP-Ar. Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 Page 3 of 11 Phase I: translation and cross-cultural adaptation After permission from the original developer (Dr. Gillian Hawker) was granted, and in accordance with the recom- mended guidelines [29], this process included the follow- ing procedures in order. Stage I, forward translation of the original ICOAP into Arabic was carried out indepen- dently by two translators; a professional translator and a musculoskeletal physical therapy consultant. Both are native Arabic speakers and bilingual in English. Stage II, based on the received translations, the principal investi- gator constructed a preliminary version of the ICOAP-Ar after consensus meeting with the two translators. Stage III, the preliminary ICOAP-Ar was translated back into English by two bilingual translators who are native Eng- lish speakers and unfamiliar with the original ICOAP. Stage IV, a consensus panel including the translators, a language professional, a methodologist, the study inves- tigators, a musculoskeletal clinician, and the original ICOAP scale developer (Dr. Gillian Hawker) reviewed all the translated versions and developed a prefinal version of the ICOAP-Ar. The consensus panel made decisions to achieve equivalence between the original ICOAP and the prefinal ICOAP-Ar. At this stage, the item and scale con- tent validity was objectively assessed by calculating the Content Validity Index (CVI). Stage V, the prefinal version of the ICOAP-Ar was pilot-tested on a sample of 30 patients with knee OA for clarity and understanding to examine its face validity. The patients were asked to complete the prefinal ICOAP-Ar. Then, the participants were interviewed to record their feedback, using written feedback reports, about the ques- tionnaire in terms of meaning of items, clarity of instruc- tions and ability to self-complete it, and relevance to their condition. Participants’ interview reports were reviewed by the consensus panel and the ICOAP-Ar was modified as necessary. Phase II: testing the Psychometric Properties Eligible patients with knee OA were invited to join the study and were informed about the purpose of the study. After signing a consent form, participants were instructed to complete the ICOAP-Ar. The same patients were asked to retake the ICOAP-Ar 72–96 hours later for assessing the test-retest reliability of the ICOAP-Ar. Finally, after receiving four weeks of physical therapy treatment, responsiveness of the ICOAP-Ar was assessed. Structural validity of the ICOAP-Ar was examined using the exploratory factor analysis (EFA) to verify its dimensionality as previously recommended [19]. Then, confirmatory factor analysis (CFA) was conducted to confirm the factor structure of the ICOAP-Ar. Construct validity (convergent validity) of the ICOAP-Ar was inves- tigated using the Hypothesis-testing method to test the relationship between the ICOAP-Ar (total and constant and intermittent pain subscales) and the pain and symp- toms scores of the KOOS measurement tool. Accord- ingly, participants were instructed to complete the KOOS along with the ICOAP-Ar. Different measures were calculated to assess the reli- ability of the ICOAP-Ar. These included the test-retest reliability, internal consistency (Cronbach’s alpha and the corrected item-total correlation), standard error of mea- surement (SEM), and smallest detectable change at 95% confidence interval (SDC95). To evaluate responsiveness of the ICOAP-Ar, par- ticipants were reevaluated after receiving standard physical therapy treatment for a period of 4 weeks. Treat- ment protocols were not controlled or standardized for patients. However, general standard physical therapy treatment protocols were administered to the partici- pants that included stretching and strengthening exer- cises for the lower extremity musculature, which were provided by licensed physical therapists in outpatient clinical settings. Participants were instructed to complete the ICOAP-Ar and Global Rating of Change (GRoC) scales after completing a physical therapy course. The GRoC was used as an external reference for calculation of measurement error and responsiveness. All study proce- dures and measures were administered and supervised by trained physical therapists. Instrumentation The ICOAP is an 11-item questionnaire. Each item is rated from 0 to 4 on a 5-point Likert scale. It has two subscales: constant pain (5 items) and intermittent pain (6 items). A score is calculated for the constant (0–20) and intermittent (0–24) subscales separately and for total pain (0–44), which are further normalized to a score of 0 (no pain) -100 (extreme pain) [19]. The KOOS is a self-reported questionnaire used to subjectively assess five knee OA-relevant domains: Pain, Symptoms, Activities of Daily Living (Function), Sport and Recreation Function (Sport/Rec) and knee-related Quality of Life (QOL). It comprises 42 questions across the 5 subscales. A 5-point Likert scale is used to answer questions, and standardized answers are assigned a score from 0 (no symptoms) to 4 (extreme symptoms). Finally, a normalized score (0–100, worst to best) is calculated for each subscale [30]. The GRoC is a 15-point Likert scale used to assess the patient perceived deterioration or improvement follow- ing an intervention, ranging from − 7 (a very great deal worse) to + 7 (a very great deal better), with 0 indicat- ing no change of condition. The GRoC includes a single question about change in health status after 4 weeks of physical therapy for knee OA [31]. Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 Page 4 of 11 Data analysis Descriptive statistics were calculated, and data presented using frequencies and percentages for categorical vari- ables and mean ± standard deviation for continuous vari- ables. The normalized scores (0-100) of the ICOAP-Ar total and subscales were used for data analysis. Statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS, Chicago, IL) version 21. The level of significance was set at P ≤ 0.05 for all the analyses. Sample size The sample size was calculated using the PASS software (Version 20.0.2) considering the Spearman’s correlation coefficient (rho) for testing the construct validity of the ICOAP-Ar. Based on acceptable rho value of at least 0.3 (R1 in PASS), a power of 90% and p value of 0.05, a sam- ple size of 92 participants was required. Validity The CVI, for the entire scale (S-CVI) and for each item independently (I-CVI), was used to evaluate the content validity of the ICOAP-Ar with an acceptable value of at least 0.8. Additionally, the S-CVI was calculated using the average I-CVI scores (S-CVI/Ave), and the propor- tion of scale items that were reported relevant by all the experts (S-CVI/UA). The EFA was implemented using the principal compo- nent analysis with varimax rotation and eigenvalue = 1, and based on acceptable values of greater than 0.5 for the Kaiser-Meyer-Olkin (KMO) [32] and less than 0.05 for the Bartlett’s test of sphericity [33]. The goodness of model fit was examined using the CFA based on a rec- ommended CMIN/df (degrees of freedom) value below 3, standardized Root Mean Squared Residual (SRMR) value below 0.07, and comparative fit index (CFI) value above 0.90 [34]. Factor loading above 0.3 was considered acceptable [35]. Construct validity was evaluated by calculating the rho between the ICOAP-Ar score and the relevant pain and symptoms score of the KOOS. The coefficient is classified as follows: rho = 0.3–0.7 moderate correlation, and > 0.7 strong correlation [36]. We hypothesized a priori that the ICOAP-Ar total and subscales (constant and intermit- tent) pain score would correlate moderately negatively with the pain and symptoms subscales of the Knee Injury and OA Outcome Score (KOOS). If five of the six prede- termined hypotheses are accepted, the construct validity of ICAOP-Ar would be considered adequate [37]. Reliability A Cronbach’s alpha value of ≥ 0.7, and corrected item- total correlation of ≥ 0.3 were considered acceptable [37, 38]. Test-retest reliability was assessed using the intra- class correlation coefficient (two-way mixed effects, absolute agreement, single rater/measurement). An ICC value of > 0.8, and 0.6 to 0.8 was considered as excel- lent and good correlation, respectively [39]. The SEM and SDC95 were calculated using the formula: SEM =(SD×[√(1-ICC)]), where SD was the sample’s standard deviation and SDC95 = SEM ×1.96 ×√2, respectively [40]. Responsiveness To assess responsiveness of the ICOAP-Ar, several approaches were implemented. We compared the pre- (baseline) and post-treatment (after four weeks) ICOAP- Ar total and subscale pain scores using the Wilcoxon Signed-Ranks test. The standardized effect size (SES) and standardized response mean (SRM) were calcu- lated to evaluate the effect size, which was interpreted as large (≥ 0.80), moderate (≥ 0.50) or small (≥ 0.20). A hypothesis-testing approach was utilized to assess the correlation between the score changes (ICOAP-Archange= ICOAP-ARfinal – ICOAP-ARbaseline) of the ICOAP- Ar total and subscale pain scores and the GRoC. The Spearman’s correlation coefficient was calculated with the same classification as described above [41]. It was hypothesized that the score change of the ICOAP-Ar total and subscales would correlate moderately negatively with the GRoC scale. Finally, anchor-based responsiveness of the ICOAP- Ar score was assessed adopting the GRoC score as the external anchor. Participants were categorized accord- ing to their reported GRoC scores to either improved (GRoC ≥ 3) or stable group (GRoC < 3 to >-3). Receiver operating characteristics (ROC) curve of the ICOAP-Ar final score was blotted to calculate the area under the curve (AUC). An AUC value of ≥ 0.7 was considered ade- quate to indicate agreement with the GRoC [37]. The cut- off score of the ICOAP-Ar scale was determined as the point on the ROC curve yielding the minimal value for (1 − sensitivity)2 + (1 − specificity)2 [42]. Floor and ceiling effects The presence of floor and ceiling effects was evaluated as it compromises the responsiveness of a measurement tool. The floor and ceiling effects were considered pres- ent if more than 15% of the sample scored the lowest or highest possible score on baseline ICOAP-Ar subscales and total pain [37]. Results Subjects The demographic data of the participants are summa- rized in Table 1. Initially, 135 knee OA patients were approached, and 36 subjects were excluded because they did not meet the inclusion requirements. The remain- ing 99 patients participated in the validity assessment and only two patients dropped out during the reliability Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 Table 1 Participants’ characteristics (n = 99) Mean (SD) Variables 52.97 (9.9) Age (years) 160.58 (7.5) Stature (cm) 83.54 (14.7) Body mass (kg) BMI (kg/cm2) 32.26 (5.7) 19.85 (25.3) Knee OA Duration (Months) Sex Level of education Affected knee Kellgren-Lawrence (K-L) grade GRoC (n = 75) ICOAP-Ar_Total visit 1 (n = 99) ICOAP-Ar_Constant visit 1 (n = 99) Men Women Secondary Bachelor’s degree Master’s degree PhD Right knee Left knee Both knees Grade 0 Grade 1 Grade 2 Grade 3 Grade 4 Improved Stable Mean (SD) 55.76 (20.3) 54.75 (25.4) 95% CI 51.0–54.9 159.1–162.1 80.6–86.5 31.1–33.4 14.8–24.9 Frequency % 27 72 55 40 27.3 72.7 55.6 40.4 4 0 15.2 13.1 71.7 0 26.3 43.4 25.3 5.1 54.7 45.3 4 0 15 13 71 0 26 43 25 5 41 34 95% CI 51.7–59.8 49.7–59.8 ICOAP-Ar_Intermittant visit 56.61 (18.7) 52.9–60.3 1 (n = 99) ICOAP-Ar_Total visit 2 (n = 97) ICOAP-Ar_Constant visit 2 (n = 97) 50.66 (20.1) 49.51 (24.4) 46.6–54.7 44.6–54.4 ICOAP-Ar_Intermittant visit 51.63 (19.0) 47.8–55.5 2 (n = 97) ICOAP-Ar_Total visit 3 (n = 75) ICOAP-Ar_Constant visit 3 (n = 75) 44.79 (21.1) 43.33 (24.5) 40.2–49.9 38.0–49.3 ICOAP-Ar_Intermittant visit 46.00 (20.8) 41.4–51.0 3 (n = 75) KOOS Symptoms (n = 99) KOOS Pain (n = 99) SD: standard deviation, CI: Confidence Interval, n: number of participants, BMI: body mass index, ICOAP: Intermittent and Constant Osteoarthritis Pain Questionnaire KOOS: Knee Injury and Osteoarthritis Outcome Score, GRoC: Global Rating of Change 58.4 (20.1) 50.5 (19.2) 54.4–62.5 46.7–54.3 assessment (n = 97). Seventy-five patients completed the third visit (75.8%), and their data were used for respon- siveness assessment. Translation and cross-cultural adaptation Few differences were identified between the original and backward-translated versions of the ICOAP. Upon meet- ing with the author of the ICOAP scale, the panel of Page 5 of 11 experts had consensus on most of the scale items. The main concerns were related to the translation of the words “frustrated or annoyed” and “upset or worried” in item 4 and 5, respectively. These terms could have nearly similar meanings when translated into Arabic. This was also reported by the participant’s feedback during pilot testing of the prefinal version of the ICOPA-Ar. Accord- ingly, the panel agreed after elaboration from the author to translate “frustrated or annoyed” to “جاعزنإ وأ طابحإ” and “upset or worried” to “قلقلا وأ قيضلا”. Other than that, the participants did not report any difficulty regarding the clarity or understanding of the ICOAP-Ar during pre- testing with a measured CVI between 0.97 and 1 for all the items. Validity Content validity The ICOAP-Ar was very clear and easily understood by both the panel of experts and participants with an excel- lent I-CVI for all the items (I-CVI = 1). Similarly, the S-CVI, S-CVI/Ave and S-CVI/UA all had a perfect value of 1 (Table 2). Structural validity The EFA showed suitability for factor analysis with KMO = 0.88 and acceptable Bartlett’s index (p < 0.001). The analysis yielded two factors that accounted for 67.8% of the variance. However, all the items loaded across the two factors. Items 1,2,3,6,7, and 8 loaded more on the first factor (loading range = 0.6–0.85), while items 4,5,10, and 11 loaded more on the second factor (loading range = 0.67–0.86). Item 9 loaded equally across factors (0.54–0.58). Therefore, we proceeded with the CFA con- sidering a one factor model. The CFA revealed an accept- able factor loading for all the items (Fig. 1) and good fit indexes (CMIN/df = 2.6, SRMR = 0.06, and CFI = 0.92). Construct validity The results revealed that all the predefined hypotheses were confirmed. The ICOAP-Ar total and its constant and intermittent pain domains had a strong (rho between − 0.71 and − 0.76, p < 0.001) and moderate (rho between − 0.57 and − 0.68, p < 0.001) negative correlation with KOOS pain and symptoms domains, respectively. The ICOAP-Ar constant pain domain had the lowest correla- tion with the KOOS pain and symptoms domains, while the ICOAP-Ar total had the highest correlation with the KOOS pain domain (Table 3). Reliability The ICOP-Ar total (α = 0.93) and its domains showed excellent internal consistency. They also had excellent test-retest reliability (ICC = 0.98 − 0.92). The SEM for the ICOAP-Ar total and its constant and intermittent pain Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 Page 6 of 11 I-CVI UA Pc K* Table 2 Evaluation of I-CVIs with expert’s agreement, scales’ items Items Experts’ rating of relevance E1 E3 E4 E2 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. Proportion relevance I-CVI = item-level content validity index; UA = Universal Agreement; Pc = probability of a chance computed by formula: Pc = [N/A(N - A)]0.5 N where N = number of experts and A = Number agreeing on good relevance; K = modified kappa coefficient k* = (I-CVI - Pc)/(1 - Pc), a As stated by (Cicchetti & Sparrow, 1981);(Fleiss, 1981) about the criteria for K, 0.40 to 0.59 = fair, 0.60 to 0.74 = good, and > 0.74 excellent 1 1 1 1 1 1 1 1 1 1 1 S-CVI/UA = 1 0.063 0.063 0.063 0.063 0.063 0.063 0.063 0.063 0.063 0.063 0.063 4 4 4 4 4 4 4 4 4 4 4 1 4 4 4 4 4 4 4 4 4 4 4 1 4 4 3 3 4 4 3 4 4 4 4 1 4 4 3 4 4 4 3 4 4 4 4 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Experts in agreement 4 4 4 4 4 4 4 4 4 4 4 S-CVI/Ave = 1 domains was 6.1, 7.1, and 6.3, respectively. The SDC95 ranged between 16.8 for the ICOAP-Ar total and 19.7 for its constant pain domain (Table 4). The corrected item-total correlations were acceptable for the ICOAP-Ar items. The corrected item-total cor- relation coefficient ranged between 0.53 and 0.82 for the total scale. For the scale domains, it ranged between 0.55 and 0.77, and 0.75 and 0.87 for the intermittent and con- stant pain domains, respectively (Table 5). Responsiveness reported (54.7%) patients Forty-one improvement according to their GRoC score. There was a significant difference between the pre- and post-treatment ICOAP- Ar total and subscale scores (P < 0.001). The mean score change for the entire sample ranged between 11.9 and − 12.9. The SES (0.51–0.65) and SRM (0.86–0.99) values for the ICOAP-Ar total and subscales were moderate and large, respectively. Additionally, there was a signifi- cant negative moderate correlation (rho between − 0.44 and − 0.48, p < 0.001) between the ICOAP-Ar total and subscale change score and the GRoC (Table 6), which confirms the predefined hypothesis. Furthermore, the ICOAP-Ar showed adequate accuracy with an AUC of 0.81 (95% confidence interval = 0.71–0.91). The optimal cutoff point was 51.1 to differentiate between patients who improved and those who were stable at a sensitivity and specificity of 0.85 and 0.71, respectively (Fig. 2). Floor/Ceiling effect No floor or ceiling effects were present (Table 4). Discussion The current study cross-culturally adapted the ICOAP scale into Arabic following the international recommen- dations [29, 43] and examined its psychometric proper- ties in patients with knee OA. The results showed that the ICOAP-Ar was acceptable and easily comprehended by the participants. This was corroborated by the perfect CVI value (I-CVI and S-CVI = 1) obtained from both the expert panel and participants, which is higher than that reported for the Chinese version (CVI = 0.8–1) [20]. The ICOAP-Ar also demonstrated acceptable validity, reli- ability, and responsiveness, which is in accordance with the previous studies [20–25]. The structural validity of the ICOAP-Ar was examined using EFA and CFA. The EFA confirmed the dimension- ality of ICOAP-Ar showing the scale to have a single con- struct of pain. This is in accordance with the previously reported factorability of the original ICOAP. Hawker et al. (2008) reported that EFA of the ICOAP scale revealed a single pain construct suggesting sufficient homogene- ity between the 11 scale items [19]. This was further confirmed by the CFA findings that showed acceptable model fit parameters with good factor loading. Since CFA is unique to the current study, we cannot compare our findings to the original or any of the previously trans- lated versions of the ICOAP scale. Our predictions regarding the construct validity of the ICOAP-Ar were all confirmed. The ICOAP-Ar total and subscales scores had strong (rho= -0.71 to -0.76) to mod- erate (rho= -0.57 to -0.68) negative correlations with the KOOS pain and symptoms subscales, respectively. This finding is in agreement with previous studies validating the Portuguese (rho= -0.61 to -0.81) [23], Persian (rho= -0.5 to -0.7) [22], and Chinese (rho= -0.65 to -0.680) [44] versions of the ICOAP. The correlation coefficient values Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 Page 7 of 11 Fig. 1 Path diagram showing factor structure of the ICOAP-Ar are also comparable with those reported by other stud- ies that used the Western Ontario and McMaster Uni- versities Osteoarthritis Index (WOMAC) instead of the KOOS [20, 21, 44]. It is important to note that the ICOAP-Ar (total and subscales) correlation coefficients were higher with the KOOS pain domain than those calculated with the KOOS symptoms domain. This interesting finding was in line with the previous studies that validated the different versions of the ICOAP scale. It was consistently observed in studies that used the KOOS pain and symp- toms subscales to examine the construct validity of the ICOAP scale [22, 23], as well as other studies that used the WOMAC pain and function subscales or the health survey (SF36 or SF12, physical component) for the same purpose [20, 25, 45]. This further confirms the divergent validity of the ICOAP-Ar as appropriate for assessing the construct of pain. Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 Table 3 Correlations between ICOAP-Ar and KOOS for construct validity ICOAP-Ar domains Instrument for correlation Correlation coefficient (rho) ICOAP-Ar Total ICOAP_Ar Constant pain domain ICOAP_Ar Intermittent pain domain ICOAP-Ar Total ICOAP_Ar Constant pain domain ICOAP_Ar Intermittent pain domain P value is significant at ˂0.001 KOOS - Symptoms domain -0.64* -0.57* -0.68* KOOS - Pain domain -0.76* -0.71* -0.74* Hpoth- eses con- firmed? Yes Yes Yes Yes Yes Yes The internal consistency of the ICOAP-Ar total and subscales was excellent (Cronbach’s alpha = 0.86 to 0.93), which indicates the consistency and homogeneity of the scale items. This corresponds well with the Cron- bach’s alpha values reported for the original (Cronbach’s Page 8 of 11 alpha = 0.93) [19] and different language versions (Cron- bach’s alpha = 0.82 to 0.97) of the ICOAP [20–23, 25, 44]. Further confirmation of the ICOAP-Ar item consistency and lack of redundancy is the acceptable corrected item- total correlation values for its total and subscales scores (ranged from 0.53 to 0.87). These correlation coefficient values are also in agreement with those reported for other language versions for the ICOAP [20, 22, 23]. The ICOAP-Ar total (ICC = 0.91) and its constant (ICC = 0.92) and intermittent (ICC = 0.89) pain subscales showed excellent test-retest reliability reflecting the ques- tionnaire’s good reproducibility. The obtained values are slightly higher than those reported for the original ver- sion (ICC = 0.85) [19], but comparable to other language versions (ICC = 0.88–0.96) [20–23, 44]. Data analysis revealed that the ICOAP-Ar had SEM values of about 6–7 points for its total and subscale pain scores. This, accordingly, resulted in a MDC95 value of 16.8 points for the ICOAP-Ar total score. This means that an ICOAP- Ar score change of at least 16.8 points is required to be interpreted as a real within-subject change of knee OA pain [46]. Table 4 Internal consistency, test–retest reliability (ICC), Standard Error of Measurement (SEM), Minimal Detectable Change (MDC95), and floor and ceiling effects for the ICOAP-Ar (total and domains) Variable MDC95 SEM Internal Consistency(α) Test re-test reli- ability (95% CI) Floor effect % 0.93 ICOAP-Ar Total 0.92 ICOAP_Ar Constant pain domain 0.86 ICOAP_Ar Intermittent pain domain α: Cronbach alpha, ICC: Intra-class correlation; CI: confidence interval. n = 99. Data for the SEM and MDC95 are normalized to scores 0-100 0.91 (0.79–0.95) 0.92 (0.84–0.96) 0.89 (0.76–0.94) 16.8 19.7 17.5 6.1 7.1 6.3 0.0 7.1 0.0 Table 5 Corrected item-total correlation (n = 99) ICOAP items Corrected item-total coefficients* Corrected item-total coefficients† Cronbach’s α if item deleted* Cron- bach’s α if item deleted† Constant pain subscale 1. How intense has your constant knee pain been? 2. How much has your constant knee pain affected your sleep? 3. How much has your constant knee pain affected your overall quality of life? 4. How frustrated or annoyed have you been by your constant knee pain? 5. How upset or worried have you been by your constant knee pain? Intermittent pain subscale 6. How intense has your most severe knee pain that comes and goes been? 7. How frequent has this knee pain that comes and goes occurred? 8. How much has your knee pain that comes and goes affected your sleep? 9. How much has your knee pain that comes and goes affected your overall quality of life? 10. How frustrated or annoyed have you been by your knee pain that comes and goes? 11. How upset or worried have you been by your knee pain that comes and goes? * Obtained for ICOAP constant and intermittent pain subscales 0.747 0.748 0.866 0.837 0.786 0.610 0.552 0.686 0.768 0.654 0.653 † Obtained for ICOAP total pain 0.726 0.745 0.822 0.813 0.758 0.600 0.528 0.746 0.731 0.635 0.656 0.913 0.913 0.889 0.895 0.906 0.845 0.854 0.832 0.816 0.921 0.920 0.916 0.917 0.920 0.926 0.929 0.920 0.921 0.837 0.925 0.838 0.924 *Ceil- ing effect % 0.0 1.0 0.0 Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 Page 9 of 11 Table 6 Responsiveness of the ICOAP-Ar total and subscale pain (n = 75) Variable Pre-treatment Mean ± SD Post-treatment Mean ± SD Change Mean ± SD SES SRM Correlation with GRoC 44.8 ± 21.1 ICOAP-Ar Total 43.3 ± 24.6 ICOAP_Ar Constant paindomain 46.0 ± 20.8 ICOAP_Ar Intermittent pain domain Change = post-treatment score - pre-treatment score; improvement if change < 0 57.2 ± 19.7 56.2 ± 25.1 57.9 ± 18.2 -12.4 ± 12.5 -12.9 ± 14.2 -11.9 ± 13.6 0.63 0.51 0.65 0.99 0.91 0.86 -0.48 -0.44 -0.44 Bald values indicate significant difference comparing pre- and post-treatment scores (P < 0.001) SES = \|mean (change) ÷ SD (pre-treatment)\| SRM = \|mean (change) ÷ SD (change)\| Hpoth- eses con- firmed? Yes Yes Yes Our findings provided evidence supporting the accuracy and sensitivity to change of the ICAOP-Ar after physi- cal therapy treatment for knee OA. The ICOAP-Ar total and subscale pain scores were significantly reduced after the four-week treatment period revealing a moderate effect size (SES = 0.51–0.65), but large SRM (0.86–0.99). Responsiveness of the original ICAOP scale was previ- ously examined, and the findings showed variable low (SES = 0.46–0.54), moderate (SES = 0.5–0.88), and large (SES = 0.93–1.71) effect size for pharmacological treat- ment, intra-articular injection, and joint replacement sur- gery, respectively [44, 45, 47–50]. Responsiveness of the ICOAP to physical therapy treat- ment was assessed for the Portuguese version [41]. Com- pared to the ICOAP-Ar, the Portuguese ICOAP has higher SES (0.83–1.42) and SRM (1.33–1.50) values after physi- cal therapy intervention for knee OA. The higher effect size could be attributed to differences of the sample crite- ria and data recording approach between the two studies. The participants’ mean knee OA duration for the Portu- guese ICOAP responsiveness study was 10.1 years, while it was 1.7 years for our ICAOP-Ar study. The shorter dura- tion of knee OA could have led to less response to treat- ment in the current study compared with the Portuguese ICOAP one. This assumption is further corroborated by the nearly similar SES value for the constant pain subscale of the ICOAP-Ar (0.51) and the Portuguese ICOAP (0.51) for the subgroup of patients with less than 5 years dura- tion of knee OA. Additionally, the ICOAP-Ar data were self-reported, while the Portuguese ICOAP data were interviewer-administered. The different approaches of data collection process could likely have contributed to the lower SES and SRM values for the ICOAP-Ar due to the absence of willingness to please the therapist influence or presence of a misunderstanding bias [48]. The predefined hypothesis regarding the correlation between the score changes of the ICOAP-Ar total and subscale pain and the GRoC were established showing a moderate negative correlation (rho= -0.44 to -0.48), which is consistent with the findings of the Portuguese version (rho= -0.56 to -0.64) [41]. The accuracy of the ICAOP-Ar was established with adequate AUC value Fig. 2 Receiver operating characteristic curve for the ICOAP-Ar (n = 75) The SEM/MDC95 values (3.7/10.3 points) of the ICOAP total pain were reported for the traditional Chinese ver- sion (tChICOAP) only [20]. However, for comparative purposes, we could easily calculate the SEM/MDC95 val- ues for the other language versions from the published data. The SEM/MDC95 values were 6.7/18.5 and 6.9/19.1 points for the Persian and Portuguese ICOAP total pain, respectively [22, 23]. It is clear that the SEM/MDC95 val- ues for the ICOAP-Ar total pain are highly comparable to those for the other language versions, except the tChI- COAP. This could be attributed to the sample criteria and baseline tChICOAP total pain score. The baseline tChICOAP total pain score (34.52) is less than that for the ICOAP-Ar (55.76). Accordingly, one could logically speculate limited score variability for the tChICOAP compared with the ICOAP-Ar upon retest, which was represented by the higher test-retest reliability for the tChICOAP (ICC2,1 = 0.96) compared with the ICOAP- Ar (ICC2,1 = 0.91) total pain score. Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 (0.81) to discriminate between patients who improved with decreased knee pain after treatment and those who did not. The AUC value for the ICOAP was previously reported for the Chinese version [44] after total knee replacement (AUC = 0.92), and the Greek version [45] after intra-articular injection treatment (AUC = 0.76). A cut-off score of 51.1 was determined for the ICOAP-Ar, which is an important indicator for clinicians as it indicates that patients with an ICOAP-Ar total score of ≥ 51 points are expected to improve, and their knee pain becomes less after physical therapy treatment. Finally, and in agreement with other versions of the ICOAP, the floor and ceiling effects were not detected for the ICOAP-Ar total and sub- scale pain scores. These findings support the notion that the ICOAP-Ar can be used to longitudinally assess knee pain changes with acceptable accuracy. Some limitations of this study should be addressed. Unfortunately, we were not able to validate the ICOAP- Ar scale for the hip OA patients as we could not recruit enough participants to perform reliable data analysis. Our sample may not be representative since it had a short knee OA duration. However, the impact knee OA duration may have on the psychometric properties of the ICOAP-Ar is expected to be minimal. We did not examine the respon- siveness characteristics of the ICOAP-Ar in patients with knee OA receiving treatments other than the physical ther- apy, which we believe is important to provide comprehen- sive data regarding the validation of the ICOAP-Ar scale. following Conclusions The ICOAP scale was successfully cross-culturally adapted into Arabic international guidelines. The ICOAP-Ar total and subscale is a valid and reliable tool to assess knee OA pain. It has also demonstrated adequate accuracy and sensitivity to knee pain change in response to physical therapy intervention. An ICOAP-Ar total score of 51.1 points could be used as a cut-off to discriminate the response of patients with knee OA to physical therapy treatment. Abbreviations OA ICOAP Ar AUC GRoC tChICOAP ICC KOOS ACR CVI KOOS SPSS SEM SDC SES SRM ROC Osteoarthritis Intermittent and Constant Osteoarthritis Pain Arabic Area Under the Curve Global Rating of Change traditional Chinese ICOAP Intraclass Correlation Coefficient Knee Injury and OA Outcome Score American College of Rheumatology Content Validity Index Knee Injury and Osteoarthritis Outcome Score Statistical Package for the Social Sciences Standard Error of Measurement Smallest Detectable Change Standardized Effect Size Standardized Response Mean Receiver Operating Characteristics. Page 10 of 11 Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12891-023-06492-w. Supplementary Material 1 Acknowledgements The authors would like to thank all the participants for their time and effort. We appreciate the assistance of Mr. Ahmed AlGhamdi, physical therapist at King Fahd Military Medical Complex, with patient recruitment and data collection. We would like to thank Dr. Eidan Alzahrani, physical therapy consultant, for his effort in forward translation of the ICOAP questionnaire to Arabic. Author Contributions AF: conceived and designed the study. DA and AH: recruited patients and collected the data. AF and WE: performed statistical analysis and interpreted the findings. AF, WE and AS: drafted the manuscript. AF, WE, DA, AH and AS: approved the final version of the manuscript. Funding This study did not receive support from any funding agency. Data Availability The datasets for the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Ethical approval was obtained from the Institutional Review Board of the King Fahd Military Medical Complex, Saudi Arabia (AFHER-IRB-2020-024). All participants received verbal and written information about the study and signed a written informed consent form. This study adheres to the principles of the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1Basic Science Department, Faculty of Physical Therapy, Cairo University, Cairo, Egypt 2Department of Physical Therapy, College of Applied Medical Sciences, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia 3Department of Physical Therapy, King Fahd Military Medical Complex, Dhahran, Saudi Arabia 4Therapy & Rehab. Center, Hurghada, Egypt 5Department of Health Rehabilitation Sciences, College of Applied Medical Sciences, King Saud University, Riyadh, Saudi Arabia Received: 6 November 2022 / Accepted: 4 May 2023 References 1. Hunter DJ, Bierma-Zeinstra S, Osteoarthritis. Lancet. 2019;393(10182):1745–59. 3. 2. Hunter DJ, Schofield D, Callander E. The individual and socioeconomic impact of osteoarthritis. Nat Rev Rheumatol. 2014;10(7):437–41. Fu K, Robbins SR, McDougall JJ. Osteoarthritis: the genesis of pain. Rheuma- tology (Oxford). 2018;57(suppl4):iv43–iv50. Brandt KD, Radin EL, Dieppe PA, van de Putte L. Yet more evidence that osteo- arthritis is not a cartilage disease. Ann Rheum Dis. 2006;65(10):1261–4. 5. Martel-Pelletier J, Barr AJ, Cicuttini FM, Conaghan PG, Cooper C, Goldring MB, 4. et al. Osteoarthr Nat Rev Dis Primers. 2016;2:16072. 6. Hunter DJ, McDougall JJ, Keefe FJ. The symptoms of osteoarthritis and the genesis of pain. Med Clin North Am. 2009;93(1):83–100. Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481 Page 11 of 11 7. Altman R, Asch E, Bloch D, Bole G, Borenstein D, Brandt K, et al. Development of criteria for the classification and reporting of osteoarthritis. Classification of osteoarthritis of the knee. Diagnostic and therapeutic Criteria Committee of the american Rheumatism Association. Arthritis Rheum. 1986;29(8):1039–49. 8. National Clinical Guideline C. National Institute for Health and Clinical Excel- 28. Mokkink LB, Terwee CB, Patrick DL, Alonso J, Stratford PW, Knol DL, et al. The COSMIN checklist for assessing the methodological quality of studies on measurement properties of health status measurement instruments: an international Delphi study. Qual Life Res. 2010;19(4):539–49. 29. Beaton DE, Bombardier C, Guillemin F, Ferraz MB. Guidelines for the process 9. lence. Guidance. Osteoarthritis: care and management in adults. London: National Institute for Health and Care Excellence (UK); 2014. Zhang W, Doherty M, Peat G, Bierma-Zeinstra MA, Arden NK, Bresnihan B, et al. EULAR evidence-based recommendations for the diagnosis of knee osteoarthritis. Ann Rheum Dis. 2010;69(3):483–9. 10. Prieto-Alhambra D, Judge A, Javaid MK, Cooper C, Diez-Perez A, Arden NK. Incidence and risk factors for clinically diagnosed knee, hip and hand osteo- arthritis: influences of age, gender and osteoarthritis affecting other joints. Ann Rheum Dis. 2014;73(9):1659–64. 11. Turkiewicz A, Petersson IF, Bjork J, Hawker G, Dahlberg LE, Lohmander LS, et al. Current and future impact of osteoarthritis on health care: a popula- tion-based study with projections to year 2032. Osteoarthritis Cartilage. 2014;22(11):1826–32. 12. Vos T, Allen C, Arora M, Barber RM, Bhutta ZA, Brown A, et al. Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. The Lancet. 2016;388(10053):1545 − 602. 13. Cui A, Li H, Wang D, Zhong J, Chen Y, Lu H. Global, regional prevalence, incidence and risk factors of knee osteoarthritis in population-based studies. EClinicalMedicine. 2020;29–30:100587. of cross-cultural adaptation of self-report measures. Spine (Phila Pa 1976). 2000;25(24):3186–91. 30. Alfadhel SA, Vennu V, Alnahdi AH, Omar MT, Alasmari SH, AlJafri Z, et al. Cross-cultural adaptation and validation of the saudi arabic version of the knee Injury and Osteoarthritis Outcome score (KOOS). Rheumatol Int. 2018;38(8):1547–55. Jaeschke R, Singer J, Guyatt GH. Measurement of health status. Ascer- taining the minimal clinically important difference. Control Clin Trials. 1989;10(4):407–15. 31. 32. Zygmont C, Smith M. Robust factor analysis in the presence of normality vio- lations, missing data, and outliers: empirical questions and possible solutions. Quant Methods Psychol. 2014;10:40–55. 33. Byrne BM. Structural equation modeling with AMOS: Basic concepts, applica- tions, and programming, 3nd ed. New York, NY, US: Routledge/Taylor & Francis Group; 2016. xix, 396-xix, p. 34. Kline R, St C. Principles and Practice of Structural Equation Modeling2022. 35. Kaiser HF. The application of electronic computers to factor analysis. Educ Psychol Meas. 1960;20(1):141–51. 36. Dancey CP, Reidy J. Statistics without Maths for psychology: using SPSS for Windows. Prentice Hall; 2004. 14. Parry EL, Thomas MJ, Peat G. Defining acute flares in knee osteoarthritis: a 37. Terwee CB, Bot SD, de Boer MR, van der Windt DA, Knol DL, Dekker J, et al. systematic review. BMJ open. 2018;8(7):e019804. 15. Cross M, Dubouis L, Mangin M, Hunter DJ, March L, Hawker G, et al. Defining flare in Osteoarthritis of the hip and knee: a systematic literature review - OMERACT virtual special interest Group. J Rheumatol. 2017;44(12):1920–7. 16. Vitaloni M, Botto-van Bemden A, Sciortino Contreras RM, Scotton D, Bibas M, Quintero M, et al. Global management of patients with knee osteoarthritis begins with quality of life assessment: a systematic review. BMC Musculoske- let Disord. 2019;20(1):493. 17. Dai Z, Neogi T, Brown C, Nevitt M, Lewis CE, Torner J, et al. Sleep quality is related to worsening knee Pain in those with widespread Pain: the Multi- center Osteoarthritis Study. J Rheumatol. 2020;47(7):1019–25. 18. Hawker GA, Stewart L, French MR, Cibere J, Jordan JM, March L, et al. Understanding the pain experience in hip and knee osteoarthritis–an OARSI/ OMERACT initiative. Osteoarthritis Cartilage. 2008;16(4):415–22. 19. Hawker GA, Davis AM, French MR, Cibere J, Jordan JM, March L, et al. Devel- opment and preliminary psychometric testing of a new OA pain measure–an OARSI/OMERACT initiative. Osteoarthritis Cartilage. 2008;16(4):409–14. 20. Sit RWS, Chan DCC, Wong W, Yip BHK, Chow LLY, Wong SYS. Translation, cross- cultural adaptation and validation of the traditional chinese intermittent and constant osteoarthritis pain (ICOAP) questionnaire for knee osteoarthritis. BMJ open. 2019;9(3):e026006. 21. Erel S, Simsek IE, Ozkan H. Analysis of the reliability and validity of the turkish version of the intermittent and constant osteoarthritis pain questionnaire. Acta Orthop Traumatol Turc. 2015;49(5):508–12. 22. Panah SH, Baharlouie H, Rezaeian ZS, Hawker G. Cross-cultural adapta- tion and validation of the Persian version of the intermittent and con- stant Osteoarthritis Pain measure for the knee. Iran J Nurs Midwifery Res. 2016;21(4):417–23. 23. Goncalves RS, Cabri J, Pinheiro JP, Ferreira PL, Gil J. Cross-cultural adaptation and validation of the portuguese version of the intermittent and constant osteoarthritis pain (ICOAP) measure for the knee. Osteoarthritis Cartilage. 2010;18(8):1058–61. 24. Maillefert JF, Kloppenburg M, Fernandes L, Punzi L, Gunther KP, Martin Mola E, et al. Multi-language translation and cross-cultural adaptation of the OARSI/ OMERACT measure of intermittent and constant osteoarthritis pain (ICOAP). Osteoarthritis Cartilage. 2009;17(10):1293–6. 25. Mehta SP, Sankar A, Venkataramanan V, Lohmander LS, Katz JN, Hawker GA, et al. Cross-cultural validation of the ICOAP and physical function short forms of the HOOS and KOOS in a multi-country study of patients with hip and knee osteoarthritis. Osteoarthritis Cartilage. 2016;24(12):2077–81. 26. Alageel M, Al Turki A, Alhandi A, Alohali R, Alsalem R, Aleissa S. Cross-cultural adaptation and validation of the Arabic Version of the intermittent and con- stant Osteoarthritis Pain Questionnaire. Sports Med Int Open. 2020;4(1):E8–e12. 27. Farrag A, AlZahrani E, Elsayed W. Letter to the editor concerning “Cross- Cultural adaptation and validation of the Arabic Version of the intermittent and constant Osteoarthritis Pain Questionnaire. Sports Med Int Open. 2020;4(3):E67. - E8. Quality criteria were proposed for measurement properties of health status questionnaires. J Clin Epidemiol. 2007;60(1):34–42. 38. Nunnally JC, Bernstein IH. Psychometric theory:McGraw-Hill; 1994. 39. Koo TK, Li MY. A Guideline of selecting and reporting Intraclass correlation coefficients for Reliability Research. J Chiropr Med. 2016;15(2):155–63. 40. Weir JP. Quantifying test-retest reliability using the intraclass correlation coef- ficient and the SEM. J Strength Cond Res. 2005;19(1):231–40. 41. Goncalves RS, Meireles AC, Gil JN, Cavalheiro LM, Rosado JO, Cabri J. Responsiveness of intermittent and constant osteoarthritis pain (ICOAP) after physical therapy for knee osteoarthritis. Osteoarthritis Cartilage. 2012;20(10):1116–9. 42. Akobeng AK. Understanding diagnostic tests 3: receiver operating character- istic curves. Acta Paediatr. 2007;96(5):644–7. 43. Prinsen CAC, Mokkink LB, Bouter LM, Alonso J, Patrick DL, de Vet HCW, et al. COSMIN guideline for systematic reviews of patient-reported outcome measures. Qual life research: Int J Qual life aspects Treat care rehabilitation. 2018;27(5):1147–57. 44. Zhang C, Liu DH, Qu YL, Jia ZY, Wang W, Li J, et al. Transcultural adaptation and validation of the chinese version of the intermittent and constant osteoarthritis pain (ICOAP) measure in patients with knee osteoarthritis. Osteoarthritis Cartilage. 2017;25(4):506–12. 45. Manolarakis GE, Kontodimopoulos N, Sifaki-Pistolla D, Niakas D. Establishing the Psychometric Properties of the ICOAP questionnaire through Intra-Artic- ular Treatment of Osteoarthritic Pain: implementation for the Greek Version. Arthritis. 2016;2016:6201802. 46. Beckerman H, Roebroeck ME, Lankhorst GJ, Becher JG, Bezemer PD, Verbeek AL. Smallest real difference, a link between reproducibility and responsive- ness. Qual Life Res. 2001;10(7):571–8. 47. Davis AM, Lohmander LS, Wong R, Venkataramanan V, Hawker GA. Evaluating the responsiveness of the ICOAP following hip or knee replacement. Osteoar- thritis Cartilage. 2010;18(8):1043–5. 48. Ruyssen-Witrand A, Fernandez-Lopez CJ, Gossec L, Anract P, Courpied JP, Dougados M. Psychometric properties of the OARSI/OMERACT osteoarthritis pain and functional impairment scales: ICOAP, KOOS-PS and HOOS-PS. Clin Exp Rheumatol. 2011;29(2):231–7. 49. Bond M, Davis A, Lohmander S, Hawker G. Responsiveness of the OARSI- OMERACT osteoarthritis pain and function measures. Osteoarthritis Cartilage. 2012;20(6):541–7. 50. Risser RC, Hochberg MC, Gaynor PJ, D’Souza DN, Frakes EP. Responsiveness of the intermittent and constant Osteoarthritis Pain (ICOAP) scale in a trial of duloxetine for treatment of osteoarthritis knee pain. Osteoarthritis Cartilage. 2013;21(5):691–4. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Farrag et al. BMC Musculoskeletal Disorders (2023) 24:481
10.1186_s12891-023-06370-5	Niu et al. BMC Musculoskeletal Disorders (2023) 24:252 https://doi.org/10.1186/s12891-023-06370-5 RESEARCH BMC Musculoskeletal Disorders Open Access Associations between blood antioxidant levels and femoral neck strength Peng Niu1†, Yongxi Liu1, Yanfeng Zhang1 and Lei Li2* Abstract Background Studies have confirmed that antioxidants contribute to a lower risk of osteoporosis, which is an independent factor for femoral neck fracture (FNF). However, the associations between blood antioxidant levels and femoral neck strength remain unclear. Objective Our aim was to test the hypothesis that levels of blood antioxidants are positively associated with com- posite indices of bone strength in femoral neck, which integrate the bending strength index (BSI), compressive strength index (CSI), and impact strength index (ISI), in a population of middle-aged and elderly individuals. Methods This cross-sectional study utilized data from the Midlife in the United States (MIDUS) study. Blood levels of antioxidants were measured and analyzed. Results In total, data from 878 participants were analyzed. Results of Spearman correlation analyses indicated that blood levels of 6 antioxidants (total lutein, zeaxanthin, alpha-carotene, 13-cis-beta-carotene, trans-beta-carotene and total lycopene) were positively associated with CSI, BSI, or ISI in middle-aged and elderly individuals. Conversely, blood gamma-tocopherol and alpha-tocopherol levels were negatively associated with CSI, BSI, or ISI scores. Furthermore, linear regression analyses suggested that only blood zeaxanthin levels remained positively associated with CSI (odds ratio, OR 1.27; 95% CI: 0.03, 2.50; p = 0.045), BSI (OR, 0.54; 95% CI: 0.03–1.06; p = 0.037), and ISI (OR, 0.06; 95% CI: 0.00, 0.13; p = 0.045) scores in the study population after adjusting for age and sex. Conclusions Our results indicated that elevated blood zeaxanthin levels were significantly and positively associated with femoral neck strength (CSI, BSI, or ISI) in a population of middle-aged and elderly individuals. These findings sug- gest that zeaxanthin supplementation may reduce FNF risk independently. Keywords Femoral neck fracture, Antioxidants, Femoral neck strength †The first author: Peng Niu The affiliation “The Quzhou Affiliated Hospital of Wenzhou Medical University, Quzhou People’s Hospital, Quzhou City, 324002, Zhejiang Province, China” of Dr. Lei Li is the main or first institution of this research. Correspondence: Lei Li lilei1085869748@163.com 1 Department of spine and joint surgery, Nan Yang Second General Hospital, Nanyang City, Henan Province 473009, China 2 The Quzhou Affiliated Hospital of Wenzhou Medical University, Quzhou People’s Hospital, Quzhou City, Zhejiang Province 324002, China Introduction Osteonecrosis of the femoral head (ONFH) is one of the most common debilitating diseases, as it increases the risk of traumatic and nontraumatic fracture occur- rence in the general population [1, 2]. Existing evidence in ONFH patients suggests that these individuals have lower bone density of the femoral head, which may be related to a reduction in the osteogenic differentiation of bone marrow stromal cells (BMSCs) and an inhibi- tion of osteogenic gene expression [2, 3]. A recent study demonstrated that nontraumatic ONFH patients had lower bone mineral density (BMD) of the lumbar spine © The Author(s) 2023, corrected publication 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Niu et al. BMC Musculoskeletal Disorders (2023) 24:252 Page 2 of 9 and femoral neck than that of healthy populations [1]. Although BMD is widely used to evaluate bone strength [4, 5], other studies indicate that BMD may only reflect 50%-70% of total bone strength [6, 7]. The comprehen- sive indices of femoral neck strength include the bending strength index (BSI), compressive strength index (CSI) and impact strength index (ISI) which are good indica- tors of femoral bone strength, a measure that is consid- ered a predictor for femoral neck fracture (FNF) [8]. Oxidative stress promotes bone loss and remodeling via its impact on the regulation of osteoblast survival and dif- ferentiation and enhancement of inflammatory responses [9]. Conversely, antioxidants can inhibit oxidative stress and prevent the pathological process [10]. Existing evi- dence indicates that the consumption of antioxidant rich fruits and vegetables, such as apples, tomatoes, and oranges, is related to attenuations in bone mass loss, a factor impacting fracture risk, in postmenopausal women [11]. However, the beneficial effects of an increased anti- oxidant intake on bone strength are controversial. On the one hand, some epidemiological studies have reported that an increased intake of certain antioxidants, includ- ing vitamin C, vitamin E, and carotenoids, may confer benefits to BMD in premenopausal and postmenopausal females [12–14]. On the other hand, several studies have reported a positive correlation between higher intakes of vitamin C and higher risk of osteoporosis [15, 16]. Fur- thermore, previous evidence also indicated that antioxi- dants may produce bone-site specific beneficial effects on bone health [17]. for example, a previous study reported that a high β-carotene consumption was related to increased BMD of the femoral neck and total hip rather than other body parts in postmenopausal women [12]. This cross-sectional study was conducted using data from the Midlife in the United States (MIDUS) study. We aimed to investigate the associations between the 10 blood antioxidants and BSI, CSI and ISI in the middle- aged and elderly individuals. Methods Study participants The MIDUS Study mainly aimed to investigate the psy- chosocial and behavioral factors involved in age-related health conditions among a national sample of Ameri- cans [18, 19]. Data used in our study were obtained from participants in the Biomarker Project of MIDUS II (N = 1,255). The Biomarker Project recruited participants from the original MIDUS I cohort and aimed to measure various biological indicators in blood, urine, saliva, and other biological samples from 2004 to 2009 [19]. Addi- tional details regarding the study methods and samples have been published elsewhere [18, 19]. Of the 1,255 participants in the Biomarker Project, we excluded data Fig. 1 Flow chart displaying the final analysis of the included samples from 377 participants who had missing data on impor- tant variables. The final analysis from 878 participant samples were seen in Fig. 1. All methods were carried out in accordance with relevant guidelines and regula- tions, approval from appropriate Institutional Review Boards at the Midlife in the United States (MIDUS) study centers [three general clinical research centers (GCRC), including Georgetown University, the University of Cali- fornia at Los Angeles and the University of Wisconsin- Madison] was granted for this study and all participants gave informed consent before participation. We retro- spectively analyzed MIDUS data from an open database (Inter-University Consortium for Political and Social Research). Blood measurements Details for the collection methods of biological samples in the Biomarker Project have been previously described [20]. In summary, we measured and analyzed blood levels Niu et al. BMC Musculoskeletal Disorders (2023) 24:252 Page 3 of 9 of 10 antioxidant markers (total lutein, zeaxanthin, beta- cryptoxanthin, 13-cis-beta-carotene, alpha-carotene, trans-beta-carotene, total lycopene, gamma-tocopherol, alpha-tocopherol, and retinol). Blood antioxidant meas- urements were obtained from fasting blood samples col- lected in the morning. Assessment of femoral neck strength During each participants visit, the BMD values of the left hip and lumbar spine (L1-L4) were measured by DXA scans via Hologic 4500 (UCLA and Georgetown sites) technology or GE Healthcare Lunar Prodigy (Madison site). Femoral neck width (FNW) and femoral neck axis length (FNAL) were also measured from the hip scans using manufacturer guidelines. The composite indices of femoral neck strength (g/kg-m), including CSI, BSI, and ISI, were calculated by the following formulas [21]: 1) CSI = BMD × FNW/Weight; 2) BSI = (BMD × FNW2)/ (FNAL × Weight); ISI = (BMD × FNW × FNAL)/ (Height × Weight). 3) Statistical analyses For research purposes, Spearman correlation analyses were preliminarily performed to examine the relation- ships between the 10 antioxidants and femoral neck strength (BSI, CSI and ISI). Then, linear regression analy- ses were conducted to assess the independent relation- ships among variables with antioxidants as independent variables and femoral neck strength indices as dependent variables (CSI, BSI and ISI) in the models. Age and sex have been reported to be related to bone mass loss [22], therefore the models were adjusted for age (Model 1) and sex (Model 2) successively. Standardized correlation coef- ficients and 95% confidence intervals (CIs) were reported for these models. We used multiple linear regression analyses to explore the relationships between the 10 measured antioxidants and femoral neck strength, stratified by age, sex and BMI. We used the median age value to form a categorical variable for age (< 53 years and ≥ 53 years) and dichoto- mized BMI as < 24 kg/m2 or ≥ 24 kg/m2. Then, we tested for effect modification of age, sex and BMI on the asso- ciations. R version 4.0 and SPSS 25.0 were used for all analyses. The P value ≤ 0.05 represented a statistically sig- nificant difference. Results Characteristics of participants A total of 878 samples were analyzed. As shown in Table 1, the age [53 (44–61) years], sex [male, 356 (40.55%)] and BMI [28.61 (25.07–32.97)] of our study sample were similar to the complete sample from the Biomarker Project. 377 Biomarker Project participants Table 1 General characteristics of the study population Variables Age (year) Gender (male), n (%) BMI (kg/m2) Femoral Neck bone mineral density (gms/cm2) Compression Strength Index (g/kg-m) Bending Strength Index (g/kg-m) Impact Strength Index (g/kg-m) Blood markers Blood total lutein (umol/L) Blood zeaxanthin (umol/L) Blood beta-cryptoxanthin (umol/L) Blood 13-cis-beta-carotene (umol/L) Blood alpha-carotene (umol/L) Blood all trans-beta-carotene (umol/L) Blood total lycopene (umol/L) Blood gamma-tocopherol (umol/L) Blood alpha-tocopherol (umol/L) Blood retinol (umol/L) BMI: Body mass index N (%) or Median (interquartile range) 53 (44–61) 356 (40.55%) 28.61 (25.07–32.97) 0.78 (0.66–1.02) 3.49 (3.05–4.03) 1.17 (1.00–1.36) 0.20 (0.16–0.23) 0.20 (0.14–0.28) 0.05 (0.04–0.08) 0.17 (0.11–0.27) 0.05 (0.03–0.09) 0.06 (0.03–0.11) 0.37 (0.19–0.75) 0.39 (0.28–0.53) 3.66 (2.23–5.62) 25.95 (20.07–34.53) 1.58 (1.26–1.92) were mainly excluded from our analyses due to the pres- ence of missing data for the femoral neck strength (BSI, CSI and ISI) and blood antioxidant variables. Hence, our study participants may have similar BDM [0.78 (0.66– 1.02) gms/cm2], CSI [3.49 (3.05–4.03) g/kg-m], BSI [1.17 (1.00–1.36) g/kg-m] and ISI [0.20 (0.16–0.23) g/kg-m] values with the Biomarker Project sample. Similarly, the blood levels of antioxidants, including total lutein, zeaxanthin, beta-cryptoxanthin, 13-cis-beta-carotene, alpha-carotene, lycopene, trans-beta-carotene, gamma-tocopherol, alpha-tocopherol, and retinol, were 0.20 (0.14–0.28) µmol/L, 0.05 (0.04–0.08) µmol/L, 0.17 (0.11–0.27) µmol/L, 0.05 (0.03–0.09) µmol/L, 0.06 (0.03– 0.09) µmol/L, 0.37 (0.19–0.75) µmol/L, 0.39 (0.28–0.53) µmol/L, 3.66 (2.23–5.62) µmol/L, 25.95 (20.07–34.53) µmol/L and 1.58 (1.26–1.92) µmol/L, respectively. total Spearman correlation analyses between blood levels of the measured antioxidants and bone strength of the femoral neck 13-cis-beta-carotene, We observed that elevated blood levels of total lutein, zeaxanthin, alpha-carotene, trans-beta-carotene and total lycopene levels were posi- tively associated with CSI (all P ≤ 0.05), and blood levels of gamma-tocopherol and alpha-tocopherol were negatively associated with CSI. Levels of circulating beta-cryptoxan- thin and retinol were not associated with CSI (P > 0 0.05). Niu et al. BMC Musculoskeletal Disorders (2023) 24:252 Page 4 of 9 Circulating concentrations of total lutein, zeaxanthin, beta-cryptoxanthin, 13-cis-beta-carotene, alpha-carotene, trans-beta-carotene and total lycopene were positively associated with BSI (all P ≤ 0.05). However, blood gamma- tocopherol and alpha-tocopherol concentrations were inversely related to BSI (all P ≤ 0.05). Similar results were present for ISI (Table 2). Adjusted associations between blood antioxidant levels and bone strength of the femoral neck The results of our age and sex adjusted linear regression analyses, as shown in Table 3, confirmed that elevated blood zeaxanthin (r = 1.27; 95% CI: 0.03, 2.50; P = 0.045) and 13-cis-beta-carotene levels (r = 0.11; 95% CI: 0.28, 1.94; P = 0.009) were associated with increased CSI in the femoral neck, whereas elevated blood gamma-tocoph- erol (r = -0.03; 95% CI: -0.05, -0.01; P = 0.006) and alpha- tocopherol (r = -0.01; 95% CI: -0.01, -0.00; P = 0.024) levels were associated with lower CSI. Moreover, we found that blood total lutein (r = 0.19; 95% CI: 0.03, 0.35; P = 0.024), zeaxanthin (r = 0.54; 95% CI: 0.03, 1.06; P = 0.037), beta-cryptoxanthin (r = 0.22; 95% CI: 0.09, 0.36; P = 0.001), 13-cis-beta-carotene (r = 0.37; 95% CI: 0.02, 0.71; P = 0.036) and alpha-carotene (r = 0.28; 95% CI: 0.02, 0.54; P = 0.035) levels were positively associated with BSI. Finally, the results of the linear regression anal- yses were similar for ISI, as shown in Table 3. Subgrouping analyses on the associations between blood zeaxanthin levels and bone strength of the femoral neck The above results showed that only elevated blood zeax- anthin levels were positively and significantly correlated with CSI, BSI, and ISI, respectively. Thus, we further analyzed scatter plots and found that higher blood zeax- anthin levels may have contributed to higher composite indices of femoral neck strength (Fig. 2). Furthermore, we conducted subgrouping analyses to evaluate the effects of age, sex, and BMI on the relationship between blood zeaxanthin levels and bone strength of the femo- ral neck. Interestingly, although the interaction effects for age, sex, and BMI were not statistically significant (all interaction P values ≥ 0.05), as shown in Table 4, we still found that elevated blood zeaxanthin levels were associ- ated with higher CSI, BSI and ISI in female participants aged ≥ 53 years or with a BMI ≥ 24 kg/m2. Discussion There are many studies investigating the associations between antioxidants and bone health. For instance, it has been reported that greater serum carotenoid and lutein concentrations are associated with higher BMD in Chinese adults [23], and an adequate intake of veg- etables may reduce the risk of osteoporotic fractures among elderly men. The antioxidation of carotenoids may counteract the mechanism of osteoporosis related to leanness [24]. Consistent with carotenoids, eleva- tions in serum levels of lutein and zeaxanthin play a role in bone health [25]. Oxidative stress is associ- ated with lower BMD, which is more pronounced in individuals with low serum levels of vitamin E, as are often observed in older men [26]. Consistently, our results also showed that increased antioxidant levels were cross-sectionally associated with elevated indices of femoral neck strength (CSI, BSI, and ISI) in a rep- resentative sample of Americans. A positive correlation between zeaxanthin levels and femoral neck strength (CSI, BSI, and ISI) was also observed after adjustment for age and sex. One recent meta-analysis concluded that the role of vitamin A or its derivatives on BMD remain unclear, although most of the included studies showed a favorable effect of vitamin A on BMD [27]. Table 2 Spearman methods for correlation between blood antioxidant levels and bone strength of femoral neck Variables Femoral Neck bone mineral density (gms/cm2) Compression Strength Index (g/kg-m) Bending Strength Index (g/kg-m) Blood total lutein (umol/L) Blood zeaxanthin (umol/L) Blood beta-cryptoxanthin (umol/L) Blood 13-cis-beta-carotene (umol/L) Blood alpha-carotene (umol/L) Blood all trans-beta-carotene (umol/L) Blood total lycopene (umol/L) Blood gamma-tocopherol (umol/L) Blood alpha-tocopherol (umol/L) Blood retinol (umol/L) Spearman correlation analysis, < 0.05; ** < 0.01 -0.086* 0.002 -0.154 -0.095 -0.145 -0.208 0.001 0.022 -0.109** -0.068* 0.122 0.110 0.055 0.129** 0.083* 0.075* 0.068* -0.106 -0.108 -0.041 0.096** 0.076* 0.094 0.093 0.083* 0.087* 0.071* -0.068* -0.066* -0.033 Impact Strength Index (g/kg-m) 0.104** 0.073* 0.049 0.083* 0.073* 0.086* 0.073* -0.093** -0.068* 0.011 Niu et al. BMC Musculoskeletal Disorders (2023) 24:252 Page 5 of 9 Table 3 Linear regression analysis for correlation between blood antioxidant levels and bone strength of femoral neck Variables Femoral Neck bone mineral density (gms/cm2) Blood total lutein (umol/L) Blood zeaxanthin (umol/L) Blood beta-cryptoxanthin (umol/L) Blood 13-cis-beta-carotene (umol/L) Blood alpha-carotene (umol/L) Blood all trans-beta-carotene (umol/L) Blood total lycopene (umol/L) Blood gamma-tocopherol (umol/L) Blood alpha-tocopherol (umol/L) Blood retinol (umol/L) Compression Strength Index (g/kg-m) Blood total lutein (umol/L) Blood zeaxanthin (umol/L) Blood beta-cryptoxanthin (umol/L) Blood 13-cis-beta-carotene (umol/L) Blood alpha-carotene (umol/L) Blood all trans-beta-carotene (umol/L) Blood total lycopene (umol/L) Blood gamma-tocopherol (umol/L) Blood alpha-tocopherol (umol/L) Blood retinol (umol/L) Bending Strength Index (g/kg-m) Blood total lutein (umol/L) Blood zeaxanthin (umol/L) Blood beta-cryptoxanthin (umol/L) Blood 13-cis-beta-carotene (umol/L) Blood alpha-carotene (umol/L) Blood all trans-beta-carotene (umol/L) Blood total lycopene (umol/L) Blood gamma-tocopherol (umol/L) Blood alpha-tocopherol (umol/L) Blood retinol (umol/L) Impact Strength Index (g/kg-m) Blood total lutein (umol/L) Blood zeaxanthin (umol/L) Blood beta-cryptoxanthin (umol/L) Blood 13-cis-beta-carotene (umol/L) Blood alpha-carotene (umol/L) Blood all trans-beta-carotene (umol/L) Blood total lycopene (umol/L) Blood gamma-tocopherol (umol/L) Blood alpha-tocopherol (umol/L) Blood retinol (umol/L) Model 1: Adjusted for age Model 2: Adjusted for age and gender * < 0.05; < 0.01 Model 1 Sβ (95% CI) -0.21 (-0.34, -0.09) -0.20 (-0.60, 0.20) -0.14 (-0.25, -0.04) -0.28 (-0.55, -0.01) -0.24 (-0.44, -0.04) -0.03 (-0.06, -0.01) -0.01 (-0.10, 0.08) 0.00 (-0.00, 0.01) -0.00 (-0.00, 0.00) -0.03 (-0.06, -0.00) 0.32 (-0.06, 0.71) 1.28 (0.05, 2.52) 0.20 (-0.12, 0.53) 1.15 (0.32, 1.98) 0.64 (0.01, 1.27) 0.05 (-0.03, 0.13) 0.07 (-0.21, 0.35) -0.02 (-0.04, -0.01) -0.01 (-0.01, -0.00) -0.07 (-0.16, 0.01) 0.18 (0.02, 0.34) 0.59 (0.09, 1.10) 0.22 (0.09, 0.35) 0.46 (0.12, 0.80) 0.35 (0.10, 0.61) 0.04 (0.01, 0.07) 0.05 (-0.06, 0.17) -0.01 (-0.01, 0.00) -0.00 (-0.00, 0.00) -0.04 (-0.07, -0.01) 0.02 (-0.00, 0.04) 0.07 (0.01, 0.13) 0.02 (0.00, 0.03) 0.05 (0.01, 0.09) 0.05 (0.02, 0.08) 0.00 (0.00, 0.01) 0.01 (-0.00, 0.03) -0.00 (-0.00, -0.00) -0.00 (-0.00, -0.00) -0.00 (-0.01, 0.00) P Value < 0.001 0.322 0.007** 0.039* 0.020* 0.007** 0.814 0.268 0.153 0.024* 0.102 0.042* 0.220 0.007** 0.046* 0.209 0.627 0.014* 0.009** 0.076 0.029* 0.021* 0.001 0.008 0.007** 0.019* 0.350 0.117 0.200 0.018* 0.060 0.025* 0.049* 0.024* 0.003 0.023 0.083 0.002 0.021 0.281 Model 2 Sβ (95% CI) -0.17 (-0.30, -0.05) -0.16 (-0.55, 0.23) -0.12 (-0.22, -0.02) -0.21 (-0.47, 0.05) -0.21 (-0.41, -0.02) -0.03 (-0.05, -0.00) -0.04 (-0.13, 0.05) 0.00 (-0.00, 0.01) -0.00 (-0.00, 0.00) -0.02 (-0.05, 0.01) 0.38 (-0.01, 0.77) 1.27 (0.03, 2.50) 0.23 (-0.09, 0.56) 1.11 (0.28, 1.94) 0.59 (-0.04, 1.21) 0.05 (-0.03, 0.13) 0.06 (-0.22, 0.34) -0.03 (-0.05, -0.01) -0.01 (-0.01, -0.00) -0.05 (-0.14, 0.03) 0.19 (0.03, 0.35) 0.54 (0.03, 1.06) 0.22 (0.09, 0.36) 0.37 (0.02, 0.71) 0.28 (0.02, 0.54) 0.03 (-0.00, 0.06) 0.08 (-0.04, 0.19) -0.01 (-0.02, 0.00) -0.00 (-0.00, 0.00) -0.03 (-0.06, 0.00) 0.02 (0.00, 0.04) 0.06 (0.00, 0.13) 0.02 (0.00, 0.03) 0.04 (-0.01, 0.08) 0.04 (0.01, 0.07) 0.00 (-0.00, 0.01) 0.02 (0.00, 0.03) -0.00 (-0.00, -0.00) -0.00 (-0.00, -0.00) -0.00 (-0.00, 0.00) P Value 0.007 0.423 0.022* 0.119 0.035* 0.048* 0.338 0.509 0.834 0.123 0.059 0.045* 0.165 0.009 0.067 0.216 0.661 0.006 0.024* 0.192 0.024* 0.037* 0.001** 0.036* 0.035* 0.080* 0.200 0.064 0.286 0.094 0.050 0.045 0.043* 0.097 0.020* 0.101 0.036 < 0.001 0.035 0.714 Niu et al. BMC Musculoskeletal Disorders (2023) 24:252 Page 6 of 9 Fig. 2 Scatter diagram depicting the associations between blood zeaxanthin and the femoral neck strength (BSI, CSI and ISI) This is also consistent with our results which showed no significant associations between blood retinol levels and BMD or bone strength of the femoral neck. Antioxidants are important substances for elimi- nating free radicals. It can reduce the oxidative stress responses in the body and increase BMD. Therefore, consuming adequate amounts of fruits and vegetables can reduce the risk of osteoporosis and its complica- tions, such as pain and fracture [28]. Recent epidemio- logical evidence concluded that relatively high intakes of antioxidants, including carotenoids [6], vitamin E, vitamin C, and flavonoids were linked to an increased BMD in postmenopausal women [12–14, 29]. A recent study reported a causal link between increased circu- lating α-tocopherol and elevated BMD [30]. Another observational study also found that total dietary antiox- idant capacity of as inversely associated with the risk of osteoporosis in postmenopausal women and positively associated with bone mass in both pre- and postmeno- pausal women [31]. These previous observational stud- ies consistently asserted that the intake of antioxidants may strongly impact BMD in femoral neck. Conversely, our study showed negative associations between blood total lutein, beta-cryptoxanthin, alpha-carotene and all trans-beta-carotene levels and BMD, after adjusting for age and sex. There were no significant associations between blood gamma-tocopherol or alpha-tocopherol concentrations and BMD. A possible reason for this finding is that we included a sample of generally healthy participants in our analyses rather than a specific pop- ulation, such as postmenopausal women or elderly individuals. Another important reason may be that we measured the participants’ circulating antioxidant lev- els, rather than assessing fruit and vegetable intake, which was a commonly used method in previous stud- ies. Our results showed that the circulating levels of antioxidants were negatively correlated or uncorrelated with the BMD of the femoral neck, which is an interest- ing phenomenon and needs to be confirmed in future investigations. Osteoporosis typically occurs in individuals who are 50 years of age or older. A decrease in bodily hormone levels, particularly in postmenopausal women, leads to the proliferation of osteoclasts and the acceleration in bone loss. Increased fruit and vegetable intake has been related to bone mineral content in premenopau- sal women [28]. Further, the increased intake of vitamin C was related to higher femur BMD in premenopau- sal women [32]. These studies suggest that antioxidant intake confers clear benefits on BMD in premenopau- sal women. Consistently, our study also suggested that blood antioxidant levels were positively associated with femoral neck strength (CSI, BSI and ISI) in female sub- jects aged ≥ 53 years. These findings confirm that the consumption of antioxidant rich food is beneficial to increase femoral neck strength and may subsequently prevent FNF. Interestingly, we also observed a positive correlation between blood antioxidant levels and femoral neck strength in participants with overweight or obesity (BMI ≥ 24 kg/m2), but not in participants with a normal BMI (BMI < 24 kg/m2). This may suggest that supple- mentation with antioxidants is beneficial to increase the strength of the femoral neck in individuals with an over- weight or obese BMI. Compared with previous observational studies, our study has several advantages. Our data were obtained from the MIDUS study, which enrolled a representa- tive sample of Americans. The MIDUS II Biomarker Project conducted high-quality assessments on blood samples, which were obtained from a large sample of the general population. Our study included measures of almost all lipid soluble antioxidants, which allowed us to Niu et al. BMC Musculoskeletal Disorders (2023) 24:252 Page 7 of 9 * P e u l a V P ) I C % 5 9 ( β S * P e u l a V P ) I C % 5 9 ( β S * P e u l a V P ) I C % 5 9 ( β S * P e u l a V P ) I C % 5 9 ( β S 1 7 1 0 . 7 9 3 0 . 1 4 7 0 . 9 2 6 0 . 6 0 0 0 . 6 6 5 0 . 4 1 0 0 . 5 9 4 0 . 5 3 0 0 . ) 2 1 0 . , . 7 0 0 - ( 2 0 0 . ) 9 1 0 . , . 3 0 0 ( 1 1 0 . ) 5 1 0 . , . 8 0 0 - ( 3 0 0 . ) 6 1 0 . , . 2 0 0 ( 9 0 0 . ) 7 1 0 . , . 8 0 0 - ( 4 0 0 . ) 3 1 0 . , . 0 0 0 ( 7 0 0 . 9 4 3 0 . 1 7 8 0 . 8 0 4 0 . 5 7 3 0 . 8 1 0 0 . 4 2 1 0 . 6 7 0 0 . 8 5 7 0 . 2 2 0 0 . ) 0 1 1 . , . 1 4 0 - ( 4 3 0 . ) 2 5 1 . , . 5 1 0 ( 3 8 0 . ) 2 5 1 . , . 8 1 0 - ( 7 6 0 . ) 2 2 1 . , . 6 0 0 - ( 8 5 0 . ) 3 0 1 . , . 5 7 0 - ( 4 1 0 . ) 3 2 1 . , . 9 0 0 ( 6 6 0 . 3 7 1 0 . 6 7 6 0 . 6 8 5 0 . 1 5 6 0 . 6 0 0 0 . 1 8 2 0 . 1 4 0 0 . 8 9 5 0 . 1 4 0 0 . ) 3 4 2 . , . 2 5 1 - ( 6 4 0 . ) 6 7 3 . , . 2 6 0 ( 9 1 2 . ) 4 8 2 . , . 2 8 0 - ( 1 0 1 . ) 9 0 3 . , . 7 0 0 ( 8 5 1 . ) 9 5 2 . , . 9 4 1 - ( 5 5 0 . ) 2 7 2 . , . 6 0 0 ( 9 3 1 . 0 2 7 0 . 7 6 7 0 . 9 7 8 0 . 1 6 5 0 . 2 4 8 0 . 6 2 9 0 . 0 5 4 0 . 5 3 8 0 . 9 9 4 0 . ) 6 4 0 . , . 6 8 0 - ( 0 2 0 - . ) 3 4 0 . , . 3 5 0 - ( 5 0 0 - . ) 6 7 0 . , . 4 8 0 - ( 4 0 0 - . ) 6 2 0 . , . 9 5 0 - ( 7 1 0 - . ) 6 6 0 . , . 2 8 0 - ( 8 0 0 - . ) 9 2 0 . , . 0 6 0 - ( 5 1 0 - . 3 5 < 3 5 ≥ l e a M l e a m e F 4 2 < 4 2 ≥ ) r a e y ( e g A r e d n e G ) 2 m / g k ( I M B r e d n e g d n a e g a r o f d e t s u d A j l e u a v P n o i t c a r e t n I * P ) m - g k / g ( x e d n I h t g n e r t S t c a p m I ) m - g k / g ( x e d n I h t g n e r t S g n d n e B i ) m - g k / g ( x e d n I h t g n e r t S n o i s s e r p m o C y t i s n e d i l a r e n m e n o b k c e N l a r o m e F ) 2 m c / s m g ( k c e n l a r o m e f f o h t g n e r t s e n o b d n a s l e v e l i l n h t n a x a e z d o o b n e e w t e b n o i t a e r r o c r o l f l s i s y a n a g n p u o r G i 4 e l b a T l s e b a i r a V Niu et al. BMC Musculoskeletal Disorders (2023) 24:252 Page 8 of 9 comprehensively analyze the relationships between blood antioxidants and bone strength. In addition to BMD, we also analyzed femoral neck strength indices (CSI, BSI, and ISI), making this the first study to analyze the rela- tionship between blood antioxidants and femoral neck fracture-related indicators. Importantly, the retrospective analyses conducted in the present study did impose some limitations. First, the inherent disadvantages of cross- sectional studies made it difficult to assess causal associa- tions between antioxidants and both BMD and femoral neck bone strength, although the pathological mecha- nism by which reducing oxidative stress may attenuate bone mass loss is well characterized. Second, the sta- tistical models were adjusted for age, sex, and BMI, but there are other known and unknown confounding vari- ables that were not measured or included in the analyses. For example, we did not include information on disease history, including such conditions as hypertension, cor- onary heart disease and diabetes, in the analyses. Addi- tionally, data related to menopausal status in females and drug treatment status (e.g., vitamin D supplementation or other drug treatments that may cause osteoporosis) were not included in the analyses. Therefore, the results may be impacted by these factors. Third, 377 samples from the MIDUS II Biomarker Project (N = 1,225) were not analyzed due to missing data, and only data from 878 subjects were included in our study, potentially produc- ing an offset of the sample section. Conclusion We comprehensively conducted a cross-sectional analy- sis of the relationship between blood levels of 10 anti- oxidants and bone strength of the femoral neck. Our results indicate that increasing blood levels of antioxi- dants, especially zeaxanthin, may increase femoral neck strength (CSI, BSI, and ISI). These findings supported that antioxidant supplementation can further reduce FNF risk. Acknowledgements Not applicable. Authors’ contributions Peng Niu wrote the main manuscript; Yongxi Liu and Yanfeng Zhang completed the validation; Lei Li revised the draft. All authors reviewed the manuscript. The author(s) read and approved the final manuscript. Author information Not applicable. Funding This study was not supported by any funding. Availability of data and materials The data used in the present study are publicly available through the Inter- University Consortium for Political and Social Research (ICPSR): www. icpsr. umich. edu/ web/ ICPSR/ studi es/ 29282. Declarations Ethics approval and consent to participate All methods were carried out in accordance with relevant guidelines and regulations, and approval from appropriate Institutional Review Boards at the Midlife in the United States (MIDUS) study centers was granted for this study and all participants gave informed consent before participation. We retrospec- tively analyzed MIDUS data from an open database (ICPSR). Consent for publication Not applicable. Competing interests The authors declare no competing interests. Received: 7 December 2022 Accepted: 24 March 2023 References 1. Gangji V, Soyfoo MS, Heuschling A, Afzali V, Moreno-Reyes R, Rasschaert J, Gillet C, Fils JF, Hauzeur JP. Non traumatic osteonecrosis of the femoral head is associated with low bone mass. Bone. 2018;107:88–92. Li L, Ding Y, Liu B, Wang Z, Carlone DL, Yu X, Wei X, Zhang F, Lineaweaver WC, Yang B, Xia W, Wang DZ, Zhao D. Transcriptome landscape of the late-stage alcohol-induced osteonecrosis of the human femoral head. Bone. 2021;150:116012. 2. 3. Han L, Wang B, Wang R, Gong S, Chen G, Xu W. The shift in the balance between osteoblastogenesis and adipogenesis of mesenchymal stem cells mediated by glucocorticoid receptor. Stem Cell Res Ther. 2019;10(1):377. 4. Cosman F, McMahon D, Dempster D, Nieves JW. Standard Versus Cyclic Teriparatide and Denosumab Treatment for Osteoporosis: A Randomized Trial. J Bone Miner Res. 2020;35(2):219–25. 5. Kanis JA, McCloskey EV, Johansson H, Oden A, Melton LJ 3rd, Khaltaev N. A reference standard for the description of osteoporosis. Bone. 2008;42(3):467–75. Lorentzon M, Cummings SR. Osteoporosis: the evolution of a diagnosis. J Intern Med. 2015;277(6):650–61. ESHRE Capri Workshop Group. Bone fractures after menopause. Hum Reprod Update. 2010;16(6):761–73. 7. 6. 8. Ahn SH, Kim H, Kim BJ, Lee SH, Koh JM. Insulin resistance and composite indices of femoral neck strength in Asians: the fourth Korea National Health and Nutrition Examination Survey (KNHANES IV). Clin Endocrinol (Oxf ). 2016;84(2):185–93. 9. Kimball JS, Johnson JP, Carlson DA. Oxidative Stress and Osteoporosis. J Bone Joint Surg Am. 2021;103(15):1451–61. 10. Wong RH, ThaungZaw JJ, Xian CJ, Howe PR. Regular Supplementation With Resveratrol Improves Bone Mineral Density in Postmenopausal Women: A Randomized. Placebo-Controlled Trial J Bone Miner Res. 2020;35(11):2121–31. 11. Hu D, Cheng L, Jiang W. Fruit and vegetable consumption and the risk of postmenopausal osteoporosis: a meta-analysis of observational studies. Food Funct. 2018;9(5):2607–16. 12. Regu GM, Kim H, Kim YJ, Paek JE, Lee G, Chang N, Kwon O. Association between Dietary Carotenoid Intake and Bone Mineral Density in Korean Adults Aged 30–75 Years Using Data from the Fourth and Fifth Korean National Health and Nutrition Examination Surveys (2008–2011). Nutri- ents. 2017;9(9):1025. 13. Odai T, Terauchi M, Hirose A, Kato K, Miyasaka N. Bone Mineral Density in Premenopausal Women Is Associated with the Dietary Intake of α-Tocopherol: A Cross-Sectional Study. Nutrients. 2019;11(10):2474. 14. Hardcastle AC, Aucott L, Reid DM, Macdonald HM. Associations between dietary flavonoid intakes and bone health in a Scottish population. J Bone Miner Res. 2011;26(5):941–7. 15. Kim YA, Kim KM, Lim S, Choi SH, Moon JH, Kim JH, Kim SW, Jang HC, Shin CS. Favorable effect of dietary vitamin C on bone mineral Niu et al. BMC Musculoskeletal Disorders (2023) 24:252 Page 9 of 9 density in postmenopausal women (KNHANES IV, 2009): discrepan- cies regarding skeletal sites, age, and vitamin D status. Osteoporos Int. 2015;26(9):2329–37. 16. Sugiura M, Nakamura M, Ogawa K, Ikoma Y, Yano M. High Vitamin C Intake with High Serum β-Cryptoxanthin Associated with Lower Risk for Osteoporosis in Post-Menopausal Japanese Female Subjects: Mikkabi Cohort Study. J Nutr Sci Vitaminol (Tokyo). 2016;62(3):185–91. 17. Wattanapenpaiboon N, Lukito W, Wahlqvist ML, Strauss BJ. Dietary carot- enoid intake as a predictor of bone mineral density. Asia Pac J Clin Nutr. 2003;12(4):467–73. 18. Brim OG, Ryff CD, Kessler RC. The MIDUS national survey: an overview. 392. In: Brim OG, Ryff CD, Kessler RC, editors. How Healthy Are We? A National 393 Study of Well-Being at Midlife. Chicago: University of Chi- cago Press; 2004. p. 1–36. 19. Ryff CD, Seeman T, Weinstein M. Midlife in the United States (MIDUS 2): Biomarker Project, 2004–2009 (ICPSR 29282; Version V9). ICPSR. 2019. https:// doi. org/ 10. 3886/ ICPSR 29282. v9. 20. Dienberg Love G, Seeman TE, Weinstein M, Ryff CD. Bioindicators in the MIDUS national study: protocol, measures, sample, and comparative context. J Aging Health. 2010;22(8):1059–80. 21. Khosla S. Pathogenesis of age-related bone loss in humans. J Gerontol A Biol Sci Med Sci. 2013;68(10):1226–35. 22. Mosekilde L. Sex differences in age-related loss of vertebral trabecu- lar bone mass and structure–biomechanical consequences. Bone. 1989;10(6):425–32. https:// doi. org/ 10. 1016/ 8756- 3282(89) 90074-4. 23. Zhang ZQ, Cao WT, Liu J, Cao Y, Su YX, Chen YM. Greater serum carot- enoid concentration associated with higher bone mineral density in Chinese adults. Osteoporos Int. 2016;27(4):1593–601. 24. Dai Z, Wang R, Ang LW, Low YL, Yuan JM, Koh WP. Protective effects of dietary carotenoids on risk of hip fracture in men: the Singapore Chinese Health Study. J Bone Miner Res. 2014;29(2):408–17. 25. Bovier ER, Hammond BR. The Macular Carotenoids Lutein and Zeaxanthin Are Related to Increased Bone Density in Young Healthy Adults. Foods. 2017;6(9):78. 26. Ostman B, Michaëlsson K, Helmersson J, Byberg L, Gedeborg R, Melhus H, Basu S. Oxidative stress and bone mineral density in elderly men: antioxi- dant activity of alpha-tocopherol. Free Radic Biol Med. 2009;47(5):668–73. 27. Khojah Q, AlRumaihi S, AlRajeh G, Aburas A, AlOthman A, Ferwana M. Vitamin A and its dervatives effect on bone mineral density, a systematic review. J Family Med Prim Care. 2021;10(11):4089–95. 28. Li JJ, Huang ZW, Wang RQ, Ma XM, Zhang ZQ, Liu Z, Chen YM, Su YX. Fruit and vegetable intake and bone mass in Chinese adolescents, young and postmenopausal women. Public Health Nutr. 2013;16(1):78–86. 29. Zeng LF, Luo MH, Liang GH, Yang WY, Xiao X, Wei X, Yu J, Guo D, Chen HY, Pan JK, Huang HT, Liu Q, Guan ZT, Han YH, Zhao D, Zhao JL, Hou SR, Wu M, Lin JT, Li JH, Liang WX, Ou AH, Wang Q, Li ZP, Liu J. Can Dietary Intake of Vitamin C-Oriented Foods Reduce the Risk of Osteoporosis, Fracture, and BMD Loss? Systematic Review With Meta-Analyses of Recent Studies. Front Endocrinol (Lausanne). 2020;3(10):844. 30. Michaëlsson K, Larsson SC. Circulating Alpha-Tocopherol Levels, Bone Mineral Density, and Fracture: Mendelian Randomization Study. Nutri- ents. 2021;13(6):1940. 31. Kim D, Han A, Park Y. Association of Dietary Total Antioxidant Capacity with Bone Mass and Osteoporosis Risk in Korean Women: Analysis of the Korea National Health and Nutrition Examination Survey 2008–2011. Nutrients. 2021;13(4):1149. 32. Macdonald HM, New SA, Golden MH, Campbell MK, Reid DM. Nutritional associations with bone loss during the menopausal transition: evidence of a beneficial effect of calcium, alcohol, and fruit and vegetable nutrients and of a detrimental effect of fatty acids. Am J Clin Nutr. 2004;79(1):155–65. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12889-023-16038-3	Chair et al. BMC Public Health (2023) 23:1081 https://doi.org/10.1186/s12889-023-16038-3 RESEARCH BMC Public Health Open Access Household air pollution from solid fuel use and depression among adults in rural China: evidence from the China Kadoorie Biobank data Sek Ying Chair1, Kai Chow Choi1, Mei Sin Chong1, Ting Liu2 and Wai Tong Chien1 Abstract Background Solid fuels are still widely used for cooking in rural China, leading to various health implications. Yet, studies on household air pollution and its impact on depression remain scarce. Using baseline data from the China Kadoorie Biobank (CKB) study, we aimed to investigate the relationship between solid fuel use for cooking and depression among adults in rural China. Methods Data on exposure to household air pollution from cooking with solid fuels were collected and the Chinese version of the World Health Organization Composite International Diagnostic Interview short-form (CIDI-SF) was used to evaluate the status of major depressive episode. Logistic regression analysis was performed to investigate the asso- ciation between solid fuel use for cooking and depression. Results Amongst 283,170 participants, 68% of them used solid fuels for cooking. A total of 2,171 (0.8%) participants reported of having a major depressive episode in the past 12 months. Adjusted analysis showed that participants who had exposure to solid fuels used for cooking for up to 20 years, more than 20 to 35 years, and more than 35 years were 1.09 (95% CI: 0.94–1.27), 1.18 (95% CI: 1.01–1.38), and 1.19 (95% CI: 1.01–1.40) times greater odds of having a major depressive episode, respectively, compared with those who had no previous exposure to solid fuels used for cooking. Conclusion The findings highlight that longer exposure to solid fuels used for cooking would be associated with increased odds of major depressive episode. In spite of the uncertainty of causal relationship between them, using solid fuels for cooking can lead to undesirable household air pollution. Reducing the use of solid fuels for cooking by promoting the use of clean energy should be encouraged. Keywords Solid fuel, Cooking, Depression, Household air pollution Background Depression is one of the most common mental health disorders, affecting more than 280 million people glob- ally [1]. A recent systematic review and meta-analysis Correspondence: Mei Sin Chong jomeisin@link.cuhk.edu.hk 1 The Nethersole School of Nursing, Faculty of Medicine, The Chinese University of Hong Kong, 6/F, Esther Lee Building, Horse Material Water, Shatin, New Territories, Hong Kong SAR, China 2 School of Nursing, Sun Yat Sen University, Guangzhou, China revealed that the 12-month and lifetime prevalence rates of major depressive disorder in China were 1.6% and 1.8%, respectively, and the percentages had been increas- ing over time [2]. If the population in China is estimated to be 1.426 billion in 2023 [3], the 12-month prevalence of major depressive disorder may reach over 22.8 million of individuals. A longitudinal population study in Aus- tralia suggested that the severity of depression is a major predictor for suicidal ideation and suicidal attempt [4]. Based on a recent meta-analysis on 15 studies, the preva- lence of suicidal attempt in a lifetime among individuals © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecom- mons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Chair et al. BMC Public Health (2023) 23:1081 Page 2 of 9 with major depressive disorder was 3.45 times higher than those without major depressive disorder [5]. A study conducted in mainland China reported that the prevalence of suicidal ideation was 16.7% among 1,916 patients (18–70 years old) with major depressive disor- der [6]. Symptoms of major depressive disorder, such as low mood, anhedonia and impaired cognition, are one of the key contributors to functional impairment [7], which could cause a great economic burden to the society. A study by Rayner et al. [8] reported that there was a signif- icant correlation between total healthcare costs (i.e., acci- dent and emergency department visits, hospitalizations, and visits to doctor) and depression. In addition, patients with multimorbidity and depression had more than twice the inpatient costs compared with those without depres- sion [9]. The estimated health burden from depression has been continuously increasing over the years. More recently, a systematic review reported that major depres- sive disorder was accounted for 49 million disability- adjusted life-years in 2020 [10]. Approximately 2.4 billion people still use solid fuels such as animal dung, wood, and coal for cooking glob- ally [11]. To date, solid fuels are extensively used in China, especially in rural households. There are half a billion people (40% of total population) in mainland China living in rural areas [12], with more than three- fourths of these rural households using solid fuels for cooking [13]. Incomplete combustion of solid fuels pro- duces compounds such as carbon monoxide, sulphur dioxide, black carbon and PM2.5(fine particulate matter with a diameter of less or equal to 2.5 μm) [14]. Con- centrations of PM2.5 from household air pollution due to cooking with solid fuels can be substantially high, causing up to 40% higher PM2.5exposure compared with the indoor and outdoor environments [15]. Previ- ous studies have confirmed that the exposure to solid fuel use contributes to adverse health effects such as sleep disturbance [16], chronic bronchitis and obstruc- tive pulmonary disease [17], hypertension [18] and an increased risk of cardiovascular disease hospitalization and stroke among rural population [19]. A nationwide prospective cohort study also reported significant asso- ciation between using solid fuels for cooking and car- diovascular mortality in China [20]. A longitudinal study conducted in the United States reported that individuals with previous 30-day exposure to ambient fine PM were 1.2 times more likely to have moderate to severe depressive symptoms [21]. Based on a national longitudinal survey in China, cooking with solid fuels was associated with a higher risk of depressive symptoms among individuals aged 60 years and above [22]. Similarly, a cohort study also reported positive association between solid fuel use and depression [21]. However, the study conducted by Pun et al. [21] in the urban areas of the United States did not focus specifically on household use of solid fuels as the source of air pollu- tion. Meanwhile, the studies conducted by Li et al. [22] and Shao et al. [23] were restricted to middle-aged and older Chinese population living in urban or rural areas. Air pollutant emissions from solid fuels are associated with adverse health effects. In recent decades, improved cookstoves and combustion technologies have been implemented but a large number of individuals remain using solid fuels for cooking. Despite the important role of solid fuels in producing energy, its potential detrimen- tal impact on mental health demands urgent attention. There is a lack of studies focusing on the use of epidemio- logical data to investigate the association between solid fuel use and depression in developing countries; con- cerns arise as China is the second most populous coun- try where 40% of the population are living in rural areas and actively using solid fuels for cooking. It is crucial to investigate the relationship between these two variables specifically in rural China with a large-scale study. There- fore, this current study aimed to investigate the associa- tion between using solid fuels for cooking and depression in rural China. Methods Study design and population This study employed a secondary data analysis using the baseline data from the China Kadoorie Biobank (CKB) study. The original CKB study was conducted between 2004 and 2008, which recruited over 0.5 million adults from 10 regions across mainland China. After provid- ing informed written consent, each participant attended a face-to-face interview and a physical examination. A total of 512,681 adults aged between 30 and 79 years (without any major disability) with permanent residence were included in this baseline survey. A standardized electronic questionnaire was used to collect participant information including sociodemographic characteristics, lifestyle habits, exposure to passive smoking and domes- tic indoor air pollution, medical history, physical activity, and mental health status. The questionnaire used can be accessed via the official website of CKB (https:// www. ckbio bank. org/ study- resou rces/ survey- data). Each par- ticipant’s resting blood pressure (BP) was measured using the A&D digital BP monitor (Model No.: UA-779). A body composition analyzer (Model No.: TBF-300GS) was used to measure body mass index (BMI), while a stand- ing height measuring instrument was used to measure weight and height. BMI is calculated by using this for- mula: the participant’s weight in kilograms (kg) divided by the square of height (H) in meters (m), (BMI = kg/ H2) [24]. BMI at 28 kg / m2is recommended as the cut-off Chair et al. BMC Public Health (2023) 23:1081 Page 3 of 9 point for obesity for the Chinese people [25, 26]. More detailed information about the original CKB study has been previously reported [27–29]. Ethics approvals for the CKB study were obtained from the Chinese Center for Disease Control and Prevention (Approval Notice 005/2004) and the Oxford Tropical Research Ethics Committee (OxTREC Ref: 025–04) of the University of Oxford. The study was conducted in line with the princi- ples outlined in the Declaration of Helsinki. Measurements Exposure to household use of solid fuels The approach of Yu et al. [20] was followed to calculate the durations of exposure to solid fuels used for cooking and heating separately. Participants were asked to pro- vide detailed information about their exposure to house- hold use of solid fuels for cooking and heating, including related information such as the duration (in years) they lived in their three most recent residences, frequency of cooking in each residence, types of fuels used for cook- ing and heating, and availability of cookstove ventila- tion (chimney or extractor). Participants who reported that they cooked less often than once a month in a resi- dence were considered as noncooking and regarded as having no exposure to solid fuel used for cooking. Par- ticipants who reported that they cooked at least once a month were then asked to provide additional infor- mation related to the types of primary fuels they used. There are two categories of primary fuels, namely “clean fuels” such as gas and electricity, and “solid fuels” such as wood and coal [30]. The total duration (in years) of household use of solid fuels for cooking was calculated by summing up the duration of using solid fuels as the primary cooking fuel in each residence. Likewise, par- ticipants who used solid fuels for heating in winter were asked further questions about the types of primary fuels they used, and the total duration (in years) of household use of solid fuels for heating was calculated by sum- ming up the corresponding duration in each residence. The level of exposure to solid fuels used for heating was estimated by multiplying a weight coefficient to years of solid fuels used for heating, of which the weight coef- ficient was calculated based on the average portion of years with temperature less than 8 degree Celsius in each of the residences from 1999 to 2013, ranging from 0.18 to 0.42, as detailed in Yu et al. [20]. Depression In this study, major depressive episode was evaluated by the Chinese version of the World Health Organization Composite International Diagnostic Interview short- form (CIDI-SF) [31]. As there is no gold standard for assessing mental disorders in the CIDI-SF, this version was calibrated rather than validated and produced simi- lar population estimates of major depressive episode to the Structured Clinical Interview for DSM-IV, which is a state-of-the-art clinical research diagnostic interview tool for mental disorders [32]. Participants were first asked whether they had any of the following symptoms lasting for ≥ 2 weeks in the past 12 months: a) feeling much saddened, or depressed than usual; b) loss of inter- est in most things like hobbies or activities that usually gave you pleasure; c) feeling so hopeless and loss of appe- tite even for your favorite food; d) feeling worthless and useless, everything that went wrong was your fault, and life was very difficult with no way out. If participants answered “yes” to any of the above-mentioned situations, they were further assessed for major depression using CIDI-SF through a face-to-face interview by trained health professionals. Participants who reported at least 3 out of 7 depression symptoms (i.e., 1) weight change, 2) difficulty in sleeping, 3) losing interest in things, 4) feeling tired or low on energy, 5) trouble concentrating, 6) feeling worthless, or 7) thoughts about death) in the CIDI-SF questionnaire were considered likely to have major depression [33]. Covariates Adjustment for covariates was performed in this analy- sis, including sociodemographic characteristics (i.e., age, gender, marital status, education level and annual house- hold income), lifestyle habits (i.e., smoking status, alcohol assumption, and physical activity), health status (i.e., BMI and blood pressure), stressful life events in the past two years, passive smoking, cookstove ventilation, and expo- sure to solid fuels used for heating. Smoking status was classified into four categories: 1) never smoke, 2) quit- ted, 3) occasional smoker, and 4) current smoker. Partici- pants were classified as a “regular alcohol drinker” if they reported that they drank alcohol “usually at least once a week.” Otherwise, they were classified as a “non-regular drinker.” Physical activity was estimated as metabolic equivalent task hours per day spent on activities related to occupation, commuting, housework, and non-sed- entary leisure-time activities. Exposure to stressful life events (Yes/No) was defined as the occurrence of com- mon major life events in the past two years, such as death of a spouse, marital separation/divorce, traffic accident and major natural disaster. Exposure to passive smok- ing was assessed by self-report responses to the question related to frequency of secondhand smoking exposure. The variable was categorized into 4 levels (none, > 0 to 2 h/week, > 2 to 12 h/week, > 12 h/week). The cut-off points were conventionally selected based on the tertile points among those who had exposure to passive smok- ing, with the three exposure categories being anticipated Chair et al. BMC Public Health (2023) 23:1081 Page 4 of 9 to reflect low, middle and high levels of exposure to pas- sive smoking. non-exposure group with over 80% power at 2-sided 5% level of significance. Statistical analysis All statistical analyses were conducted using the IBM SPSS 25.0 (IBM Corp., Armonk, NY). Data were sum- marized descriptively using statistics including means, standard deviations, frequencies and percentages. For continuous variables, skewness statistics and normal- ity probability plots were used to assess normality. In this study, the outcome of interest was status of major depressive episode in the past year (Yes/No). The pri- mary exposure of interest was duration of solid fuels used for cooking which was categorized into four lev- els. Specifically, those participants who had no previ- ous exposure to solid fuels used for cooking or always used clean fuels were categorized as the reference group. The remaining participants were conventionally stratified into three tertiles to characterize low, mid- dle and high levels of exposure with totally four levels for the exposure factor: (i) none, (ii) > 0 to 20 years, (iii) > 20 to ≤ 35 years, (iv) > 35 years. Likewise, the expo- sure to solid fuels used for heating was categorized into four levels: (i) none, (ii) > 0 to 8.2 years, (iii) > 8.2 to ≤ 13.5 years, (iv) > 13.5 years. The association between major depression in the past year and exposure to solid fuels used for cooking was examined by logistic regres- sion analysis. Unadjusted and adjusted logistic regres- sion analyses were conducted with adjustment for the covariates of sociodemographic characteristics and life- style habits, presence of stressful life events in the past two years, presence of cookstove ventilation, exposure to passive smoking, and level of exposure to solid fuels used for heating. As the time scope of the outcome of major depressive episode was the past 12 months from the time of survey, it was possible that some partici- pants might have a major depressive episode prior to exposure to solid fuel usage. A sensitivity analysis was therefore conducted by excluding those participants who had no more than one year of solid fuel usage before the survey. All tests involved were 2-sided at 5% level of significance. A total of 283,170 participants were included in this secondary data analysis study. Among them, 2,171 par- ticipants were classified as having major depressive episode in the past year, and there were totally 91,611 participants without exposure to solid fuels used for cooking and 61,873 to 65,612 participants with differ- ent levels of exposure to solid fuels used for cooking. Such a sample size is adequate to detect an odds ratio of having major depressive episode of as small as 1.17 when comparing anyone of the exposure groups with the Results Characteristics of the study population Amongst 283,170 participants who were included in the baseline survey of the CKB study, the average age was 51.4 (SD = 10.5) years, and 58.2% of them were female. About 68% of them used solid fuels for cook- ing, with a 27-year median. More than half of the study sample (67%) had at least some cookstove ventilation. Nearly 23% participants had exposure to passive smok- ing for more than 12 h per week. A total of 2,171 (0.8%) participants reported major depressive episode in the past year. Characteristics of the study population strati- fied by levels of exposure to solid fuels used for cooking are shown in Table 1. Association between household use of solid fuels for cooking and major depressive episode Based on their duration of exposure to solid fuels used for cooking, participants were categorized into four lev- els: (i) none, (ii) > 0 to 20 years, (iii) > 20 to ≤ 35 years, (iv) > 35 years. Those participants who had no previous exposure to solid fuels used for cooking or always used clean fuels for cooking were categorized as the refer- ence group (none exposure). The remaining partici- pants were conventionally stratified into three tertiles to characterize low, middle and high levels of exposure. Unadjusted logistic regression analysis showed that an increased level of exposure to solid fuels used for cooking was associated with an increased odds of hav- ing a major depressive episode (unadjusted model in Table 2). After adjusting for sociodemographic charac- teristics, obesity and lifestyle habits, presence of stress- ful life events, presence of cookstove ventilation, passive smoking exposure, and level of exposure to solid fuels used for heating, the pattern of association between an increased odds of having a major depressive epi- sode and an increased level of exposure was also noted. Participants who had exposure to solid fuels used for cooking for up to 20 years, more than 20 to 35 years, and more than 35 years were 1.09 (95% CI 0.94–1.27), 1.18 (95% CI: 1.01–1.38) and 1.19 (95% CI: 1.01–1.40) times greater odds of having a major depressive episode, respectively, compared with those who had no previous exposure to solid fuel used for cooking or always used clean fuels for cooking (adjusted model 1 in Table 2). A sensitivity analysis was conducted by excluding those participants who had no more than one year of solid fuel usage before the survey, the results were similar to the primary analysis one (adjusted model 2 in Table 2). Chair et al. BMC Public Health (2023) 23:1081 Page 5 of 9 Table 1 Characteristics of the study population by level of exposure to solid fuels used for cooking (N = 283,170) Level of exposure to solid fuels used for cooking All (N = 283,170) None (n = 91,611) > 0 to 20 years (n = 64,524) > 20 to 35 years (n = 65,162) > 35 years (n = 61,873) 46,577 (16.4%) 84,487 (29.8%) 88,697 (31.3%) 48,441 (17.1%) 14,968 (5.3%) 15,809 (17.3%) 20,298 (31.5%) 5053 (7.8%) 27,860 (30.4%) 21,798 (33.8%) 23,623 (36.3%) 27,153 (29.6%) 13,257 (20.5%) 28,678 (44.0%) 15,640 (17.1%) 6924 (10.7%) 5149 (5.6%) 2247 (3.5%) 6181 (9.5%) 1627 (2.5%) 118,260 (41.8%) 79,634 (86.9%) 20,576 (31.9%) 9111 (14.0%) 164,910 (58.2%) 11,977 (13.1%) 43,948 (68.1%) 56,051 (86.0%) 5417 (8.8%) 11,206 (18.1%) 19,609 (31.7%) 19,696 (31.8%) 5945 (9.6%) 8939 (14.4%) 52,934 (85.6%) 259,280 (91.6%) 87,231 (95.2%) 60,204 (93.3%) 60,226 (92.4%) 51,619 (83.4%) 21,671 (7.7%) 2219 (0.8%) 3666 (4.0%) 714 (0.8%) 3906 (6.1%) 414 (0.6%) 4551 (7.0%) 385 (0.6%) 9548 (15.4%) 706 (1.1%) Characteristics Demographics Age (years) 30 – < 40 40 – < 50 50 – < 60 60 – < 70 ≥ 70 Sex Male Female Marital status Married Widowed / separated / divorced Never married Highest education attainment No formal school Primary school 67,740 (23.9%) 13,190 (14.4%) 12,506 (19.4%) 18,147 (27.8%) 118,593 (41.9%) 37,707 (41.2%) 24,332 (37.7%) 28,739 (44.1%) Middle school / high school 93,744 (33.1%) 38,923 (42.5%) 26,722 (41.4%) 18,032 (27.7%) Technical school / college/ university 3093 (1.1%) 1791 (2.0%) 964 (1.5%) 244 (0.4%) Household income in last year (Yuan) < 5,000 5,000 – 9,999 10,000 – 19,999 20,000 – 34,999 ≥ 35,000 Obesity status and lifestyle characteristics Obesity status 41,918 (14.8%) 70,752 (25.0%) 81,352 (28.7%) 54,285 (19.2%) 34,863 (12.3%) 10,238 (11.2%) 6634 (10.3%) 8255 (12.7%) 20,794 (22.7%) 14,517 (22.5%) 16,663 (25.6%) 25,492 (27.8%) 19,283 (29.9%) 20,737 (31.8%) 19,522 (21.3%) 14,783 (22.9%) 12,725 (19.5%) 15,565 (17.0%) 9307 (14.4%) 6782 (10.4%) Normal weight (18.5 ≤ BMI < 23.9) Overweight (24.0 ≤ BMI < 27.9) Obese (BMI ≥ 28.0) Under weight (BMI < 18.5) 162,222 (57.3%) 56,475 (61.6%) 36,709 (56.9%) 34,507 (53.0%) 82,596 (29.2%) 24,858 (27.1%) 19,271 (29.9%) 20,821 (32.0%) 22,982 (8.1%) 15,368 (5.4%) 5442 (5.9%) 4835 (5.3%) 5379 (8.3%) 3165 (4.9%) 6813 (10.5%) 3021 (4.6%) 23,897 (38.6%) 27,815 (45.0%) 10,067 (16.3%) 94 (0.2%) 16,791 (27.1%) 18,778 (30.3%) 15,840 (25.6%) 7255 (11.7%) 3209 (5.2%) 34,531 (55.8%) 17,646 (28.5%) 5348 (8.6%) 4347 (7.0%) Smoking status Never smoke Quitted Occasional smoker Current smoker Regular alcohol drinker No Yes Physical activity, MET – hours/daya Stressful life event & sleep disturbance Stressful life event in the past two years 175,263 (61.9%) 23,540 (25.7%) 46,351 (71.8%) 55,238 (84.8%) 50,134 (81.0%) 16,919 (6.0%) 12,954 (4.6%) 9841 (10.7%) 7445 (8.1%) 3181 (4.9%) 2438 (3.8%) 1614 (2.5%) 1455 (2.2%) 78,034 (27.6%) 50,785 (55.4%) 12,554 (19.5%) 6855 (10.5%) 2283 (3.7%) 1616 (2.6%) 7840 (12.7%) 245,061 (86.5%) 69,241 (75.6%) 57,499 (89.1%) 60,967 (93.6%) 57,354 (92.7%) 38,109 (13.5%) 22,370 (24.4%) 7025 (10.9%) 23.2 (14.5) 24.8 (16.2) 24.4 (14.7) 4195 (6.4%) 22.2 (13.0) 4519 (7.3%) 20.7 (12.4) No Yes 261,388 (92.3%) 85,582 (93.4%) 59,660 (92.5%) 59,988 (92.1%) 56,158 (90.8%) 21,782 (7.7%) 6029 (6.6%) 4864 (7.5%) 5174 (7.9%) 5715 (9.2%) Cook stove ventilation & passive smoking exposure Had at least some cook stove ventilation No 94,242 (33.3%) 27,318 (29.9%) 23,802 (36.9%) 24,207 (37.2%) 18,915 (30.6%) Chair et al. BMC Public Health (2023) 23:1081 Page 6 of 9 Table 1 (continued) Characteristics Yes Passive smoking exposure None > 0 to 2 h/week > 2 to 12 h/week > 12 h/week Level of exposure to solid fuels used for cooking All (N = 283,170) None (n = 91,611) > 0 to 20 years (n = 64,524) > 20 to 35 years (n = 65,162) > 35 years (n = 61,873) 188,570 (66.7%) 64,034 (70.1%) 40,661 (63.1%) 40,934 (62.8%) 42,941 (69.4%) 91,358 (32.3%) 62,116 (21.9%) 65,490 (23.1%) 64,206 (22.7%) 28,650 (31.3%) 22,290 (34.5%) 19,727 (30.3%) 18,898 (20.6%) 14,138 (21.9%) 15,068 (23.1%) 22,473 (24.5%) 14,402 (22.3%) 14,921 (22.9%) 21,590 (23.6%) 13,694 (21.2%) 15,446 (23.7%) Solid fuels usage for heating Level of exposure to solid fuels used for heatingb None > 0 to 8.2 years > 8.2 to 13.5 years > 13.5 years Major depression Major depression episode in the past year 108,153 (38.5%) 38,143 (42.0%) 26,444 (41.3%) 22,064 (34.1%) 57,408 (20.4%) 59,153 (21.1%) 56,252 (20.0%) 13,163 (14.5%) 18,968 (29.6%) 16,993 (26.3%) 19,568 (21.5%) 11,307 (17.6%) 14,500 (22.4%) 19,975 (22.0%) 7385 (11.5%) 11,155 (17.2%) 20,691 (33.4%) 14,012 (22.6%) 13,694 (22.1%) 13,476 (21.8%) 21,502 (35.1%) 8284 (13.5%) 13,778 (22.5%) 17,737 (28.9%) No Yes 280,999 (99.2%) 91,161 (99.5%) 64,032 (99.2%) 64,548 (99.1%) 61,258 (99.0%) 2171 (0.8%) 450 (0.5%) 492 (0.8%) 614 (0.9%) 615 (1.0%) Variables with data marked with a are presented as mean (standard deviation), all others are presented as frequency (%) b There were less than 0.8% (n = 2204) of participants without detailed information about fuels used for heating Table 2 Risk of major depression episode by level of exposure to solid fuels used for cooking Sensitivity analysis Unadjusted model Adjusted model 1 Adjusted model 2 Exposure factor Odds ratio (95% CI) p-value Odds ratio (95% CI) p-value Odds ratio (95% CI) p-value Level of exposure to solid fuels used for cooking None (ref ) 1 > 0 to 20 years 1.56 (1.37 – 1.77) > 20 to 35 years 1.93 (1.71 – 2.18) > 35 years 2.03 (1.80 – 2.30) < 0.001 < 0.001 < 0.001 1 1.09 (0.94 – 1.27) 1.18 (1.01 – 1.38) 1.19 (1.01 – 1.40) 0.249 0.034 0.033 1 1.08 (0.93 – 1.25) 1.18 (1.01 – 1.38) 1.18 (1.01 – 1.39) 0.316 0.041 0.043 Model 1: with adjustment for demographic, obesity and lifestyle characteristics, presence of stressful life event, presence of cook stove ventilation, passive smoking exposure, and level of exposure to solid fuels used for heating as listed in Table 1 Model 2: with adjustment for covariates in model 1 and excluding those participants who used solid fuels for cooking but had no more than one year of solid fuel used for cooking into analysis (n = 622 excluded) ref reference category for calculating odds ratios of other comparison categories Discussion Approximately 46% of the population in China used solid fuels as a household energy source, leading to household air pollution; and the proportion was sub- stantially higher in rural areas [13, 23]. In fact, the pre- sent study found that 68% of rural residents used solid fuels for cooking. To the best of our knowledge, this is the largest national study to explore the relationship between solid fuel use and depression in rural China. The results revealed an association between household use of solid fuels for cooking and major depression, particularly for those who had used solid fuels for more than 20 years, after controlling for potential confound- ing covariates, including sociodemographic charac- teristics, lifestyle habits, health status, presence of stressful life events, presence of cookstove ventilation, passive smoking exposure, and exposure to solid fuels used for heating. Although participants with longer exposure generally associated with an increased odds of having a major depressive episode, the odds ratio of the longest exposure group (> 35 years, OR = 1.19) was unexpectedly similar to the second longest exposure Chair et al. BMC Public Health (2023) 23:1081 Page 7 of 9 group (> 20 to 35 years, OR = 1.18). A possible explana- tion may be owing to the fact that people with longer exposure were more likely subject to a competing risk of death, which may diminish the strength of associa- tion, particularly in the longest exposure group. There is a growing body of evidence that solid fuel use is associated with a high risk of depression [22, 23], which is consistent with the current findings. Individu- als (N= 8637) with exposure to solid fuel combustion for over 4 years had 1.12 times greater odds of having depres- sive symptoms [23]. Supported by the following longitu- dinal survey (N= 7005) [22], individuals using solid fuels in cooking for more than 7 years had 1.36 times greater odds of depression risk than those who always used clean fuels. This study, together with the aforementioned previous studies, provides evidence on the association between the exposure to solid fuels and the prevalence of depres- sion. However, only limited evidence exists on the mech- anisms linking the use of solid fuels for cooking with depression. The incomplete combustion of solid fuels generates various air pollutants including PM, carbon monoxide, sulfur oxides, and polycyclic aromatic hydro- carbons [34, 35]. One possible explanation may be that inhalation of air pollutants can trigger associated oxida- tive stress, cerebrovascular damage, neuroinflammation, and neurodegenerative pathology, which all might cause or exacerbate the risk of depression [36–38]. Animal experience revealed that PM might cause neurotoxic- ity by inducing microglia activation characterized by the release of TNFα, which damages the olfactory bulb and increases depression risk [39]. Moreover, studies indicated that PM causes elevated levels of cortisol [40], which has been related to the development of depres- sion [41]. Furthermore, domestic cooking with solid fuels could increase the risk of chronic diseases, such as cancer and cardiorespiratory diseases [19, 42], which are strongly associated with depression [43, 44]. In rural China, solid fuels are reported to be the domi- nant cooking fuel, with biomass and coal accounting for 47.6% [45] and 13.5% [46], respectively. Our study gives valuable insights into the potential hazardous effects of using solid fuels for cooking on mental health. It indicates household solid fuels used for cooking is a critical public health issue and that policy makers must take responsi- bility to make the needed policy changes. It is necessary to encourage people to switch to cleaner fuels and tech- nologies when cooking to reduce exposure to household air pollution. Moreover, in this study, depressive episode was more prevalent in those without cookstove ventila- tion. This result is in line with those of the previous stud- ies [22, 47], showing that cooking ventilation may weaken the relationship of cooking with solid fuel and long duration cooking with depressive symptoms, suggest- ing that improvements in cooking ventilation should be strongly encouraged. As a remark, although people with longer exposure to solid fuels used for cooking generally associated with an increased odds of having a major depression episode, the odds ratio of the longest exposure group (> 35 years, OR = 1.19) was unexpectedly similar to the second long- est exposure group (> 20 to 35 years, OR = 1.18). A possi- ble explanation may be owing to the fact that people with longer exposure were more likely subject to a competing risk of death, which may diminish the strength of associa- tion, particularly in the longest exposure group. Despite the significance of the findings, there are sev- eral limitations in this study that may impact the gener- alisability of this study. First, the cross-sectional study design assesses both outcome of interest and exposure simultaneously. Therefore, it may not be able to estab- lish a cause-and-effect relationship between household solid fuels used for cooking and depression. In addi- tion, self-reported information is prone to recall bias when participants fail to accurately remember an event in the past. Nonetheless, the overestimation or underes- timation of association between cause and effect may be resolved through a longitudinal cohort study in which an event may be observed first, followed by the effects. On the other hand, different cooking practices and chemical properties of fuel such as density, volatility and thermal capacity which could affect the indoor air pollution were not examined in the CKB study. Hence, this could result in imprecision of actual exposure to solid fuels used for cooking. Although this study had controlled for poten- tial confounders (e.g., sociodemographic characteristics, obesity status and lifestyle habits, presence of stressful life events, presence of cookstove ventilation and passive smoking exposure), the results might be confounded by other unmeasured covariates. This is because our study was a secondary data analysis where the adjusted analysis was only able to be performed based on existing available variables. Conclusion This study demonstrates the significant association between the use of household solid fuels for cooking and the prevalence of depression in rural China; and the longer duration of exposure, the higher odds of having a depressive episode. Further studies are warranted to examine if there is a causal relationship between them. Nevertheless, reducing the use of solid fuels for cook- ing by promoting the use of clean energy should be encouraged. Acknowledgements Not applicable. Chair et al. BMC Public Health (2023) 23:1081 Page 8 of 9 Authors’ contributions Sek Ying Chair contributed to the conceptualization of the manuscript, writing of the original draft, reviewing and editing the manuscript. Kai Chow Choi contributed to the formal analysis, data curation, writing, reviewing and editing the manuscript. Mei Sin Chong contributed to writing, reviewing and editing the manuscript. Ting Liu contributed to writing, reviewing and editing the manuscript. Wai Tong Chien contributed to writing, reviewing and editing the manuscript. All authors read and approved the final manuscript. Funding This study was funded by grants from the National Key Research and Development Program of China (2016YFC0900500, 2016YFC0900501, 2016YFC0900504), the Kadoorie Charitable Foundation in Hong Kong, and Wellcome Trust (088158/Z/09/Z, 104085/Z/14/Z, 104085/Z/14/Z) in the UK. Availability of data and materials The datasets used and analyzed during this current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The CKB study was conducted in line with the principles outlined in the Declaration of Helsinki; ethics approvals were obtained from the Chinese Center for Disease Control and Prevention and the Oxford Tropical Research Ethics Committee of the University of Oxford; and an informed consent was obtained from each participant [26]. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Received: 16 March 2023 Accepted: 1 June 2023 References 1. World Health Organization. Depression. 2021. https:// www. who. int/ news- room/ fact- sheets/ detail/ depre ssion. Accessed 20 Nov 2022. 2. Zhao YJ, Jin Y, Rao WW, et al. Prevalence of major depressive disorder among adults in China: a systematic review and meta-analysis. Front Psychiatry. 2021;12:659470. https:// doi. org/ 10. 3389/ fpsyt. 2021. 659470. 3. United Nations. India overtakes China as the world’s most populous country. 2023. https:// www. un. org/ devel opment/ desa/ pd/ sites/ www. un. org. devel opment. desa. pd/ files/ undesa_ pd_ 2023_ policy- brief- 153. pdf . Accessed 3 May 2023. 4. Handley T, Rich J, Davies K, Lewin T, Kelly B. The challenges of predicting suicidal thoughts and behaviours in a sample of rural Australians with depression. Int J Environ Res Public Health. 2018;15(5):928. https:// doi. org/ 10. 3390/ ijerp h1505 0928. 5. Cai H, Xie XM, Zhang Q, Cui X, Lin JX, Sim K, Ungvari GS, Zhang L, Xiang YT. Prevalence of suicidality in major depressive disorder: a systematic review and meta-analysis of comparative studies. Front Psychiatry. 2021;12:690130. https:// doi. org/ 10. 3389/ fpsyt. 2021. 690130. 6. Ge F, Jiang J, Wang Y, Yuan C, Zhang W. Identifying suicidal ideation among Chinese patients with major depressive disorder: evidence from a real-world hospital-based study in China. Neuropsychiatr Dis Treat. 2020;16:665–72. https:// doi. org/ 10. 2147/ NDT. S2382 86. 7. Chow TK, Bowie CR, Morton M, Lalovic A, McInerney SJ, Rizvi SJ. Contribu- tors of functional impairment in major depressive disorder: a biopsycho- social approach. Curr Behav Neurosci Rep. 2022;9:59–72. https:// doi. org/ 10. 1007/ s40473- 022- 00247-y. 8. Rayner L, Hotopf M, Petkova H, Matcham F, Simpson A, McCracken LM. Depression in patients with chronic pain attending a specialised pain treatment centre: prevalence and impact on health care costs. Pain. 2016;157(7):1472–9. https:// doi. org/ 10. 1097/j. pain. 00000 00000 000542. 9. Bock JO, Luppa M, Brettschneider C, et al. Impact of depression on health care utilization and costs among multimorbid patients–from the Multi- Care Cohort Study. PLoS One. 2014;9(3):e91973. https:// doi. org/ 10. 1371/ journ al. pone. 00919 73. 10. COVID-19 Mental Disorders Collaborators. Global prevalence and burden of depressive and anxiety disorders in 204 countries and territories in 2020 due to the COVID-19 pandemic. Lancet (London, England). 2021;398:1700–12. https:// doi. org/ 10. 1016/ S0140- 6736(21) 02143-7. 11. World Health Organization. Household air pollution: key facts. 2022. https:// www. who. int/ news- room/ fact- sheets/ detail/ house hold- air- pollu tion- and- healt h#: ~: text= World wide% 2C% 20aro und% 202.4% 20bil lion% 20peo ple,and% 20i . Accessed 22 Jan 2023. 12. NBSC. China statistical yearbook-2019. Beijing: China Statistics Press; 2019. 13. Tang X, Liao H. Energy poverty and solid fuels use in rural China: analysis based on national population census. Energy Sustain Dev. 2014;23:122–9. https:// doi. org/ 10. 1016/J. ESD. 2014. 08. 006. 14. Shen G, Xing R, Zhou Y, et al. Revisiting the proportion of clean house- hold energy users in rural China by accounting for energy stacking. Sustainable Horizons. 2022;1:100010. https:// doi. org/ 10. 1016/j. horiz. 2022. 100010. 15 Xu H, Li Y, Guinot B, et al. Personal exposure of PM2. 5 emitted from solid fuels combustion for household heating and cooking in rural Guanzhong Plain, northwestern China. Atmos Environ. 2018;185:196–206. https:// doi. org/ 10. 1016/j. atmos env. 2018. 05. 018. 16. Chair SY, Choi KC, Cao X, et al. Association between household solid fuel use for cooking and sleep disturbance in rural China: findings from the China Kadoorie Biobank data. Sleep Med. 2021;83:13–20. https:// doi. org/ 10. 1016/J. SLEEP. 2021. 04. 029. 17. Po JY, FitzGerald JM, Carlsten C. Respiratory disease associated with solid biomass fuel exposure in rural women and children: systematic review and meta-analysis. Thorax. 2011;66:232–9. https:// doi. org/ 10. 1136/ THX. 2010. 147884. 18. Yan Z, Liu Y, Yin Q, et al. Impact of household solid fuel use on blood pres- sure and hypertension among adults in China. Air Qual Atmos Health. 2016;9:931–40. https:// doi. org/ 10. 1007/ S11869- 016- 0395-2/ TABLES/5. 19. Hystad P, Duong M, Brauer M, et al. Health effects of household solid fuel use: Findings from 11 Countries within the prospective urban and rural epidemiology study. Environ Health Perspect. 2019;127(5):57003. https:// doi. org/ 10. 1289/ EHP39 15. 20. Yu K, Qiu G, Chan KH, et al. Association of solid fuel use with risk of cardio- vascular and all-cause mortality in rural China. JAMA. 2018;319(13):1351– 61. https:// doi. org/ 10. 1001/ jama. 2018. 2151. 21. Pun VC, Manjourides J, Suh H. Association of ambient air pollution with depressive and anxiety symptoms in older adults: Results from the NSHAP study. Environ Health Perspect. 2017;125(3):342–8. https:// doi. org/ 10. 1289/ EHP494. 22. Li C, Zhou Y, Ding L. Effects of long-term household air pollution expo- sure from solid fuel use on depression: evidence from national longitudi- nal surveys from 2011 to 2018. Environ Pollut. 2021;283:117350. https:// doi. org/ 10. 1016/j. envpol. 2021. 117350. 23. Shao J, Ge T, Liu Y, Zhao Z, Xia Y. Longitudinal associations between household solid fuel use and depression in middle-aged and older Chi- nese population: a cohort study. Ecotoxicol Environ Saf. 2021;209:111833. https:// doi. org/ 10. 1016/j. ecoenv. 2020. 111833. 24. Lemmens HJ, Brodsky JB, Bernstein DP. Estimating ideal body weight–a new formula. Obes Surg. 2005;15(7):1082–3. 25. Xu W, Zhang H, Paillard-Borg S, Zhu H, Qi X, Rizzuto D. Prevalence of over- weight and obesity among Chinese adults: role of adiposity indicators and age. Obes Facts. 2016;9(1):17–28. https:// doi. org/ 10. 1159/ 00044 3003. 26. Zhou B, Coorperative Meta-Analysis Group of China Obesity Task Force. Predictive values of body mass index and waist circumference to risk fac- tors of related diseases in Chinese adult population. Zhonghua Liu Xing Bing Xue Za Zhi. 2002;23(1):5–10. 27. Chen Z, Lee L, Chen J, et al. Cohort profile: the Kadoorie Study of Chronic Disease in China (KSCDC). Int J Epidemiol. 2005;34(6):1243–9. https:// doi. org/ 10. 1093/ ije/ dyi174. 28 Chen Z, Chen J, Collins R, et al. China Kadoorie Biobank of 0.5 million peo- ple: survey methods, baseline characteristics and long-term follow-up. Int J Epidemiol. 2011;40(6):1652–66. https:// doi. org/ 10. 1093/ ije/ dyr120. Chair et al. BMC Public Health (2023) 23:1081 Page 9 of 9 Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. 29. Li LM, Lv J, Guo Y, et al. The China Kadoorie Biobank: Related methodol- ogy and baseline characteristics of the participants. Zhonghua Liu Xing Bing Xue Za Zhi. 2012;33(3):249–55. 30. Wang S, Luo K. Life expectancy impacts due to heating energy utilization in China: Distribution, relations, and policy implications. Sci Total Environ. 2018;610–611:1047–56. https:// doi. org/ 10. 1016/j. scito tenv. 2017. 08. 195. 31. Kessler RC, Andrews G, Mroczek D, et al. The World Health Organization Composite International Diagnostic Interview short-form (CIDI-SF). Int J of Methods Psychiatr Res. 1998;7:171–85. https:// doi. org/ 10. 1002/ MPR. 47. 32. Sun X, Zheng B, Lv J, Guo Y, Bian Z, Yang L, Chen Y, Fu Z, Guo H, Liang P, Chen Z, Chen J, Li L, Yu C, China Kadoorie Biobank (CKB) Collaborative Group. Sleep behavior and depression: findings from the China Kadoorie Biobank of 0.5 million Chinese adults. J Affect Disord. 2018;229:120–4. https:// doi. org/ 10. 1016/j. jad. 2017. 12. 058. 33 Meng R, Yu C, Liu N, et al. Association of depression with all-cause and cardiovascular disease mortality among adults in China. JAMA Netw Open. 2020;3(2):e1921043. https:// doi. org/ 10. 1001/ jaman etwor kopen. 2019. 21043. Published 2020 Feb 5. 34. Shen G, Chen Y, Du W, et al. Exposure and size distribution of nitrated and oxygenated polycyclic aromatic hydrocarbons among the population using different household fuels. Environ Pollut. 2016;216:935–42. https:// doi. org/ 10. 1016/j. envpol. 2016. 07. 002. 35. Zhang H, Zhu T, Wang S, et al. Indoor emissions of carbonaceous aerosol and other air pollutants from household fuel burning in Southwest China. Aerosol Air Qual Res. 2014;14:1779–88. https:// doi. org/ 10. 4209/ aaqr. 2013. 10. 0305. 36. Bhatt S, Nagappa AN, Patil CR. Role of oxidative stress in depression. Drug Discov Today. 2020;25(7):1270–6. https:// doi. org/ 10. 1016/j. drudis. 2020. 05. 001. 37. Calderón-Garcidueñas L, Calderón-Garcidueñas A, Torres-Jardón R, Avila- Ramírez J, Kulesza RJ, Angiulli AD. Air pollution and your brain: what do you need to know right now. Prim Health Care Res Dev. 2015;16(4):329– 45. https:// doi. org/ 10. 1017/ S1463 42361 40003 6X. 38. Hurley LL, Tizabi Y. Neuroinflammation, neurodegeneration, and depression. Neurotox Res. 2013;23(2):131–44. https:// doi. org/ 10. 1007/ s12640- 012- 9348-1. 39. Ji X, Liu R, Guo J, et al. Olfactory bulb microglia activation mediated neuronal death in real-ambient particulate matter exposure mice with depression-like behaviors. Sci Total Environ. 2022;821:153456. https:// doi. org/ 10. 1016/j. scito tenv. 2022. 153456. 40 Li H, Cai J, Chen R, et al. Particulate matter exposure and stress hormone levels: a randomized, double-blind, crossover trial of air purification [published correction appears in Circulation. 2017 Sep 12;136(11):e199]. Circulation. 2017;136(7):618–27. https:// doi. org/ 10. 1161/ CIRCU LATIO NAHA. 116. 026796. 41. Stetler C, Miller GE. Depression and hypothalamic-pituitary-adrenal acti- vation: a quantitative summary of four decades of research. Psychosom Med. 2011;73(2):114–26. https:// doi. org/ 10. 1097/ PSY. 0b013 e3182 0ad12b. 42. Balmes JR. Household air pollution from domestic combustion of solid fuels and health. J Allergy Clin Immunol. 2019;143(6):1979–87. https:// doi. org/ 10. 1016/j. jaci. 2019. 04. 016. 43. Carnevali L, Montano N, Statello R, Sgoifo A. Rodent models of depression-cardiovascular comorbidity: bridging the known to the new. Neurosci Biobehav Rev. 2017;76(Pt A):144–53. https:// doi. org/ 10. 1016/j. neubi orev. 2016. 11. 006. 44. Carvalho AF, Hyphantis T, Sales PM, et al. Major depressive disorder in breast cancer: a critical systematic review of pharmacological and psy- chotherapeutic clinical trials. Cancer Treat Rev. 2014;40(3):349–55. https:// doi. org/ 10. 1016/j. ctrv. 2013. 09. 009. 45. Hou B, Liao H, Huang J. Household cooking fuel choice and economic poverty: Evidence from a nationwide survey in China. Energy and Buil. 2018;166:319–29. https:// doi. org/ 10. 1016/j. enbui ld. 2018. 02. 012. 46. Duan X, Jiang Y, Wang B, et al. Household fuel use for cooking and heating in China: results from the first Chinese Environmental Exposure- Related Human Activity Patterns Survey (CEERHAPS). Appl Energy. 2014;136:692–703. https:// doi. org/ 10. 1016/j. apene rgy. 2014. 09. 066. 47. Liao W, Liu X, Kang N, et al. Effect modification of kitchen ventilation on the associations of solid fuel use and long-duration cooking with the increased prevalence of depressive and anxiety symptoms: the Henan Rural Cohort Study. Indoor Air. 2022;32(3):e13016. https:// doi. org/ 10. 1111/ ina. 13016. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12889-023-15632-9	Joseph et al. BMC Public Health (2023) 23:748 https://doi.org/10.1186/s12889-023-15632-9 BMC Public Health Who are the vulnerable, and how do we reach them? Perspectives of health system actors and community leaders in Kerala, India Jaison Joseph1, Hari Sankar1, Gloria Benny1 and Devaki Nambiar1,2,3 Abstract Background Among the core principles of the 2030 agenda of Sustainable Development Goals (SDGs) is the call to Leave no One behind (LNOB), a principle that gained resonance as the world contended with the COVID-19 pandemic. The south Indian state of Kerala received acclaim globally for its efforts in managing COVID-19 pandemic. Less attention has been paid, however, to how inclusive this management was, as well as if and how those “left behind” in testing, care, treatment, and vaccination efforts were identified and catered to. Filling this gap was the aim of our study. Methods We conducted In-depth interviews with 80 participants from four districts of Kerala from July to October 2021. Participants included elected local self-government members, medical and public health staff, as well as community leaders. Following written informed consent procedures, each interviewee was asked questions about whom they considered the most “vulnerable” in their areas. They were also asked if there were any special programmes/schemes to support the access of “vulnerable” groups to general and COVID related health services, as well as other needs. Recordings were transliterated into English and analysed thematically by a team of researchers using ATLAS.ti 9.1 software. Results The age range of participants was between 35 and 60 years. Vulnerability was described differentially by geography and economic context; for e.g., fisherfolk were identified in coastal areas while migrant labourers were considered as vulnerable in semi-urban areas. In the context of COVID-19, some participants reflected that everyone was vulnerable. In most cases, vulnerable groups were already beneficiaries of various government schemes within and beyond the health sector. During COVID, the government prioritized access to COVID-19 testing and vaccination among marginalized population groups like palliative care patients, the elderly, migrant labourers, as well as Scheduled Caste and Scheduled Tribes communities. Livelihood support like food kits, community kitchen, and patient transportation were provided by the LSGs to support these groups. This involved coordination between health and other departments, which may be formalised, streamlined and optimised in the future. Conclusion Health system actors and local self-government members were aware of vulnerable populations prioritized under various schemes but did not describe vulnerable groups beyond this. Emphasis was placed on the broad range of services made available to these “left behind” groups through interdepartmental and multi- Correspondence: Jaison Joseph jjoseph@georgeinstitute.org.in Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessPage 2 of 11 stakeholder collaboration. Further study (currently underway) may offer insights into how these communities – identified as vulnerable – perceive themselves, and whether/how they receive, and experience schemes designed for them. At the program level, inclusive and innovative identification and recruitment mechanisms need to be devised to identify populations who are currently left behind but may still be invisible to system actors and leaders. Keywords Vulnerable Population, Health Equity, Sex Differences, Universal Health Coverage, Primary Health Care, Health Systems, Primary Care Cost, Primary Care Utilization Introduction The core aim of the 2030 agenda of Sustainable Develop- ment Goals (SDGs) is to bring in transformation through Sustainable Development which requires nations to Leave no One behind (LNOB) [1]. Populations left behind are defined as being “at greater risk of poor health status and healthcare access, who experience significant disparities in life expectancy, access to and use of health- care services, morbidity and mortality” [2]. These popu- lations sometimes experience multiple morbidities which results in complex health care needs which are further exacerbated by intersecting deleterious social and eco- nomic conditions [2] Globally, each nation has the prerogative to define “left behind” groups or communities based on the social, economic, cultural and political factors, which in turn may vary across geographies subnationally [3]. In India, groups face vulnerability or marginalization on the basis of age, disability, socio-economic status, which in turn restricts the access of these communities to health and healthcare [4]. Groups that are officially considered vul- nerable in India according to the country’s main think tank, the NITI Aayog, include persons who are clas- sified as those in Scheduled Castes (SCs), Scheduled Tribes (STs), Other Backward Classes (OBCs), Economi- cally Backward Classes (EBCs), Religious Minorities, Nomadic, Semi-Nomadic and De-Notified Tribes (NT, SNT & DNTs), people who work in sanitation, known in Hindi as Safai karmacharis (SKs), Senior Citizens/ the elderly, Transgendered persons, Persons engaging in Substance Abuse, as well as those who are destitute and involved with begging[4–6]These population subgroups are prioritised for various government welfare schemes. Across the country, participation of under-represented groups in planning an decision-making is instituted through affirmative action: SC, ST and Other Back- ward Classes (OBCs) are provided reservations in public service. In the health domain, Below Poverty Line (BPL) house- holds are covered under Ayushman Bharat Pradhan Mantri Jan Arogya Yojana (AB-PMJAY) providing insur- ance coverage in the amount of 500,000 INR (~ 6,050 USD) per family for secondary and tertiary care hospi- talization expenditure through empanelled health care providers [7, 8]. In the Southern Indian state of Kerala, Ayushman Bharat benefits are extended to a broader beneficiary group, comprising Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA) house- holds, households of unorganized workers and additional population subgroups recognised as facing disadvantage by the state. Kerala has the lowest level of multidimensional poverty according to the NITI Aayog, which suggests that the population of “vulnerable” may be relatively lower in this setting [9]. Overall, this bears out: the state’s develop- ment pattern also indicates relatively low inequalities in health and education outcomes [10]. The state nonethe- less takes seriously the process of identifying and cater- ing to “vulnerable” population groups. It has a range of programmes for people recognised as having Scheduled Caste (SC) and Scheduled Tribe (ST) status, women, children, elderly and persons living with disabilities [11]. We identified no less than around 35 schemes and population-specific programs introduced by the state in the past half decade to support groups facing disadvan- tage: these include earmarked funds, subsidy schemes, as well as reservations in education and employment [3, 12]. Health programs have also been put in place by non-health departments and agencies. For example, the Scheduled Tribes Development Department implements many programs to address the general healthcare needs of tribal populations, which include allopathic health care institutions, medical reimbursement through hos- pitals, a tribal relief fund for emergency expenditure, assistance for sickle-cell anaemia patients, assistance to traditional tribal healers and mobile medical units [13]. One of the objectives of the Health and Family Welfare Department’s recently launched Aardram mission was to improve access of marginalized/vulnerable popula- tions to comprehensive health services [14]. The state is also implementing free health insurance scheme called “Awaz” for interstate migrant workers, covering Rs.15,000/- (~ 181.82 USD) for medical treatment per year and an amount Rs.200,000/- Lakhs (~ 2424 USD) for accident deaths [15] Although the state has several welfare measures and schemes to improve healthcare access for vulnerable groups, challenges remain. For one, impoverishment due to health is a major barrier that disproportionately affects those already facing marginalisation: such groups cannot rely on the public sector for services and end up impov- erished due to health expenditures in the private sector Joseph et al. BMC Public Health (2023) 23:748 Page 3 of 11 [16]. In fact, high Out-of-Pocket-Expenditure (OOPE) and rising health care cost for hospitalization have resulted in reducing health seeking [17]. Vulnerabilities therefore, are changing almost continuously. This makes the task of identifying vulnerable groups difficult – given the dynamic, complex, historically, and contextually con- tingent nature of vulnerability [18]. And yet, both global and national goals call for identification, responses and monitoring of outcomes in these population groups [1, 19]. As part of a larger health systems study, we placed emphasis on how vulnerability is defined in the state, and how vulnerabilities are addressed through schemes and equity-oriented reforms introduced in the state. It is important to understand the perspective of primary care health system actors on vulnerability and who are vul- nerable, as they are at the forefront of delivering essen- tial health care services and identification and catering to the needs of vulnerable population. Such an exercise has been carried out, for example in other regions with the support of the World Health Organization, [20]. as well as in other projects focused on equity integration in health programming and planning [21–23]. Barring a rare example published in 2015 [24], we were not able to identify such initiatives or studies in the Indian context, particularly ones that viewed “vulnerability” and efforts at inclusion from an implementer’s perspective. Seek- ing to fill this gap, we undertook a qualitative analysis of perspectives from Kerala’s health system actors, local self-government representatives and community leaders involved with Primary Healthcare Reforms (PHCR) in Kerala about their definitions and understandings of who is vulnerable in the state, what is being done to address their vulnerabilities, both within and outside of the con- text of COVID-19. Methods This study is the qualitative component of a larger health system research study in Kerala; our detailed methodol- ogy is reported elsewhere[25]. In summary, Kerala’s 14 districts were grouped into four categories using princi- pal components analysis, using indicators from the fourth round of the National Family Health Survey (NFHS) (2015–16) [26]. One district was randomly selected from each group, within which catchment areas served by two randomly selected primary health facilities (one recently upgraded by Aardram and one slated for later upgrada- tion) were also randomly selected. In-depth interviews (IDIs) were carried out in the four selected districts between July and October 2021. Participants for this study were staff from two primary healthcare facilities per district and elected representa- tives from their corresponding Local Self Governments (LSGs). We adopted purposive criterion sampling technique for the selection and recruitment of study participants. For the identification and selection of par- ticipants we employed a two-pronged strategy. As an ini- tial step we line-listed the potential health system actors (HSAs) and community leaders who could be part of this study. From each facility we enrolled HSAs includ- ing medical and public health staff, community leaders and Local Self Government representatives to obtain a comprehensive HSAs perception of vulnerable popula- tion their area. Medical and public health staff included, Medical Officer (MO), Staff Nurse/Nursing Officer, Health Inspector (HI), Junior Health Inspector (JHI), Public Health Nurse (PHNs), Junior Public Health Nurse (JPHNs), Palliative Care Nurse and Accredited Social Health Activists (ASHAs). Community members eligible for recruitment included Panchayat Presidents and Vice Presidents, Health Standing Committee member and Ward Members. We identified additional community leaders from these areas through the HSAs, LSG mem- bers and non-governmental organizations to capture the perspective of the community. On an average we enrolled 10 HSA per facility, a total of 83 HSAs were contacted for this study and three of them could not participate due to their busy schedule. The Institutional Ethics Committee of the George Insti- tute for Global Health (Project Number 05/2019) issued ethical approval for this study. In each facility area, in- depth interviews for this study were carried out by three researchers trained in qualitative research methods (HS, JJ & GB). The research team comprised of two male research fellows and a female research assistant and was supervised by a senior health systems researcher (DN). Administrative approval was taken from the Depart- ment of Health and Family Welfare, Government of Ker- ala. The team met the District Medical Officers (DMO) of four districts, shared the departmental permissions, outlined the study objectives, and shared findings of an earlier primary survey carried out in the same catch- ment areas. After the permissions were issued from the DMOs, the team of three researchers (HS, GB, JJ) took appointments with Medical Officers and briefed them about the study and sought their permission for conduct- ing IDIs with the staff under their institutions. Further, each of the HSAs were met in person and appointments for interviews were sought based on their convenience. As per their convenience IDIs were carried out in-person or through online platforms (i.e. Zoom). For carrying out the IDIs with LSG representatives, the team met with the panchayat presidents of the respective LSGs and briefed on the purpose of study and sought their permission to carry out the IDIs with other identified LSG mem- bers. Community leaders were contacted over phone, to brief them on the purpose of the study and as per their Joseph et al. BMC Public Health (2023) 23:748 convenience the researcher met them in person to carry out the interviews. All the participants were handed over with a hard copy of the topic guides and Participant Information Sheet (PIS) in English and Malayalam before the in-person interviews. Each participant’s signed informed consent was taken for participating in the study and for record- ing interviews. For those interviews conducted over online platforms, a soft copy of the topic guide, PIS and consent form were shared in advance with the partici- pants. Before commencing the interview, the participants shared the dully signed consent form with the research- ers. Malayalam was the medium of conversation and each of the IDIs lasted between 20 and 60 min. To obtain context and perspectives of HSAs in various capacities and geographies pertaining to each of the study sites across four districts the interviews with all the pre-set list of participants were completed even though achieving early data saturation was reached with some of the study topics. Three participants could not participate in the inter- view due to their busy schedules and after multiple failed attempts to schedule, we decided to remove them from the study. All IDIs were recorded; interview record- ings and field notes were stored and secured in a pass- word protected database after the completion of each interview and were accessible only to the research team members. Recordings were transliterated into English by a third-party agency empanelled by The George Institute for Global Health, India, which signed confidentiality agreements prior to accessing data. All the transliterated transcripts were reviewed by a three-member research team to ensure quality. Table 1 Participant characteristics Category Local Self Government Representatives Health System Actors Designation Panchayat President Panchayat Vice-President Health Standing Com- mittee Member Ward Member Community Leader Medical Officer Health Inspector (HI) Public Health Nurse (PHN) Junior Health Inspector (JHI) Junior Public Health Nurse (JPHN) Nursing Officer Palliative Nurse Community Health Worker Total Participants Female 3 0 3 0 1 5 1 4 0 11 3 1 16 48 Male 4 1 5 Total 7 1 8 1 6 3 5 7 0 0 0 0 1 7 8 6 4 7 11 3 1 16 32 80 Page 4 of 11 Transliterated transcripts were thematically analysed using ATLAS.ti 9 software by a four-member research team (DN, HS, JJ, GB). An inductive approach was used: the thematic structure and code book were finalized after multiple discussions among the four-member team. Finally, the coded manuscripts from the team members were merged using ATLAS.ti 9 software. Codes of inter- est for this analysis were indexed and themes consoli- dated based on further discussions and core questions of interest (i.e., who is left behind? How are they reached? and impact of COVID-19 among those left behind). A narrative was then constructed around these questions and compiled by the lead author with inputs, edits, and review by other authors. Results Participant characteristics Data for a total of 80 participants was included in the study, of which more than half (60%) were women (see Table 1). From this group of participants, we received information on who they considered was being left behind from health programming in Kerala, as well as what was being done to support them and/or address their needs (in general, and in the COVID context). Who is left behind? Participants in all districts would often first identify Scheduled Caste and Scheduled Tribe communities as vulnerable; these are nationally established catego- ries defined as facing vulnerability. Apart from this, we observed geographical variation across districts in who was described as vulnerable population by stakeholders (see Table 2). Migrant labourers were identified as vul- nerable in the semi-urban areas, while fisherfolk in the coastal areas (inland and seafaring). It was found that most of the places where the vulner- able population were identified, faced challenges related to living and working conditions - social determinants of health like sanitation, nutrition, crowding/housing were raised. According to a Medical Officer, …there is the SC/ST community- they have colo- nies1here… they have drinking water issues, food issues, improper waste management, and crowded places. It is a dengue hotspot and communicable diseases (hotspot). Also, COVID is a big issue there, 1 While system actors often mentioned colonies of SC and ST communi- ties, in subsequent fieldwork, SC communities in particular felt offended by the label of “colony” used to describe their places of residence. This could be seen as being akin to what Wacquant has called “territorial stigma,” which automatically assigns ignominy to a geographic category.(27) Although Wacquant’s theorization referred to the urban context in Chicago and Paris alone, we saw resonance of the concept for urban and rural residents of “col- onies.” The concept of the “colony,” of course, has other problematic histories and legacies. Joseph et al. BMC Public Health (2023) 23:748 Page 5 of 11 This view was held by another JPHN as well who took the view that There are no marginalised communities in my area. All the people here are from similar backgrounds since it is a coastal area. I do not know if they have any issues. Most of the people over there depend on their daily income and even when they must undergo quarantine, the authorities have delivered them essential commodities and resolved the prob- lems that came up. So, there were no issues, all such troubles were taken care of. Programs to support those left behind We found that schemes and programmes targeting vul- nerable populations were being implemented across the state in most cases. The possible exception we found was the case of fisherfolk and farmers, who were defined as vulnerable, but were not described as being covered by many government health schemes. Recently imple- mented primary health care reforms had reportedly improved access to healthcare for vulnerable groups in some areas. In many cases this involved interdepartmen- tal coordination. A Panchayat president took the follow- ing view: Our Family Health Centre works from 7 AM till 8 PM even now. The service of a gynaecology specialist is provided twice a week. Then, we have an eye spe- cialist. We have been getting the services of a phys- iotherapy specialist. People from the rural areas, including the Adivasi community, were able to ben- efit from these changes. The Tribal Department has been conducting camps in the places where Adivasis [tribal persons] live According to a Health Inspector, there was empha- sis placed on going to where communities were to offer them care/support and the role of labour department and private employers in health service delivery: We have a lot of migrants around here. The labour office is holding special camps for them. Their employers also sometimes book slots in bulk and get the workers vaccinated. As far as we are concerned, we go to their companies and conduct tests and pro- vide other services there. We also found that joint programs implemented by LSGs and the Department of Social Justice, such as the Table 2 Vulnerable Population Identified by Participants across Districts Thiruvananthapuram Kollam Alappuzha X X X People from Sched- uled Tribe People from Sched- uled Caste Pal- liative Care pa- tients Fisher- folk Farm- ers Mi- grants X X X X X (inland) X (seafaring) X X Kasara- god X X X X X because if it affects one person, the spread will be too much…because even the children run around and enter all the houses. We also found that climate change (subsequent floods in the state) and COVID-19 pandemic had affected popu- lation subgroups and added to their vulnerability. Farm workers were affected by the consequent floods in the state and fisherfolk were affected by the COVID-19 pan- demic. One Community Leader noted this: …Especially when there were floods, farm work- ers were there…. the one who is mostly engaged with paddy fields. Last financial year was a time when the yield was maximum but there was a technical difficulty in harvesting it. During such a situation, the farmers had to face a lot of trouble.People turn out to be marginalised when they cannot har- vest their crop. The situation is similar in the case of fisheries as well. Due to COVID, they could not go fishing for several days. Even if they went, there was a situation that people turned COVID posi- tive because there were about 40 people in a fishing boat... On the other hand, a few people we spoke to also men- tioned that nobody was vulnerable, because the needs of all were catered to, as per need. A Junior Public Health Nurse said: “I don’t think such a marginalised community exists anymore in this era. We all are equal. I do not think any community is being sidelined nowadays.” Joseph et al. BMC Public Health (2023) 23:748 Page 6 of 11 Kudumbasree2-self help program for women, as well as programs focussing on the elderly population, migrants, destitute and palliative care patients were intended to increase access to healthcare and to improve quality of life for groups facing these forms of disadvantage. A Health Standing Committee Member added: …for palliative patients, we provide support from Panchayat and the FHC. Other than this, we have a scheme called Ashraya for the destitute. We pro- vide them with kits through Kudumbasree. We have another scheme called Santhwanam. Under this, through Kudumbasree we conduct an event once a year. Ashraya scheme falls under the ambit of this one. Ashraya is for people with no means of support. According to a Community Health worker, the Panchayat placed emphasis on palliation and also on the health and welfare of guest or migrant workers: Yes, Panchayat provides it. Even medicines and hospital-related services are arranged by the Pan- chayat. Similarly, the Panchayat has appointed a nurse for palliative care. We visit their homes along with the palliative nurse and provide all possible services to them. If any guest workers come here, we treat them like our own people, and both the Pan- chayat and the FHC provide them with all kinds of assistance. This was corroborated by a Panchayat President in another district as well: We have proper facilities for ensuring the health of people including migrant labourers. …. Grama Pan- chayat has facilitated the treatment for numerous cancer patients in the area as well as for those with other related diseases. The area has around 250 pal- liative patients. We have implemented various pro- grams for helping all such patients. There was seen to be, therefore, responsibility taken by local leaders for vulnerable groups and the idea that these were “our own people,” whose needs related to health and beyond, were given due attention. COVID Outreach for vulnerable populations Many study participants felt that during the COVID-19 pandemic and consequent lockdowns, vulnerable popu- lations were prioritised. Various health service design 2 Kudumbashree is the poverty eradication and women empowerment pro- gramme implemented by the State Poverty Eradication Mission (SPEM) of the Government of Kerala.[28]. More information is available at: https:// www.kudumbashree.org. changes were described as being introduced to ensure the delivery of essential health care and related services under the stewardship of LSGs. A Junior Public Health Nurse described them as follows: We used to provide food to these side-lined people from the community kitchen, and provide medicines from our Tele-OP [out-patient services], when the first wave of COVID started. When COVID started and there were strict lockdowns, from the side of the health department, every day there was one or two vehicles that were arranged from the side of LSGD and in that vehicle, our staff would take details from each area of the positive cases, and create a calcu- lation on how many of them need medicine, and how many homes we need to put a sticker etc, and both these vehicles would cover two different areas without overlapping and delivered, medicine kit is, NCD medicines and Tele OP medicines everywhere promptly. Another Panchayat President noted the greater risks of exposure in certain populations and how they were pri- oritised commensurably, saying that “we have distributed kits in every ward. Due to COVID and lockdown, people were not able to go outside so we distributed kits to every- one. We especially distributed masks and sanitisers in the S[cheduled] T[ribe] colonies and other marginalised colo- nies. Because they were residing in a densely populated area and there is a high chance of spreading, we provided the kits.” A Nursing Officer also noted the role played by pan- chayat leaders in mobilising support during lockdowns, “when migrants could not go back to their homes, vol- unteers intervened and helped them. Whatever needed, from food to shelter was provided from the side of the Panchayat.” Vulnerable populations were prioritized for receiv- ing COVID-19 vaccinations. There were efforts from the health systems and LSGs to deliver vaccines at the door- steps of these population. A community health worker described how separate, priority vaccination drives were held for fisherfolk, SC and ST groups. She said simply: “They were given more preference.” A Medical Officer noted that in their area, SC, ST, persons living with dis- abilities and migrants were the first to achieve complete vaccination. This was echoed by a frontline worker in another district who noted that Bedridden patients were given vaccination doses at their houses. Palliative patients were given the vaccination at their places. We have also vac- cinated people above 80 years of age after visiting their houses. We visited the houses of those who Joseph et al. BMC Public Health (2023) 23:748 Page 7 of 11 cannot come and got them inoculated. We also con- duct health camps in colonies. A class on vaccina- tion programs was also given for them and all these were organised by the PHC. Discussion Our study sought to identify who was defined as vul- nerable by health system and LSG actors in the state of Kerala and what schemes and arrangements were in place to address their health issues. In the current study, we observed that a number of groups identified at the national level as vulnerable were also identified by our study participants, alongside other population groups that were uniquely identified in Kerala. This is consis- tent with the findings of Kerala State Poverty Eradication Plan presented to NITI Aayog, which reported that SC populations were concentrated in colonies (including in urban areas), ST populations continued to be sequestered in remote and rural locations, consistent with nationally identified groups in need [29]. However, this report also indicated the need to support coastal populations like fisherfolk who for economic reasons were also confined to particular, hard to reach geographies [29]. Decen- tralized planning in Kerala has helped keep the issue of inclusion and marginalisation on the agenda of decision- makers and implementers, even as newer groups facing vulnerability were being identified, like migrant workers [11]. Migrant workers also faced confinement in their work settings, while palliative care patients were confined due to their health situation. This distance – physical or social – was a defining feature of vulnerability from the perspective of these supply side actors. This kind of a dis- tance based vulnerability has been found in a national studies from Uttar Pradesh, Madhya Pradesh, Bihar Assam and Jharkhand during pre and post COVID-19 periods [30], although the view of health system actors or decision-makers on this was not specifically indicated in the literature. Other studies in LMICs have identi- fied vulnerability on the basis of racial, ethnic and gen- der minoritization, economically disadvantage, having chronic health issues, as well as those at extremes of age [1, 31, 32] It was also observed that it was not merely in the con- text of health, but the larger social determinants that vul- nerable populations were “hard to reach.” The residential areas of the marginalized population were underdevel- oped: providing quality health service delivery remained challenging without addressing the social determinants of health. This is consistent with the findings of the 6th Kerala Administrative Reforms Commission report (2020) which noted lack of land, improper housing, inad- equate infrastructure, poor quality of education, lack of sanitation services and unsafe drinking water among the marginalized population [33]. This report also gave spe- cial emphasis on the condition of SC and other “back- ward” communities who continue to live and work in highly dangerous and pathogenic conditions [33]. It has been deemed vital to address social determinants among the marginalized to improve their health status as they are important factor in management and prevention of communicable and non-communicable diseases alike [34]. Studies conducted in LMICs have reported lower access to safe drinking water, sanitation, and hygiene (WASH), conditions which are fundamental to living and working, are both reflective of vulnerability and are what drive disparities in health burdens, health seeking, and health outcomes [35–37] We found that natural disasters (floods) and COVID-19 pandemic added to the vulnerabilities faced by farmers and fisherfolk, suggesting that vulnerability is not a static phenomenon. A study conducted by a panel of experts in Kerala immediately after the 2018 floods reported that the vulnerable population who were the victims of floods lagged behind their peer groups in levels of human development, in part because they faced differential and layered exposures and vulnerabilities compared to other groups [38]. Another study by the Palliative Care Con- sortium on the effect of 2018 floods on elderly living alone found serious after effects of the disaster especially among the elderly women, also the palliative care ser- vices and medications were disrupted [39]. COVID-19 lockdowns imposed by the Government during the first wave (2020) affected the coastal community in the state in accessing healthcare and in resourcing the essential commodities. Along with it the declaration of some of the overcrowded coastal regions as containment zones, with restriction of movement leading to reduced work- ing hours and income further increased their vulner- ability [40]. A study conducted by Kattungi et al. (2020) assessing the impact of COVID-19 on the livelihood of fishermen in Puducherry found loss of employment among many fishermen which has resulted in increas- ing inequities and poverty [41]. Aura CM et al. (2020), in their study which assesses the consequences of flood- ing and COVID-19 Pandemic among inland fisherfolk in Kenya in East Africa, found that natural calamities and pandemic affected the livelihood of fisherfolk, reduced fishing time and trips, decline in consumables such as boat fuel resulting low fish catches etc [42, 43]. COVID- 19 has negatively affected small scale farmers in LMICs which resulted in low production, low income and higher food insecurity which has increased their vulnerability [44, 45] There has been a fairly high degree of multisectoral action and coordination to reaching the “vulnerable” in Kerala. We found a fascinating convergence in the views of those who identified vulnerable groups and those Joseph et al. BMC Public Health (2023) 23:748 Page 8 of 11 who did not. Both noted that schemes existed and that vulnerable groups (or everyone!) were taken care of the state through schemes implemented by government departments. This includes multisectoral action led by the State government in prevention and control of Non- communicable Diseases (NCDs) [46, 47], convergence to support awareness of and enrolment in the Depart- ment of Labour’s health insurance scheme (supported greatly by LSG leaders and Kudumbasree mission work- ers under Department of Social Justice), [48]. as well as other schemes introduced by the Kerala Social Security Mission [49–51] The state’s response in handling the COVID-19 pan- demic was another example of multi-sectoral coordina- tion backed by decentralized governance, along with whole of society approaches where community action complemented the work of health system actors [52, 53]. During COVID-19, a community kitchen initia- tive was introduced through LSGs with the support of Kudumbasree, which provided free meals to labourers, people who were under quarantine, the destitute and other needy marginalized population [54]. Grassroots agencies were also involved with delivering free food kits universally, which required a special focus on vul- nerable population typically excluded from social secu- rity benefit programmes like transgender persons [53]. In a scoping review by Hasan et al. (2021) about the response of LMICs in management of COVID-19 found that decentralized governance coupled with stewardship and multisectoral collaboration facilitated the delivery of integrated health service delivery[55] ,which was found through our study in Kerala. Another interesting feature in Kerala was seen dur- ing COVID-19 in the context of vaccination. Initially COVID-19 vaccination in Kerala followed global norms by prioritising health workers followed by frontline work- ers [56], then national norms prioritising citizens above the age of 60 years and citizens aged between 45 and 59 with specified comorbidities [57]. However, by April 2021 Kerala created state specific norms by way of 32 prior- ity categories in the age group of 18–45 which included other frontline workers, seafarers, field staff, teachers, students and more [58]. This demonstrates the possibility of defining and redefining those in need in the context of a crisis. It is less clear, however, if such prioritization of populations in need could be done on an ongoing basis, helping the state to identify those who may face unique disadvantages and may need to be reached by program- ming beyond the existing ambit. This is a clear area for further research. Beyond this, there are other areas warranting further research: greater attention to how multi-sectoral policy processes for the “vulnerable” take place, in what con- texts, could offer lessons for their replication in other contexts, and also for their enhancement in Kerala. Moreover, it is unclear, at present, how intersections of vulnerability may be addressed in current programming, for e.g. SC or ST populations receiving palliative care, women involved with the fishing industry. Whether or not such programs are catering to these intersectional needs would be a critical area for future policymaking. Finally, there is a very little understanding of those fac- ing vulnerability as being more than “target populations” or “beneficiaries” of services. Other research on UHC has shown that just producing interventions and consid- ering communities passive recipients can easily alienate and exclude them from health reform processes[59]. Fur- ther study is needed – across all these and more groups facing vulnerability – on how they perceive themselves, and how they receive, and experience schemes designed for them, and in the absence of such schemes, how they manage their health and related needs. This would have to be given more attention in research and policymaking and is a limitation in the framing of our study as well. Limitations This analysis is based on the perceptions of government health system actors. It therefore does not include the perceptions of the general population as well as those who constitute “those left behind.” Research is currently underway to understand the care seeking experiences of these, “demand side” actors and is a crucial part of our understanding of vulnerability. Conclusion Our analysis sought to understand supply side perspec- tives in the health sector on who is left behind in the southern Indian state of Kerala. Health system actors and local self-government members were aware of vulnerable population prioritized under various schemes but did not describe vulnerable groups beyond this. Emphasis was placed on the broad range of services available to these “left behind” groups. Further study (currently underway) may offer insights into how these communities – identi- fied as vulnerable – perceive themselves, and how they receive, and experience schemes designed for them. Innovative sampling and recruitment mechanisms need to be devised to identify populations who are currently left behind but may also be invisible to system actors and leaders. While the Kerala government has shown initiative in carrying out a mapping of poorest households in the state, there are other critical forms of vulnerability that affect residents in the state; continuous monitoring of “who is being left behind,“ in partnership with academic and civil society institutions, could help enhance such initiatives. Joseph et al. BMC Public Health (2023) 23:748 List of abbreviations SDGs LNOB SC STs OBCs EBCs SKs BPL AB-PMJAY MGNREGA OOPE PHCR IDIs HSAs FHC LSG MO HI JHI PHN JPHN ASHAs PIS Sustainable Development Goals Leave No One Behind Schedule Caste Schedule Tribes Other Backward Castes Economically Backward Castes Safai Karmacharis Below Poverty Line Ayushman Bharat Pradhan Mantri Jan Arogya Yojana Mahatma Gandhi National Rural Employment Guarantee Act Out-of-Pocket Expenditure Primary Health Care Reform In-depth Interviews Health System Actors Family Health Centre Local Self-Government Medical Officer Health Inspector Junior Health Inspector Public Health Nurse Junior Public Health Nurse Accredited Social Health Activists Participation Information Sheet Acknowledgements We are grateful to Mr. Santosh Sharma, Research Fellow, The George Institute for Global Health, India, for his key reflections and critical inputs. Author contributions Conceptualization: JJ Methodology: JJ, HS, DN Formal analysis and investigation: JJ, GB Writing - original draft preparation: JJ, HS, GB Writing - review and editing: JJ, HS, GB, DN Funding acquisition: DN Supervision: DN. Funding We wish to indicate that this work was supported by the Wellcome Trust/DBT India Alliance Fellowship(https://www.indiaalliance.org) Grant number IA/ CPHI/16/1/502653) awarded to Dr. Devaki Nambiar. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The funder provided support in the form of salaries and research materials and field work support for authors DN, HS, GB and JJ but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section. Data availability All datasets used for supporting the conclusions of this paper are available from the corresponding author on request. Declarations Ethics approval of the study was received from the institutional ethics committee of George Institute for Global Health (Project Number 05/2019). All participants gave written informed consent before taking part in the study including Illiterate participants in the survey who were read out and explained the consent form in the local language. Thereafter, they were able to sign their names. The ethics committee that approved the study also approved this procedure of obtaining written informed consent from these participants. All methods were carried out in accordance with relevant guidelines and regulations. Consent to publish Not applicable. Competing interests The authors declare no competing interests. Author details 1The George Institute for Global Health, New Delhi, India 2Faculty of Medicine, University of New South Wales, Sydney, Australia Page 9 of 11 3Prasanna School of Public Health, Manipal Academy of Higher Education, Manipal, India Received: 6 September 2022 / Accepted: 7 April 2023 References 1. United Nations Sustainable Development Group. Leave No One Behind [Internet]. 2022 [cited 2022 Jun 9]. Available from: https://unsdg. un.org/2030-agenda/universal-values/leave-no-one-behind 3. 2. No authors listed. Vulnerable Populations: Who Are They? The American Jour- nal of Managed Care [Internet]. 2006 Nov 1 [cited 2022 May 30]; Available from: https://www.ajmc.com/view/nov06-2390ps348-s352 Balan PP, George S, Kunhikannan TP, Marginalisation. and Deprivation Studies in Multiple Vulnerabilities [Internet]. Thrissur, Kerala: Kerala Institute of Local Administration; 2016 [cited 2022 Jun 1]. Available from: http://dspace.kila. ac.in:8080/jspui/bitstream/123456789/276/1/Marginalisation%20and%20 Deprivation%20Studies%20in%20Multiple%20Vulnerabilities.pdf Agarwal M. Vulnerable Groups in India: Status, Schemes, Constitution of India [Internet]. 2022 [cited 2022 May 30]. Available from: http://www.legalser- vicesindia.com/article/1079/Vulnerable-Groups-in-IndiaStatus,-Schemes,- Constitution-of-India.html Aayog N. Role of Social Justice & Empowerment Division [Internet]. NITI Aayog; 2019 [cited 2022 Jul 9]. Available from: https://www.niti.gov.in/sites/ default/files/2019-08/Role-of-SJE-Division-in-NITI.pdf 4. 5. 6. George AA, Mechanisms. Laws, Institutions and Bodies for Vulnerable Sec- 7. tions [Internet]. Clear IAS. 2020 [cited 2022 Jul 9]. Available from: https://www. clearias.com/mechanisms-laws-institutions-bodies-vulnerable-sections/ Karunya Arogya Suraksha Padahathi, State Health Agency Kerala. Schedules for Service Contract of AB-PMJAY-KASP [Internet]. Department of Health and Family Welfare, Government of Kerala; 2020 [cited 2022 Jun 9]. Available from: https://arogyakeralam.gov.in/wp-content/uploads/2020/03/Schedules.pdf 8. Government of India. DPE-GM-15/0001/2016-GM-FTS-5921, Brochure on Reservation for SCs/STs and OBCs [Internet]. Ministry of Heavy Industries & Public Enterprises; 2016 [cited 2022 Jun 9]. Available from: https://dpe.gov.in/ sites/default/files/Reservation_Brochure-2.pdf Board SP. Govt. of Kerala. Economic Review 2021 [Internet]. Thiruvanantha- puram; 2021 [cited 2022 Jun 9]. Available from: https://spb.kerala.gov.in/sites/ default/files/2022-03/ECNO_%20ENG_21_%20Vol_1.pdf 9. 10. Ghosh M. Regional Disparities in Education, Health and Human Develop- ment in India. Indian Journal of Human Development [Internet]. 2011 Jan [cited 2022 Sep 6];5. Available from: https://www.researchgate.net/publica- tion/302964390_Regional_Disparities_in_Education_Health_and_Human_ Development_in_India 11. Marginalisation. and Deprivation Studies in Multiple Vulnerabilities.pdf [Inter- net]. [cited 2022 Jun 1]. Available from: http://dspace.kila.ac.in:8080/jspui/ bitstream/123456789/276/1/Marginalisation%20and%20Deprivation%20 Studies%20in%20Multiple%20Vulnerabilities.pdf 12. Chatterjee CB, Sheoran G. Centre for Enquiry into Health & Allied Themes (Mumbai I. vulnerable groups in India. Mumbai: Centre for Enquiry into Health and Allied Themes; 2007. 13. Scheduled Tribes Development Department., Govt. of Kerala. Healthcare Schemes [Internet]. 2022 [cited 2022 Jun 9]. Available from: https://www. stdd.kerala.gov.in/healthcare-schemes 14. Aarogyakeralam National Health Mission. Aardram [Internet]. 2018 [cited 2022 Jun 9]. Available from: https://arogyakeralam.gov.in/2020/04/01/ aardram/ 15. Staff Reporter. 5.13 lakh migrant workers registered under Awaz scheme. The Hindu [Internet]. 2022 Mar [cited 2022 Jun 10]; Available from: https://www. thehindu.com/news/national/kerala/513-lakh-migrant-workers-registered- under-awaz-scheme/article65223711.ece 16. Haddad S, Baris E, Narayana D. Safeguarding the Health Sector in Times of Macroeconomic Instability: Policy Lessons for Low- and Middle-Income Countries [Internet]. African World Press & International Development Research Centre; 2008 [cited 2022 May 30]. Available from: https://www.idrc. ca/en/book/safeguarding-health-sector-times-macroeconomic-instability- policy-lessons-low-and-middle Joseph et al. BMC Public Health (2023) 23:748 Page 10 of 11 17. Mukherjee S, Haddad S. Social class related inequalities in household health expenditure and economic burden: evidence from Kerala, South India. International journal for equity in health. 2011 Jan7;10:1. 18. Gallardo M. Identifying vulnerability to poverty: a critical survey. J Economic Surveys. 2018;32(4):1074–105. 19. Ministry of Health and Family Welfare, Government of India. National Health Policy India [Internet]. New Delhi; 2017 [cited 2022 Aug 4]. Available from: https://www.nhp.gov.in/nhpfiles/national_health_policy_2017.pdf 20. Solar O, Valentine N, Castedo A, Brandt GS, Sathyandran J, Ahmed Z et al. Action on the social determinants for advancing health equity in the time of COVID-19: perspectives of actors engaged in a WHO Special Initiative. International Journal for Equity in Health [Internet]. 2023 Jan 24 [cited 2023 Feb 8];21(3):193. Available from: https://doi.org/10.1186/s12939-022-01798-y 21. Brixi H, Mu Y, Targa B, Hipgrave D. Engaging sub-national governments in addressing health equities: challenges and opportunities in China’s health system reform. Health Policy and Planning [Internet]. 2013 Dec 1 [cited 2023 Feb 8];28(8):809–24. Available from: https://doi.org/10.1093/heapol/czs120 22. Chanchien Parajón L, Hinshaw J, Sanchez V, Minkler M, Wallerstein N. Practicing Hope: Enhancing Empowerment in Primary Health Care through Community-based Participatory Research. American Journal of Community Psychology [Internet]. 2021 [cited 2023 Feb 8];67(3–4):297–311. Available from: https://onlinelibrary.wiley.com/doi/abs/https://doi.org/10.1002/ ajcp.12526 23. Chin MH, King PT, Jones RG, Jones B, Ameratunga SN, Muramatsu N, et al. Lessons for achieving health equity comparing Aotearoa/New Zealand and the United States. Health Policy. 2018 Aug;122(8):837–53. 24. Nambiar D, Muralidharan A, Garg S, Daruwalla N, Ganesan P. Analysing implementer narratives on addressing health inequity through convergent action on the social determinants of health in India. Int J Equity Health [Internet]. 2015 Dec [cited 2021 Aug 2];14(1):133. Available from: http://www. equityhealthj.com/content/14/1/133 25. Nambiar D, Sankar H, Negi J, Nair A, Sadanandan R. Field-testing of primary health-care indicators, India. Bull World Health Organ. 2020 Nov;98(1):747–53. 26. Ministry of Health and Family Welfare, Govt. of India. National Family Health Survey (NFHS-4) [Internet]. IIPS Mumbai. ; 2015 [cited 2023 Feb 9]. Available from: http://rchiips.org/nfhs/factsheet_nfhs-4.shtml Middle-Income Countries during the COVID-19 Pandemic through a Gen- dered Lens. Int J Environ Res Public Health. 2022 Sep 20;19(19):11845. 38. Thummarukudy M, Peter B. Leaving no one behind: Lessons from the Kerala disasters - India [Internet]. Perumbavoor, Kerala: Centre for Migration and Inclusive Development; 2019 Oct [cited 2022 Jun 10]. Available from: https:// reliefweb.int/report/india/leaving-no-one-behind-lessons-kerala-disasters 39. Kerala State Disaster Management Authority, India P, Indian Associa- tion of Palliative Care. Addressing the Needs of Elderly and People Under Palliative Care in Disaster Management [Internet]. Trivandrum; 2019 Jun [cited 2022 Jun 10]. Available from: https://sdma.kerala.gov.in/wp-content/ uploads/2020/11/Report-Palliative-Care-Consultation-cum-Workshop.pdf 40. Sumitha TS, Thelly AS, Medona B, Lijimol AS, Rose MJ, Rajagopal MR. Response to COVID-19 Crisis with Facilitated Community Partnership among a Vulnerable Population in Kerala, India - A Short Report. Indian J Palliat Care. 2022;28(1):115–9. 41. Sekhar KV, Sutha DAI, Devi DRU, IMPACT OF COVID-19 ON THE LIVELIHOODS OF FISHERMEN COMMUNITY IN YANAM., PUDUCHERRY: AN ANALYSIS. Euro- pean Journal of Molecular & Clinical Medicine. 2020 Nov 30;7(8):869–80. 42. Aura CM, Nyamweya CS, Odoli CO, Owiti H, Njiru JM, Otuo PW, et al. Consequences of calamities and their management: the case of COVID-19 pandemic and flooding on inland capture fisheries in Kenya. J Great Lakes Res. 2020 Dec;46(6):1767–75. 43. Love DC, Allison EH, Asche F, Belton B, Cottrell RS, Froehlich HE, et al. Emerg- ing COVID-19 impacts, responses, and lessons for building resilience in the seafood system. Glob Food Sect. 2021 Mar;28:100494. 44. Löhr K, Mugabe P, Turetta APD, Steinke J, Lozano C, Bonatti M, et al. Assessing impacts of COVID-19 and their responses among smallholder farmers in Brazil, Madagascar and Tanzania. Outlook Agric. 2022 Dec;51(4):460–9. 45. Hammond J, Milner D, van Wijk M. How COVID controls hit farm- ers in 7 low-income countries, most in Africa [Internet]. 2022 [cited 2023 Mar 30]. Available from: https://www.preventionweb.net/news/ how-covid-controls-hit-farmers-7-low-income-countries-most-africa 46. Sen S, Gautam P, Kerala Wins UN, Award For. “Outstanding Contribution” Towards Control of Non-Communicable Diseases. The Logical Indian Crew [Internet]. 2020 Sep [cited 2022 Jul 29]; Available from: https://thelogicalin- dian.com/uplifting/kerala-un-award-23980 27. Wacquant L, Slater T, Pereira VB. Territorial Stigmatization in Action. Environ 47. Arora M, Chauhan K, John S, Mukhopadhyay A. Multi-Sectoral Action for Plan A [Internet]. 2014 Jun 1 [cited 2023 Mar 31];46(6):1270–80. Available from: https://doi.org/10.1068/a4606ge 28. State Poverty Eradication Mission, Government of Kerala. Kudumbashree \| What is Kudumbashree [Internet]. 2023 [cited 2023 Mar 31]. Available from: https://www.kudumbashree.org/pages/171 29. Kerala State Planning Board. Kerala State Poverty Eradication Plan [Internet]. 2018 [cited 2022 Jun 10]. Available from: https://www.niti.gov.in/writeread- data/files/Kerala%20presentation.pdf 30. Bhattacharya M, Banerjee P. COVID-19: Indices of economic and health vulnerability for the Indian states. Social Sciences & Humanities Open. 2021 Jan 1;4(1):100157. 31. 5 Vulnerable Populations in Healthcare [Internet]. AJMC. 2018 [cited 2023 Mar 30]. Available from: https://www.ajmc.com/ view/5-vulnerable-populations-in-healthcare 32. Peters DH, Garg A, Bloom G, Walker DG, Brieger WR, Hafizur Rahman M. Poverty and Access to Health Care in developing countries. Ann N Y Acad Sci. 2008;1136(1):161–71. 33. Administrative Reforms Commission, Government of Kerala. Welfare to Rights II, Implementation of Select Legislations, Review Phase II [Internet]. Kerala; 2020 Oct [cited 2022 Jun 10]. Report No.: 6. Available from: https://arc.kerala. gov.in/sites/default/files/inline-files/Welfare%20to%20Rights%202%20-%20 File%20Copy%20-%2001-01-2020.pdf 34. Marmot M. Social determinants of health inequalities. The Lancet. 2005 Mar;19(9464):1099–104. 35. Hutton G, Chase C. Water Supply, Sanitation, and Hygiene. In: Mock CN, Nugent R, Kobusingye O, Smith KR, editors. Injury Prevention and Environ- mental Health [Internet]. 3rd ed. Washington (DC): The International Bank for Reconstruction and Development / The World Bank; 2017 [cited 2023 Mar 30]. Available from: http://www.ncbi.nlm.nih.gov/books/NBK525207/ 36. Prüss-Ustün A, Wolf J, Bartram J, Clasen T, Cumming O, Freeman MC, et al. Burden of disease from inadequate water, sanitation and hygiene for selected adverse health outcomes: an updated analysis with a focus on low- and middle-income countries. Int J Hyg Environ Health. 2019 Jun;222(5):765–77. 37. Sahoo KC, Dubey S, Dash GC, Sahoo RK, Sahay MR, Negi S et al. A System- atic Review of Water, Sanitation, and Hygiene for Urban Poor in Low- and 48. Addressing Social Determinants of Noncommunicable Diseases and Main- streaming Health Promotion in National Health Programmes in India. Indian Journal of Community Medicine. 2011 Dec 1;36(5):43. Joy J. The Impact of RSBY-CHIS on utilisation of Healthcare Services in Kerala [Internet]. Cochin: SCMS Cochin School of Business; 2019 [cited 2022 Jul 29]. Available from: https://spb.kerala.gov.in/sites/default/files/inline-files/RSBY- CHIS.pdf 49. Kerala Social Security Mission., Govt. of Kerala. Schemes [Internet]. 2022 [cited 2022 Jul 9]. Available from: https://socialsecuritymission.gov.in/scheme_info. php?id=2 50. Kerala Social Security Mission., Govt. of Kerala. Samashwasam [Internet]. 2020 [cited 2022 Jul 9]. Available from: https://socialsecuritymission.gov.in/ scheme_info.php?id=MTk= 51. Madhavan M, Ravindran RM, Shinu KS. Functioning of Arogyakiranam programme in Kerala: a qualitative study. J Family Med Prim Care. 2021 Nov;10(11):4117–23. 52. Rahim AA, Chacko TV. Replicating the Kerala State’s successful COVID-19 Containment Model: insights on what worked. Indian J Community Med. 2020;45(3):261–5. 53. Solomon N, K A G, Krishnan A, Cicily Joseph A, Dhanuraj D. Kerala Model of Respone to COVID-19 [Internet]. Kochi: Centre for Public Policy Research (CPPR); 2020 [cited 2022 Jun 10]. Available from: https://www.cppr.in/wp- content/uploads/2020/10/KERALA-MODEL-OF-RESPONSE-TO-COVID-19.pdf 54. World Health Organization. Responding to COVID-19 - Learn- ings from Kerala [Internet]. 2020 [cited 2022 Jun 10]. Available from: https://www.who.int/india/news/feature-stories/detail/ responding-to-covid-19---learnings-from-kerala 55. Hasan MZ, Neill R, Das P, Venugopal V, Arora D, Bishai D et al. Integrated health service delivery during COVID-19: a scoping review of published evidence from low-income and lower-middle-income countries. BMJ Global Health. 2021 Jun 1;6(6):e005667. IANS. Kerala vaccination drive: Focus now on frontline workers in 2nd phase. Business Standard India [Internet]. 2021 Feb 11 [cited 2022 Jun 10]; Available from: https://www.business-standard.com/article/current-affairs/ 56. Joseph et al. BMC Public Health (2023) 23:748 Page 11 of 11 kerala-vaccination-drive-focus-now-on-frontline-workers-in-2nd- phase-121021100593_1.html 57. Health and Family Welfare Department. Guidelines for COVID-19 Vaccination for the priority groups [Internet]. Government of Kerala; 2021 [cited 2022 Jun 10]. Available from: https://dhs.kerala.gov.in/wp-content/uploads/2021/03/ Guideline-for-COVID19-Vaccination-for-the-priority-groups_compressed.pdf 58. Health and Family Welfare Department. Government of, Kerala GO. (Rt) No. 1114/2021/H&FWD, Health & Family Welfare Department: Prioritization for Vaccination in the age group of 18–45 years, Modified Orders issued [Inter- net]. 2021 [cited 2022 Jun 10]. Available from: https://arogyakeralam.gov.in/ wp-content/uploads/2020/03/Prioritization-of-vaccinnation-in-18-45.pdf 59. George MS, Davey R, Mohanty I, Upton P. “Everything is provided free, but they are still hesitant to access healthcare services”: why does the indigenous community in Attapadi, Kerala continue to experience poor access to healthcare? International Journal for Equity in Health [Internet]. 2020 Jun 26 [cited 2022 Jul 22];19(1):105. Available from: https://doi.org/10.1186/ s12939-020-01216-1 Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Joseph et al. BMC Public Health (2023) 23:748
10.1186_s12889-023-15561-7	Zheng et al. BMC Public Health (2023) 23:749 https://doi.org/10.1186/s12889-023-15561-7 BMC Public Health Dissecting the causal relationship between household income status and genetic susceptibility to cardiovascular-related diseases: Insights from bidirectional mendelian randomization study Xifeng Zheng1†, Yu Yang2†, Jianying Chen1 and Bing Lu2* Abstract Objectives Observational studies have revealed that socioeconomic status is associated with cardiovascular health. However, the potential causal effect remains unclear. Hence, we aimed to investigate the causal relationship between household income status and genetic susceptibility to cardiovascular-related diseases using a bidirectional Mendelian randomization (MR) study. Methods An MR study based on a large-sample cohort of the European population from a publicly available genome-wide association study datasets was conducted using a random-effects inverse-variance weighting model as the main standard. Simultaneously, MR-Egger regression, weighted median, and maximum likelihood estimation were used as supplements. Sensitivity analysis, consisting of a heterogeneity test and horizontal pleiotropy test, was performed using Cochran’s Q, MR-Egger intercept, and MR-PRESSO tests to ensure the reliability of the conclusion. Results The results suggested that higher household income tended to lower the risk of genetic susceptibility to myocardial infarction (OR: 0.503, 95% CI = 0.405–0.625, P < 0.001), hypertension (OR: 0.667, 95% CI = 0.522–0.851, P = 0.001), coronary artery disease (OR: 0.674, 95% CI = 0.509–0.893, P = 0.005), type 2 diabetes (OR: 0.642, 95% CI = 0.464–0.889, P = 0.007), heart failure (OR: 0.825, 95% CI = 0.709–0.960, P = 0.013), and ischemic stroke (OR: 0.801, 95% CI = 0.662–0.968, P = 0.022). In contrast, no association was evident with atrial fibrillation (OR: 0.970, 95% CI = 0.767–1.226, P = 0.798). The reverse MR study suggested a potentially negative trend between heart failure and household income status. A sensitivity analysis verified the reliability of the results. Conclusions The results revealed that the population with higher household income tended to have a lower risk of genetic susceptibility to myocardial infarction and hypertension. †Xifeng Zheng and Yu Yang were listed as co-first authors. *Correspondence: Bing Lu lubing8564@gdmu.edu.cn Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessPage 2 of 8 Keywords Household income status, Cardiovascular health, Causal relationship, Instrumental variable, Mendelian randomization study Introduction According to a report by the World Health Organiza- tion, ischemic heart disease and stroke were the top two causes of mortality worldwide in 2019, accounting for 16% and 11% of deaths, respectively. Cardiovascular- related diseases (i.e., ischemic heart disease and stroke) are a major health threat to the aging population. In addi- tion, they cause a trend of increasing morbidity in young individuals due to obesity, diabetes, and drug abuse. Over the past two decades, this trend has been reflected by an increasing incidence of ischemic stroke among young people in the United States, Sweden, France, and Denmark. Although cardiovascular-related diseases are highly prevalent, it is gratifying that with the continuous popularization of health education and secondary pre- vention using medications, the total number of deaths in developed countries declined from 2000 to 2019, espe- cially among high-income populations [1, 2]. Further- more, multiple observational studies have reported that people with superior socioeconomic status usually have a lower risk of morbidity in cardiovascular diseases or better prognosis [3]. In contrast, low-income popula- tions may be exposed to higher risk or worse prognosis [4, 5]. This social phenomenon requires further research on the potential causal relationship between household income and cardiovascular-related disease morbidity. Unfortunately, reverse causation, measurement error, and potential bias are inherent disadvantages of obser- vational studies that prevent clarification of the potential causal relationship. To the best of our knowledge, limited evidence exists on the causal relationship between house- hold income status and cardiovascular disease-related morbidity, especially the lack of large-sample cohort studies. Mendelian randomization (MR) is a method that applies valuable genetic variants, such as single nucleo- tide polymorphisms (SNPs), as instrumental variables (IVs) to evaluate the causal effects between modifiable, non-genetic exposure factors and genetic susceptibil- ity to diseases. In the absence of randomized controlled trials (RCTs), MR studies represent an alternative strat- egy for causal inference because genetic variants are randomly assigned during meiosis to simulate the RCT process. Compared with traditional observational stud- ies, the greatest advantage of MR studies is that they are less likely to be influenced by unmeasured confounding factors because genetic variants are identified at the time of conception [6, 7]. MR studies have been successfully applied to various causal relationship analyses between behavior exposure, education, socioeconomic conditions, and various diseases [8, 9]. Hence, this research aims to identify the bidirectional causal relationship between household income status and genetic susceptibility to common cardiovascular-related diseases using an MR study. Materials and methods Study design and GWAS datasets information To achieve impartial results, an MR study depends on three fundamental assumptions: (1) the selected genetic IVs must be significantly associated with the exposure factor, (2) the IVs should be independent of potential confounders associated with exposure factors and out- comes, and (3) the IVs should affect the outcomes only through the exposure factor [6]. This study conducted the MR analysis 14 times to explore the bidirectional associa- tion between annual household income status and seven cardiovascular-related diseases. The research is based on a large-sample cohort of the European population from publicly available genome- wide association study (GWAS) datasets. The variable genetic information involved in this study was extracted from the Integrative Epidemiology Unit (IEU) GWAS database [10] (https://gwas.mrcieu.ac.uk/), which is a publicly available GWAS summary database. There- fore, the requirement for ethical committee approval was waived. The GWAS summary dataset “average total household income before tax” represented the house- hold income status of 397,751 samples originally from the UK biobank database. The annual household income was divided into five intervals: less than 18,000 pounds, 18,000 to 30,999 pounds, 31,000 to 51,999 pounds, 52,000 to 100,000 pounds, and greater than 100,000 pounds. In contrast, cardiovascular-related diseases were repre- sented by coronary artery disease, myocardial infarction, heart failure, atrial fibrillation, hypertension, ischemic stroke, and type 2 diabetes, respectively. Detailed infor- mation on all the GWAS datasets is listed in Table 1. The household income GWAS dataset and GWAS datasets of cardiovascular diseases originated from different con- sortiums to decrease the potential bias caused by sample overlap. In addition, all GWAS datasets involved in this study included populations of European ancestry to miti- gate bias from population stratification. Zheng et al. BMC Public Health (2023) 23:749 2018 2021 397,751 GWAS ID European European ukb-b-7408 finn-b-I9_CHD Years Population Sample size Total sample Case/Control 21,012/197,780 Table 1 Basic information of the GWAS datasets involved in the study Traits Exposure factor Household income status [11] Outcomes Coronary artery dis- ease [12] Myocardial infarction [13] Heart fail- ure [14] Atrial fibril- lation [15] Hyperten- sion [12] Ischemic stroke [16] Type 2 diabetes [17] finn-b-I9_HYP- TENS ebi-a-GCST006908 2018 ebi-a-GCST006867 2018 ebi-a-GCST011365 2021 ebi-a-GCST009541 2020 ebi-a-GCST006414 2018 60,620/970,216 61,505/577,716 62,892/596,424 34,217/406,111 55,917/162,837 47,309/930,014 European European European European European European 2021 Selection criteria for IVs The IVs were single nucleotide polymorphisms (SNPs) filtered according to the three afore mentioned piv- otal assumptions of the MR study. First, the SNPs were matched with a genome-wide statistical significance − 8). Second, the corresponding link- threshold (P < 5 × 10 age disequilibrium was tested to confirm the presence of SNPs in the linkage disequilibrium state and to confirm that these SNPs were independent by trimming SNPs within a 0–10,000 kb window at a threshold of r2 < 0.001. Third, to evaluate the assumption that the IVs affect the outcomes only through the exposure factor, the potential phenotypes that may be relevant to the IVs were inves- tigated by searching the human genotype-phenotype association database (PhenoScanner-V2, http://www. phenoscanner.medschl.cam.ac.uk/) [18]. Fourth, SNPs identified as IVs were further matched with those in the outcome GWAS dataset to establish genetic associations. The summary SNP-phenotype and SNP-outcome statis- tics were harmonized to ensure effect size alignment, and the palindromic SNPs were excluded. Finally, F-statistics (> 10) were used to evaluate the strength of the IVs to avoid the influence of weak instrumental bias [19]. Mendelian randomization study and sensitivity analysis The MR study was performed using a random-effects inverse-variance weighting (IVW) model [20] as the pri- mary standard and three other models (MR-Egger regres- sion [21], weighted median [22] and maximum likelihood Page 3 of 8 [23]) as supplements to evaluate the potential causal relationship between household income status and seven common cardiovascular-related diseases. The reverse MR study evaluated the potential causal relationship between seven common cardiovascular-related diseases and household income status using the same methods. In addition, sensitivity analysis was performed to measure the reliability and stability of the conclusion. The sensi- tivity analysis consisted of (1) Cochran’s Q test (accord- ing to the IVW model or MR-Egger regression model), (2) horizontal pleiotropy test using MR-Egger intercept [24] and MR-PRESSO test [25], and (3) “leave-one-out” test (each SNP was abandoned successively to repeat the IVW analysis to identify whether any specific SNP drives the causal relationship estimate). The results are reported as odds ratios (OR) with corresponding 95% confidence intervals (CI) and P-values, illustrated as scatter plots. The evidential threshold in MR analysis was defined as P-value < 0.004(0.05/14) according to the Bonferroni cor- rection method. P-value < 0.05 but above the Bonferroni corrected evidential threshold was regarded as a poten- tial association. Meanwhile, a P-value < 0.05 was also con- sidered significant in the sensitivity analysis. The R 4.0.3 software, with “TwoSampleMR” [26] and “MR-PRESSO” [25] packages were used to process and visualize the study. Results Results of the MR study The sample overlapping of the seven diseases GWAS dataset and UK-biobank database were as follows: coro- nary artery disease 0%, myocardial infarction: 73.8%, heart failure: 36.96%, atrial fibrillation: 38.4%, hyperten- sion: 0%, stroke: 0%, and type 2 diabetes: 9.8%, respec- tively. Even though the sample overlapping rate of myocardial infarction is comparatively high, we have fol- lowed the sample size and timeliness priority to make the best choice as much as possible. The numbers of SNPs that ultimately identified as IVs in different outcome datasets were 35 (type 2 diabetes), 43 (coronary artery disease, myocardial infarction, heart failure and hypertension), 44 (ischemic stroke), and 45 (atrial fibrillation) respectively. The F-statistic score of all these selected SNPs were more than 10 (coronary artery disease:57.76, myocardial infarction:57.87, heart failure:57.76, hypertension:57.77, atrial fibrillation:57.52, ischemic stroke:57.76, type 2 diabetes:58.44, respec- tively), indicating a low risk of weak-instrument bias. According to the random-effects IVW model results, as the primary standard, higher household income tended to lower the risk of genetic susceptibility to cor- onary artery disease (OR: 0.674, 95% CI = 0.509–0.893, P = 0.005), myocardial (OR: 0.503, 95% CI = 0.405–0.625, P < 0.001), heart failure (OR: 0.825, 95% infarction Zheng et al. BMC Public Health (2023) 23:749 CI = 0.709–0.960, P = 0.013), hypertension (OR: 0.667, 95% CI = 0.522–0.851, P = 0.001), type 2 diabetes (OR: 0.642, 95% CI = 0.464–0.889, P = 0.007) and ischemic stroke (OR: 0.801, 95% CI = 0.662–0.968, P = 0.022). How- ever, no evidence was reported on the potential causal relationship between household income status and atrial fibrillation (OR: 0.970, 95% CI = 0.767–1.226, P = 0.798). In addition, the results of the weighted median and maxi- mum likelihood estimation models supported these con- clusions. Unfortunately, the MR-Egger regression model results did not show statistically significant differences. In summary, according to the Bonferroni correction standard, the MR study revealed that the higher house- hold income population tended to have a lower risk of genetic susceptibility to myocardial infarction and hyper- tension. It also suggested a potentially negative relation- ship between coronary artery disease, heart failure, type 2 diabetes, and ischemic stroke. Detailed information is displayed in the forest map in Fig. 1 and illustrated as a scatter plot (supplement Figure-1). Results of sensitivity analyses in the MR study The result of Cochran’s Q test indicated certain heteroge- neity among the IVs in coronary artery disease, myocar- dial infarction, atrial fibrillation, hypertension, and type 2 diabetes (Table 2). Hence, the random effects IVW model was applied to minimize the effect of heterogeneity in the MR study as much as possible. More importantly, no horizontal pleiotropy was detected using the MR-Egger Page 4 of 8 intercept and MR-PRESSO tests (Table 2). In addition, the “leave-one-out” method indicated that no specific SNP among the IVs significantly affected the overall result (supplement Figure-2). In general, the sensitivity analysis verified the robustness of the conclusions. Table 2 The results of heterogeneity and horizontal pleiotropy tests Diseases Heterogeneity test MR-Egger regression IVW model Horizontal plei- otropy test MR-Egger intercept Coronary artery disease Myocardial infarction Heart failure Atrial fibrillation Hypertension Ischemic stroke Type 2 diabetes Note: IVW: Inverse Variance Weighting. P-value < 0.05 was considered as with statistical differences in both heterogeneity and horizontal pleiotropy tests 0.036 < 0.001 0.197 < 0.001 < 0.001 0.223 < 0.001 0.034 < 0.001 0.173 < 0.001 < 0.001 0.195 < 0.001 0.462 0.402 0.722 0.755 0.457 0.780 0.291 Results of reverse MR study and sensitivity analyses The numbers of SNPs that ultimately identified as IVs in different cardiovascular-related diseases were 115 (type 2 diabetes), 24 (coronary artery disease), 74(myo- cardial infarction), 9 (heart failure), 53 (hypertension), 7 MR- PRES- SO test - 0.988 0.265 0.685 0.663 0.202 0.831 Fig. 1 The result of MR study illustrated by forest plot Note: The causal relationship between household income status and cardiovascular-related diseases by MR study. OR = Odds Ratio, CI = Confidence Interval Zheng et al. BMC Public Health (2023) 23:749 (ischemic stroke), and 111 (atrial fibrillation) respectively in the reverse MR study. Based on the random-effects IVW model results, there were potential indications of causal association with myocardial infarction (OR: 0.982, 95% CI = 0.967–0.997, P = 0.025), heart failure (OR: 0.964, 95% CI = 0.931–0.998, P = 0.037), type 2 diabetes (OR: 0.986, 95% CI = 0.976– 0.997, P = 0.013) and household income status. The results of the maximum likelihood estimation model support these conclusions. However, the results of the MR-Egger regression model and the weighted median model did not show significant statistical differences. The detailed information is displayed in Fig. 2. However, horizontal pleiotropy was detected in SNPs of myocar- dial infarction (P = 0.027) and type 2 diabetes (P = 0.034) using the MR-Egger intercept test in the sensitivity analy- sis. This resulted in a lack of stability in the conclusions. Thus, the reverse MR study’s results only suggested a potentially negative association between heart failure and household income status. Page 5 of 8 Discussion Epidemiological studies have revealed that socioeco- nomic status has a non-negligible effect on cardiovas- cular health. Hence, household income status, as an indispensable part of socioeconomic status, has been consistently considered to be associated with the risk of cardiovascular disease [27]. However, in-depth causality needs further exploration. To the best of our knowledge, this research is the first that focuses on the causal effect between household income status and cardiovascular health using a bidirectional two-sample MR study. The results revealed that the population with higher house- hold income status tended to have a lower risk of genetic susceptibility to myocardial infarction and hypertension, and suggested a potentially negative relationship with coronary artery disease, heart failure, type 2 diabetes, and ischemic stroke. Moreover, a reverse MR study indi- cated a potential negatively association between heart failure and household income status. These conclusions were supported by accumulating evidence from cohort studies and meta-analyses. For example, Kucharska-Newton et al. summarized two inde- pendent cohorts from the US and Finland with 14 years of follow-up. They demonstrated that low-income status Fig. 2 The result of reverse MR study illustrated by forest plot Note: The causal relationship between cardiovascular-related diseases and household income status by reverse MR study. OR = Odds Ratio, CI = Confi- dence Interval Zheng et al. BMC Public Health (2023) 23:749 Page 6 of 8 rs1455350, rs784256). The result enlightens us that household income status may affect the progress of car- diovascular diseases through GABA-mediated mental or psychological-related biological mechanisms. Never- theless, more direct evidence is needed to support this assumption. The greatest advantage of this bidirectional MR study is that our findings effectively avoided the influence of reverse causes and minimized residual confounding. However, some limitations of the study should be men- tioned. First, the heterogeneity test results suggested cer- tain heterogeneity among the IVs. Although the random effects IVW model has been applied to minimize the effect of heterogeneity in MR study as much as possible. Second, models based on different assumptions involved in MR studies have distinctive advantages and disadvan- tages that may increase the likelihood of obtaining incon- sistent or contradictory results. Thus, this study’s results should be cautiously interpreted. Third, the sample over- lapping of the myocardial infarction GWAS dataset and UK-biobank database was comparatively high, which may increase potential bias. Conclusion This study explored the causal relationship between household income status and cardiovascular health using a bidirectional MR study based on datasets with million samples. The results revealed that the population with higher household income status tended to have a lower risk of genetic susceptibility to myocardial infarction and hypertension and suggested a potentially negative relationship with coronary artery disease, heart failure, type 2 diabetes, and ischemic stroke. These findings strengthen that medical reimbursement should consider household income inequalities and continuously improve the fairness and accessibility to medical services for the population in low household income status. Abbreviation MR SNP GWAS IV IVW OR CI Mendelian Randomization Single Nucleotide Polymorphism Genome-Wide Association Study Instrumental Variable Inverse Variance Weighting Odds Ratio Confidence Interval Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12889-023-15561-7. Supplementary Material 1 Supplementary Material 2 was significantly associated with an increased risk of major adverse cardiac events [28]. Similarly, in America and England, the aging population in low socioeconomic levels has been reported to have nearly double the car- diovascular disease-related mortality rates than that at superior socioeconomic levels [29, 30]. Moreover, Wang et al. conducted a study with 8,989 samples concentrated between income changes and cardiovascular disease risk. They concluded that changes in income might result in negative regulation of cardiovascular health over 17 years [31]. Additionally, it is worth noting that there is a bidi- rectional negative causal relationship between house- hold income status and heart failure. On the one hand, low-income populations lack advanced medical treat- ment and quality care [32, 33]. Moreover, they neglect necessary health maintenance behaviors, such as annual medical checkups or adherence to secondary preven- tion medications. This may result in a worse prognosis of myocardial infarction and likely development of heart failure [34–36]. On the other hand, patients with heart failure have limited mobility to various degrees or dis- abilities in daily work that cause their income and qual- ity of life to decline. Overall, the significant bidirectional negative causal relationship between low household income and heart failure is a complex public health con- cern. Therefore, policymakers for medical reimburse- ment should consider household income inequalities and continuously improve the fairness and accessibility to medical services for people [37]. Several relevant mechanisms which may explain the association between household income status and genetic susceptibility to cardiovascular diseases. They include: (1) low-income populations’ likelihood to con- sume fast foods that are more energy-dense and yield more calories for a given price [38], which creates a long- term tendency to obesity and metabolic syndromes. (2) low-income populations face greater living pressures that may increase depression and anxiety, which are ulti- mately linked to cardiovascular diseases [39, 40]. (3) low- income populations are more likely to face poor living environments, smoke, and engage in lifestyle behaviors such as drinking, which are already identified as risk fac- tors for cardiovascular health [41, 42]. The biological functions of SNPs as IVs and how they affect cardiovascular health deserve further discussion. Hill et al. reported 30 independent loci associated with individual income, which may involve the biological pro- cess of GABAergic and serotonergic neurotransmission [43]. The current research indicated that dysregulation of the GABA signaling pathway was linked to psychi- atric disorders, such as depression, autism, and anxi- ety [44, 45]. Among the 30 reported genetic loci, seven SNPs share common with the selected IVs in our study (rs11588857, rs6699397, rs32940, rs10429582, rs2332719, Zheng et al. BMC Public Health (2023) 23:749 Acknowledgements We would like to acknowledge the participants and investigators of the IEU OpenGWAS project, the UK Biological Bank and the FinnGen studies for their contributions. Author Contribution Xifeng Zheng was responsible for conception and article writing, Yu Yang was responsible for data mining, Jianying Chen was involved in literature review and Bing Lu was responsible for scientific supervision. All authors reviewed and approved the final manuscript. Funding The research was funded by “Affiliated Hospital of Guangdong Medical University Clinical Research Program, No. LCYJ2020A002”. Data Availability The genetic variable information of single nucleotide polymorphisms (SNPs) was obtained from the IEU GWAS database (https://gwas.mrcieu.ac.uk/ datasets/), a publicly available GWAS summary database. Declarations Ethics approval and consent to participate Not Applicable. Competing interests The authors declared that they have no conflicts of interests to this work. We declare that we do not have any commercial or associative interests that represents a conflict of interests in connection with the work submitted. Consent for publication Not Applicable. Author details 1Department of Cardiology, Affiliated Hospital of Guangdong Medical University, No.57 South of Renming Road, Zhanjiang, Guangdong, China 2Department of Geriatrics, Affiliated Hospital of Guangdong Medical University, No.57 South of Renming Road, Zhanjiang, Guangdong, China Received: 26 December 2022 / Accepted: 29 March 2023 2. 3. References 1. WHO. Global Health estimates 2020: deaths by cause, Age, Sex, by Country and by Region, 2000–2019. World Health Organization Geneva; 2020. Andersson C, Vasan RS. Epidemiology of cardiovascular disease in young individuals. Nat Reviews Cardiol. 2018;15:230–40. Jaquet E, Gencer B, Auer R, et al. Association between income and control of cardiovascular risk factors after acute coronary syndromes: an observational study. Swiss Med Wkly. 2019;149:w20049. Stirbu I, Looman C, Nijhof GJ, Reulings PG, Mackenbach JP. Income inequali- ties in case death of ischaemic heart disease in the Netherlands: a national record-linked study. J Epidemiol Commun Health. 2012;66:1159–66. Zang Y, Zhu Z, Shi M, et al. Association between annual household income and adverse outcomes in patients who had ischaemic stroke. J Epidemiol Community Health. 2022;76:293–300. 4. 5. 7. 6. Davies NM, Holmes MV, Davey Smith G. Reading mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ. 2018;362:k601. Richmond RC, Davey Smith G. Mendelian Randomization: Concepts and Scope.Cold Spring Harb Perspect Med2022;12. Tillmann T, Vaucher J, Okbay A, et al. Education and coronary heart disease: mendelian randomisation study. BMJ. 2017;358:j3542. 8. 9. Harrison S, Davies AR, Dickson M, et al. The causal effects of health conditions and risk factors on social and socioeconomic outcomes: mendelian random- ization in UK Biobank. Int J Epidemiol. 2020;49:1661–81. 10. Elsworth B, Lyon M, Alexander T et al. The MRC IEU OpenGWAS data infra- structure. bioRxiv 2020:2020.08.10.244293. Page 7 of 8 11. Bycroft C, Freeman C, Petkova D, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562:203–9. 12. Kurki MI, Karjalainen J, Palta P et al. FinnGen: Unique genetic insights from combining isolated population and national health register data. medRxiv 2022. 13. Hartiala JA, Han Y, Jia Q, et al. Genome-wide analysis identifies novel suscepti- bility loci for myocardial infarction. Eur Heart J. 2021;42:919–33. 14. Shah S, Henry A, Roselli C, et al. Genome-wide association and mendelian randomisation analysis provide insights into the pathogenesis of heart failure. Nat Commun. 2020;11:163. 15. Nielsen JB, Thorolfsdottir RB, Fritsche LG, et al. Biobank-driven genomic discovery yields new insight into atrial fibrillation biology. Nat Genet. 2018;50:1234–9. 16. Malik R, Chauhan G, Traylor M, et al. Multiancestry genome-wide association study of 520,000 subjects identifies 32 loci associated with stroke and stroke subtypes. Nat Genet. 2018;50:524–37. 17. Xue A, Wu Y, Zhu Z, et al. Genome-wide association analyses identify 143 risk variants and putative regulatory mechanisms for type 2 diabetes. Nat Com- mun. 2018;9:2941. 18. Yuan S, Carter P, Bruzelius M, et al. Effects of tumour necrosis factor on cardio- vascular disease and cancer: a two-sample mendelian randomization study. EBioMedicine. 2020;59:102956. 19. Wu F, Huang Y, Hu J, Shao Z. Mendelian randomization study of inflammatory bowel disease and bone mineral density. BMC Med. 2020;18:312. 20. Burgess S, Dudbridge F, Thompson SG. Combining information on multiple instrumental variables in mendelian randomization: comparison of allele score and summarized data methods. Stat Med. 2016;35:1880–906. 21. Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015;44:512–25. 22. Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent estimation in mendelian randomization with some Invalid Instruments using a weighted median estimator. Genet Epidemiol. 2016;40:304–14. 23. Nguyen LT, Schmidt HA, von Haeseler A, Minh BQ. IQ-TREE: a fast and effec- tive stochastic algorithm for estimating maximum-likelihood phylogenies. Mol Biol Evol. 2015;32:268–74. 24. Burgess S, Thompson SG. Interpreting findings from mendelian randomiza- tion using the MR-Egger method. Eur J Epidemiol. 2017;32:377–89. 25. Verbanck M, Chen CY, Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from mendelian randomization between complex traits and diseases. Nat Genet. 2018;50:693–8. 26. Hemani G, Zheng J, Elsworth B et al. The MR-Base platform supports system- atic causal inference across the human phenome.Elife2018;7. 27. Schultz WM, Kelli HM, Lisko JC, et al. Socioeconomic Status and Cardiovascu- lar Outcomes: Challenges and Interventions. Circulation. 2018;137:2166–78. 28. Kucharska-Newton AM, Harald K, Rosamond WD, Rose KM, Rea TD, Salomaa V. Socioeconomic indicators and the risk of acute coronary heart disease events: comparison of population-based data from the United States and Finland. Ann Epidemiol. 2011;21:572–9. 29. Stringhini S, Zaninotto P, Kumari M, Kivimäki M, Lassale C, Batty GD. Socio- economic trajectories and cardiovascular disease mortality in older people: the English Longitudinal Study of Ageing. Int J Epidemiol. 2018;47:36–46. 30. Faselis C, Safren L, Allman RM, et al. Income disparity and incident cardiovas- cular disease in older Americans. Prog Cardiovasc Dis. 2022;71:92–9. 31. Wang SY, Tan ASL, Claggett B, et al. Longitudinal Associations between Income Changes and Incident Cardiovascular Disease: the atherosclerosis risk in Communities Study. JAMA Cardiol. 2019;4:1203–12. 32. Yong CM, Abnousi F, Asch SM, Heidenreich PA. Socioeconomic inequalities in quality of care and outcomes among patients with acute coronary syndrome in the modern era of drug eluting stents. J Am Heart Assoc. 2014;3:e001029. 33. Parikh PB, Yang J, Leigh S, et al. The impact of financial barriers on access to care, quality of care and vascular morbidity among patients with diabetes and coronary heart disease. J Gen Intern Med. 2014;29:76–81. Johnson AE, Swabe GM, Addison D, et al. Relation of Household Income to Access and Adherence to Combination Sacubitril/Valsartan in Heart failure: a retrospective analysis of commercially insured patients. Circulation Cardio- vasc Qual outcomes. 2022;15:e009179. 34. 35. Teng TK, Tay WT, Richards AM, et al. Socioeconomic status and outcomes in heart failure with reduced ejection Fraction from Asia. Circulation Cardiovasc Qual outcomes. 2021;14:e006962. Zheng et al. BMC Public Health (2023) 23:749 Page 8 of 8 36. Eberly LA, Yang L, Eneanya ND, et al. Association of Race/Ethnicity, gender, and socioeconomic status with sodium-glucose cotransporter 2 inhibitor use among patients with diabetes in the US. JAMA Netw open. 2021;4:e216139. 37. Ma C, Song Z, Zong Q. Urban-Rural Inequality of Opportunity in Health Care: Evidence from China.International journal of environmental research and public health2021;18. 38. Mellis AM, Athamneh LN, Stein JS, Sze YY, Epstein LH, Bickel WK. Less is more: negative income shock increases immediate preference in cross commodity discounting and food demand. Appetite. 2018;129:155–61. 39. Freeman A, Tyrovolas S, Koyanagi A, et al. The role of socio-economic status in depression: results from the COURAGE (aging survey in Europe). BMC Public Health. 2016;16:1098. 40. Ridley M, Rao G, Schilbach F, Patel V. Poverty, depression, and anxiety: Causal evidence and mechanisms.Science2020;370. 41. Barber S, Hickson DA, Kawachi I, Subramanian SV, Earls F. Neighborhood Disadvantage and Cumulative Biological Risk among a socioeconomically diverse sample of african american adults: an examination in the Jackson Heart Study. J racial ethnic health disparities. 2016;3:444–56. 42. Maglalang DD, Avila JC, Ahluwalia JS, Murphy CM, Alexander AC, Nollen NL. The role of Financial strain and Educational Attainment on Smoking abstinence of African Americans and Whites who smoke. Ethn Dis. 2022;32:223–30. 43. Hill WD, Davies NM, Ritchie SJ, et al. Genome-wide analysis identifies molecu- lar systems and 149 genetic loci associated with income. Nat Commun. 2019;10:5741. 44. Hernández-Vázquez F, Garduño J, Hernández-López S. GABAergic modula- tion of serotonergic neurons in the dorsal raphe nucleus. Rev Neurosci. 2019;30:289–303. 45. Padmakumar M, Van Raes E, Van Geet C, Freson K. Blood platelet research in autism spectrum disorders: in search of biomarkers. Res Pract Thromb haemostasis. 2019;3:566–77. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Zheng et al. BMC Public Health (2023) 23:749
10.1186_s12915-020-00859-4	Burns et al. BMC Biology (2020) 18:133 https://doi.org/10.1186/s12915-020-00859-4 R E S E A R C H A R T I C L E Open Access Retargeting azithromycin analogues to have dual-modality antimalarial activity Amy L. Burns1, Brad E. Sleebs2,3, Ghizal Siddiqui4, Amanda E. De Paoli4, Dovile Anderson4, Benjamin Liffner1, Richard Harvey1, James G. Beeson5,6,7, Darren J. Creek4, Christopher D. Goodman8, Geoffrey I. McFadden8 and Danny W. Wilson1,5* Abstract Background: Resistance to front-line antimalarials (artemisinin combination therapies) is spreading, and development of new drug treatment strategies to rapidly kill Plasmodium spp. malaria parasites is urgently needed. Azithromycin is a clinically used macrolide antibiotic proposed as a partner drug for combination therapy in malaria, which has also been tested as monotherapy. However, its slow-killing ‘delayed-death’ activity against the parasite’s apicoplast organelle and suboptimal activity as monotherapy limit its application as a potential malaria treatment. Here, we explore a panel of azithromycin analogues and demonstrate that chemical modifications can be used to greatly improve the speed and potency of antimalarial action. Results: Investigation of 84 azithromycin analogues revealed nanomolar quick-killing potency directed against the very earliest stage of parasite development within red blood cells. Indeed, the best analogue exhibited 1600-fold higher potency than azithromycin with less than 48 hrs treatment in vitro. Analogues were effective against zoonotic Plasmodium knowlesi malaria parasites and against both multi-drug and artemisinin-resistant Plasmodium falciparum lines. Metabolomic profiles of azithromycin analogue-treated parasites suggested activity in the parasite food vacuole and mitochondria were disrupted. Moreover, unlike the food vacuole-targeting drug chloroquine, azithromycin and analogues were active across blood-stage development, including merozoite invasion, suggesting that these macrolides have a multi-factorial mechanism of quick-killing activity. The positioning of functional groups added to azithromycin and its quick-killing analogues altered their activity against bacterial-like ribosomes but had minimal change on ‘quick-killing’ activity. Apicoplast minus parasites remained susceptible to both azithromycin and its analogues, further demonstrating that quick-killing is independent of apicoplast-targeting, delayed-death activity. Conclusion: We show that azithromycin and analogues can rapidly kill malaria parasite asexual blood stages via a fast action mechanism. Development of azithromycin and analogues as antimalarials offers the possibility of targeting parasites through both a quick-killing and delayed-death mechanism of action in a single, multifactorial chemotype. Keywords: Plasmodium, Malaria, Antimalarial, Macrolide * Correspondence: danny.wilson@adelaide.edu.au 1Research Centre for Infectious Diseases, School of Biological Sciences, The University of Adelaide, Adelaide 5005, Australia 5Burnet Institute, Melbourne, Victoria 3004, Australia Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Burns et al. BMC Biology (2020) 18:133 Page 2 of 23 Background Malaria is a mosquito-borne disease caused by proto- the genus Plasmodium. In 2017, zoan parasites of there were ~ 219 million cases of malaria that resulted in ~ 435,000 deaths [1, 2], with most deaths as the re- sult of Plasmodium falciparum infection in children under 5 years of age within sub-Saharan Africa. Current control strategies include use of insecticide treated bed- nets and indoor residual spraying, which target mosquito transmission, chemoprophylaxis in high-risk groups, and artemisinin combination therapies (ACTs) to both cure patients and limit their transmission. Widespread use of these control measures has resulted in significant de- creases in malaria mortality over the past two decades [1, 2]. However, there is growing concern that artemisinin- resistant P. falciparum parasites may spread from the Greater Mekong sub-region and Eastern India, where they have previously been identified, and will lead to the loss of our most effective drug treatments [3–6]. Furthermore, there is also substantial resistance to some of the current partner drugs used in ACTs, most notably piperaquine and mefloquine [7]. Therefore, new antimalarials with novel mechanisms of action that rapidly clear blood-stage parasites are urgently needed [8, 9]. Clinically used macrolide antibiotics, in particular azi- thromycin, have been proposed as partner drugs for ACTs [10, 11]. Macrolide antibiotics have been shown to target the malaria parasite’s remnant plastid (apico- plast), which has a bacterium-like ribosomal complex es- sential for protein translation and organelle biogenesis [12–14]. The apicoplast is essential for synthesis of iso- pentenyl pyrophosphate (IPP) precursors required for protein prenylation, ubiquinone biosynthesis and doli- chols required for N-glycosylation and production of GPI anchors (reviewed in [15] and [16]). Indeed, IPP synthesis is the sole essential function of the apicoplast in blood stages, but apicoplast biogenesis and house- keeping activity is essential for IPP production, making the apicoplast ribosome an attractive antimalarial target [13, 14, 17]. P. falciparum parasites treated with clinic- ally relevant (nanomolar) concentrations of macrolide antibiotics exhibit a ‘delayed-death’ phenotype in which parasite growth is arrested during the second replication cycle after treatment (~ 4 days post-treatment) [13, 14]. Azithromycin exhibits three favourable properties as an antimalarial: a half-life > 50 hrs making it suitable for infrequent dosing [18], good in vivo safety profile [19] and high potency against P. falciparum in vitro [20, 21]. Azithromycin also shows efficacy as a prophylactic [22] (reviewed in [23]), improved clinical outcomes in com- bination with pyrimethamine during intermittent pre- ventative treatment for malaria in pregnancy (IPTp) trials [24] and led to a significant decrease in P. falcip- arum infections following mass drug administrations of azithromycin monotherapy for trachoma infection [25]. Evidence also suggests that azithromycin inhibits the de- velopment of mosquito transmissible parasites and liver stages in rodent models [22, 26, 27]. However, when azi- thromycin was trialled for treatment of clinical malaria, it exhibited sub-optimal activity as a monotherapy and was generally less effective than the similarly acting anti- biotic clindamycin when used in combination with other antimalarials [28]. Crucially, the delayed-death activity of azithromycin has limited its use as a treatment for clinical disease. Currently, azithromycin is not used as a these first-line considerations. for malaria because of treatment We previously demonstrated that azithromycin can also cause rapid parasite death when tested at higher concentrations (IC50 ~ 10 μM) [27, 29]. Most strikingly, azithromycin can rapidly inhibit P. falciparum merozoite invasion of RBCs at these higher concentrations. In addition, azithromycin kills parasites within one intracel- lular blood-stage lifecycle (from immediately post- merozoite invasion to final schizont maturation at 48 hrs, in-cycle) at a similar IC50 as the drug’s invasion in- hibitory activity. Testing of a small panel of azithromycin analogues showed that these ‘quick-killing’ IC50s could be enhanced through chemical modification. Import- antly, parasites selected for resistance to azithromycin’s delayed-death activity (120 hr post-invasion) remained susceptible to both invasion-inhibition and intracellular parasite quick-killing activities (invasion, in-cycle and 72 hr inhibition), indicating that azithromycin has a second- ary, apicoplast-independent, mechanism of action [27, 29]. Therefore, chemical modification of azithromycin presents a unique opportunity to develop a dual-acting antimalarial with two independent mechanisms of action that combines both quick-killing (for rapid clearance of clinical infection) and delayed-death activities, providing an element of resistance proofing and improving longer- term protection from recrudescence or reinfection. In this study, we screened 84 azithromycin analogues and defined their efficacy against different stages of the blood-stage lifecycle. A high proportion of analogues ex- hibited improved quick-killing activity over azithromycin against both P. falciparum and P. knowlesi, a model for P. vivax and human pathogen of developing importance in Southeast Asia [30], and were equally effective against lacking an apicoplast. The parasites containing or analogues acted rapidly at inhibitory concentrations with only short treatment times required to kill parasites throughout the blood-stage low cost of established safety profile, manufacture, and previous evaluation in ACTs, the re- development of azithromycin-like compounds into an antimalarial with dual mechanisms of action provides a novel strategy to develop new antimalarials. development. Given life, long-half Burns et al. BMC Biology (2020) 18:133 Page 3 of 23 Results Azithromycin analogues show improvement in quick- killing activity against P. falciparum We characterised the activity of 84 azithromycin ana- logues across the malaria parasites asexual blood-stage development in fine detail, including their activity against early ring stages. The IC50 values for 72 hr growth- inhibition assays (drug treatment assays represented in Fig. 1; 1 cycle assay Fig. 1c) and their toxicity against mammalian cells for analogues presented in this study have been published previously [31–35]. Here, we tested in-cycle assays, for quick-killing activity using 44 hr wherein 10 μM of drug was added to early ring-stage D10- PfPHG parasites within a few hours of invasion and para- site development quantified at late schizont stage with no exposure of invading merozoites to the drug. This initial screen identified 65 of 84 analogues that inhibited growth by > 30% (Fig. 1b, Additional file 1: Tables S1a-c). The in- cycle IC50 values for these 65 analogues were determined (Additional file 1: Tables S1a-c) with all but two analogues showing improved potency over azithromycin (azithromy- cin IC50 with 44 hr in-cycle treatment, 11.3 μM) with the most potent compound exhibiting a 1615-fold lower IC50 than azithromycin (GSK-66 IC50 0.007 μM). Notably, 39 analogues showed > 10-fold improvement over azithromy- cin (IC50 < 1 μM), with 16 exhibiting a > 55-fold improve- ment (IC50 < 0.2 μM). Summary inhibitory assay data and structure for 19 of the most potent analogues featuring different added functional groups is available in Table 1 and Fig. 2. Published cytotoxicity data against mammalian cells is available for 13 of the most potent analogues [31, 33–35] with the IC50 against the HepG2 cell line ranging between 3 and 83 μM and the selectivity index (SI; IC50 against HepG2/44 hr D10-PfPHG IC50 from this study) ranging between 15 to 415 fold. Eleven of these analogues had a SI > 50, indicating low mammalian cell toxicity. The analogues with the low nanomolar 44 hr in-cycle activity often featured quinoline or chloroquinoline modi- fications (Table 1, Fig. 2, Additional file 1: Tables S1a-c). However, there were exceptions including a number of phenyl-substituted analogues (GSK-5, GSK-6, GSK-9, GSK-11, GSK-14, GSK-16, GSK-17, GSK-19) and naphthalene-substituted analogues (GSK-3, GSK-4, GSK- 15, GSK-18), which all displayed IC50 values < 1 μM. There was no structural difference between the most po- tent analogues and the analogues with activity > 1 μM that could explain the observed activity discrepancy. Consist- ently, chloroquinoline analogues (GSK-1, GSK-2, GSK-56 and GSK-66) were more potent than their respective unsubstituted quinoline counterparts (GSK-7, GSK-10, GSK-58 and GSK-71). Analogues GSK-6 and GSK-9 with thiourea aryl substitution displayed comparable potency (IC50 0.2 and 0.44 μM) to naphthalene analogues GSK-3 and GSK-4 (IC50 0.18 and 0.19 μM). However, a large Fig. 1. Schematic of drug treatment regimens outlining the times of treatment and stage/time of parasitaemia measurement for assays used in this study. a Merozoite invasion of RBCs: Merozoites were drug treated prior to addition of RBCs. RBC invasion was measured at early ring stages (< 1 hr rings). b In-cycle: highly synchronous, early ring-stage parasites (0–4 hrs post-invasion) were treated with drug, with the resulting growth inhibition analysed at schizont stage (44 hrs post-invasion for P. falciparum and 26 hrs for P. knowlesi). c One cycle (0–72 hrs): highly synchronous, early ring-stage parasites (0–4 hrs post-invasion) were drug-treated and the resulting growth inhibition was measured after ~ 72 hrs of growth, post one cycle of re-invasion, at schizont stages. d 2 cycle (delayed death); highly synchronous, early ring-stage parasites (0–4 hrs post-invasion) were drug-treated and allowed to grow for 92 hrs before washing drug with fresh media (post second invasion cycle). Growth inhibition was assessed approximately 30 hrs later, at schizont stages (0–120 hrs post-invasion for P. falciparum and 0–92 hrs for P. knowlesi) Burns et al. BMC Biology (2020) 18:133 Page 4 of 23 Table 1 In vitro efficacy of antimalarials and azithromycin analogues against Plasmodium spp. parasites cmpnd R3 class R4 class R5 class In-cycle (44 hr) growth DD2 IC50 (μMb, ±SEM) Invasion inhibition D10-PfPHG IC50 (μMc, ±SEM) AZR Me Me CQ QN DHA 1 56 59 66 69 70 72 8 10 58 71 73 3 4 15 5 6 9 17 H H Chloroquinoline Me Chloroquinoline H Chloroquinoline H Me Me Me Me Me Me Quinoline Quinoline Me Me Me Naphthalene Naphthalene Naphthalene Substituted phenyl thiourea Substituted phenyl thiourea Substituted phenyl thiourea Substituted phenyl thiourea Me Me Me Me Me Me Quinoline Me Me Me Me Me Me Me Me Me In-cycle (44 hr) growth D10-PfPHG IC50 (μMa, ±SEM) 11.31 (0.49) 0.052 (0.006) 0.39 (0.07) 0.0008 (0.0001) 0.019 (0.004) 0.011 (0.002) 0.073 (0.02) 15.6 (2.1) 0.31 (0.31) ND ND 0.082 (0.02) 0.093 (0.02) 0.049 (0.005) 0.043 (0.002) ND ND Chloroquinoline 0.007 (0.001) Chloroquinoline 0.031 (0.004) Chloroquinoline 0.05 (0.006) Chloroquinoline 0.27 (0.01) 0.065 (0.004) H H H Quinoline Quinoline H H H H H H H 0.41 (0.02) 0.48 (0.04) 0.048 (0.004) 0.053 (0.005) 0.31 (0.02) 0.183 (0.02) 0.19 (0.01) 0.67 (0.07) 0.2 (0.01) 0.52 (0.1) 0.748 (0.1) 0.056 (0.01) 0.16 (0.02) 0.48 (0.2) 0.32 (0.07) ND 0.4 (0.1) 0.4 (0.05) 0.28 (0.05) 0.27 (0.07) 0.44 (0.07) 0.24 (0.04) 0.7 (0.05) 0.54 (0.06) 10 (1.4) ND ND ND ND 3.2 (0.39) ND ND ND ND 1.7 (0.02) 4.4 (1.2) ND ND ND ND 1.8 (0.5) 2.0 (0.2) 3.6 (0.4) 1.61 (0.02) ND ND ND In-cycle (24 hr) growth PkYH1 IC50 (μMd, ±SEM) 16 (1.8) 0.017 (0.005) ND 0.0024 (0.001) 0.2 (0.005) 0.031 (0.008) ND 0.012 (0.002) ND ND 0.15 (0.06) 0.15 (0.01) 0.1 (0.005) 0.071 (0.013) 0.041 (0.005) 0.248 (0.07) 0.095 (0.02) ND 0.32 (0.12) 0.082 (0.02) 0.16 (0.03) 0.016 (0.005) 0.36 (0.01) a Drug treatment of intracellular growth, from rings to late schizonts, with no rupture cycle for D10-PfPHG (P. falciparum, 0–44 hrs). Data represents the mean of 3 or more experiments b Drug treatment of intracellular growth, from rings to late schizonts, with no rupture cycle for DD2 (P. falciparum, 0–44 hrs). Data represents the mean of 3 or more experiments c Drug treatment of D10-PfPHG merozoites prior to addition of RBCs. Parasitemia was measured by flow cytometry ~ 30 min post invasion. Data represents the mean of 2 (for compounds 4 and 5) or 3 experiments d Drug treatment of intracellular growth, from rings to late schizonts, with no rupture cycle for P. knowlesi YH1 (P. knowlesi, 0–24 h). Data represents the mean of 2 or more experiments number of analogues supporting thiourea and urea aryl substitutions were significantly less active, with no clear distinction between the activity and substitution pattern on the thiourea- or urea-substituted analogues. ring of aryl Analogues with aliphatic substitution on the urea or thiourea (GSK-31, GSK-35, GSK-38, GSK-45, GSK-47, GSK-51) generally had reduced activity compared to ana- logues with pendant aryl moieties (Table 1, Fig. 2, Add- itional file 1: Tables S1a-c), suggesting the aryl substituent was important for modulating potency. Consistent with this observation, analogues that did not terminate with an aromatic substituent and were only decorated with small aliphatic functionality (analogues GSK-34, GSK-44, GSK- 46, GSK-50, GSK-52, GSK-53, GSK-54, GSK-55, GSK-62, GSK-64, GSK-83, GSK-84) were either weakly active (> 3 μM) or inactive. These data suggested that the type of functionality and the length of the carbon-chain linking the aromatic group to the macrolactone was not important for activity. However, analogues GSK-56, GSK- 57, GSK-58, GSK-66, GSK-67 and GSK-71, with short 3- carbon linkers between the macrolactone and the quinoline group, were amongst the most potent. Overall, there was no consistent trend between the type of func- tionality and the length of the carbon-chain linking the aromatic group to the macrolactone. The position of the pendant quinoline or aromatic system attached to the macrolactone—either N6-, O- Burns et al. BMC Biology (2020) 18:133 Page 5 of 23 N R4 X X n R3 N HO OH O O OH OR1 OR2 R5O O Desosaminyl O OH O Cladinosyl R R N quinoline X X n naphthalene H N H N n X substituted phenyl urea R Modifications Azithromycin R1 = desosaminyl, R2 = cladinosyl, R3 and R4 = CH3, R5 = H Analogues in this study R1 = desosaminyl, R2 = cladinosyl or H, R3, R 4, R5 = H, Me or modifications Fig. 2. Structure of azithromycin and analogues. Outline of the structure of the parent molecule azithromycin, structural side-chains and sites of attachment of functional groups (R1–5) for compounds shown in Table 1. Structure of functional groups added is listed in Table 1 desosaminyl or N-desosaminyl—did not affect the in- cycle 44 hr activity of analogues (Table 1, Fig. 2, Add- itional file 1: Tables S1a-c). For example, analogues with the same quinoline functionality, GSK-1, GSK-56 and GSK-66 attached to either N6-, N-desosaminyl or O- desosaminyl positions, displayed similar IC50 values be- tween 7 and 19 nM. This trend was observed amongst other analogues for which there were matched pairs. The cladinosyl group did not affect 44 hr in-cycle activ- ity, for example respective analogues with the cladinosyl group, GSK-1, GSK-10, GSK-56 and GSK-66, possessed similar activity compared to analogues without the cladi- nosyl group, GSK-67, GSK-7 and GSK-57. This observa- tion is consistent with our previous findings on the azalide structure activity relationship [29]. Azithromycin analogues show improved activity against merozoite RBC invasion We previously showed that azithromycin and analogues inhibit merozoite invasion, with merozoites found to con- tact and briefly deform the RBC membrane, and then de- tach when examined in the presence of azithromycin [29]. We investigated whether the 39 analogues that had an in- cycle (44 hr) IC50 < 1 μM could inhibit merozoite invasion at a concentration of 1 μM and identified eight analogues that inhibited invasion by > 20% at 1 μM (Fig. 1a, Table 1, Fig. 2, Additional file 1: Tables S1a-c). The invasion inhibi- tory IC50 for seven of these analogues with sufficient avail- able sample were determined; there was a 2- to 6-fold reduction in the invasion inhibitory IC50 over azithromy- cin (range GSK-8 4.4 μM to GSK-5 1.6 μM) (Table 1, Additional file 2: Figure S1). Importantly, azithromycin analogues with improved in-cycle activity also had im- proved potency against merozoite invasion, confirming previous observations that both invasion and in-cycle quick-killing activities can be improved with a single chemical modification [29]. We next tested whether azi- thromycin analogue invasion inhibitory activity was di- rected against treating purified the merozoite by merozoites with 10 μM of GSK-72 (invasion inhibitory IC50 1.7 μM), followed by washing drug off the merozoites, and then mixing merozoites with RBCs (Additional file 3 Figure S2). GSK-72-treated merozoites were stopped from invading RBCs after washing off the drug, suggesting that the invasion inhibitory activity of azithromycin analogues is irreversible and directed towards the merozoite. Quick-killing activity is independent of apicoplast targeting We previously showed that quick-killing activity is main- tained against delayed-death-resistant parasites [29], sug- gesting that quick-killing occurs through a mechanism of Burns et al. BMC Biology (2020) 18:133 Page 6 of 23 action independent of the apicoplast. However, the fact that the apicoplast and apicoplast-ribosome were still present in these drug-treated parasites left open the possi- bility that quick-killing activity could still be linked to the apicoplast [36]. To confirm quick-killing is completely in- dependent of the apicoplast, we generated apicoplast minus (PfPHGapicoplast-null) parasites through prolonged treatment with azithromycin and then rescued with media supplementation with the isoprenoid precursor, isopente- nyl pyrophosphate (IPP) [17, 36]. PfPHGapicoplast-null para- sites showed a complete loss of sensitivity to azithromycin in 120 hr delayed-death assays, confirming that the apico- plast had been removed (Additional file 4: Figure S3a) [17, 36]. In contrast, there was no difference in growth inhib- ition for the PfPHGapicoplast-null and PfPHGwildtype parasites when treated with azithromycin (Additional file 4: Figure S3b) and 15 lead analogues at the in-cycle D10- PfPHGwildtype IC90 concentration for 44 hrs (Additional file 4: Figure S3c; Additional file 1: Table S1a-b). These data confirm that quick-killing activity is independent of the apicoplast, indicating that there is a secondary mechanism of action for azithromycin and analogues. Azithromycin is a rapid and irreversible inhibitor across blood-stage parasite growth After confirming that azithromycin and analogues have both invasion (Table 1, Additional file 2: Figure S1) and intracellular (Fig. 3a) blood-stage quick-killing activity that is independent of apicoplast-targeting delayed death (Add- itional file 4: Figure S3a-c), we next determined drug ac- tivity across early rings (0–12 hrs post invasion), early trophozoites (12–24 hrs post invasion), late trophozoites (24–36 hrs post invasion) and schizonts (36–44 hrs post invasion). Azithromycin demonstrated a similar IC50 across each pulsed treatment stage (0–12 hr IC50 14 μM, 12–24 hr IC50 16 μM, 24–36 hr IC50 15 μM) with these values similar to the IC50 values obtained for 44 hr (IC50 11.3 μM) and invasion inhibition (IC50 10 μM) treatments (Fig. 3b, c). We confirmed that azithromycin’s quick- killing activity works rapidly by assessing the morpho- logical effects of pulsed treatment with a 2× IC90 drug concentration. Ring-stage treatments (0–12 hrs) showed pronounced vacuolation of the cytoplasm, a typical sign of parasite stress. Trophozoite stages (12–24 hrs and 24–36 hrs) appeared either pyknotic or severely vacuolated with indicative of rapid cell death only a 12-hr treatment, (Fig. 3b, e). Although azithromycin treated schizont stages (36–44 hrs post-invasion) did not show potent growth in- hibitory activity when assessed by flow cytometry, light microscopy smears showed late-stage parasites with severe vacuolation and minimal merozoite maturation, indicating this population was indeed killed by azithromycin treat- ment (Fig. 3e). These data, together with our earlier data, provide direct evidence that azithromycin acts broadly across invasion and throughout the entire blood-stage life- cycle, including early ring stages. Azithromycin and analogues rapidly kill early ring-stage parasites Our finding that azithromycin could kill ring-stage para- sites (0–12 hrs post invasion) with similar efficacy to 44 hrs of drug treatment is of major interest since the ma- jority of clinically used antimalarials, with the notable exception of the artemisinins [37, 38], have relatively poor activity against newly invaded ring stages [39–42]. early To provide further insights into how quickly azithro- mycin and analogues act against early ring stages, we ex- amined activity of 6- and 12-hr treatments of early ring stages (0–6 hrs and 0–12 hrs post-invasion treatments) for azithromycin and a panel of diverse analogues that had activity at nanomolar concentrations in parallel. Azi- thromycin and the analogues tested showed < 2-fold re- duction in potency with a 6-hr ring-stage treatment compared to a 12-hr ring-stage or full 1 cycle (44 hr) treatment, highlighting the drug efficacy against early ring stages (Fig. 3d, e, Fig. 4a, b, Additional file 5: Table S2). Consistent with previous publications, dihy- droartemisinin (DHA) resulted in severe growth retardation with early ring-stage treatment [37– 39]. DHA is considered to be one of the few clinically used antimalarials with reasonable efficacy against early ring-stage parasites [37–39], making the ability of azi- thromycin and analogues to also cause rapid death of these stages a promising finding. In contrast, chloro- quine had comparatively poor activity for early ring- stage treatments, which is as expected since chloroquine is known to lack potency against ring-stage parasites. treatment Microscopy analysis was performed for parasites treated with a 2× IC90 (0–44 hrs) of azithromycin and analogues to examine the phenotypic changes associated with early ring-stage drug treatment (Fig. 4b). Early (0– 6 hrs) ring stages treated with azithromycin GSK-66 and GSK-3 exhibited vacuolation, with evidence of pyknotic cells developing with extended treatment for GSK-71 and GSK-3 (0–12 hrs). Notably, GSK-5 resulted in a large number of pyknotic parasites within only 6 hrs of drug treatment, highlighting the speed with which these compounds can act. DHA treatment of early (0–6 hrs) ring stages did not lead to a clear change in parasite morphology. However, after extended ring-stage treat- ment (0–12 hrs) pyknotic cells became prominent. No aberrant growth phenotype was observed with chloro- quine with treatment of early ring stages (0–6 hrs), with evidence of vacuolation only occurring after extended ring-stage treatment (0–12 hrs). Short-term pulse treat- ments confirmed that azithromycin and analogues rap- idly kill early ring-stage parasites, the growth inhibitory of effects and modification reversible, not are Burns et al. BMC Biology (2020) 18:133 Page 7 of 23 A ) l o r t n o C % ( h t w o r G 120 100 80 60 40 20 0 C ) l o r t n o C % ( h t w o r G 120 100 80 60 40 20 B 0-44 hrs 0-72 hrs 0-120 hrs 0-6 hrs 0-12 hrs 12-24 hrs 24-36 hrs 36-48 hrs 0.01 0.1 1 10 100 Azithromycin (µM) 0-12 hrs 12-24 hrs 24-36 hrs 36-44 hrs 0 0.1 1 100 Azithromycin (µM) 10 D ) l o r t n o C % ( h t w o r G 120 100 80 60 40 20 0 0-6 hrs 0-12 hrs 0-44 hrs 0.01 0.1 1 10 100 Azithromycin (µM) Fig. 3. Azithromycin has broad activity against blood-stage parasites. a Early ring-stage P. falciparum parasites (0–4 hrs post-invasion) were treated with doubling dilutions of azithromycin and inhibition of growth measured for in-cycle (44 hr, IC50, 11 μM), 1-cycle (72 hr, IC50, 14 μM) and 2-cycle (delayed death, 120 hr, IC50, 0.07 μM) assays (44 hr vs 72 hr, P=NS; 120 hr vs 44 hr P < 0.0001; 120 hr vs 72 hr P < 0.0001). b Schematic of drug washout treatment scheme to assess azithromycin’s quick-killing stage of activity. Early ring-stage parasites (0–4 hrs post-invasion) were aliquoted to a 96-well plate and doubling dilutions of azithromycin added between 0-12 hrs, 12–24 hrs, 24–36 hrs and 36–44 hrs post invasion prior to drug removal by washing with fresh media. c Growth inhibition of azithromycin across 0–12 hrs, 12–24 hrs, 24–36 hrs and 36–44 hrs post invasion prior to drug removal by washing with fresh media. There was no significance between treatment times for 0–12 hrs, 12–24 hrs, 24–36 hrs, but there was for 0–12 hrs vs 36–44 hrs (P = 0.005), 12–24 hrs vs 36–44 hrs (P = 0.01) and 24–36 hrs vs 36–44 hrs (P = 0.01). d Growth inhibition of azithromycin with very early ring-stage treatment across 0–6 hrs and 0–12 hrs post-invasion compared to a full in-cycle (0–44 hr) treatment. Treatments showed significant difference (P < 0.0001) with the exception of 0–12 hrs vs 0–44 hrs (P = 0.19). For all growth curves, parasitemia was measured at 44 hrs post invasion at schizont stage via flow cytometry. Data represents the means of 3 or more experiments expressed as a percentage of non-inhibitory control and error bars represent ± SEM. Dose response IC50s compared using extra sum of squares F-test. Repeat measure data is available in Additional file 15 Supporting Value Data. e Representative Giemsa-stained thin blood smears showing the growth phenotypes seen for non-inhibitory media controls (top panels) and in the presence of 2× IC90 concentration of azithromycin (bottom panels) across different stages of intraerythrocytic blood-stage development (0–6 hrs, 0–12 hrs, 12–24 hrs, 24–36 hrs and 36–44 hrs) Burns et al. BMC Biology (2020) 18:133 Page 8 of 23 Fig. 4. (See legend on next page.) Burns et al. BMC Biology (2020) 18:133 Page 9 of 23 (See figure on previous page.) Fig. 4. Growth inhibition profiles of azithromycin analogues and control drugs with short-term and in-cycle drug treatments. a Early ring-stage P. falciparum parasites (0–4 hrs post-invasion) were treated with doubling dilutions of azithromycin analogues/control drugs for 0–6 hrs and 0–12 hrs prior to washing the drug out of cultures allowing growth to continue until parasites were 44 hrs old. A 0–44 hr continuous drug control treatment was also included. a Growth inhibition profile of GSK-3 (naphthalene), GSK-5 (substituted phenyl), GSK-66 (chloroquinoline), GSK-71 (quinoline), dihydroartemisinin (DHA) and chloroquine with very early ring-stage treatment across 0–6 hrs and 0–12 hrs post-invasion compared to a full in-cycle treatment. There was no significant difference in drug efficacy between the treatment times of GSK-5 or GSK-71 (P > 0.01). GSK- 66 showed a significant difference between 0-6 hr vs 0–12 hr treatments (P < 0.0079) and 0–6 hr vs 0–44 hr (P = 0.001), but there was no significant difference in drug efficacy between 0-12 hr vs 0–44 hr treatments (P = 0.96). GSK-3 and DHA showed no significant difference in efficacy between treatment times (P > 0.01), with the exception of 0–6 hr vs 0–44 hr (P = 0.005 and P = 0.01, respectively). In contrast, chloroquine demonstrated a significant difference in drug efficacy between all treatment times (0–6 hr vs 0–12 hr P < 0.0001; 0–6 hr vs 0–44 hr P < 0.0001; 0– 12 hr vs 0–44 hr P < 0.0001). Parasitemia was measured via flow cytometry 44 hrs post-invasion. Data represents the means of 3 or more experiments expressed as a percentage of non-inhibitory control and error bars represent ± SEM. Dose response IC50s compared using extra sum of squares F-test. Repeat measure data is available in Additional file 15 Supporting Value Data. b Representative Giemsa-stained thin blood smears showing the growth phenotypes seen for non-inhibitory media controls, and treatment with 2× IC90 of azithromycin analogues GSK-3 (0.74 μM), GSK-5 (0.62 μM), GSK-66 (0.034 μM), GSK-71 (0.18 μM) and control drugs DHA (0.003 μM) and chloroquine (0.222 μM) (bottom panels) 0–6 hrs post treatment and 0–12 hrs post treatment azithromycin can produce analogues with broad and po- tent efficacy across blood-stage parasite growth. the against Quick-killing azithromycin analogues maintain activity against drug-resistant P. falciparum and P. knowlesi We next investigated whether analogues retained po- chloroquine/mefloquine/pyrimeth- tency falciparum DD2 line [43], and an amine-resistant P. artemisinin-resistant P. falciparum Cambodian isolate [44–46] (Table 1). Relative to the chloroquine sensitive D10-PfPHG line, DD2 parasites exhibited a 0.24- to 8.4- fold loss of sensitivity to azithromycin and analogues. Of note, analogues featuring a chloroquinoline moiety (GSK-1, GSK-56, GSK-66, GSK-72) were 4.77-fold less chloroquine-resistant DD2, whereas potent and quinoline-, analogues phenyl-substituted moieties were on average 1.35-fold less (Table 1, Add- itional file 6: Table S3). sensitive (n = 11 compounds) naphthalene- featuring against We next tested the efficacy of azithromycin ana- logues against the P. falciparum artemisinin-resistant clinical isolate Cam3.II, which has a mutation within the Kelch13 (PF3D7_1343700) propeller gene (R539T, Cam3.IIDHA resistant(R539T)) associated with increased early ring-stage (0–3 hrs) survival in vitro with DHA treat- ment [44–46]. Early ring-stage Cam3.IIDHA resistant(R539T) resistant and a reverted sensitive line (Cam3.IIsensitive) were pulsed for 4 hrs before the drug was washed off, with growth determined 66 hrs later via flow cytometry [46, 47]. Since comparison of IC50 has limited relevance in ring-stage survival assays, we compared instead the per- centage (%) parasite growth of Cam3.IIDHA resistant(R539T) parasites at the drug concentration that inhibited 95% of growth for the Cam3.IIsensitive line. As expected, ~ 41% Cam3.IIDHA resistant(R539T) parasites survived DHA treat- ment killed 95% of concentration that Cam3.IIsensitive parasites (Fig. 5, Table 2). In contrast, the at growth of both the Cam3.IIDHA resistant(R539T) and the Cam3.IIsensitive lines were equally inhibited at the con- centration that killed 95% of DHA-sensitive parasites for azithromycin and analogues GSK-56, GSK-71, GSK-3 and GSK-5. for Of note, the IC50 of 4 hr ring-stage treatments ob- served for the Cam3.IIsensitive line was similar to that of 6 hr ring-stage treatment seen in D10-PfPHG line upon treatment of azithromycin (Cam3.IIsensitive IC50 31 μM, D10-PfPHG IC50 30 μM) and GSK-5 (Cam3.IIsensitive IC50 0.20 μM, D10-PfPHG IC50 0.3 μM). Furthermore, activity against early ring-stage Cam3.IIsensitive parasites was also similar to the in-cycle (44 hr) treatment activity against D10-PfPHG parasites azithromycin (Cam3.IIsensitive IC50 31 μM, D10-PfPHG IC50 14 μM), GSK-5 (Cam3.IIsensitive IC50 0.20 μM, D10-PfPHG IC50 0.26 μM), GSK-56 (Cam3.IIsensitive IC50 0.006 μM, D10- PfPHG IC50 0.010 μM), GSK-58 (Cam3.IIsensitive IC50 0.075 μM, D10-PfPHG IC50 0.048 μM) and GSK-4 0.04 μM, D10-PfPHG IC50 (Cam3.IIsensitive 0.19 μM). Despite the much more stringent drug wash- out procedure employed for the Cam3.IIsensitive ring- stage survival assays, activity against early ring stages was equivalent to that seen for 6 hr treatment of D10- PfPHG ring stages and similar to in-cycle treatments of D10-PfPHG. These results support that azithromycin and analogues have rapid activity against early ring-stage parasites of different P. falciparum lines. IC50 We next tested the activity of azithromycin and ana- logues against the zoonotic malaria parasite P. knowlesi, which is a significant human pathogen in regions of Southeast Asia [48] and an in vitro culturable model for P. vivax [49]. We found that azithromycin maintains po- tency against P. knowlesi in both in-cycle (28 hr for P. knowlesi, Pk) and delayed-death (92 hr) assays compared to P. falciparum (Pf) (Pk in-cycle IC50 13 μM, delayed- death IC50 0.08 μM, Pf in-cycle IC50 11.3 μM, delayed- Burns et al. BMC Biology (2020) 18:133 Page 10 of 23 Fig. 5. Activity of azithromycin analogues against artemisinin-resistant parasites. Lead azithromycin analogues were tested against artemisinin- resistant Cam3.IIDHA resistant(R539T) parasites containing the K13 propeller mutation and reverted, artemisinin-sensitive, Cam3.IIsensitive parasites in ring-stage survival assays (4 hr drug pulse of very early rings 0–3 hrs post invasion) prior to washing off drug and assessment of parasitaemia (66 hrs later by flow cytometry). Dihydroartemisinin (DHA), azithromycin, GSK-3 (naphthalene), GSK-5 (substituted phenyl), GSK-56 (chloroquinoline) and GSK-71 (quinoline). Parasitemia was measured via flow cytometry ~ 72 hrs post-invasion. Data represents the mean of 2 or more experiments expressed as a percentage of non-inhibitory control and error bars represent ± SEM. Repeat measure data is available in Additional file 15 Supporting Value Data Table 2 Ring-stage survival assay percent survival values from drug treated artemisinin-resistant and artemisinin-sensitive parasites Modification Compound DHA Cam3.IIsensitive 72 hr growth IC50 (μM, ±SEM) 0.007 (0.002) Cam3.IIresistant 72 hr growth IC50 (μM, ±SEM) 0.011 (0.001) Azithromycin 30 (5.5) 0.035 (0.004) 0.2 (0.02) 0.004 (0.001) Naphthalene 2-Chlorophenyl 7- Chloroquinoline Quinolone 3 5 56 71 30 (0.005) 0.04 (0.004) 0.28 (0.005) 0.009 (0.001) 0.07 (0.007) 0.15 (0.03) Concentration of drug = 5% growth of Cam3.IIsensitive (μM) Growth Cam3.IIDHA resistant(R539T) (%) 0.05 100 0.4 0.5 0.055 0.6 41 1 8 6 7 6 The μM concentration of drug (DHA, azithromycin, GSK-3, GSK-5, GSK-56 and GSK-71) that resulted in a 5% survival value for artemisinin-sensitive Cam3.IIsensitive parasites was then used to treat artemisinin-resistant Cam3.IIDHA resistant(R539T) parasites, and the resulting % parasite survival value for the resistant parasites is displayed in the table. The IC50 value of the drugs against Cam3.IIsensitive and Cam3.IIresistant strains is also shown to indicate their overall potency against artemisinin-sensitive and artemisinin-resistant parasites. Parasites were incubated for one cycle (72 hrs) after pulsed drug treatment and washing prior to measurement of parasitaemia by flow cytometry. Data represent the mean of 2 experiments Burns et al. BMC Biology (2020) 18:133 Page 11 of 23 this divergent parasite species death IC50 0.07 μM) (Table 1) as previously shown [50]. We next tested a panel of azithromycin analogues that had potent quick-killing activity against P. falciparum for their efficacy against P. knowlesi and identified that the majority of analogues had similar quick-killing po- tency against (Add- itional file 7: Table S4). Of interest, the analogue GSK-9 exhibited a significant 33.1-fold improvement in activity against P. knowlesi when compared to activity against P. falciparum, suggesting that some species-specific differ- ences in drug activity can occur. Together, these data efficacy that azithromycin analogues have support against diverse human malaria parasites and across DHA and multi-drug-resistant parasites. bacteria pneumoniae Analogues modified at the macrolactone-ring maintain dual mechanisms of action We next sought to define whether the more potent quick- killing azithromycin analogues maintained apicoplast- targeting delayed-death activity. As quick-killing IC50s for a number of analogues (GSK-1, GSK-4, GSK-5, GSK-29, GSK-57, GSK-66, GSK-71, GSK-78) approached that of the delayed-death IC50 values of azithromycin (120 hr IC50 0.07 μM), the measurement of apicoplast targeting delayed- death activity (i.e. activity after 120 hrs of treatment, Fig. 1d) would likely be compromised by quick-killing potency. Therefore, we assessed the activity of azithromycin and a panel of quick-killing analogues against the azithromycin- Streptococcus (Add- sensitive itional file 8: Table S5) on the basis that this Gram-positive bacteria’s ribosome could serve as a proxy for the malaria parasite bacterium-like apicoplast ribosome [12, 51]. Con- sistent with previously published results, limited inhibition of bacterial growth was observed for analogues with an N- substitution on the desosamine sugar moiety [34, 35, 52]. Indeed, N-substituted analogues of azithromycin have been deliberately designed to reduce off-target drug activity against bacteria for use in alternative drug applications [34, 35, 52]. In contrast, all analogues with N6-substitutions on the macrolactone backbone (GSK-1, GSK-4, GSK-5, GSK- 6, GSK-9, GSK-11, GSK-12, GSK-16, GSK-17, GSK-21, GSK-25) had activity against S. pneumoniae similar to azi- thromycin. Thus, selecting the site of azithromycin modifi- cation can allow improved quick-killing activity while maintaining apicoplast targeting delayed-death activity, or delayed-death activity can be removed along with off-target antibacterial effects to produce a quick-killing specific antimalarial. Analysis of the quick-killing mechanism of action suggests a multi-factorial mechanism of action In an attempt to identify the molecular target of quick- killing activity, we selected for in vitro drug resistance by subjecting an azithromycin delayed-death-resistant D10 line (D10-AZRr) with a stepwise increase [12] of the quick-killing azithromycin analogue GSK-59 featuring a chloroquinoline-substituted desosamine moiety that lacks delayed-death activity. After three attempts, we failed to select for resistant parasites > 3 months after drug removal, suggesting that the mechanism of quick- killing cannot be readily selected for in vitro. We next undertook an untargeted metabolomics screen to identify changes in the metabolomic signature of azithromycin and the quick-killing analogues GSK-5 (substituted phenyl), GSK-66 (chloroquinoline) and GSK-71 (quinoline) and to compare changes during treatment with these analogues to known antimalarials, such as chloroquine and DHA (Fig. 6, Additional file 9: Figure S4, Additional file 10: Table S6, Additional file 11: Table S7, Additional file 12: Table S8). Following a 2 hr treatment of trophozoite-stage parasites at a 5× IC50 (44 hr) concentration, supervised multivariate ana- lysis (partial least squares-discriminate analysis) and heat map showed that the most prominent metabolomic sig- nature shared between azithromycin and analogues was a series of short peptides that were increased for all of azithromycin, GSK-71, GSK-5 and the food vacuole- targeting control drug chloroquine (Fig. 6, Add- itional file 10: Table S6a&b). Since increases in these peptides have previously been demonstrated for chloro- quine- and piperaquine-treated trophozoites [53], it is possible that this signature indicates a mechanism of ac- tion similar to the 4-aminoquinolines, which are thought to act by inhibiting crystallisation of haemoglobin- derived haem to form haemozoin within the parasite’s food vacuole. However, it was also noted in the study by Creek et al. that the sequences for the majority of these peptides are not derived from degraded haemoglobin, in- dicating that the metabolomic signature shared between chloroquine, azithromycin, GSK-71 and GSK-5 are likely due to disruption of proteolytic processes other than haemoglobin digestion. In addition, GSK-66 which has the most chloroquine-like functional group in terms of structure and was the most potent analogue tested in this study, showed little in the way of changed metabo- lites and gave a profile most similar to untreated control. Since chloroquine is known to disrupt the haemoglobin digestion pathway by inhibition of haemozoin formation [54–56], we next measured the levels of haemoglobin, haem and haemozoin in the parasites following treat- ment with analogues GSK-66 (chloroquinoline) and GSK-71 (quinoline) [57] (Fig. 7). Trophozoite-stage par- asites were treated with CQ, GSK-71, and GSK-66 at 10× IC50 for 5 hrs. There was an increase in measurable haemoglobin and a reduction in haemozoin formation for parasites treated with chloroquine, as expected for this known inhibitor of haemoglobin digestion and hae- mozoin formation. A similar build-up in haemoglobin Burns et al. BMC Biology (2020) 18:133 Page 12 of 23 Fig. 6. Hierarchical clustering of the different sample groups, treated with chloroquine (CQ) (blue), DHA (green), azithromycin (Az) (light blue), GSK-5 (purple), GSK-71 (yellow), GSK-66 (grey) and ethanol control (red). Vertical clustering displays similarities between sample groups, while horizontal clusters reveal the relative abundances of the 50 most significantly different metabolites from experiment 1. The significantly differentially regulated metabolites are further classified into three different groups, the CQ-like peptides (blue line), TCA cycle (red line) and haemoglobin-derived peptides (orange lines). All compounds were tested with three technical replicates. White indicates no change, while red and blue indicates increased and decreased abundances respectively. Ward’s minimum variance method algorithm was used to generate the hierarchical cluster analysis was seen for GSK-71; however, there was no decrease in haemozoin, supporting that this drug may have activity in the food vacuole, but this did not involve measurable inhibition of haemozoin formation. Again, GSK-66 treatment had no effect on haemoglobin or haemozoin levels, supporting the non-targeted metabolomics data which suggests that this drug has limited effects on para- the concentration and duration site metabolism at Burns et al. BMC Biology (2020) 18:133 Page 13 of 23 Fig. 7. Haemoglobin fractionation of GSK-71, chloroquine and GSK-66-treated Plasmodium falciparum (3D7) parasites compared to an ethanol control. Scatter dot plots representing the relative levels of a haemoglobin, b free haem and c haemozoin in trophozoite-stage parasites following a 5 hr incubation with 10× IC50 (44 hr) concentration of GSK-71 (1100 nM), chloroquine (520 nM) and GSK-66 (70 nM) expressed as the fold change when compared to an EtOH control. Data is represented as the mean of > 3 paired replicates from three independent experiments with the error bars expressed as SEM. Significant differences were assessed using Student’s t test. Repeat measure data is available in Additional file 15 Supporting Value Data. d A panel of representative Giemsa-stained parasites treated with 10× IC50 (44 hr) concentration of GSK- 71, chloroquine, GSK-66 and the ethanol negative control after 5 hrs. tested. These data support that azithromycin and ana- logues have activity in the food vacuole of drug-treated trophozoites, but also indicate additional activity outside of haemoglobin digestion. A second shared metabolomic signature was observed for azithromycin and the phenyl-substituted analogue GSK-5, with a major reduction in key metabolites (in- cluding succinate, fumarate, malate) of the mitochon- drial tricarboxylic acid (TCA) cycle (Additional file 11: Table S7a&b, Additional file 13: Figure S5). The reduc- tion in TCA metabolites was evident across repeat ex- periments for azithromycin, but was less prominent for GSK-5 in the second experiment (Additional file 11: Table S7a&b). Although several steps in the Plasmodium TCA cycle are considered dispensable in blood-stage parasites, the fumarate hydratase conversion of fumarate to malate followed by the malate quinone oxidoreduc- to (MQO) mediated conversion of malate tase oxaloacetate are thought to have important roles in the parasite’s purine salvage pathway [58, 59]. Reduced bio- availability of fumarate and malate, two key metabolites required for efficient purine salvage, would negatively impact on purine production and parasite growth over time and offers a novel drug development strategy. In- deed, a recent paper has identified blood-stage inhibitors of MQO in the Pathogen Box [60] suggesting that this pathway is a viable drug target against asexual-stage par- asites. These data implicate a second membrane-bound organelle as a potential target during trophozoite stages of the parasite lifecycle, underlining the potential for multifactorial mechanisms of action. Azithromycin and GSK-5 also caused a reduction in haemoglobin-derived peptides across both experiments to levels lower than seen for chloroquine and DHA, two food vacuole targeting drugs (Additional file 12: Table S8a&b). Thus, treatment with azithromycin and GSK-5 Burns et al. BMC Biology (2020) 18:133 Page 14 of 23 caused an increase in specific non-haemoglobin-derived peptides similar to that seen for chloroquine, a consist- ent decrease in haemoglobin-derived peptides (most prominently for GSK-5 in this data set) and a decrease in TCA cycle metabolites. In contrast, GSK-71 was most notably associated with an increase in non-haemoglobin chloroquine-like peptides, while GSK-66 and DHA had minimal impact on the metabolic profile under the con- ditions analysed here. This highlights the potential abil- ity of azithromycin analogues with different structures to interrupt normal metabolic functions across the cell and in different organelles, even when used at the same fold- IC50 and against the same lifecycle stages. Given the metabolomics evidence suggesting that azithromycin and analogues may target the food vacuole, we investigated whether the rapid ring-stage killing activity of the chloroquinoline analogue GSK-66 (Fig. 4a, b, Table 1) may be a result of azithromycin pre- sensitising ring stages to the chloroquinoline moiety. We treated early ring-stage D10-PfPHG parasites (0–6 hrs) with azithromycin at an IC10 concentration and added a dilution series of chloroquine. Addition of azithromycin did not potentiate chloroquine’s activity against early ring stages, with the IC50 of azithromycin+chloroquine remaining well above the activity of GSK-66 (Add- itional file 14: Figure S6). In addition, a range of func- tional groups were found to potentiate azithromycin’s quick-killing activity. These combined data suggest that to azithromycin does not chloroquinoline-like moieties nor act through disruption of haem polymerisation per se as chloroquine is believed to, but rather may act more broadly within the parasite’s food vacuole as well as potentially other cellular and organellar targets such as the parasite’s mitochondrion. pre-sensitise parasites Discussion The spread of parasites resistant to artemisinin combin- ation therapies (ACTs) in Southeast Asia, India and other regions highlights the need for novel antimalarial drug treatment strategies to ensure timely and effective treatment of clinical disease [3–6, 8]. Despite limited use against clinical cases of malaria, macrolide antibiotics re- main of interest as potential partner drugs in antimalar- ial combinations due to their activity against malaria parasites and well-established safety profile in children and pregnant women [10, 11, 24, 61]. Recently, we iden- tified that high concentrations of clinically used macro- lides inhibit merozoite invasion in vitro and showed that this mechanism of action was independent of apicoplast- targeting delayed death [29]. Here, we demonstrate the potential for the antibiotic azithromycin to be repur- posed as an antimalarial with two potent mechanisms of action with the identification of azithromycin analogues that have potent activity throughout intra-erythrocytic parasite development and against merozoite invasion. We established that this activity is through a mechanism independent of the known activity of azithromycin against the parasite apicoplast, revealing potential new pathways for development of novel antimalarials. We investigated the activity of a panel of the analogues and identified 65 with improved in-cycle activity (44 hr early rings to schizont treatment) compared to azithro- mycin. Of these, 39 analogues with diverse functional groups IC50 0.02 μM), naphthalene (GSK-3, IC50 0.183 μM), quin- IC50 0.048 μM) and chloroquinoline oline (GSK-58, (GSK-66, IC50 0.007 μM) had nanomolar IC50s, provid- ing between an 11- to 1615-fold improvement over azithromycin. including substituted phenyl (GSK-5, Azithromycin and analogues exhibited equipotent intracellular blood-stage quick-killing activity across parasite growth. This included rapid activity against early ring-stage development (both 0–6 and 0–12 hrs post in- vasion) at a similar potency to 0–44 hr (one cycle) treat- ments. Therefore, azithromycin and analogues have a similar efficacy profile to the artemisinins [37, 38], being effective against early ring stages and across the blood- stage lifecycle, but with additional potential to be active against liver and transmission-stage parasites [22, 26, 27]. We found that the azithromycin analogues with the best activity in 44 hr assays (GSK-3, GSK-5, GSK-56 and GSK-72) also exhibited the greatest improvement in in- vasion inhibitory activity over azithromycin, highlighting that both quick-killing activities can be improved over azithromycin. However, the ability to push potency of merozoite invasion inhibition into clinically relevant concentrations below 1 μM may be limited. Importantly, assays where merozoites were treated directly prior to compound removal and addition of RBCs to begin inva- sion show that the invasion inhibitory activity of azithro- mycin and analogues is directed against the merozoite and not against the RBC. A number of invasion inhibi- tory antimalarial strategies are being pursued globally (reviewed in [62]), and there remains the possibility that further improvements in azithromycin analogue invasion inhibitory additional development. achievable with IC50 are and quinoline chloroquinoline It is interesting to note that improved quick-killing ac- tivity is ubiquitous across analogues with phenyl, naph- functional thalene, groups. It has previously been hypothesised that the high potency of several analogues featuring quinoline and chloroquinoline moieties was due to these analogues act- ing like hybrid azithromycin (apicoplast ribosome target- ing) and chloroquine (food vacuole target) activity [33, 34] molecules. Interestingly, azithromycin analogues with the four functional groups display properties dis- similar to chloroquine, these being (i) improved invasion Burns et al. BMC Biology (2020) 18:133 Page 15 of 23 for lines against activity chloroquine-resistant analogues inhibitory activity compared to azithromycin, whereas chloroquine does not inhibit invasion [39, 63], and (ii) and similar featuring chloroquine-sensitive substituted phenyl, naphthalene and quinoline moieties. Activity against chloroquine-resistant DD2 for analogues with chloroquinoline functional groups was variable with two analogues showing improved potency against the chloroquine-resistant line over the chloroquine-sensitive line, while three compounds were less potent against the resistant line; and (iii) potent inhibition of very early ring stages (0–6 hrs post invasion), which are largely insensi- tive to chloroquine. However, additional evidence from this study does support the idea that azithromycin and analogues quick-killing activity may, in part, be acting against the parasite’s food vacuole. trends were observed with the Although our ability to perform comprehensive and detailed SAR comparison was limited by compound availability impacting on matched-pair analysis, some general analogues available. Analogues with chloroquinoline and quinoline substituents were generally the most potent in one-cycle 44 hr assays. Naphthalene had modest potency and is a close bioisostere of quinoline. In general, analogues with a short carbon linking the amino quinoline to the N6- position of the macrocycle or the O- or N-position of the desosamine group were the most active. Appending functional moieties to the N6-position of the macrolac- tone, or to the desosamine sugar, both conferred signifi- cantly improved in-cycle activity, with a slight tendency for improved quick-killing activity when the functional group was either attached to the N- or the O- of the des- osamine sugar as opposed to the N6-position of the macrolactone (i.e. chloroquinoline GSK-66desos (IC50 0.007 μM) and GSK-1macro (IC50 0.019 μM); naphthalene GSK-78desos (IC50 0.59 μM)). Thus, the position of the functional group on the macrocyclic did not greatly impact activity, suggest- ing the macrocycle may be acting as a vehicle for trans- portation of the active functionality. (IC50 0.51 μM) and GSK-12macro Within the parasite, it is possible that analogues are metabolised and then release the pendant quinoline or aromatic system as the active component of compound. This is possible either by an oxidative mechanism hydro- lysing amine-linked substituents, or by proteolytic or hydrolytic degradation of the amide and urea functional- ity linking the pendant quinoline or aromatic group to the macrolactone. In this study, we could not conclu- sively address whether metabolism was occurring, but this will be an important facet to address in a future mechanistic study of these azalide analogues. The possibil- ity of the macrolactone acting as a delivery vehicle with subsequent metabolic release of the active payload in the parasite raises the prospect for the azithromycin scaffold to be tethered to and act as a delivery vehicle for other an- timalarials that act at a similar asexual killing rate to chloroquine, akin to antimalarial candidates undergoing clinical trials such as KAF156 or MMV048 [64]. Such a strategy to improve dual target efficacy of azithromycin analogues, and delay the onset of resistance, is an attract- ive option. Furthermore, while it has been demonstrated that these analogues have efficacy in in vivo rodent models [31, 33, 35], the effective contribution of quick-killing has not been assessed. In addition, whether these analogues would be stable to first pass metabolism in the liver is an- other important aspect to consider in future development of the azalide analogue class. Although the azithromycin analogues identified as having improved quick-killing activity in this study fea- ture a range of added functional groups, compounds with quinoline and chloroquinoline moieties feature prominently amongst the most potent quick-killing ana- logues. Hybrid molecules featuring quinolines fused to a second chemotype with antimalarial properties such as endoperoxides [65] or reversed chloroquine drugs that are linked to a reversal agent, a molecule known to in- hibit or circumvent the activity of the chloroquine resist- ance transporter PfCRT [66, 67], have been developed and shown to have efficacy in rodent malaria models (reviewed in [68]). The current lead reversed chloro- quine compound, DM1157 [69], has shown low nano- molar potency against chloroquine-resistant parasites, demonstrated efficacy against P. chabaudi rodent mal- aria parasites and has recently undergone Phase I trials in humans (NCT03490162, [70]). Despite the potential of DM1157, hybrid molecules have faced hurdles in de- velopment including examples of endoperoxide hybrids unable to overcome existing resistance mechanisms [71] and the high MW of the compounds impacting on desir- able drug-like properties. In this regard, it is interesting to note that the ketolide antibiotics solithromycin and telithromycin, semi-synthetic derivatives of erythromycin which both feature a large functional group added to the macrolactone ring, have been progressed for clinical use. This highlights that modified macrolides can be devel- oped that maintain favourable drug-like properties des- pite their high MW. Metabolomic analysis of azithromycin and analogue- treated parasites suggests one potential site of drug activity in trophozoite stages is the parasite’s food vacu- ole, with a similar build-up of largely non-haemoglobin peptides observed for azithromycin, analogues GSK-5 and GSK-71 as seen for chloroquine. However, a num- ber of differences to chloroquine were also observed in- cluding the chloroquinoline-modified analogue GSK-66 causing minimal change in parasite metabolism, azithro- mycin and GSK-5 having activity against mitochondrial metabolism and GSK-5 also causing a reduction in Burns et al. BMC Biology (2020) 18:133 Page 16 of 23 haemoglobin-derived peptides. Previous studies have shown that trophozoite-stage treatment with the mito- chondrial targeting drug atovaquone, alone and in com- bination with proguanil, leads to a build-up of the TCA metabolite fumarate [53, 72]. It was postulated that this could be a result of the TCA enzyme malate-quinone oxidoreductase complex also having a role in the mito- chondrial electron transport chain (the target of atova- quone) that may be affected by atovaquone, leading to off-target disruption of the TCA cycle. In contrast, azi- thromycin and GSK-5 treatment caused a reduction in fumarate and other TCA metabolites, a signature differ- ent to that of atovaquone. Interestingly, treatment with the membrane-bound glucose transporter inhibitor 3361 led to a reduction in TCA and haemoglobin-derived peptides after 6 hrs of drug treatment [72], similar to that seen for azithromycin and GSK-5 here. The mul- tiple changes in parasite metabolic networks seen when inhibiting glucose uptake supports data generated in this study that suggests azithromycin and analogues quick- through multifactorial occur killing mechanisms. activity may While there are limitations in this analysis, including only one lifecycle stage and drug concentration (5× the 44 hr IC50) tested for each analogue, these data clearly demonstrate that azithromycin and analogues likely have multi-factorial mechanisms of action even against a sin- gle lifecycle stage. Given the apparent site of activity for azithromycin and analogues includes the membrane- bound food vacuole and mitochondrion, it is possible that additional membrane-bound organelles in other life- cycle stages (i.e. the rhoptry in merozoites) could also be the target of these drugs. Additional experimental valid- ation for the site of activity across a range of analogues and lifecycle stages will need to be undertaken in order to detail the potential promiscuity of these drugs in stopping parasite growth. Previous studies have suggested that azithromycin ana- logues may act through a chloroquine-like mechanism [33–35] (reviewed in [73]), and evidence presented in this study from metabolomic experiments and haemoglobin fractionation assays supports that one of the sites of activ- ity for azithromycin and analogues is the parasite’s food vacuole. If a chloroquine-like targeting of the food vacuole is an important component of azithromycin and analogue quick-killing activity, these modified analogues have two major advantages over chloroquine and quinine for clin- ical and quinoline-substituted analogues maintained reasonable activity against chloroquine-resistant DD2 parasites. The maintenance of potency against chloroquine-resistant par- asites could be explained by the different properties of the drug limiting the ability of the mutated chloroquine- the drug from the to expel resistant Firstly, phenyl-, naphthalene- transporter treatment. developing vacuole [74, 75]. Secondly, azithromycin and analogues have rapid activity against early ring-stage para- sites. Rapid activity against ring stages is in stark contrast to the poor activity of chloroquine and quinine against these early parasites and it is certainly possible that azi- thromycin and analogues could access the site of the ini- tial stages of haemoglobin digestion, similar to artemisinin [37, 38, 76], via superior lipophilic properties [33, 34]. [43] D10-PfPHG chloroquine- and Azithromycin and analogues display several other properties of interest. The majority of quick-killing analogues tested against chloroquine/pyrimethamine- artemisinin-resistant DD2 resistant and Cam3.IIDHA resistant(R539T) [44, 45] retained potency and artemisinin- compared to the sensitive artemisinin-sensitive Cam3.IIsensitive lines. While there were examples of chloroquinoline containing analogues being less potent against DD2 parasites, these data broadly indicate that a wide range of azithromycin analogue modifications can significantly improve quick-killing activity in a way that overcomes a number of established resistance mecha- nisms. Azithromycin and analogue invasion blocking ac- tivity is shared across distantly related Apicomplexan parasites such as Toxoplasma gondii [29, 77], P. berghei [29] and the zoonotic human malaria parasite P. knowlesi. Since neither T. gondii nor Plasmodium spp. merozoites contain a food vacuole, the target of chloroquine, it seems likely that azithromycin and analogues have additional mechanisms of action, with properties such as modulation of intraerythrocytic calcium (Ca2+), interference of kinase signalling pathways, cationic trapping and sequestration within acidic environments, as well as decreasing mobility of phospholipid bilayers demonstrated for azithromycin in other eukaryotic cell systems, all potential alternative MOAs contributing to quick-killing [78–82]. Finally, the influence of the site of modification to azi- functional thromycin and the addition of different groups was investigated in the context of delayed-death activity. Previous studies have demonstrated that the desosamine sugar is critical for binding to bacterial ribo- somes, and we anticipated that modifications to this re- gion would stop apicoplast-targeting delayed-death activity [12, 51, 52]. However, the potent quick-killing activity of azithromycin analogues (GSK-4, GSK-5, GSK- 12, GSK-16, GSK-57, GSK-71, etc.) precluded assess- ment of delayed-death activity using traditional 120 hr parasite assays. Therefore, we assessed whether a fo- cused set of azithromycin analogues maintained their ac- tivity against prokaryotic ribosomes by determining the minimum inhibitory concentration (MIC) activity of the gram-positive bacteria, S. pneumoniae. Comparison of P. falciparum quick-killing IC50 and S. pneumoniae MIC confirmed that attaching the functional group to the desosamine sugar (GSK-57, GSK-66, GSK-71 and GSK- Burns et al. BMC Biology (2020) 18:133 Page 17 of 23 78) abrogated activity against bacterial ribosomes as ex- pected. In contrast, analogues with the functional group attached to the N6-positon of the macrolactone (GSK-1, GSK-4, GSK-5, GSK-6, GSK-9, GSK-11, GSK-12, GSK- 16, GSK-17, GSK-21, GSK-25) maintained activity against S. pneumoniae, suggesting that delayed-death ac- tivity via targeting the bacterium-like ribosome of the apicoplast is maintained in analogues featuring modifica- tion to the N6-positon of the macrolactone (GSK-1, GSK-4, GSK-5, GSK-6, GSK-9, GSK-11, GSK-12, GSK- 16, GSK-17, GSK-21, GSK-25). Thus, analogues could be modified to act through either single (i.e. quick- killing) or dual (i.e. quick-killing and delayed-death) mechanisms of action depending on the properties de- sired (i.e. quick parasite clearance and/or long-term prophylaxis) and whether removal of non-selective anti- biotic activity is preferred over apicoplast-targeting delayed-death prophylaxis. azithromycin analogues intracellular blood-stage development, Conclusion We have shown that azithromycin and analogues have a quick-killing mechanism of action that kills parasites throughout in- cluding inhibition of merozoite invasion of RBCs. Add- exhibit promising itionally, potency against very early ring-stage parasites, which is a rare feature amongst existing antimalarials. Importantly, quick-killing can be improved without losing activity against protein synthesis by the apicoplast ribosome (de- layed death). Conversely, the option to engineer azithro- mycin to remove activity against a bacterium-like ribosome and thereby avoid selection for ‘bystander’ bacterial resistance is available. Further development of azithromycin analogues offers the prospect of designing compounds with either quick-killing (quick-parasite clearance) mode of action or both quick-killing and slow-killing prophylactic activity. This design strategy should also retard resistance acquisition by hitting two targets. Fine-tuning the quick-killing activity of azithro- mycin analogues significantly broadens its clinical appli- cations and offers resistance proofing through two independent mechanisms of action. Therefore, the iden- tification of potent azithromycin analogues with rapid killing phenotypes and dual mechanisms of action (de- layed-death and quick-killing activity) provide a new av- enue for anti-malarial drug development. Sigma), azithromycin (100 mM, AK-Scientific) and GSK analogues (10 mM, GSK-1–84) were made up in ethanol as vehicle. Chloroquine diphosphate salt (10 mM, Sigma- Aldrich) was dissolved in 10% acetic acid in H2O. Dihy- droartemisinin (10 mM, DHA, Sigma-Aldrich) were dis- solved in dimethyl sulfoxide (DMSO). Drugs were added such that the vehicle was diluted > 100-fold for merozo- ite invasion assays and > 1000-fold for intracellular growth assays to minimise non-specific inhibition. Culture and synchronisation of Plasmodium spp. parasites falcip- Green fluorescent protein (GFP) expressing P. arum D10-PfPHG parasites [84], DD2 [43], artemisinin- (Cam3.IIDHA resistant(R539T)) and artemisinin- resistant sensitive (Cam3.IIsensitive) Cambodian isolates [45] and P. knowlesi PkYH1 [85] were cultured in human O+ eryth- rocytes (RBCs) (Australian Red Cross Blood Service). Parasites were cultured in RPMI-HEPES culture medium (pH 7.4, 50 μg/mL hypoxanthine, 25 mM NaHCO3, 20 μg/mL gentamicin, 0.5% Albumax II (Thermo Fisher Scientific)) and maintained in an atmosphere of 1% O2, 4% CO2 and 95% N2 according to established protocols [86]. Tight synchronisation of D10-PfPHG parasites was achieved using sodium heparin [63, 87]. P. falciparum DD2, the Cambodian isolates and P. knowlesi (PkYHI), were synchronised with continuous passage over a gradi- ent of 70% Percoll (Sigma-Aldrich) for purification of late-stage schizonts and 5% w/v sorbitol (Sigma-Aldrich) treatments for ring stages. Drug inhibition assays A diagram outlining the different Plasmodium spp. drug inhibition assays used in this study is available in Fig. 1 and has been described previously [29, 63]. Stage specifi- city assessment of azithromycin or analogues during blood-stage P. falciparum development was undertaken through the addition of the drug at the specified time points (0–6 hrs, 0–12 hrs, 12–24 hrs, 24–36 hrs or 36– 44 hrs post merozoite invasion) and the subsequent re- moval through three consecutive washes with 200 μl of medium (centrifuged at 300×g for 2 min) before resus- pending in a final volume of 200 μl. Parasite growth was quantified at late schizont stages (44–48 hrs post inva- sion) by flow cytometry of parasites stained with 10 μg/ mL ethidium bromide (EtBr) for 1 hr prior to washing with PBS. Methods Antimalarial drugs Azithromycin analogues (GSK-1–84) were a gift from GlaxoSmithKline and were synthesised as described pre- viously [31–35, 83]. Additional file 1: Tables S1a-c pro- vides further details of chemical structure and analogue (3075 mM origin. Stock concentrations of quinine Invasion inhibition assays Purification of viable merozoites and merozoite invasion inhibition assays has been described previously [29, 63, 87]. Briefly, 300 mL of D10-PfPHG schizont culture, 3% haematocrit and 4–5% parasitaemia tightly synchronised to a 6 hr window of invasion with heparin were magnet purified (Mitenyi Biotech) away from RBCs at 40–46 hrs Burns et al. BMC Biology (2020) 18:133 Page 18 of 23 post-invasion. Purified schizonts were eluted in up to 30 mL of media, 10 μM of E64 (Sigma-Aldrich) was added and the parasites were left to mature for 5 hrs. Schizonts were filtered through a 1.2-μm syringe filter (Minisart, Sartorius) in incomplete media with NaHCO3 to release merozoites and 22.5 μl of filtrate was added to 2.5 μl of drug prior to addition of RBC (0.5% final haem- atocrit). Plates were agitated at 400 rpm for 10 min at 37 °C to promote invasion. For drug washout, 90 μL of purified merozoites was added to 10 μL of either incomplete media (no serum) or incomplete media plus drug before transfer to a 0.22- μm Ultrafree-MC centrifugal filter (Thermo Fisher). Fil- ter columns were centrifuged at 750 rcf for 1 min and washed with incomplete media twice. Free merozoites were resuspended off the filter in 45 μL of incomplete media and transferred to 96-well U-bottom plates con- taining 5 μL of RBCs at 1% haematocrit (final haemato- crit of 0.1%). Plates were agitated at 400 rpm for 10 min at 37 °C and cultures were incubated at 37 °C for 30 min. Cells were treated with 5 μg/mL EtBr for 10 min prior to being washed in 1 x PBS and ring-stage parasitemia measured by flow cytometry. Ring-stage survival assays (RSA0-3h) For ring-stage survival assays [44–46], tightly synchro- nised artemisinin-resistant Cam3.IIDHA resistant(R539T) and artemisinin-sensitive Cam3.IIsensitive late schizont stage parasites were concentrated over a gradient of 70% Per- coll (Sigma-Aldrich), washed once in complete medium and incubated for 3 hrs with fresh RBCs to allow inva- sion. Cultures were sorbitol treated to eliminate the remaining schizonts. The 0–3 hr post-invasion rings were adjusted to 1% parasitemia and 1% haematocrit be- fore exposure to a serial dilution of DHA, azithromycin and azithromycin analogue concentrations for 4 hrs. Plates were washed five times with 200 μl of medium be- fore parasites were transferred into a new 96-well plate to ensure the complete removal of drug [47]. Parasites were grown for a further 66 hrs, before parasitemia was assessed by flow cytometry. Apicoplast-null inhibition assays Apicoplast-null (D10-PfPHGapicoplast-null) [17, 36] parasites were generated through supplementation of culture media with 200 μM isopentenyl pyrophosphate (IPP) and apico- plast removal through treatment with 0.35 μM (5× IC50) of azithromycin for 6 days, with parasites cultured con- tinuously thereafter with IPP. Removal of the apicoplast was confirmed by growing D10-PfPHGwildtype and D10- PfPHGapicoplast-null (+IPP) parasites with reducing concen- trations of azithromycin for ~ 120 hrs which identified a ~ 64 fold-change in the IC50 values between the parasite populations (D10-PfPHGapicoplast-null IC50, 4.5 μM; D10- PfPHGwildtype IC50, 0.07 μM) confirming apicoplast re- moval. To test for azithromycin analogue activity against the apicoplast, tightly synchronised ring-stage D10-PfPHGapicoplast-null (+IPP) and D10-PfPHGwildtype parasites were treated with the in-cycle 90% inhibitory concentration (IC90) of drugs obtained for D10- PfPHGwildtype for ~ 44 hrs (in-cycle) and the resulting growth inhibition determined by flow cytometry. Flow cytometry and microscopy analysis of inhibition Parasitaemia was measured on an LSR Fortessa (Becton Dickinson) with a 96-well plate reader. Mature (> 36 hr post-invasion) P. falciparum D10-PfPHG parasites were counted using Fl-1-high (GFP; excitation wavelength, 488 nm) and Fl-2-high (EtBr; excitation wavelength, 488 nm). D10-PfPHG ring-stage parasites (< 6 hrs post inva- sion) were counted using a Fl-1-high (GFP) and Fl-2-low (EtBr) gate [63]. Mature parasites of the remaining lines were gated with a forward scatter (FSC) and FL-2-high (EtBr) gate [63]. Typically, 20,000–40,000 RBCs were counted in each well. Samples were analysed using FlowJo software (TreeStar Inc) with growth of drug treatments normalised against media control wells to calculate the percentage survival. Thin smears for mi- croscopy were fixed with fresh methanol and stained in 10% Giemsa (Merck) for 10 min. IC50s and IC90s were determined for each drug using GraphPad Prism (GraphPad Software) according to the recommended protocol for nonlinear regression (constrained to top = 100 and bottom = 0) of a log-(inhibitor)-versus-response curve. Selection of azithromycin-resistant P. falciparum lines In vitro selection of quick-killing-resistant lines was car- ried out using a P. falciparum (D10-PfPHG) line featur- ing a G91D mutation in the apicoplast ribosomal gene, rpl4, resulting in a ~ 57-fold loss of sensitivity to azithro- mycin’s delayed-death activity (2 cycles, Fig. 1d) (D10- AZRr). To select for quick-killing resistance [12], D10- AZRr parasites were first exposed to 3× IC50 of GSK-59 (chloroquinoline moiety, delayed-death inactive drug) for 3 days, followed by a 5× IC50 concentration for 4 days then 3× IC50 for an additional 2 days prior to re- moval of the drug. After treatment, parasites were fed once every 2 days, and once a week, 30–40% of culture was replaced with fresh RBCs. Parasites were examined every 2 to 3 days by Giemsa-stained thin blood films for between 3 (90 days) and 5 months (150 days) with no re- crudescent parasites observed. Antibacterial screen Antibacterial activity of azithromycin and analogues against Streptococcus pneumoniae was determined using 96-well minimum inhibitory concentration (MIC) assays Burns et al. BMC Biology (2020) 18:133 Page 19 of 23 [88]. Two-fold serial dilutions were added to macrolide- sensitive D39 S. pneumoniae in 100 μL Mueller Hinton Broth supplemented with 5% lysed horse blood. Bacterial growth was assessed after 24 hrs incubation with drug by estimating the MIC where bacterial growth, as indicated by a media colour change, could be identified (MIC expressed as μM). Sample extraction for metabolomics analysis For metabolomics experiments, two 150-mL flasks at 6% haematocrit containing tightly synchronised ~ 30–34 hr D10-PfPHG trophozoites were harvested via magnet purification (Miltenyi Biotech). Infected RBC density was quantitated by flow cytometry [89], and 2 mL of 3 × 107 parasites were added to and incubated in 24-well mi- crotiter plates for 1 hr at 37 °C to stabilise the culture. Drugs (5× IC50) were added and incubated for a further 2 hrs prior to removal of the supernatant, 2× washes with 800 μL ice-cold 1× PBS with cells pelleted via cen- trifugation at 400×g for 5 min at 4 °C. The cell pellets were resuspended in 150 μL of ice-cold extraction buffer (MeOH) containing 1 μM internal standards; CHAPS and PIPES, and incubated on ice for 1 hr with shaking at 200 rpm. Insoluble material was pelleted with centrifuga- tion at 14,800×g for 10 min at 4 °C and 120 μL of super- natant was collected and stored at − 80 °C until analysis. spectrometry LC-MS analysis (LC-MS) Liquid chromatography-mass data was acquired on a Q-Exactive Orbitrap mass spectrometer (Thermo Scientific) coupled with high- performance liquid chromatography system (HPLC, Dionex Ultimate® 3000 RS, Thermo Scientific) as previ- ously described [53]. Briefly, chromatographic separation was performed on a ZIC-pHILIC column equipped with a guard (5 μm, 4.6 × 150 mm, SeQuant®, Merck). The mobile phase (A) was 20 mM ammonium carbonate (Sigma Aldrich), and (B) acetonitrile (Burdick and Jack- son) and needle wash solution was 50% isopropanol. The column flow rate was maintained at 0.3 ml/min with temperature at 25 °C and the gradient programme was as follows: 80% B decreasing to 50% B over 15 min, then to 5% B at 18 min until 21 min, increasing to 80% B at 24 min until 32 min. Total run time was 32 min with an injection volume of 10 μL. A mass spectrometer was op- erated in full scan mode with positive and negative po- larity switching at 35k resolution at 200 m/z, with detection range of 85 to 1275 m/z, AGC target was 1e6 ions with a maximum injection time of 50 ms. Electro- spray ionisation source (HESI) was set to 4.0 kV voltage for positive and negative mode, and sheath gas was set to 50, aux gas to 20 and sweep gas to 2 arbitrary units, capillary temperature to 300 °C and probe heater temperature to 120 °C. The samples were analysed as a single batch to avoid batch-to-batch variation and ran- domised to account for LCMS system drift over time. Repeated analysis of pooled quality control samples was the batch to confirm signal performed throughout reproducibility. Data processing using IDEOM The acquired LCMS data was processed in untargeted fashion using open source software, IDEOM [90] (http:// Initially, Proteo- mzmatch.sourceforge.net/ideom.php). Wizard was used to convert raw LC-MS files to mzXML format and XCMS was used to pick peaks. Mzmatch.R was used to convert to peakML files, align samples and filter peaks using minimum detectable intensity of 100, 000, relative standard deviation (RSD) of < 0.5 (reprodu- cibility), and peak shape (codadw) of > 0.8. Mzmatch was also used to retrieve missing peaks and annotate related peaks. Default IDEOM parameters were used to elimin- ate unwanted noise and artefact peaks. Loss or gain of a proton was corrected in negative and positive ESI modes, respectively, followed by putative identification of metabolites by accurate mass within 3 ppm mass error searching against common metabolite databases includ- ing the Kyoto Encyclopedia of Genes and Genomes (KEGG), MetaCyc and LIPIDMAPS. To reduce the number of false positive identifications, retention time error was calculated for each putatively identified metab- olite using IDEOM’s build-in retention time model which uses actual retention time data of authentic standards (~ 350 standards). Metabolites identified by comparison to authentic standards (including TCA cycle metabolites) are level 1 identifications according to the other Metabolomics putatively identified metabolites (including all peptides) are assigned as level 2. Statistical analysis on filtered data was performed using the Metaboanalyst web interface [91]. Standards Initiative, and all Haemoglobin fractionation The haemaglobin fractionation assay was adapted from [57]. Aliquots of 6.5 mL of 30–32 hr post invasion para- site cultures were adjusted to 8% parasitaemia and 2% haematocrit and then incubated with chloroquine, GSK- 66, GSK-71 or ethanol (vehicle control) for 5 hrs. Treat- ments were performed in triplicate. Following incuba- tion, the media was aspirated off and the culture was incubated with 2.3 mL of 0.1% saponin in 1× PBS with protease inhibitors (complete mini protease inhibitor cocktail (Roche)) for 10 min at 4 °C in order to lyse the iRBCs. The parasites were washed three times with PBS and stored at − 80 °C. For the haemoglobin fractionation, lysed parasites were resuspended in 50 μL of Milli-Q water and soni- cated for 5 min in a water bath sonicator. Following Burns et al. BMC Biology (2020) 18:133 Page 20 of 23 sonication, 50 μL of 0.2 M HEPES (pH 7.5) was added and the samples were centrifuged at 4000 rpm for 20 min. The supernatant containing the haemoglobin frac- tion was carefully transferred to new tubes and 50 μL of 4% of SDS was added before the samples were incubated at 95 °C for 5 min. Following heating, 50 μL of 0.3 M NaCl and 50 μL of 25% (v/v) pyridine (Sigma) in 0.2 M HEPES was added, the sample containing the haemoglo- bin fraction were vortexed and transferred to a 96-well plate. The remaining pellets were treated with 50 μL of MilliQ water and 50 μL of 4% SDS and resuspended be- fore being sonicated for 5 min and incubated at 95 °C for 5 min in order to solubilise the free haem. Following in- cubation, 50 μL of 0.2 M HEPES, 0.3 M NaCl and 25% pyridine were added to the samples. The samples were then subsequently centrifuged at 4000 rpm for 20 min. The supernatant was transferred to the 96-well plate, corresponding to the free haem fraction. The remaining pellet containing the haemozoin frac- tion was solubilised by resuspending with 50 μL of MilliQ water and 50 μL of 0.3 M NaOH. The samples were sonicated for 15 min before 50 μL of 0.2 M HEPES, 0.3 M HCl and 25% pyridine was added. The samples were then transferred to the 96-well plate, corresponding to the haemozoin fraction. The total amount of haem in each fraction was quantified using a haem standard curve prepared from a 100 μg/mL standard solution of haematin in 0.3 M NaOH. Serial dilution of the standard curve was carried out in a 96-well plate in triplicate, and 50 μL of 0.2 M HEPES, 4% SDS, 0.3 M NaCl, 0.3 M HCl and 25% pyridine was added. The absorbance of the standard curve and each fraction was measured at a 405- nm wavelength using a Perkin Elmer Ensight Plate Reader. The samples were normalised via a paired ana- lysis to the ethanol control and graphed as their fold change vs ethanol ± SEM. All fractions had > 2 replicates from 2 independent experiments. Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s12915-020-00859-4. Additional file 1 : Table S1. Activities of azithromycin analogues. Additional file 2 : Figure S1. Azithromycin analogues show improvement in invasion inhibitory activity. (A) Screening a panel of azithromycin analogues identified 7 with up to 6-fold lower invasion in- hibitory IC50 activity in contrast to the parental azithromycin. IC50 Azithro- mycin 10 μM; GSK-4, 2.0 μM (Azithromycin vs GSK-4 P < 0.0001); GSK-5, 1.61 μM (Azithromycin vs GSK-5 P < 0.0001); GSK-56, 3.2 μM (Azithromycin vs GSK-56 P < 0.0001); GSK-8, 4.4 μM (Azithromycin vs GSK-8 P = 0.2); GSK- 3, 1.8 μM (Azithromycin vs GSK-3 P < 0.0001); GSK-15, 3.6 μM (Azithromy- cin vs GSK-15 P < 0.001); GSK-72, 1.7 μM (Azithromycin vs GSK-72 P < 0.0001). Newly invaded ring-stage parasitemia was measured at 1 hr post invasion via flow cytometry. Data represents the mean of 2 (GSK 5) or more experiments expressed as percentage of non-inhibitory control. Error bars represent ± SEM. Dose response IC50s compared using extra sum of squares F-test. (B) The food-vacuole targeting antimalarial drugs chloroquine and quinine showed minimal invasion inhibitory activity at 10 μM while merozoite invasion was blocked by the invasion inhibitory control heparin (25 μg/mL). Data represents the mean of 3 experiments expressed as percentage of non-inhibitory control. Error bars represent ± SEM. Repeat measure data is available in Additional file 15 Supporting Value Data. Additional file 3 : Figure S2. Azithromycin analogues inhibit merozoite invasion irreversibly. Whether azithromycin analogues inhibited invasion through a direct effect on the merozoite, rather than an effect on the RBC, was assessed by directly treating and then washing the drug off purified merozoites. Analogue GSK-72 was chosen as a compound with improved invasion inhibitory activity over azithromycin with merozoites treated at 10 μM. The actin inhibitor cytochalasin D (cytoD, 500 μM) was included as an irreversible washout control. The RON2 binding peptide R1 (100 μg/mL) was included as a reversible control. Ring-stage parasit- aemia of newly invaded parasites was determined ~ 30 min post invasion by flow cytometry, with results presented as % parasitaemia relative to a media control. Results represent the mean of 2 experiments and the error bars represent the ± SEM. Repeat measure data is available in Add- itional file 15 Supporting Value Data. Additional file 4 : Figure S3. Growth inhibition profiles of azithromycin and analogues in parasites lacking the apicoplast. Early ring-stage P. fal- ciparum parasites (0–4 hrs post-invasion) were treated with doubling dilu- tions of azithromycin and inhibition of growth measured for (A) 2 cycle (delayed death, 120 hr) assays (D10-PfPHGapicoplast-null IC50, 4.5 μM; D10- PfPHGwildtype IC50, 0.07 μM. P = < 0.0001) or (B) 44 hr (in-cycle) (D10- PfPHGapicoplast-null IC50, 16 μM; D10-PfPHGwildtype IC50, 11.3 μM. P = 0.24) as- says. Parasitemia was measured at 120 hrs or 44 hrs post invasion, re- spectively, at schizont stage via flow cytometry. Data represents the mean of 3 (or more) experiments expressed as percentage of non- inhibitory control and error bars represent ± SEM. (C) There was no differ- ence in 44 hr IC50s between D10-PfPHGapicoplast-null and D10-PfPHGwildtype parasites when treated with the azithromycin analogues GSK 1 (D10- PfPHGapicoplast-null IC50, 0.028 μM; D10-PfPHGwildtype IC50, 0.023 μM. P = 0.36) and GSK 66 (D10-PfPHGapicoplast-null IC50, 0.009 μM; D10-PfPHGwildtype IC50, 0.007 μM. P = 0.08). Data represents the mean of 2 (D10-PfPHGapicoplast-null) or 3 (D10-PfPHGwildtype) experiments expressed as percentage of non- inhibitory control and error bars represent ± SEM. Dose response IC50s compared using extra sum of squares F-test. Repeat measure data is available in Additional file 15 Supporting Value Data. Additional file 5 : Table S2. Azithromycin analogue activity across different age ranges of D10-PfPHG blood stage development. Additional file 6 : Table S3. Azithromycin analogue inhibition for chloroquine sensitive and resistant lines. Additional file 7 : Table S4. Azithromycin analogue activity against P. falciparum D10-PfPHG and P. knowlesi YH1 parasites. Additional file 8 : Table S5. Azithromycin analogue activity against the bacterial pathogen Streptococcus pneumoniae compared to P. falciparum D10-PfPHG. Additional file 9 : Figure S4. Sparse partial least square-discriminant ana- lysis (SPLS-DA) of Plasmodium falciparum (D10-PfPHG)-infected red blood cells following treatment with DHA (green), chloroquine (blue), azithro- mycin (light blue), GSK-5 (purple), GSK-71 (yellow), GSK-66 (grey), and ethanol control (red) from experiment 1. sPLS-DA showing scores plot for components one and two, the plots were generated using the top 10 metabolites for each component. Points represent individual sample rep- licates while the 95% confidence interval is represented by the shaded re- gion. (File format .pdf). Additional file 10 : Table S6. Changes in metabolites upon azithromycin and analogue treatment shared with chloroquine treated parasites. Additional file 11 : Table S7. Changes in metabolites upon azithromycin and analogue treatment associated with the parasite TCA cycle. Additional file 12 : Table S8. Changes in metabolites upon azithromycin and analogue treatment mapping to haemoglobin after drug treatment. Burns et al. BMC Biology (2020) 18:133 Page 21 of 23 Additional file 13 : Figure S5. Model for TCA metabolism following treatment of Plasmodium falciparum (D10-PfPHG)-infected red blood cells. Relative abundance of the TCA metabolites from infected red blood cells treated with DHA (blue), chloroquine (red), azithromycin (green), GSK-5 (purple), GSK-71 (orange), GSK-66 (black), compared with the Ethanol con- trol from experiment 1. Data are represented as mean fold change from triplicate treatments multiplied by corresponding RSD values. Abbrevia- tions: OAA, oxaloacetate; PEP, phosphoenolpyruvate. Additional file 14 : Figure S6. Azithromycin does not pre-sensitise early- ring stages to chloroquine. Early ring-stage P. falciparum parasites (0–4 hrs post-invasion) were treated with doubling dilutions of chloroquine (IC50; 0–6 hrs, 0.73 μM), chloroquine + IC10 of azithromycin (IC50; 0–6 hrs, 1.1 μM) or highly potent analogue GSK-66 which features a chloroquino- line moiety (IC50; 0–6 hrs, 0.004 μM) for 0–6 hrs, prior to removal of drugs by washing. Comparison of the resulting in-cycle growth shows a small change between growth of chloroquine vs chloroquine + azithromycin treated parasites (P = 0.0041). This compares to a large difference be- tween growth inhibitory IC50 of GSK-66 and chloroquine (P < 0.0001) and chloroquine + azithromycin (P < 0.0001), indicating that azithromycin does not potentiate ring stage activity of chloroquine. Parasitemia was measured at 44 hrs post invasion at schizont stage via flow cytometry. Data represents the mean of 3 (or more) experiments expressed as per- centage of non-inhibitory control and error bars represent ± SEM. Dose response IC50s compared using extra sum of squares F-test. Repeat meas- ure data is available in Additional file 15 Supporting Value Data. (File for- mat .pdf). Additional file 15 : Supporting data values. Excel Spreadsheet containing repeat measure data for Figs. 3, 4, 5 and 7, and Additional files 2, 3, 4 and 15. Abbreviations ACT: Artemisinin combination therapies; RBC: Red blood cell; Ca2+: Calcium; DMSO: Dimethyl sulfoxide; EtBr: Ethidium bromide; FIC: Fractional Inhibitory Concentration; FSC: Forward scatter; GFP: Green fluorescent protein; GSK: GlaxoSmithKline; HPLC: High-performance liquid chromatography; IC: Inhibitory concentration; IPP: Isoprenoid pyrophosphate; IPTp: Intermittent preventative treatment for malaria in pregnancy; KEGG: Kyoto Encyclopedia of Genes and Genomes; LC-MS: Liquid chromatography-mass spectrometry; MAPK: Mitogen-activated protein kinase; MIC: Minimum inhibitory concentration; pf: Plasmodium falciparum; pk: Plasmodium knowlesi; RSD: Relative standard deviation; RSA: Ring-stage survival assay; SEM: Standard error of the mean; WHO: World Health Organization Acknowledgements Dr. Francisco Javier Gamo and Dr. Noemi Bahamontes Rosa (GlaxoSmithKline, Tres Cantos facility, Spain) for the provision of modified azalides. Dr. Jeremy Burrows, Medicines for Malaria Venture, for helpful discussion and advice. David Fidock and Leann Tilley for providing the laboratory adapted Cam3.IIDHA resistant(R539T) and Cam3.IIsensitive lines. We thank Juan Miguel Balbin for help in generating the diagrams. Human erythrocytes were kindly provided by the Red Cross Blood Bank (Adelaide, Australia). Metabolomics analysis was performed at the Monash Proteomics and Metabolomics Facility. Authors’ contributions DW, BS, CG, JB, DC, GS, and GM contributed to the conceptualization. AB, GS, AD, DA, BL, RH, and DW contributed to the experiments and validation. GS, BS, and DC contributed to the specialised analysis. AB, BS, GS, AD, DA, BL, RH, JB, DC, CD, GM, and DW contributed to the writing, reviewing and editing of the manuscript. All authors read and approved the final manuscript. Funding This work was made possible through the National Health and Medical Research Council of Australia (Project Grant 1143974 to D.W.W., G.I.M, B.E.S. and C.D.G; Development Grant 1113712 to B.E.S.; Senior Research Fellowship 1077636 to JGB; Career Development (II) Fellowship 1148700 to DJC) and the Victorian State Government Operational Infrastructure Support and Australian Government NHMRC IRIISS. D.W.W. is a University of Adelaide Beacon Fellow and Hospital Research Foundation Fellow. B.E.S. is a Corin Centenary Fellow. Availability of data and materials All data generated or analysed during this study are included in this published article, its supplementary information files and publicly available repositories. The metabolomics spectrometry data and search results [92] supporting the conclusions of this article are available at the NIH Common Fund’s National Metabolomics Data Repository (NMDR) website, the Metabolomics Workbench, https://www.metabolomicsworkbench.org where it has been assigned Project ID (ST001315). The data can be accessed directly via it’s Project DOI: (https://doi.org/10.21228/M8CX0M). This work is supported by NIH grant U2C-DK119886. Supporting data values for other experiments are included in Additional file 15 Supporting Data Values. Other datasets used and/or analysed during the current study are available from the corresponding author on request. Ethics approval and consent to participate Human RBCs were provided by the Australian Red Cross Blood Bank with ethics approval for use of the cells obtained from the University of Adelaide Human Ethics Committee. Consent for publication Not Applicable. Competing interests The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The authors have declared that no conflict of interest exists. Author details 1Research Centre for Infectious Diseases, School of Biological Sciences, The University of Adelaide, Adelaide 5005, Australia. 2Walter and Eliza Hall Institute of Medical Research, Melbourne, Victoria 3050, Australia. 3Department of Medical Biology, University of Melbourne, Melbourne, Victoria 3050, Australia. 4Monash Institute of Pharmaceutical Sciences, Monash University, Melbourne, Victoria 3052, Australia. 5Burnet Institute, Melbourne, Victoria 3004, Australia. 6Department of Medicine, University of Melbourne, Melbourne, Australia. 7Central Clinical School and Department of Microbiology, Monash University, Melbourne, Australia. 8School of Biosciences, University of Melbourne, Melbourne, Victoria 3010, Australia. Received: 9 March 2020 Accepted: 28 August 2020 References 1. Murray CJ, Ortblad KF, Guinovart C, Lim SS, Wolock TM, Roberts DA, et al. Global, regional, and national incidence and mortality for HIV, tuberculosis, and malaria during 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2014;384(9947):1005–70. 2. World Malaira Report. Geneva: World Health Organization (WHO). 2018. 3. Ariey F, Witkowski B, Amaratunga C, Beghain J, Langlois AC, Khim N, et al. A molecular marker of artemisinin-resistant Plasmodium falciparum malaria. Nature. 2014;505(7481):50–5. 5. 6. 4. Miotto O, Sekihara M, Tachibana S-I, Yamauchi M, Pearson RD, Amato R, et al. Emergence of artemisinin-resistant Plasmodium falciparum with kelch13 C580Y mutations on the island of New Guinea. bioRxiv. 2019: 621813. Das S, Saha B, Hati AK, Roy S. Evidence of artemisinin-resistant Plasmodium falciparum malaria in Eastern India. N.Engl. J. Med. 2018;379(20):1962–4. Ashley E, Dhorda M, Fairhurst R, Amaratunga C, Lim P, Suon S, et al. Spread of Artemisinin resistance in Plasmodium falciparum malaria. N Engl J Med. 2014;371(5):411–23. Nsanzabana C. Resistance to artemisinin combination therapies (ACTs): do not forget the partner drug! Int J Infect Dis. 2019; 4(1): 26. Blasco B, Leroy D, Fidock DA. Antimalarial drug resistance: linking Plasmodium falciparum parasite biology to the clinic. Nat Med. 2017;23(8): 917–28. Burrows JN, Duparc S, Gutteridge WE, Hooft van Huijsduijnen R, Kaszubska W, Macintyre F, et al. New developments in anti-malarial target candidate and product profiles. Malar J. 2017;16(1):26. 7. 9. 8. Burns et al. BMC Biology (2020) 18:133 Page 22 of 23 10. Noedl H, Krudsood S, Chalermratana K, Silachamroon U, Leowattana W, 32. Hutinec A, Rupcic R, Ziher D, Smith KS, Milhous W, Ellis W, et al. An Tangpukdee N, et al. Azithromycin combination therapy with artesunate or quinine for the treatment of uncomplicated Plasmodium falciparum malaria in adults: a randomized, phase 2 clinical trial in Thailand. Clin Infect Dis. 2006;43(10):1264–71. Sykes A, Hendriksen I, Mtove G, Mandea V, Mrema H, Rutta B, et al. Azithromycin plus artesunate versus artemether-lumefantrine for treatment of uncomplicated malaria in Tanzanian children: a randomized, controlled trial. Clin Infect Dis. 2009;49(8):1195–201. Sidhu AB, Sun Q, Nkrumah LJ, Dunne MW, Sacchettini JC, Fidock DA. In vitro efficacy, resistance selection, and structural modeling studies implicate the malarial parasite apicoplast as the target of azithromycin. J Biol Chem. 2007;282(4):2494–504. 11. 12. 13. Dahl EL, Rosenthal PJ. Multiple antibiotics exert delayed effects against the Plasmodium falciparum apicoplast. Antimicrob Agents Chemother. 2007; 51(10):3485–90. 14. Goodman CD, Su V, McFadden GI. The effects of anti-bacterials on the malaria parasite Plasmodium falciparum. Mol Biochem Parasitol. 2007;152(2): 181–91. Imlay L, Odom AR. Isoprenoid metabolism in apicomplexan parasites. Curr Clin Microbiol Reo. 2014;1(3–4):37–50. 15. 16. Guggisberg AM, Amthor RE, Odom AR. Isoprenoid biosynthesis in Plasmodium falciparum. Eukaryot Cell. 2014;13(11):1348–59. 17. Yeh E, DeRisi JL. Chemical rescue of malaria parasites lacking an apicoplast defines organelle function in blood-stage Plasmodium falciparum. PLoS Biol. 2011;9(8):e1001138. Lode H, Borner K, Koeppe P, Schaberg T. Azithromycin--review of key chemical, pharmacokinetic and microbiological features. J Antimicrob Agents Chemother. 1996;37(Suppl C):1–8. 18. 19. Hopkins S. Clinical toleration and safety of azithromycin. Am J Med. 1991; 20. 21. 22. 91(3a):40s–5s. Sagara I, Oduro AR, Mulenga M, Dieng Y, Ogutu B, Tiono AB, et al. Efficacy and safety of a combination of azithromycin and chloroquine for the treatment of uncomplicated Plasmodium falciparum malaria in two multi- country randomised clinical trials in African adults. Malar J. 2014;13:458. Kimani J, Phiri K, Kamiza S, Duparc S, Ayoub A, Rojo R, et al. Efficacy and safety of azithromycin-chloroquine versus sulfadoxine-pyrimethamine for intermittent preventive treatment of Plasmodium falciparum malaria infection in pregnant women in Africa: an open-label, Randomized Trial. PloS One. 2016;11(6):e0157045. Friesen J, Silvie O, Putrianti ED, Hafalla JC, Matuschewski K, Borrmann S. Natural immunization against malaria: causal prophylaxis with antibiotics. Sci Transl Med. 2010;2(40):40ra9. 23. Gaillard T, Madamet M, Pradines B. Tetracyclines in malaria. Malar J. 2015;14: 24. 25. 26. 445. Luntamo M, Kulmala T, Mbewe B, Cheung YB, Maleta K, Ashorn P. Effect of repeated treatment of pregnant women with sulfadoxine-pyrimethamine and azithromycin on preterm delivery in Malawi: a randomized controlled trial. Am J Trop Med Hyg. 2010;83(6):1212–20. Schachterle SE, Mtove G, Levens JP, Clemens E, Shi L, Raj A, et al. Short-term malaria reduction by single-dose azithromycin during mass drug administration for trachoma, Tanzania. J Emerg Infect Dis. 2014;20(6):941–9. Shimizu S, Osada Y, Kanazawa T, Tanaka Y, Arai M. Suppressive effect of azithromycin on Plasmodium berghei mosquito stage development and apicoplast replication. Malar J. 2010;9:73. automated, polymer-assisted strategy for the preparation of urea and thiourea derivatives of 15-membered azalides as potential antimalarial chemotherapeutics. Bioorg Med Chem. 2011;19(5):1692–701. 33. Peric M, Fajdetic A, Rupcic R, Alihodzic S, Ziher D, Bukvic Krajacic M, et al. Antimalarial activity of 9a-N substituted 15-membered azalides with improved in vitro and in vivo activity over azithromycin. J Med Chem. 2012; 55(3):1389–401. 34. Pesic D, Starcevic K, Toplak A, Herreros E, Vidal J, Almela MJ, et al. Design, synthesis, and in vitro activity of novel 2′-O-substituted 15-membered azalides. J Med Chem. 2012;55(7):3216–27. Starcevic K, Pesic D, Toplak A, Landek G, Alihodzic S, Herreros E, et al. Novel hybrid molecules based on 15-membered azalide as potential antimalarial agents. Eur J Med Chem. 2012;49:365–78. 35. 36. Uddin T, McFadden GI, Goodman CD. Validation of putative apicoplast- targeting drugs using a chemical supplementation assay in cultured human malaria parasites. Antimicrob Agents Chemother. 2018;62(1): e01161-17. 37. Dogovski C, Xie SC, Burgio G, Bridgford J, Mok S, McCaw JM, et al. Targeting 38. the cell stress response of Plasmodium falciparum to overcome artemisinin resistance. PLoS Biol. 2015;13(4):e1002132. Klonis N, Xie SC, McCaw JM, Crespo-Ortiz MP, Zaloumis SG, Simpson JA, et al. Altered temporal response of malaria parasites determines differential sensitivity to artemisinin. Proc Natl Acad Sci U S A. 2013;110(13):5157–62. 39. Zhang Y, Asante KS, Jung A. Stage-dependent inhibition of chloroquine on Plasmodium falciparum in vitro. J Parasitol. 1986;72(6):830–6. 41. 40. Cowman AF, Morry MJ, Biggs BA, Cross GA, Foote SJ. Amino acid changes linked to pyrimethamine resistance in the dihydrofolate reductase- thymidylate synthase gene of Plasmodium falciparum. Proc Natl Acad Sci U S A. 1988;85(23):9109–13. Fry M, Pudney M. Site of action of the antimalarial hydroxynaphthoquinone, 2-[trans-4-(4′-chlorophenyl) cyclohexyl]-3-hydroxy-1,4-naphthoquinone (566C80). Biochem Pharmacol. 1992;43(7):1545–53. Egan TJ, Ross DC, Adams PA. Quinoline anti-malarial drugs inhibit spontaneous formation of beta-haematin (malaria pigment). FEBS Lett. 1994;352(1):54–7. 43. Wellems TE, Panton LJ, Gluzman IY. do Rosario VE, Gwadz RW, Walker-Jonah 42. A, et al. Chloroquine resistance not linked to mdr-like genes in a Plasmodium falciparum cross. Nature. 1990;345(6272):253–5. 45. 44. Witkowski B, Amaratunga C, Khim N, Sreng S, Chim P, Kim S, et al. Novel phenotypic assays for the detection of artemisinin-resistant Plasmodium falciparum malaria in Cambodia: in-vitro and ex-vivo drug-response studies. Lancet Infect Dis. 2013;13(12):1043–9. Straimer J, Gnadig NF, Witkowski B, Amaratunga C, Duru V, Ramadani AP, et al. K13-propeller mutations confer artemisinin resistance in Plasmodium falciparum clinical isolates. Science. 2015;347(6220):428–31. Straimer J, Gnadig NF, Stokes BH, Ehrenberger M, Crane AA, Fidock DA. Plasmodium falciparum K13 Mutations Differentially Impact Ozonide Susceptibility and Parasite Fitness In Vitro. mBio. 2017;8(2): e00172-17. 47. Yang T, Xie SC, Cao P, Giannangelo C, McCaw J, Creek DJ, et al. Comparison of 46. the exposure time dependence of the activities of synthetic ozonide antimalarials and dihydroartemisinin against K13 wild-type and mutant Plasmodium falciparum strains. Antimicrob Agents Chemother. 2016;60(8): 4501–10. Singh B, Kim Sung L, Matusop A, Radhakrishnan A, Shamsul SS, Cox-Singh J, et al. A large focus of naturally acquired Plasmodium knowlesi infections in human beings. Lancet. 2004;363(9414):1017–24. 48. 27. Goodman CD, Useglio M, Peiru S, Labadie GR, McFadden GI, Rodriguez E, 49. Moon RW, Sharaf H, Hastings CH, Ho YS, Nair MB, Rchiad Z, et al. et al. Chemobiosynthesis of new antimalarial macrolides. Antimicrob Agents Chemother. 2013;57(2):907–13. van Eijk AM, Terlouw DJ. Azithromycin for treating uncomplicated malaria. Cochrane database Syst Rev. 2011;2011(2):Cd006688. 28. 29. Wilson DW, Goodman CD, Sleebs BE, Weiss GE, de Jong NW, Angrisano F, et al. Macrolides rapidly inhibit red blood cell invasion by the human malaria parasite, Plasmodium falciparum BMC Biol 2015;13:52. 30. Cooper DJ, Rajahram GS, William T, Jelip J, Mohammad R, Benedict J, et al. Plasmodium knowlesi malaria in Sabah, Malaysia, 2015-2017: ongoing increase in incidence despite near-elimination of the human-only Plasmodium species. Clin Infect Dis. 2020;70(3):361–7. 31. Bukvic Krajacic M, Peric M, Smith KS, Schonfeld ZI, Ziher D, Fajdetic A, et al. Synthesis, structure-activity relationship, and antimalarial activity of ureas and thioureas of 15-membered azalides. J Med Chem. 2011;54(10):3595– 605. 50. Normocyte-binding protein required for human erythrocyte invasion by the zoonotic malaria parasite Plasmodium knowlesi. Proc Natl Acad Sci U S A. 2016;113(26):7231–6. van Schalkwyk DA, Moon RW, Blasco B, Sutherland CJ. Comparison of the susceptibility of Plasmodium knowlesi and Plasmodium falciparum to antimalarial agents. J Antimicrob Chemother. 2017;72(11):3051–8. 51. Poehlsgaard J, Douthwaite S. The bacterial ribosome as a target for 52. antibiotics. Nat Rev Microbiol. 2005;3(11):870–81. Schlunzen F, Zarivach R, Harms J, Bashan A, Tocilj A, Albrecht R, et al. Structural basis for the interaction of antibiotics with the peptidyl transferase centre in eubacteria. Nature. 2001;413(6858):814–21. 53. Creek DJ, Chua HH, Cobbold SA, Nijagal B, MacRae JI, Dickerman BK, et al. Metabolomics-based screening of the malaria box reveals both novel and established mechanisms of action. Antimicrob Agents Chemother. 2016; 60(11):6650–63. Burns et al. BMC Biology (2020) 18:133 Page 23 of 23 77. 76. Abu Bakar N, Klonis N, Hanssen E, Chan C, Tilley L. Digestive-vacuole genesis and endocytic processes in the early intraerythrocytic stages of Plasmodium falciparum. J Cell Sci. 2010;123(Pt 3):441–50. Lee Y, Choi JY, Fu H, Harvey C, Ravindran S, Roush WR, et al. Chemistry and biology of macrolide antiparasitic agents. J Med Chem. 2011;54(8):2792–804. Kanoh S, Rubin BK. Mechanisms of action and clinical application of macrolides as immunomodulatory medications. Clin Microbol Rev. 2010; 23(3):590–615. 78. 79. Wasserman M, Alarcon C, Mendoza PM. Effects of Ca++ depletion on the 80. 81. asexual cell cycle of Plasmodium falciparum. Am J Trop Med Hyg. 1982; 31(4):711–7. Sicard A, Semblat JP, Doerig C, Hamelin R, Moniatte M, Dorin-Semblat D, et al. Activation of a PAK-MEK signalling pathway in malaria parasite- infected erythrocytes. Cell Microbiol. 2011;13(6):836–45. Stepanic V, Kostrun S, Malnar I, Hlevnjak M, Butkovic K, Caleta I, et al. Modeling cellular pharmacokinetics of 14- and 15-membered macrolides with physicochemical properties. J Med Chem. 2011;54(3):719–33. 82. Berquand A, Mingeot-Leclercq MP, Dufrene YF. Real-time imaging of drug- membrane interactions by atomic force microscopy. Biochim Biophys Acta. 2004;1664(2):198–205. 83. Bukvic Krajacic M, Novak P, Dumic M, Cindric M, Paljetak HC, Kujundzic N. Novel ureas and thioureas of 15-membered azalides with antibacterial activity against key respiratory pathogens. Eur J Med Chem. 2009;44(9): 3459–70. 84. Wilson DW, Crabb BS, Beeson JG. Development of fluorescent Plasmodium 85. 86. falciparum for in vitro growth inhibition assays. Malar J. 2010;9:152. Lim C, Hansen E, DeSimone TM, Moreno Y, Junker K, Bei A, et al. Expansion of host cellular niche can drive adaptation of a zoonotic malaria parasite to humans. Nat Commun. 2013;4:1638. Trager W, Jensen JB. Human malaria parasites in continuous culture. Science. 1976;193(4254):673–5. 87. Boyle MJ, Wilson DW, Richards JS, Riglar DT, Tetteh KK, Conway DJ, et al. Isolation of viable Plasmodium falciparum merozoites to define erythrocyte invasion events and advance vaccine and drug development. Proc Natl Acad Sci U S A. 2010;107(32):14378–83. 88. Wiegand I, Hilpert K, Hancock RE. Agar and broth dilution methods to determine the minimal inhibitory concentration (MIC) of antimicrobial substances. Nat Protoc. 2008;3(2):163–75. Tham WH, Wilson DW, Lopaticki S, Schmidt CQ, Tetteh-Quarcoo PB, Barlow PN, et al. Complement receptor 1 is the host erythrocyte receptor for Plasmodium falciparum PfRh4 invasion ligand. Proc Natl Acad Sci U S A. 2010;107(40):17327–32. 89. 90. Creek DJ, Jankevics A, Burgess KE, Breitling R, Barrett MP. IDEOM: an excel interface for analysis of LC-MS-based metabolomics data. Bioinformatics. 2012;28(7):1048–9. 91. Chong J, Soufan O, Li C, Caraus I, Li S, Bourque G, et al. MetaboAnalyst 4.0: towards more transparent and integrative metabolomics analysis. Nucleic Acids Res. 2018;46(W1):W486–w94. 92. Burns AL, Sleebs, BE., Siddiqui, G De Paoli, AE, Anderson, D, Liffner, B, Harvey, R, Beeson, JG, Creek, DJ, Goodman, CD, McFadden, GI, Wilson, DW. Project ID (ST001315) Retargeting azithromycin-like compounds as antimalarials with dual modality. Metabolmics Workbench. 2020; doi: https://doi.org/10.21228/M8CX0M. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 54. Olafson KN, Ketchum MA, Rimer JD, Vekilov PG. Mechanisms of hematin crystallization and inhibition by the antimalarial drug chloroquine. Proc Natl Acad Sci U S A. 2015;112(16):4946–51. 55. Dorn A, Vippagunta SR, Matile H, Jaquet C, Vennerstrom JL, Ridley RG. An assessment of drug-haematin binding as a mechanism for inhibition of haematin polymerisation by quinoline antimalarials. Biochem Pharmacol. 1998;55(6):727–36. 56. Olafson KN, Nguyen TQ, Rimer JD, Vekilov PG. Antimalarials inhibit hematin crystallization by unique drug-surface site interactions. Proc Natl Acad Sci U S A. 2017;114(29):7531–6. 57. Combrinck JM, Fong KY, Gibhard L, Smith PJ, Wright DW, Egan TJ. Optimization of a multi-well colorimetric assay to determine haem species in Plasmodium falciparum in the presence of anti-malarials. Malar J. 2015;14:253. 58. Bulusu V, Jayaraman V, Balaram H. Metabolic fate of fumarate, a side 59. product of the purine salvage pathway in the intraerythrocytic stages of Plasmodium falciparum. J Biol Chem. 2011;286(11):9236–45. Ke H, Lewis IA, Morrisey JM, McLean KJ, Ganesan SM, Painter HJ, et al. Genetic investigation of tricarboxylic acid metabolism during the Plasmodium falciparum life cycle. Cell Rep. 2015;11(1):164–74. 60. Wang X, Miyazaki Y, Inaoka DK, Hartuti ED, Watanabe YI, Shiba T, et al. Identification of Plasmodium falciparum mitochondrial malate: quinone oxidoreductase inhibitors from the pathogen box. Genes. 2019;10(6): 471. 61. Rosenthal PJ. Azithromycin for malaria? Am J Trop Med Hyg. 2016;95(1):2–4. 62. Burns AL, Dans MG, Balbin JM, de Koning-Ward T, Gilson PR, Beeson JG, et al. Targeting malaria parasite invasion of red blood cells as an antimalarial strategy. FEMS Microbiol Rev. 2019;43(3):223–38. 63. Wilson DW, Langer C, Goodman CD, McFadden GI, Beeson JG. Defining the timing of action of antimalarial drugs against Plasmodium falciparum. Antimicrob Agents Chemother. 2013;57(3):1455–67. 64. Ashton TD, Devine SM, Mohrle JJ, Laleu B, Burrows JN, Charman SA, et al. The development process for discovery and clinical advancement of modern antimalarials. J Med Chem. 2019; 62(23): 10526-10562. 65. Coslédan F, Fraisse L, Pellet A, Guillou F, Mordmüller B, Kremsner PG, et al. Selection of a trioxaquine as an antimalarial drug candidate. Proc Natl Acad Sci U S A. 2008;105(45):17579–84. 66. Martin SK, Oduola AM, Milhous WK. Reversal of chloroquine resistance in Plasmodium falciparum by verapamil. Science. 1987;235(4791):899–901. 67. Burgess SJ, Selzer A, Kelly JX, Smilkstein MJ, Riscoe MK, Peyton DH. A chloroquine-like molecule designed to reverse resistance in Plasmodium falciparum. J Med Chem. 2006;49(18):5623–5. 68. Oliveira R, Miranda D, Magalhães J, Capela R, Perry MJ, O'Neill PM, et al. From hybrid compounds to targeted drug delivery in antimalarial therapy. Bioorganic Med Chem. 2015;23(16):5120–30. 69. Burgess SJ, Kelly JX, Shomloo S, Wittlin S, Brun R, Liebmann K, et al. Synthesis, structure-activity relationship, and mode-of-action studies of antimalarial reversed chloroquine compounds. J Med Chem. 2010;53(17): 6477–89. 70. ClinicalTrials.gov. Safety and Pharmacokinetics Study of DM1157 to Treat Malaria. (NCT03490162) NIH, U.S. National Library of Medicine: National Institute of Allergy and Infectious Diseases (NIAID); 2018 (updated June 17, 2019). Available from: https://clinicaltrials.gov/ct2/show/NCT03490162. Accessed July 2020. 71. Paloque L, Witkowski B, Lelièvre J, Ouji M, Ben Haddou T, Ariey F, et al. Endoperoxide-based compounds: cross-resistance with artemisinins and selection of a Plasmodium falciparum lineage with a K13 non-synonymous polymorphism. J Antimicrob Chemother. 2018;73(2):395–403. 72. Cobbold SA, Chua HH, Nijagal B, Creek DJ, Ralph SA, McConville MJ. Metabolic dysregulation induced in Plasmodium falciparum by dihydroartemisinin and other front-line antimalarial drugs. J Infect Dis. 2016; 213(2):276–86. 73. Paljetak HC, Tomaskovic L, Matijasic M, Bukvic M, Fajdetic A, Verbanac D, et al. Macrolide hybrid compounds: drug discovery opportunities in anti- infective and anti-inflammatory area. Curr Top Med Chem. 2017;17(8):919– 40. Juge N, Moriyama S, Miyaji T, Kawakami M, Iwai H, Fukui T, et al. Plasmodium falciparum chloroquine resistance transporter is a H+-coupled polyspecific nutrient and drug exporter. Proc Natl Acad Sci U S A. 2015; 112(11):3356–61. 74. 75. Martin RE, Marchetti RV, Cowan AI, Howitt SM, Broer S, Kirk K. Chloroquine transport via the malaria parasite's chloroquine resistance transporter. Science. 2009;325(5948):1680–2.
10.1186_s12913-021-06257-y	Wiedenmayer et al. BMC Health Services Research (2021) 21:272 https://doi.org/10.1186/s12913-021-06257-y R E S E A R C H A R T I C L E Open Access Adherence to standard treatment guidelines among prescribers in primary healthcare facilities in the Dodoma region of Tanzania Karin Wiedenmayer1,2* Selemani Sungi3 and Manfred Stoermer1,2 , Eva Ombaka3, Baraka Kabudi4, Robert Canavan1,2, Sarah Rajkumar1,2, Fiona Chilunda5, Abstract Background: Tanzania’s primary healthcare system suffers from a scarcity of financial and human resources that impedes its effectiveness to deliver dependable and uniform quality healthcare. Adherence to standard treatment guidelines (STG) can help provide more consistent and correct diagnoses and treatments and limit the irrational use of medicines and the negative health consequences that can occur as a result. The purpose of this study was to investigate prescribers’ adherence of their diagnoses and respective treatments to national STG and to identify potential areas for planning interventions. Methods: A cross-sectional study on prescribers’ adherence to diagnosis and treatment, according to national STG, was conducted in 2012 in public primary healthcare facilities (HCF) in the Dodoma region of Tanzania. Information on 2886 patients was collected, prospectively and retrospectively, from 120 HCF across the Dodoma region using a structured questionnaire. Twenty-five broadly defined main illness groups were recorded and the nine most prevalent and relevant conditions were statistically analysed in detail. Results: Diagnoses and related treatments were recorded and analysed in 2872 cases. The nine most prevalent conditions were upper respiratory tract infections (25%), malaria (18%), diarrhoea (9.9%), pneumonia (6.1%), skin problems (5.8%), gastrointestinal diagnoses (5%), urinary tract infections (4%), worm infestations (3.6%) and eye problems (2.1%). Only 1.8% of all diagnoses were non-communicable diseases. The proportion of prescribers’ primary diagnoses that completely adhered to national STG was 599 (29.9%), those that partially adhered totalled 775 (38.7%), wrong medication was given in 621 cases (30.9%) and no diagnosis or medication was given in nine cases (0.5%). Sixty-one percent of all patients received an antibiotic regardless of the diagnoses. Complete adherence was highest when worms were diagnosed and lowest for diarrhoea. The proportion of cases that did not adhere to STG was highest with patients with skin problems and lowest for malaria. (Continued on next page) * Correspondence: karin.wiedenmayer@swisstph.ch 1Swiss Tropical and Public Health Institute, P.O. Box, CH-4002 Basel, Switzerland 2University of Basel, P.O. Box, CH-4003 Basel, Switzerland Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Wiedenmayer et al. BMC Health Services Research (2021) 21:272 Page 2 of 10 (Continued from previous page) Conclusion: Prescribers’ general adherence to national STG in primary HCF in the public sector in Dodoma region is sub-optimal. The reasons are multifaceted and focused attention, directed at improving prescribing and pharmacotherapy, is required with a view of improving patient care and health outcomes. Keywords: Standard treatment guidelines, Adherence, Prescribers, Primary healthcare facilities, Antibiotics prescription, Rational use of medicines, Tanzania Background Considerable global efforts are required to achieve the commitments and targets of the United Nations Sustain- able Development Goal (SDG) 3, to ‘Ensure healthy lives and promote well-being for all at all ages,’ [1]. Achieving and maintaining universal healthcare (UHC), in particu- lar, is an increasing challenge in the face of a continuing global economic crisis precipitated by the Corona Virus pandemic and its effects on the global healthcare budget, maximum efficiency among limited resources is there- fore required. The rational use of medicines is an essen- tial strategy to achieve optimal outcomes in healthcare by minimizing the threats posed by inappropriate treat- ment such as antimicrobial resistance, excessive use of intravenous medication that can lead to blood-borne dis- eases, illnesses due to under-prescribing, wastage and sickness due to polypharmacy [2]. Responsible prescrib- ing and the adherence to standard treatment guidelines (STG) therefore needs optimizing in order to reduce dis- ease burden. This is especially true for low- and middle- income countries (LMICs) where the irrational use of medicines is more widespread than in higher-income settings [3–7]. Yet, LMICs are often ill equipped to effectively manage the challenges attributable to the irrational use of medi- cines and prescribers non-adherence to STG. The decision-making process of PHC workers can be facili- tated, e.g., by offering clinical assessment assisting soft- ware, as seen in the Netherlands, one of Europe’s most conservative antibiotic prescribers [8]. Although, the same approach is not yet an option in countries with a less developed digital environment such as the United Republic of Tanzania (hereinafter Tanzania). Increased consultation times, technical challenges using the de- vices, lack of qualified staff and financial motivation are some of the reasons for low uptake of such devices [9]. Nevertheless, several small promising efficacy and effect- iveness studies of electronic clinical diagnostic algorithm tools, which have the potential to improve diagnosis and treatment to further advance the rational use of medi- cines in low resource settings, have been carried out [10]. However, what functions well in one clinical situ- ation and health system context is not necessarily effect- ive in another. As a point of reference, STG therefore need to be current, evidence-based and adapted to the local context. In addition, they need to be disseminated in combination with training, supervision and feedback [6, 11–13]. The Tanzanian STG and National Essential Drugs List was first printed in 1991 with further editions printed in 1997, 2007, 2013 and 2017. There are a limited number of studies on prescribers’ adherence to national STG; however, prescribers’ low usage, ownership or know- ledge of existing STG was recorded in 2012 with add- itional negative feedback regarding, among others, their complexity and the complex language used [14]. A later study in 2017, concerning the adherence to malaria STG among healthcare workers in Meatu, Tanzania, demonstrated slightly better prescriber aware- ness and access to STG. Nevertheless, usage remained low and fewer strictly adhered to them [15]. Despite the introduction of newer STG editions, the inappropriate and irrational use of medicines and prescribing habits persist. Irunde and colleagues revealed in 2017 that the overprescribing of antibiotics, among others, continued to be an issue [3]. In addition, adherence appeared to be lower in rural healthcare facilities (HCF) compared to that of urban HCF. Furthermore, public HCF appeared to be slightly more compliant than private ones. How- ever, a promising way forward could be to expand anti- microbial stewardship (AMS) programmes to improve the utilization of antibiotics across Tanzania. They have been established in some LMICs with limited but prom- ising results. Interventions such as clinician education, protocol development and continuous reviews of guide- line compliance have, inter alia, led to an increase in prescriber compliance with STG and reductions in anti- biotic prescribing [16]. The Tanzanian Government aims to strengthen the health system country-wide so it can progress towards its Development Vision 2025 and SDG 3, including UHC. In 2011, the Health Promotion and System Strengthening (HPSS) project was introduced in Tanzania to support the Tanzanian Government with this aim by applying a comprehensive approach to health system strengthening within the health financing, medicines, health promotion and technology management sectors [17–19]. This study was part of the initial phase of the HPSS project. The aim was to explore the adherence of PHC workers diagnoses and respective treatments to national STG with the intent Wiedenmayer et al. BMC Health Services Research (2021) 21:272 Page 3 of 10 to inform any future interventions. The main research questions to ascertain adherence were: i) do prescribers comply with good prescribing practices; ii) do prescribers comply with the national STG; iii) are there differences in adherence to STG for different target groups; and iv) are there differences in adherence to STG for different diseases? Methods Study area A cross-sectional study was carried out between August and October 2012 across six districts within the Dodoma Region in Tanzania, namely, Kondoa, Bahi, Dodoma Municipal, Chamwino, Kongwa and Mpwapa. Twenty sample HCF per district were identified through a mix- ture of simple and systematic random sampling. From all 270 public HCF in the six districts of Dodoma, 120 facilities were randomly included. Data collection and sample size Data for this study were collected either prospectively, on the day of the visit to the facility, or retrospectively, from facility records dating back up to 1 year. Prospect- ive data were collected by reviewing patients’ notebooks (which serve as both patient files and prescriptions) when they were handed in for dispensing after consult- ation. In addition to observing the standard recording of the patient’s name, age, sex, location, date, prescriber’s signature, treatment and diagnosis, any evidence of his- tory taking, physical examination and laboratory investi- gation was also recorded. Recording the diagnosis was a prerequisite in order to assess whether the treatment followed the national STG. The same information was collected from randomly selected retrospective data in patient ledgers over the one-year period. However, in this case, it was not possible to get evidence on history, physical examination or laboratory investigations. The intended sample size for prospective and retro- spective data was 30 per facility. Three groups were clas- sified according to the level of adherence to Tanzania’s 2007 STG: i) complete adherence to STG; ii) partial ad- herence; and iii) non adherence. An analgesic prescribed as additional medication was not considered wrong (in this work wrong medicines were defined as those pre- scribed that do not conform to the STG for the relevant illness). Further, adherence to the STG was defined by comparing a given diagnosis with the indicated medi- cine(s) as per the STG. Doses, dosing intervals and dur- ation of treatments were not explored. A total of 2886 patient cases were recorded and analysed. The most fre- quently diagnosed illnesses, as documented in patient notebooks and patient records, were summarized in 25 broadly defined main illness groups that were coded ac- cordingly (see supplementary information). However, and upper only a selection of the nine most prevalent and locally relevant diagnoses were analysed in detail. The acute re- spiratory infection (ARI) category was subdivided into infections pneumonia (URTI). Bronchitis was assigned to URTIs. Adherence to national STG was assessed for these nine groups. Pri- mary diagnoses (the main concern and diagnosis for a patient’s visit) were analysed in detail. Further secondary and tertiary diagnoses, for lesser accompanying ailments, were not compared for the purposes of this paper. respiratory tract Twelve pharmacy graduates from St John’s University in Dodoma were trained to carry out the study in a three-day training session which included pilot testing and revising the tools. Regular monitoring sessions as- sured quality standards in data collection as per instruc- tions. The assessment on adherence to STG was reviewed and guided by a senior pharmacist experienced in similar assessments. Data processing and analysis The manually filled and completed study tools and questionnaires were collected and double-entered into an Access 2010 database then transformed, checked, cleaned and summarized in Epi Info™ 7 and Stata/IC 12.1. Ethical considerations The ethical clearance for the study was given by St John’s University of Tanzania, Directorate of research and consultancy, internal review committee on 16 Octo- ber 2012. Results Overview of study sample and recorded data In total, 2886 patient cases were recorded by data collec- tors, 784 (27.2%) prospectively, 1609 (55.8%) retrospect- ively and for 493 cases (17.1%) this information was missing. Of the 2886 patient cases recorded, 2554 in- cluded information on both diagnosis and treatment. The remaining 332 cases consisted of 318 with informa- tion on treatment but no diagnosis and 14 cases without either information. Just under a half of all patients recorded (n = 1377; 47.7%) were children, 1231 (42.7%) were adults and for 278 (9.6%) data in this category were missing. The pro- portion of records per district was also calculated with Kondoa producing 551 (19.1%) study records, Bahi 377 (13.1%), Dodoma Municipal 252 (8.7%), Chamwino 553 (19.2%), Mpwapwa 553 (19.2%) and Kongwa 600 (20.8%). According to data collected from clinicians’ records, females appeared to represent the majority of primary diagnoses with 1202 (41.7%) cases, while 927 (32.1%) were males. For an additional 631 cases (21.9%) sex was not recorded by the clinician and for the Wiedenmayer et al. BMC Health Services Research (2021) 21:272 Page 4 of 10 remaining 126 (4.4%) these data were missing from the data collectors’ questionnaires (Table 1). In addition, 399 (13.8%) patients underwent a physical examination and 288 (10.0%) patients underwent a laboratory investiga- tion (Table 1). These were mostly malaria rapid diagnos- tic tests or urine and stool examinations. diagnoses in children vs one third in adults with up to a 10% proportion of ages unknown. Primary diagnoses were distributed among the six dis- tricts (Fig. 2). In five out of six districts URTIs were the most frequent diagnosis, only in Bahi was malaria diag- nosed marginally more. Distribution of diagnoses The 2886 cases recorded by the clinicians and data col- lectors were classified into one of the 25 illness groups (Fig. 1 and supplementary information). Of the 2554 diagnoses specified by the clinicians 2502 (86.7%) were primary diagnoses and 52 (1.8%) were, for unknown rea- sons, not classified. Nine conditions were selected for deeper analysis based on prevalence and local significance such as eye diseases, a very common diagnosis in Dodoma region due to arid weather conditions. Musculoskeletal diagno- ses and injuries which included a variety of conditions poorly addressed in the STG were excluded from this in depth analysis. Similarly, sexually transmitted infections were also excluded as they were addressed by a special programme for syndromic case management. Therefore, the nine selected diseases analysed in detail were, in order of prevalence, URTI, malaria, diarrhoea, pneumo- nia, skin problems, gastrointestinal problems (GI), urin- ary tract infections (UTI), worms and eye problems. Comparing the number of all nine diagnoses using a females were diagnosed significantly more Chi2 test, than males (p < 0.001). There was no significant differ- ence (p = 0.6) testing GI problems and UTI separately. However, a trend was observed with females suffering more often from GI problems compared with men (65% vs 35% respectively) and from UTI (71% vs 29%). When comparing adults and children’s primary diagnoses, GI problems were more prevalent in adults (68%) than in children (25%); 7% of the subjects’ ages were unknown. The percentage of primary diagnoses of UTIs in adults was 59% compared with 31% of children and 9.7% of the subjects’ ages were unknown. Similar proportions ap- plied to the other seven disease groups, generally dem- onstrating a distribution of ca. two thirds of the primary Table 1 Overview of information recorded by clinician (N = 2886) Adherence to national standard treatment guidelines For the 2502 primary diagnoses, information on adher- ence was available for 2004 cases. Complete adherence to STG for all 25 illness groups was recorded in 599 (29.9%) cases. Partial adherence was observed in just over a third (775 or 38.7%) of cases where patients re- ceived the correct medication but additional unnecessary or wrong medicines. Non-adherence to STG was found in 621 (30.9%) cases. In nine (0.5%) cases no diagnosis/ medication was given (Fig. 3). Complete adherence of primary diagnoses to STG for the nine selected conditions was highest when worms were diagnosed and lowest for diarrhoea. The propor- tion of cases that were incorrectly treated was highest in patients with skin problems and lowest for malaria (Fig. 4). The adherence to STG for age and sex were stratified and no important differences could be found. High pre- scribing of antibiotics was observed with 61.2% of all pa- tients having an antibiotic prescribed either as a sole treatment or as an additional treatment (Fig. 5). Diagnosis and distribution for non-communicable diseases Only 51 (2%) non-communicable diseases (NCDs) were recorded among the 2554 diagnoses. Adherence to STG was high in three categories: epilepsy, asthma and psy- chiatric disorders; however, not a single case of cardio- vascular disease was treated correctly. Epilepsy was diagnosed in 17 cases. In 13 of those cases the STG were completely adhered to. There were no instances of par- tial adherence; however, three cases did not adhere at all and in one case the information was missing. Cardiovas- cular diseases were diagnosed in 13 cases. None of the diagnoses completely adhered to the STG, six partially Recorded by clinician Not recorded by clinician Age Sex Location Date History Physical Examination Lab investigation aData missing from data collection sheets 2556 (88.6) 2129 (73.8%) 2522 (87.4%) 2819 (97.7%) 822 (28.5%) 399 (13.8%) 288 (10.0%) 281 (9.7%) 631 (21.9%) 320 (11.1%) 32 (1.1%) 1975 (68.4%) 2353 (81.5%) 2475 (85.8%) Unknowna 49 (1.7%) 126 (4.4%) 44 (1.5%) 35 (1.2%) 89 (3.1%) 134 (4.6%) 123 (4.3%) Wiedenmayer et al. BMC Health Services Research (2021) 21:272 Page 5 of 10 Fig. 1 Distribution of all diagnoses over 25 illness groups and their relative percentage. Legend: URTI Upper respiratory tract infections; UTI Urinary tract infections Fig. 2 Distribution of primary diagnoses among the six districts. Legend: URTI Upper respiratory tract infections; UTI Urinary tract infections Wiedenmayer et al. BMC Health Services Research (2021) 21:272 Page 6 of 10 Fig. 3 Adherence to standard treatment guidelines for primary diagnosis of 25 illness groups adhered but seven did not at all. Asthma was recorded in 13 cases. In 10 of those cases there was complete ad- herence, there was partial adherence in only one case and non-adherence in another two cases. Complete ad- herence to STG was logged in three out of the five psy- chiatric disorders diagnosed, one case of non-adherence was noted and in one case the information was missing. In the only case of diabetes that was diagnosed the STG were not adhered to. Discussion The three most prevalent diagnoses in this study were ARI, malaria and diarrhoea. URTI and diarrhoea diagno- ses and treatments adhered least to STG. These findings coincide with studies that seem to suggest that low adherence to STG, especially for the management of childhood diseases such as diarrhoea and respiratory tract infections, is common not only in LMICs but worldwide [20–22]. Partial adherence to STG for URTI was mainly due to the high prescription of antibiotics for bronchitis which is generally caused by a virus and would therefore not benefit from them. In accordance with STG, antibiotics were indicated for the treatment of pneumonia. They were not, however, indicated for URTIs and GI problems. Fifty-six percent of diarrhoea cases, 25% of malaria cases and 21% of worm cases also received an antibiotic. However, in rural areas diagnosis is complicated by the scarcity of diagnostic tools and fa- cilities, which can lead to prescribing antibiotics pre- sumptively when rapid decisions are a matter of Fig. 4 Adherence to standard treatment guidelines for primary diagnosis. Legend: URTI Upper respiratory tract infections; UTI Urinary tract infections Wiedenmayer et al. BMC Health Services Research (2021) 21:272 Page 7 of 10 Fig. 5 Prescribing of antibiotics. Legend: URTI Upper respiratory tract infections; UTI Urinary tract infections mortality. This was observed, at least, for malaria treat- ment in rural areas, where patients presenting with fever were more likely to receive co-prescription of antibiotics and antimalarial when diagnosis was clinical and not backed by malaria rapid testing [23]. Nevertheless, the fact that 61% of patients in this study received an anti- biotic, disregarding the diagnoses, is of grave concern. It is also a far cry from the WHO prescribing indicator that recommends their use in less than 30% of cases [2, 24]. Tanzania has a history of overprescribing antibiotics and according to an assessment conducted in four re- gions of Tanzania in 2014 it does not appear to have improved [3]. AMS programmes that have positively affected the behaviour of clinicians’ antibiotic utilization and proven to be effective in high-income countries have also been introduced to African countries with some success [16, 25–27]. One such programme was initiated in Mbeya Zonal Referral Hospital of Tanzania [28, 29]. A baseline assessment of antibiotic prescription in this training hospital revealed that prescribers adhered to the Tanzania STG recommendations for antibiotic choice 63% of the time, but it decreased to 15% when treatment duration was taken into account. Successful AMS relies on a commitment from facility management leadership and accountability. AMS programmes usually comprise a multidisciplinary prudent committee antibiotic use [16, 30, 31]. To implement an AMS programme in PHC facilities, however, will be more of a challenge in Tanzania considering the shortage of staff, insufficient guidance and supervision and frequent stock-outs of medicines. ensuring The total of all malaria diagnoses in all districts in this is interesting to note the lower study was 18%. It prevalence of malaria, 1.8% of all diagnoses, in the Dodoma Municipal which is a more urban district. Complete adherence to STG for malaria for primary diagnoses was 65%, partial 28% and non-adherence 7.5%. Budimu et al. reported, from a study conducted in the Meatu district of Tanzania in 2017, 54.6% of all 196 healthcare workers there strictly adhered to the STG for case management. Ten (5.1%) healthcare malaria workers partially adhered when they chose antimalarials without confirmed cases of malaria and 79 (40.3%) health workers did not adhere [15]. Although the study in the Meatu district was on a much smaller scale and in a different region, comparing these studies, the adher- ence to STG seems to have regressed. In a study in the Kilosa district of Tanzania back in 2010, concerns were raised that the STG for administering the malaria treat- ment, artemisinin combination therapy, were not clear enough and thereby probably contributed to the pre- scribers’ non-conformity with STG [32]. Complete adherence to diarrhoea STG was extremely low. Partial adherence was mainly due to prescribing oral rehydration solution (ORS) plus antibiotics and the lack of prescribed Zinc. Extremely low adherence to STG for the management of acute diarrhoea in children under 12 was also found in a study in Ujjain, Madhya Pradesh, India and the high rate of prescribing non- recommended medication was discussed [20]. The dur- ation and volume of diarrhoea is not lessened with ORS, therefore, many practitioners look for alternative therap- ies to shorten its time span [33]. As vomiting can be caused and aggravated by incorrectly prepared ORS, par- ents and caregivers may be discouraged to continue the therapy leading to a failure in oral hydration [34]. Thus, Wiedenmayer et al. BMC Health Services Research (2021) 21:272 Page 8 of 10 the perceived ineffectiveness of ORS therapy may have then led to an increase in prescribing other non-recom- mended medications such as antibiotics. Pathak et al. also considered that accompanying symptoms like the pres- ence of fever, pain, blood in the stool and vomiting signifi- cantly increased antibiotic prescribing even though most diarrhoeal episodes are of a viral origin [20]. Another matter of interest in the present study was the low number of diagnoses of NCDs with just 51 cases (2%) despite NCDs such as heart disease, stroke, cancer, chronic respiratory diseases and diabetes being the lead- ing cause of mortality in the world today. In fact, recent data shows that NCDs are estimated to account for 33% of all deaths in Tanzania [35]. Thus, the low number of diagnoses in the present study may be explained with ei- ther actual low prevalence in the Dodoma region at the time or, what is more likely, low awareness and insuffi- cient diagnostic skills of what are often, initially, silent diseases. Similarly, mental disease appears not to have been a problem in Dodoma region. This again may be due to low prevalence but more likely to be an unaware- ness and underdiagnoses of mental conditions. The implementation of STG provides a point of refer- ence by which practitioners can review, compare and ad- vance the quality of care that they deliver. They are packaged so as to contain statements that provide ex- pected standards of practice in order to diminish varia- tions in clinical practice and to reduce costly and avoidable mistakes and adverse events [36]. It is of con- cern then that, approximately, only a third of primary diagnoses in this study were prescribed and treated com- pletely in accordance with the national STG. A little over a third (38.7%) of primary diagnoses prescriptions partially adhered to them, thus, in these cases patients at least received the correct medicine but also further un- necessary or incorrect medicines, which is a waste of limited resources. In addition, approximately a third of prescribers diagnosed and treated patients incorrectly and not in accordance with STG; therefore the quality of care and patient outcome may have been seriously com- promised. Deviation from clinical guidelines may be due to various factors such as conscious clinical decision- making, empirical prescribing habits, resource availabil- ity limitations caused by supply challenges and lack of familiarity or availability of guidelines. As printed materials alone seem to have little effect in changing the prescribing behaviour of clinical health workers, STG need to be accompanied by reminders, educational outreach and feedback in order to be effect- ive [6, 11–13]. Notwithstanding, not implementing ef- fective training and supportive supervision, the shortage of healthcare workers, together with high clinical and administrative workloads negatively impacts the quality In Tanzania, between the of patient care delivered. period 2007 and 2013 the physician to population ratio per 10,000 was 0.3. This was far lower than the WHO African Region average of 2.7 and the global average of 13.9. During the same period the nursing and midwifery personnel to population ratio per 10,000 was 4.4. Again, this was far lower than the WHO African region average of 12.4 per 10,000 and the global average of 28.6 [37]. The low ratio of healthcare professionals, together with an expanding population, shortage of medical commod- ities and increasing health burdens from chronic and emerging diseases, cause undue pressure on healthcare staff. Health personnel, already operating at (or beyond) their limit, will find it even harder to find the time for comprehending a complex 450-page document such as Tanzania’s STG, which further contributes to the wider lack of reference and adherence [38]. Study limitations Diagnoses were accepted as written by clinicians and were not assessed for correctness. Thus, it is assumed that the diagnoses were correct. In many cases symp- toms rather than diagnoses were noted, as for instance pain or fever. Therefore it was impossible to assess the underlying illness and corresponding suitability of ther- apy. In two cases of fever and pain, correct treatment according to STG was assumed to be an analgesic or antipyretic respectively regardless of underlying path- ology. Moreover, as some patients had multiple diagno- ses, wrong medications could not be clearly assigned to a specific diagnosis. They were usually assumed to be- long to the secondary or tertiary diagnoses. In some facilities, the number of patients was very small and the target number of 30 could not be reached. Also, due to communication limitations, three facilities in Kondoa district and one facility in Mpwapwa district had to be replaced on the day of visit. The nearest dis- pensaries were visited instead. Finally, there are limitations regarding the evidence collected in the framework of this study. First, this is a descriptive report and only such conclusions can there- fore be drawn; second, the study was conducted in 2012 and some data is missing, therefore, the article presents data that may not accurately reflect the current situation. Nevertheless, other subsequent studies and the authors’ continued experience in Tanzania indicates that the situ- the ation has not current study fills an important gap owing to the scarcity of relevant studies in this area in Tanzania. To the best of our knowledge this is still the only study that covers adherence to STG in Dodoma region in such detail. significantly changed. Moreover, Conclusion Prescribers’ general adherence to national STG in PHC facilities in the public sector in Dodoma Region was Wiedenmayer et al. BMC Health Services Research (2021) 21:272 Page 9 of 10 found to be very low. Clinical evidence-based guidelines such as STG are of little value if not implemented and adhered to. Tanzania’s underfunded health system does not provide for printing, delivering and mentoring of new STG editions for all prescribers in all health facil- ities in Tanzania, especially in remote rural areas. Fur- thermore, PHC facilities are strained from an excess of patients from an expanding population. The complex STG, often written for specialists in secondary or tertiary level care, are difficult to interpret and implement for lower cadre prescribers in the peripheral healthcare level. These are among many concerns likely to have an effect on prescribers’ noncompliance with the clinical guidance STG can provide. Poor prescribing practice di- minishes quality of care and increases the chances of poor health outcomes. Multicomponent interventions, similar to the AMS programmes for antibiotics, incorp- orating prescriber education, complemented with supportive supervision should be implemented. Greater commitments from the govern- ment and stakeholders are consequently required to strengthen the health system and expand the financial and human resources available. training and patient Abbreviations UHC: Universal healthcare; SDG: Sustainable development goal; STG: Standard treatment guidelines; LMICs: Low- and middle-income coun- tries; PHC: Primary healthcare; HCF: Healthcare facilities; HPSS: Health Promotion and System Strengthening Project; ARI: Acute respiratory infection; URTI: Upper respiratory tract infections; GI: Gastrointestinal; UTI: Urinary tract infections; NCDs: Non-communicable diseases; ORS: Oral rehydration solution; AMS: Antimicrobial stewardship Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12913-021-06257-y. Additional file 1: Supplement Table. Illness group and Diagnoses as recorded by prescriber/clinician. Acknowledgements We would like thank the whole team of the HPSS project in Dodoma Tanzania who greatly supported the survey logistically and administratively. We extend our gratitude to the enthusiastic pharmacy graduates from St. Johns’ University Tanzania for the diligent and careful field work and data collection. Authors’ contributions KW and EO participated in the concept, study design, planning, data collection and data interpretation. EO acted as the principal investigator in Tanzania, supervised data collection and contributed to reporting. BK contributed to training and management of data collection. SR managed data entry and data analysis. SS co-supervised data collection. FC supported planning and organized logistics. MS advised and supported the study as project director. RC drafted the manuscript for input by the other authors. All authors read and approved the final manuscript. Funding This study was funded by The Swiss Agency for Development and Cooperation (Grant number Phase 7F-07381.01.02, Contract No. 81013135). The funder did not contribute to the study design, data collection, analysis, interpretation or the manuscript writing. Availability of data and materials The data sets used and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The ethical clearance for the study was given in writing by St John’s University of Tanzania, directorate of research and consultancy, internal review committee on 16 October 2012. Study participants were provided with information explaining the rationale for the study, their rights, confidentiality and who will have access to information provided. Informed consent was obtained verbally as described in the approved study protocol. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1Swiss Tropical and Public Health Institute, P.O. Box, CH-4002 Basel, Switzerland. 2University of Basel, P.O. Box, CH-4003 Basel, Switzerland. 3St. John’s University Tanzania, Dodoma, Tanzania. 4Mission for Essential Medical Supplies, P.O. Box 1005, Arusha, Tanzania. 5Health Promotion and System Strengthening project, Dodoma, Tanzania. Received: 20 December 2020 Accepted: 8 March 2021 References 1. World Health Organization. SDG 3: Ensure healthy lives and promote 2. 3. 4. 5. 6. 7. 8. 9. wellbeing for all at all ages. https://www.who.int/sdg/targets/en/. Accessed 9 July 2020. Ofori-Asenso R, Brhlikova P, Pollock AM. Prescribing indicators at primary health care centers within the WHO African region: a systematic analysis (1995-2015). BMC Public Health. 2016;16(1):724. https://doi.org/10.1186/s12 889-016-3428-8. Irunde H, Minzi O, Moshiro C. Assessment of rational medicines prescribing in healthcare facilities in four regions of Tanzania. J Pharm Pract Comm Med. 2017;3(4):225–31. https://doi.org/10.5530/jppcm.2017.4.64. Goossens H, Ferech M, Stichele RV, Elseviers M, Grp EP. Outpatient antibiotic use in Europe and association with resistance: a cross-national database study. Lancet. 2005;365(9459):579–87. https://pubmed.ncbi.nlm.nih.gov/1 5708101/. Adriaenssens N, Coenen S, Kroes ACM, Versporten A, Vankerckhoven V, Muller A, Blix HS, Goossens H, on behalf of the ESAC Project Group, Mittermayer H, Vaerenberg S, Markova B, Andrašević A, Kontemeniotis A, Vlček J, Frimodt-Møller N, Rootslane L, Vuopio-Varkila J, Cavalie P, Kern W, Giamarellou H, Ternák G, Briem H, Cunney R, Raz R, Folino P, Dumpis U, Valinteliene R, Bruch M, Borg M, Natsch S, Blix HS, Hryniewicz W, Ribeirinho M, Băicuş A, Ratchina S, Foltán V, Čižman M, Campos J, Skoog G, Zanetti G, Ünal S, Davey P. European Surveillance of Antimicrobial Consumption (ESAC): systemic antiviral use in Europe. J Antimicrob Chemoth. 2011;66(8): 1897–905. https://doi.org/10.1093/jac/dkr190. Kruk ME, Gage AD, Arsenault C. High-quality health systems in the sustainable development goals era: time for a revolution (vol 6, pg e1196, 2018). Lancet Glob Health. 2018;6(11):E1162. Holloway KA, Henry D. WHO essential medicines policies and use in developing and transitional countries: an analysis of reported policy implementation and medicines use surveys. Plos Med. 2014;11(9). Ivanovska V, Hek K, Teeuwisse AKM, Leufkens HGM, Nielen MMJ, van Dijk L. Antibiotic prescribing for children in primary care and adherence to treatment guidelines. J Antimicrob Chemoth. 2016;71(6):1707–14. https:// doi.org/10.1093/jac/dkw030. Shao AF, Rambaud-Althaus C, Swai N, Kahama-Maro J, Genton B, D'Acremont V, et al. Can smartphones and tablets improve the management of childhood illness in Tanzania? A qualitative study from a primary health care worker’s perspective. BMC Health Serv Res. 2015;15(1):135. https://doi.org/10.1186/s12 913-015-0805-4. Wiedenmayer et al. BMC Health Services Research (2021) 21:272 Page 10 of 10 29. Hall JW, Bouchard J, Bookstaver PB, Haldeman MS, Kishimbo P, Mbwanji G, et al. The Mbeya antimicrobial stewardship team: Implementing antimicrobial stewardship at a zonal-level hospital in southern tanzania. MDPI Pharm. 2020;8(2). 30. Pulcini C, Binda F, Lamkang AS, Trett A, Charani E, Goff DA, Harbarth S, Hinrichsen SL, Levy-Hara G, Mendelson M, Nathwani D, Gunturu R, Singh S, Srinivasan A, Thamlikitkul V, Thursky K, Vlieghe E, Wertheim H, Zeng M, Gandra S, Laxminarayan R. Developing core elements and checklist items for global hospital antimicrobial stewardship programmes: a consensus approach. Clin Microbiol Infect. 2019;25(1):20–5. https://doi.org/10.1016/j. cmi.2018.03.033. 31. Centre for Disease Control and Prevention (CDC). Core elements of hospital antibiotic stewardship programs. Atlanta: US Department of Health and Human Services; 2014. Simba DO, Kakoko D. Primacy of effective communication and its influence on adherence to artemether-lumefantrine treatment for children under five years of age: a qualitative study. BMC Health Serv Res. 2012;12(1). https:// doi.org/10.1186/1472-6963-12-146. 32. 33. Howteerakul N, Higginbotham N, Freeman S, Dibley MJ. ORS is never enough: physician rationales for altering standard treatment guidelines when managing childhood diarrhoea in Thailand. Soc Sci Med. 2003;57(6): 1031–44. https://doi.org/10.1016/S0277-9536(02)00478-1. 34. Albano F, Lo Vecchio A, Guarino A. The applicability and efficacy of guidelines for the management of acute gastroenteritis in outpatient children: a field-randomized trial on primary care pediatricians. J Pediatr-Us. 2010;156(2):226–30. https://doi.org/10.1016/j.jpeds.2009.07.065. 35. World Health Organization. Noncommunicable Diseases (NCD) country 36. profiles. 2018. Kredo T, Bernhardsson S, Machingaidze S, Young T, Louw Q, Ochodo E, Grimmer K. Guide to clinical practice guidelines: the current state of play. Int J Qual Health C. 2016;28(1):122–8. https://doi.org/10.1093/ intqhc/mzv115. 37. World Health Organization. Atlas of African Health Statistics 2016, 38. Health situation analysis of the African region: WHO Regional Office for African; 2016. Kwesigabo G, Mwangu MA, Kakoko DC, Warriner I, Mkony CA, Killewo J, Macfarlane SB, Kaaya EE, Freeman P. Tanzania's health system and workforce crisis. J Public Health Pol. 2012;33(S1):S35–44. https://doi.org/1 0.1057/jphp.2012.55. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 10. Keitel K, D'Acremont V. Electronic clinical decision algorithms for the integrated primary care management of febrile children in low-resource settings: review of existing tools. Clin Microbiol Infec. 2018;24(8):845–55. https://doi.org/10.1016/j.cmi.2018.04.014. 11. Meyer JC, Summers RS, Moller H. Randomized, controlled trial of prescribing training in a south African province. Med Educ. 2001;35(9):833–40. https:// doi.org/10.1046/j.1365-2923.2001.01000.x. 12. World Health Organization. WHO handbook for guideline development. In. Geneva; 2012. 13. Holloway KA, Ivanovska V, Wagner AK, Vialle-Valentin C, Ross-Degnan D. Have we improved use of medicines in developing and transitional countries and do we know how to? Two decades of evidence. Tropical Med Int Health. 2013;18(6):656–64. https://doi.org/10.1111/tmi.12123. 14. Urassa DP, Mwangu M, Makwi JK. Utilization of standard treatment guidelines (STG) at primary health facilities, Magu district, Tanzania. East Afr J Pub Health. 2013;10(1):235–45. 15. Budimu A, Emidi B, Mkumbaye S, Kajeguka DC. Adherence, awareness, access, and use of standard diagnosis and treatment guideline for malaria case management among healthcare workers in Meatu, Tanzania. J Trop Med-Us. 2020;2020. 16. Akpan MR, Isemin NU, Udoh AE, Ashiru-Oredope D. Implementation of antimicrobial stewardship programmes in African countries: a systematic literature review. J Glob Antimicrob Re. 2020;22:317–24. https://doi.org/10.1 016/j.jgar.2020.03.009. Swiss Tropical and Public Health institute. Swiss-Tanzania cooperation: Health Promotion and System Strengthening (HPSS) project. In.: issuu. https://issuu.com/communications.swisstph/docs/hpss_brochure. 17. 18. Wiedenmayer K, Mbwasi R, Mfuko W, Mpuya E, Charles J, Chilunda F, 19. Mbepera D, Kapologwe N. Jazia prime vendor system- a public-private partnership to improve medicine availability in Tanzania: from pilot to scale. J Pharm Policy Pract. 2019;12(1):4. https://doi.org/10.1186/s40545- 019-0163-4. Kapologwe NA, Kagaruki GB, Kalolo A, Ally M, Shao A, Meshack M, Stoermer M, Briet A, Wiedenmayer K, Hoffman A. Barriers and facilitators to enrollment and re-enrollment into the community health funds/Tiba Kwa Kadi (CHF/TIKA) in Tanzania: a cross-sectional inquiry on the effects of socio- demographic factors and social marketing strategies. BMC Health Serv Res. 2017;17(1):308. https://doi.org/10.1186/s12913-017-2250-z. 20. Pathak D, Pathak A, Marrone G, Diwan V, Lundborg CS. Adherence to treatment guidelines for acute diarrhoea in children up to 12 years in Ujjain, India - a cross-sectional prescription analysis. BMC Infect Dis. 2011;11(1). https://doi.org/10.1186/1471-2334-11-32. 21. Cheraghali AM, Idries AM. Availability, affordability, and prescribing pattern of medicines in Sudan. Pharm World Sci. 2009;31(2):209–15. https://doi.org/1 0.1007/s11096-009-9282-3. 22. Hoekstra JH, Diarrhoe EWGA. Acute gastroenteritis in industrialized countries: compliance with guidelines for treatment. J Pediatr Gastr Nutr. 2001;33:S31–5. https://doi.org/10.1097/00005176-200110002-00006. 23. Njozi M, Amuri M, Selemani M, Masanja I, Kigahe B, Khatib R, Kajungu D, Abdula S, Dodoo AN. Predictors of antibiotics co-prescription with antimalarials for patients presenting with fever in rural Tanzania. BMC Public Health. 2013;13(1). https://doi.org/10.1186/1471-2458-13-1097. 24. Ofori-Asenso R. A closer look at the World Health Organization's prescribing indicators. J Pharmacol Pharmacother. 2016;7(1):51–4. https://doi.org/10.41 03/0976-500X.179352. 25. Pollack LA, van Santen PL, Weiner LM, Dudeck MA, Edwards JR, Srinivasan A. Antibiotic stewardship programs in US acute care hospitals: findings from the 2014 national healthcare safety network annual hospital survey. Clin Infect Dis. 2016;63(4):443–9. https://doi.org/10.1093/cid/ciw323. Tonna AP, Gould IM, Stewart D. A cross-sectional survey of antimicrobial stewardship strategies in UK hospitals. J Clin Pharm Ther. 2014;39(5):516–20. https://doi.org/10.1111/jcpt.12181. 26. 27. Cairns KA, Roberts JA, Cotta MO, Cheng AC. Antimicrobial stewardship in Australian hospitals and other settings. Infect Dis Ther. 2015;4(S1):27–38. https://doi.org/10.1007/s40121-015-0083-9. 28. Haldeman MS, Kishimbo P, Seddon M, Sangare A, Mwasomola D, Hall J, Shaffer M, Leclair R, Caulder C, Bookstaver PB, Nsojo A. Evaluation of antimicrobial utilization and concordance with national guidelines at a tertiary hospital in the Southern Highlands Zone of Tanzania. Am J Trop Med Hyg. 2020;102(2):370–6. https://doi.org/10.4269/ajtmh.19-0231.
10.1172_JCI152042	SARS-CoV-2–specific memory B cells can persist in the elderly who have lost detectable neutralizing antibodies Anna Jeffery-Smith, … , Laura E. McCoy, Mala K. Maini J Clin Invest. 2022;132(2):e152042. https://doi.org/10.1172/JCI152042. Research Article COVID-19 Immunology Graphical abstract Find the latest version: https://jci.me/152042/pdf The Journal of Clinical Investigation SARS-CoV-2–specific memory B cells can persist in the elderly who have lost detectable neutralizing antibodies Anna Jeffery-Smith,1,2,3 Alice R. Burton,1 Sabela Lens,1 Chloe Rees-Spear,1 Jessica Davies,1 Monika Patel,2 Robin Gopal,2 Luke Muir,1 Felicity Aiano,4 Katie J. Doores,5 J. Yimmy Chow,6 Shamez N. Ladhani,4 Maria Zambon,2 Laura E. McCoy,1 and Mala K. Maini1 1Division of Infection and Immunity, Institute of Immunity and Transplantation, University College London (UCL), London, United Kingdom. 2Virus Reference Department, Public Health England (now called UK Health Security Agency [UKHSA]), London, United Kingdom. 3Blizard Institute, Queen Mary University of London, London, United Kingdom. 4Immunisation and Countermeasures Division, Public Health England (now called UKHSA), London, United Kingdom. 5Department of Infectious Diseases, School of Immunology and Microbial Sciences, King’s College London, London, United Kingdom. 6London Coronavirus Response Cell, Public Health England (now called UKHSA), London, United Kingdom. Memory B cells (MBCs) can provide a recall response able to supplement waning antibodies (Abs) with an affinity-matured response better able to neutralize variant viruses. We studied a cohort of elderly care home residents and younger staff (median age of 87 years and 56 years, respectively), who had survived COVID-19 outbreaks with only mild or asymptomatic infection. The cohort was selected because of its high proportion of individuals who had lost neutralizing antibodies (nAbs), thus allowing us to specifically investigate the reserve immunity from SARS-CoV-2–specific MBCs in this setting. Class- switched spike and receptor-binding domain (RBD) tetramer–binding MBCs persisted 5 months after mild or asymptomatic SARS-CoV-2 infection, irrespective of age. The majority of spike- and RBD-specific MBCs had a classical phenotype, but we found that activated MBCs, indicating possible ongoing antigenic stimulation or inflammation, were expanded in the elderly group. Spike- and RBD-specific MBCs remained detectable in the majority of individuals who had lost nAbs, although at lower frequencies and with a reduced IgG/IgA isotype ratio. Functional spike-, S1 subunit of the spike protein– (S1-), and RBD- specific recall was also detectable by enzyme-linked immune absorbent spot (ELISPOT) assay in some individuals who had lost nAbs, but was significantly impaired in the elderly. Our findings demonstrate that a reserve of SARS-CoV-2–specific MBCs persists beyond the loss of nAbs but highlight the need for careful monitoring of functional defects in spike- and RBD-specific B cell immunity in the elderly. Introduction The human coronavirus SARS-CoV-2 has had a particularly devastating impact on the elderly, who are at much greater risk of morbidity and mortality (1, 2). Understanding the nature of a successful immune response in those who have avoided these outcomes and cleared SARS-CoV-2 after a mild infection, despite advanced age, is key to protecting this vulnerable group in the future. Whether older survivors of SARS-CoV-2 infection are able to mount robust and durable responses with the potential to provide long-term protection from reinfection, and from emerg- ing viral variants, remains to be understood. Insights into the strengths and limitations of the immune response in those who have had a successful outcome of natural infection can inform the future optimization of vaccines. It is also crucial to under- stand the nature of the immune protection afforded to previously Authorship note: MZ, LEM, and MKM are co–senior authors. Conflict of interest: The authors have declared that no conflict of interest exists. Copyright: © 2022, Jeffery-Smith et al. This is an open access article published under the terms of the Creative Commons Attribution 4.0 International License. Submitted: June 4, 2021; Accepted: November 24, 2021; Published: January 18, 2022. Reference information: J Clin Invest. 2022;132(2):e152042. https://doi.org/10.1172/JCI152042. infected individuals while they await vaccination, especially with the ongoing delays in rollout and the lag in provision of vaccines to low- and middle-income countries. Antibodies (Abs), in particular the neutralizing fraction, provide a vital frontline defense to achieve protective immunity against viruses. An initial waning of antibody titers is typically seen after resolution of an acute viral infection (3, 4). In the case of some viruses, long-lived plasma cells are then able to main- tain Abs for decades (5–7). By contrast, in the months following infection with other viruses, including human coronaviruses like SARS-CoV-2, neutralizing Abs (nAbs) continue to wane and can drop below the threshold of detection in a proportion of indi- viduals (3, 8–13). Even if Abs are maintained, they may fail to provide sufficient functional flexibility to cross-recognize viral variants (14–16). However, inadequate Ab titers or Abs that are unable to cross-recognize variants can be compensated by a second line of defense provided by antigen-specific memory B cells (MBCs) that are poised to react rapidly upon pathogen reen- counter or vaccine boosting (17–19). Not only can MBCs provide a faster response upon reexposure to the virus, they are also able to diversify in the face of a mutating virus, resulting in a more potent, affinity-matured Ab response and enhanced resistance to viral mutations (9, 20). 1 RESEARCH ARTICLEIn this study, we therefore sought to determine whether MBCs develop in elderly individuals following the resolution of SARS- CoV-2 infection and whether they can maintain functionality once the nAbs have waned. To address these questions, we studied elderly individuals with mild or asymptomatic SARS-CoV-2 infec- tion who had recovered from the infection following outbreaks in 3 elder care homes in the United Kingdom. A substantial pro- portion of these individuals had lost detectable nAbs 5 months after infection. We compared MBCs between the elderly care home residents and younger staff members to assess the impact of aging. We identified MBCs specific for SARS-CoV-2 spike and receptor-binding domain (RBD) proteins that persisted when serum nAbs had waned below detectable limits. Their frequency, phenotype, isotype, and function were analyzed according to the individual’s age and/or nAb loss, to inform the assessment and boosting of durable immunity in the elderly. Results SARS-CoV2 spike- and RBD-specific MBCs can persist after loss of nAbs. To study the role of MBCs, we obtained PBMCs from a sub- set of individuals (n = 42) from a large cohort who had survived COVID-19 with mild or asymptomatic infection after outbreaks in 3 elder care homes in April 2020 (see Methods, Supplemental Table 1, and refs. 21, 22). The care home cohort subset was select- ed in order to have a wide range of detectable titers of nAbs against live virus at the first sampling time point (month 1, May 2020, Figure 1), while all maintained detectable binding Abs by at least 1 assay (Supplemental Table 2; supplemental material available online with this article; https://doi.org/10.1172/JCI152042DS1). By the end of September 2020 (month 5), 29% of all partici- pants sampled had stable (or in 2 cases increasing) nAbs against live virus. In contrast, 17% had declining titers, and 52% had lost detectable nAbs (Figure 1, A and B). One individual never had detectable nAb titers. To compare MBC frequencies among individuals who had maintained or lost nAbs, we stained PBMCs with SARS-CoV-2 spike trimer tetramers, made by preincubating recombinant biotinylated trimeric spike protein with fluorescence-conjugated streptavidin (15). Dual staining with spike tetramers with 2 distinct fluorochromes was used to enhance the discrimination of true antigen-specific MBCs (Figure 1C), as described previously (23– 25). We calculated the frequencies of antigen-specific responses within the memory fraction of B cells (CD19+CD20+ excluding IgD+, CD38hi, and CD21+CD27– naive fractions; see gating strategy in Supplemental Figure 1A, as previously described in ref. 26). A threshold for background nonspecific staining was set at the mean ± 2 SD of staining seen in pre-pandemic healthy donor samples (Supplemental Figure 1B). Results were also compared with the control cohort derived from the same care homes (seronegative at both time points, Supplemental Table 1). Spike-specific MBCs were detectable in 41 of the 42 tested individuals 5 months after infection, compared with 2 of 11 indi- viduals of the care home control group who remained negative for binding Abs (Figure 1D). The frequency of spike-specific MBCs was reduced in those who had lost nAbs compared with those in whom they were still detectable (Figure 1E). Of note, however, most of those (96%) who had lost detectable nAbs still had some The Journal of Clinical Investigation persistent spike-specific MBCs, a proportion comparable to that in the group that maintained nAb levels (Figure 1F). The frequency of spike-specific MBCs correlated significantly with the strength of the nAb response (nAb titer against live virus) at 5 months (Fig- ure 1G); however, there was partial discordance due to detection of spike-specific MBCs in most individuals with no nAbs (dot- outlined box, Figure 1G). Next, we analyzed the MBC response specifically direct- ed against RBD, since this is the region within spike that many SARS-CoV-2–specific nAbs target (15, 27–29). RBD-specific MBCs were identified by gating on dual spike-tetramer staining cell populations that were also stained with a tetramer formed from recombinant biotinylated RBD protein preincubated with fluores- cence-conjugated streptavidin (Figure 1H). RBD-specific respons- es were detectable in 38 of the 41 with a sufficient magnitude of spike-specific MBC responses (>20 dual spike+ cells) to allow analysis of the RBD-costained cells (Figure 1I). The frequency of RBD-specific MBCs was significantly reduced in the group of individuals who had lost nAbs compared with those with stable (or waning but still detectable) nAbs (Figure 1I). However, as noted with spike-specific MBCs, some RBD-specific MBCs remained detectable in most of the cohort, irrespective of whether they had lost nAbs (Figure 1J). Overall, the magnitude of RBD-specific MBCs correlated with nAb titers, although, again, there was some discordance due to the persistence of RBD-specific MBCs in those who had lost nAbs (dot-outlined box in Figure 1K). Important- ly, both the RBD-positive and RBD-negative components of the spike-specific B cell response significantly correlated with nAb titers (Figure 1K and Supplemental Figure 1C). This highlights the importance of the RBD as the major target for nAbs, while also underscoring the contribution of Ab-targeting regions outside of the RBD (for example, the N-terminal domain [NTD] of the spike protein; refs. 15, 29–31) to the nAb response at the 5-month time point in this cohort. These data therefore revealed the persistence of detectable, albeit reduced, MBCs specific for both spike and RBD proteins in most individuals whose nAb titers against live virus had fall- en below the threshold of detection. Thus, loss of detectable nAbs 5 months after asymptomatic/mild infection is frequently compensated by the presence of a memory response primed to respond upon reexposure. Comparable persistence of spike- and RBD-specific MBCs in elderly care home residents and younger staff. The elderly care home cohort was constructed to sample 2 comparator groups: elderly residents (median age, 87 years; range, 66–100 years) and a con- trol group of younger staff (median age, 56 years; range 41–65 years). Five months after asymptomatic or mild infection, the proportion of elderly residents who had lost detectable nAbs was nonsignificantly lower than that of the care home staff members (Figure 2A), and those who maintained detectable nAbs had simi- lar titers (Figure 2B). We postulated that there may, nevertheless, be a defect in the maintenance of spike- and RBD-specific MBCs in the elderly compared with the younger age group. However, spike-specific MBCs were maintained at similar frequencies and in comparable proportions in the elderly residents and young- er staff (Figure 2, C and D). There were no clear trends toward a decrease in spike-specific MBCs with increasing age, even in resi- 2 J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 RESEARCH ARTICLEThe Journal of Clinical Investigation Figure 1. Spike- and RBD-specific MBCs persist 5 months after SARS-CoV-2 infection despite waning nAbs. (A) Paired live virus nAb titers 1 month and 5 months after infection (n = 42). (B) Proportion of infected individuals with a change in nAbs between months 1 and 5: increase = 4-fold or greater rise; stat- ic = less than a 4-fold increase/decrease; decline = 4-fold or higher decrease; loss = undetectable at 5 months; never detectable = undetectable at 1 month and 5 months (n = 42). (C) Representative FACS plots of dual staining of MBCs with SARS-CoV-2 spike tetramers for infected and uninfected individuals. (D and E) Frequency of spike-specific MBCs (D) in infected (n = 42) and uninfected (n = 11) individuals and in (E) infected individuals with nAbs (n = 19) or no nAbs (n = 13) at 5 months. Dashed lines indicate the threshold for spike-specific responses determined by pre-pandemic controls (see also Supplemental Figure 1B). (F) Proportion of infected individuals with detectable spike-specific MBCs with nAbs (n = 19) or no nAbs (n = 23) at month 5. (G) Correlation between the frequency of spike-specific MBCs and nAb titers (n = 42). (H) Representative FACS plots showing dual staining of MBCs with SARS-CoV-2 spike tetramers (top) and RBD tetramer staining of dual spike-specific MBCs (bottom) from an infected individual. (I) Frequency of RBD-specific MBCs in infected individuals with spike-specific responses stratified by nAbs (n = 18) and no nAbs (n = 20) at 5 months. (J) Proportion of infected individuals with RBD-specific MBCs with nAbs (n = 18) or no nAbs (n = 20) at 5 months. (K) Correlation between RBD-specific MBC frequency and live virus nAb titers (n = 38). (A) Wilcoxon matched-pairs test, P ≤ 0.0001. (D, E, and I) Bars indicate the median and IQR; Mann-Whitney U test; (D) P ≤ 0.0001, (E) P = 0.0039, (I) P = 0.003. (F and J) Fisher’s exact test; (F) P > 0.9999, (J) P = 0.6135. (P < 0.005 and *P < 0.0001.) (G and K) Dot-outlined boxes indicate individuals with discordant MBC and nAb responses. Significance was determined by Spearman’s rank correlation. Analysis of RBD-specific MBCs was done only for those with 20 or more cells in the spike-specific gate. J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 3 RESEARCH ARTICLEdents in their nineties (Figure 2E). The percentage of global B cells (among live cells) was significantly lower in the elderly residents (in line with previous reports of B cell lymphopenia with aging; Supplemental Figure 2A) (32). However, residents had higher pro- portions of IgD– MBCs than did staff members (Supplemental Fig- ure 2B), such that spike-specific MBCs were still not significantly lower in residents than in staff when calculated as a percentage of all live cells acquired (Supplemental Figure 2C). Similarly, RBD-specific MBCs were equally well maintained in the elderly residents and staff (Figure 2, F and G), with no decline in their frequencies (as a fraction of total MBCs) with increasing age (Figure 2H). RBD-specific MBCs comprised a vari- able proportion of the total spike-specific MBC response (4.6%– 1.0%; median, 19.3%), the remainder representing B cells tar- geting non-RBD regions of the spike protein. The proportions of RBD- and non-RBD-binding, spike-specific MBCs again showed no changes with age (Figure 2I). Skewed isotype and activated memory phenotype of spike- and RBD-specific B cells. Having identified and quantified antigen- specific B cells with tetramer staining, we were able to apply high- dimensional multiparameter flow cytometry to phenotype these low-frequency cell populations without any in vitro manipulation. We investigated the Ig isotype, memory phenotype, homing mark- ers, and transcription factor usage of spike- and RBD-specific B cells and global B cells (Figure 3). The vast majority of SARS-CoV-2 MBCs expressed IgG, with a similar isotype distribution observed between spike- and RBD-specific MBCs (Figure 3, A–C). However, individuals with persistent nAbs had a higher frequency of IgG isotype–expressing spike- and RBD-specific MBCs than did their counterparts who had lost nAbs (Figure 3, B and C), indicating the establishment of a robust, class-switched memory response in these individ- uals. In contrast, those whose nAbs had waned below detect- able limits had lost more IgG and had a relative preservation of IgA class–switched, RBD-specific MBCs (Figure 3, B and C). IgM-expressing cells represented a small proportion of spike- or RBD-specific MBCs 5 months after infection, and their frequen- cies were comparable between groups (Figure 3, B and C). In the elderly residents, we observed a similar trend toward less IgG on the spike-specific MBCs but, overall, no significant skewing of Ig class–switching compared with younger staff (Figure 3, B and C). Global B cells showed the same pattern of expression of different Ig isotypes on their surface in SARS-CoV-2–resolved donors as in uninfected controls, with roughly equal proportions of IgG and IgA and less than 15% IgM (Supplemental Figure 3A). We examined MBC subsets using the combination of CD27 and CD21. The majority of spike- and RBD-specific B cells had a classical resting memory phenotype (CD27+CD21+), characteristic of functional responses and comparable to the global MBC com- partment, in both the elderly resident and staff groups (Figure 3, D–F, and Supplemental Figure 3B). Double-negative (DN) B cells have been associated with B cell dysfunction in aging (33–35) and the DN2 subset with an extrafollicular short-lived plasmablast response in the acute phase of a cohort with severe COVID-19 (36). However, at the 5-month time point following mild or asymptom- atic infection in our cohort, neither the elderly nor those who had lost nAbs showed any expansion of CD27–CD21– B cells (Figure 3, The Journal of Clinical Investigation E and F) or the DN2 subset (CD27–CD21–CXCR5loCD11chi, Sup- plemental Figure 3C). Instead, we found a selective enrichment of the activated MBC subset (CD27+CD21–, previously described to be expanded in HIV and Ebola infection or after vaccination; refs. 37–39) in the RBD-binding fraction in elderly residents, with the same trend observed in those who had lost nAbs (Figure 3G). Those who had lost nAbs also had reduced expression of the B cell homing molecules CXCR3 and CXCR5 on spike-specific and glob- al MBCs (nonsignificant trend and significant, respectively, Sup- plemental Figure 3, D–F). T-bet, a transcription factor critical for acute antiviral function in B cells but associated with dysfunction in chronic infections and autoimmunity (40–43), also tended to be expressed at lower levels in the spike-specific MBCs of those losing nAbs (Figure 3H). Taken together, the isotype and memory phenotype of global and antigen-specific B cells was largely preserved in the elderly care home population, apart from a notable increase in spike-specific activated MBCs. Individuals who maintained nAbs had predomi- nantly IgG-expressing antigen-specific MBCs. In contrast, in those who had lost nAbs by 5 months, whether staff or residents, residual antigen-specific B cells showed preferential preservation of IgA. Elderly residents maintain functional spike- and RBD-specific B cells but at reduced frequency compared with younger care home staff members. Having found that antigen-specific MBCs could persist following loss of all detectable circulating nAbs, we wanted to confirm their potential for functional recall upon reencountering SARS-CoV-2. We therefore used cultured B cell enzyme-linked immune absorbent spots (ELISPOTs) to examine the in vitro capacity of persistent SARS-CoV-2–specific MBCs to differentiate into plasmablasts capable of secreting IgG-binding recombinant trimeric spike, S1 subunit of the spike protein (S1), or RBD proteins. ELISPOTs were performed using PBMCs from 32 seropositive elderly care home residents and staff members (n = 23 residents, n = 9 staff), with the threshold for detection set at the highest observed value in an uninfected control group (n = 5 seronegative elderly care home residents and n = 5 pre-pandemic controls). Only individuals with responses detectable in a control total IgG well were included in the analysis. Where responses were too numerous to count (TNTC), we used the highest number of spot-forming cells (SFCs) observed in the maximal response to the respective protein (Supplemental Figure 4A). We observed functional recall responses to SARS-CoV-2 tri- meric spike protein in 26 of the 32 seropositive individuals tested, with ELISPOTs tending to be positive in a larger number of those who had maintained nAbs (Figure 4A). However, the majority of individuals who had lost detectable nAbs still had a spike-specif- ic response by ELISPOT, with no significant difference in their magnitude compared with the nAb group (Figure 4A). ELISPOTs revealed similar results for binding of IgG to S1 and RBD, with a trend toward a lower proportion of positive results in individuals who had lost nAbs, but no significant difference in the magnitude of B cell recall responses in those who did or did not maintain serum nAbs (Figure 4, B and C). The magnitude of the RBD recall response assessed by ELIS- POT showed a significant correlation with detection of both spike and RBD proteins in MBCs by tetramer staining (Figure 4, D and E). However, there was some discordance due to individuals who 4 J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 RESEARCH ARTICLEThe Journal of Clinical Investigation Figure 2. Comparable persistence of spike- and RBD-specific MBCs in elderly care home residents and younger staff. (A) Proportion of staff (n = 10) and residents (n = 32) with detectable nAbs at 1 month, who continued to have detectable nAbs at 5 months. (B) nAb titers at month 5 for all infected individ- uals stratified by staff (n = 10) and residents (n = 32). Dashed line indicates the assay threshold for detection; undetectable titers were assigned a value of 10. (C) Frequency of dual spike-specific MBCs in staff (n = 10) and residents (n = 32). (D) Proportion of infected individuals with detectable spike-specific MBCs stratified by staff (n = 10) and residents (n = 32). (E) Frequency of dual spike-specific MBCs for staff (gray) and residents (blue) ordered by age from youngest on the left to oldest on the right. (F) Frequency of RBD-specific MBCs in staff (n = 10) and residents (n = 28) with detectable spike-specific responses. (G) Proportion of infected individuals with detectable RBD-specific MBCs stratified by staff (n = 10) and residents (n = 32). (H) Frequency of RBD-specific MBCs for staff (gray) and residents (blue) ordered by age from youngest (left) to oldest (right). (I) Proportion of dual spike-specific cells with specificity for RBD (staff = dark gray; residents = dark blue) or the non-RBD region (staff = light gray; residents = light blue) in staff (n = 10) and residents (n = 28). (A, D, and G) Fisher’s exact test was used to determine statistical significance: (A) P > 0.9999, (D) P > 0.9999, and (G) P = 0.5569. (B, C, and F) Bars indicate the median and IQR. A Mann-Whitney U test was used to determine significance: (B) P = 0.4367, (C) P = 0.2552, and (F) P = 0.1068. (C and E) Dashed line indicates the threshold for spike-specific responses determined by pre-pandemic controls (Supplemental Figure 1B). had tetramer-binding spike or RBD B cells that did not produce detectable IgG by ELISPOT (dot-outlined boxes, Figure 4, D and E), mainly in those who had lost nAbs. IgM- and IgA-specif- ic ELISPOTs detected minimal numbers of S1 and RBD IgM- and IgA-expressing cells in the subset of individuals who had nega- tive IgG ELISPOTs despite detectable tetramer binding (Supple- mental Figure 4, B and C), suggesting that isotype specificity was not the main factor accounting for this discrepancy. Importantly, J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 5 RESEARCH ARTICLEThe Journal of Clinical Investigation Figure 3. Preserved memory phenotype but skewed isotype of spike- and RBD-specific B cells with loss of nAbs. (A) Representative FACS plots of IgM/IgG on spike-specific, RBD-specific, and global MBCs from an infected individual. (B) Frequency of IgG+, IgA+ (IgD–IgG–IgM–), and IgM+ spike–specific MBCs by nAbs (n = 17) and no nAbs (n = 18) at 5 months and staff (gray, n = 10) and resident (blue, n = 25) status. (C) Frequency of IgG+, IgA+, and IgM+ RBD–specific MBCs by nAbs (n = 11) and no nAbs (n = 7) at 5 months and by staff (gray, n = 6) and resident (blue, n = 12) status. (D) Representative FACS plots of CD21 and CD27 on spike-specific, RBD-specific, and global MBCs from an infected individual. (E) Frequency of CD21–CD27+, CD21+CD27+, and CD21–CD27– subsets of spike- specific MBCs by nAbs (n = 17) and no nAbs (n = 19) at 5 months, ordered by increasing age. (F) Frequency of CD21–CD27+, CD21+CD27+, and CD21–CD27– subsets of RBD-specific MBCs with nAbs (n = 13) or no nAbs (n = 8) at 5 months, ordered by increasing age. (G) Frequency of CD21–CD27+ RBD-specific MBCs by nAbs (n = 13) and no nAbs (n = 8) at 5 months and by staff (gray, n = 7) and resident (blue, n = 14) status. (H) Representative plots and summary data showing the frequency of spike-specific and global MBCs expressing T-bet, by nAbs (n = 19) and no nAbs (n = 13) at 5 months, by staff (gray, n = 10) and resident (blue, n = 22) status, and by uninfected controls (n = 13). (B, C, and G) Bars indicate the median and IQR. A Mann-Whitney U test was used to determine statistical significance (P < 0.05, **P < 0.005). (B) IgG P = 0.0382; NS, IgA P = 0.0045; NS, IgM. (C) IgG P = 0.0220; NS, IgA P = 0.0055; NS, IgM. (G) NS, P = 0.0180. (H) Bars indicate the median and IQR. Statistical significance was assessed by Kruskal-Wallis test with Dunn’s correction between nAb, no nAb, and uninfected subgroups and staff, resident, and uninfected subgroups, respectively, on global cell populations (all NS). Analysis was performed on all individuals with 50 or more cells in the parent gate for all phenotypic analyses. 6 J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 RESEARCH ARTICLEThe Journal of Clinical Investigation Figure 4. Elderly individuals maintain some functional spike- and RBD-specific B cells at reduced frequency compared with younger care home staff. (A–C) Left panels: SFCs/106 PBMCs for infected individuals with nAbs or no nAbs at 5 months and for uninfected controls. Right panels: Proportion of infected individuals with detectable recall responses for (A) spike (nAbs, n = 14; no nAbs, n = 18; uninfected, n = 10), (B) S1 (nAbs, n = 14; no nAbs, n = 18; uninfected, n = 10), and (C) RBD protein (nAbs, n = 14; no nAbs, n = 16; uninfected, n = 10). (D and E) Correlation between RBD SFCs/106 PBMCs and frequen- cy of (D) spike-specific and (E) RBD-specific MBCs for those with (green) or without (black) nAbs. (F) SFCs/106 PBMCs indicating recall responses for spike (gray), S1 (dark blue), and RBD (pale blue) per individual, stratified by nAb (n = 14) and no nAb (n = 18) status at 5 months and ordered by increasing age. (G–I) SFCs/106 PBMCs for infected staff members and residents indicating recall responses for (G) spike (staff, n = 9; residents, n = 23), (H) S1 (staff, n = 9; residents, n = 23), and (I) RBD (staff, n = 9; residents, n = 21). (J) RBD SFCs/106 PBMCs for infected residents with nAbs (n = 8) or no nAbs (n = 12) at 5 months. (K) Summary heatmap of the proportion of staff members and residents with nAbs, spike- and RBD-specific MBCs by flow cytometry, and spike, S1, and RBD recall by ELISPOT at 5 months. Bars in A–C (left panels) and G–J indicate the median and IQR; dashed lines indicate the threshold for sero- negative and pre-pandemic controls. Statistical significance in A–C (left panels) was determined by Kruskal-Wallis test with Dunn’s correction. Statistical significance in A–C (right panels) was determined by Fisher’s exact test; (A) P = 0.3413, (B) P = 0.3926, and (C) P = 0.1870. (D and E) Dot-outlined boxes indicate individuals with a discordant MBC and ELISPOT response. Statistical significance was determined by Spearman’s rank correlation. (G–J) Statistical significance was evaluated by Mann-Whitney U test. (A, B, G, and H) Inverted triangles indicate TNTC responses, with the maximal response observed assigned. (F) Δ indicates S1 SFCs TNTC; # indicates spike SFCs TNTC; and θ indicates that RBD counts were unavailable. Individuals with zero response were assigned a value of 1 for logarithmic plotting; statistical analysis was performed using original values. these data revealed that circulating antigen-specific B cells can be detected in the absence of functional recall. Next, we compared functional responses to all 3 proteins for each individual, ranked according to nAb status and age. Individu- als with strong recall responses to spike (as measured by ELISPOT) tended to also have strong responses to S1 and RBD, whereas others had weak responses to all 3 antigens (Figure 4F). Functional MBC recall responses decreased with increasing age in both the groups, J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 7 RESEARCH ARTICLEregardless of whether serum nAbs were maintained (Figure 4F). Thus, elderly residents had significantly lower ELISPOT MBC responses against spike, S1, and RBD than did the younger staff group (Figure 4, G–I), which was confirmed with a negative correla- tion between age and ELISPOT response to all 3 antigens (Supple- mental Figure 4, D–F). Focusing on elderly residents, we found that those who had lost nAbs tended to have undetectable or reduced MBCs capable of a functional recall response to RBD (Figure 4J). Overall, the measurement of nAbs against live virus, combined with the assessment of spike- and RBD-specific MBCs by tetram- er staining and functional ELISPOTs, provided complementary insights into B cell immunity (Figure 4K). A substantial proportion of those who had lost detectable neutralizing activity against live virus maintained spike- and RBD-specific MBCs detectable with 1 or both assays, regardless of age. However, some of those with persistent antigen-specific MBCs could not mount a detectable functional response, particularly the elderly (Figure 4K). Discussion In this study, we sampled a cohort of very elderly residents and younger staff who developed mild or asymptomatic SARS-CoV-2 infection during care home outbreaks, a high proportion of whom had lost nAbs by 5 months (despite the maintenance of spike-bind- ing Abs). This allowed us to dissect the potential for B cell memory to persist beyond detectable serum nAb levels, providing a backup reserve for humoral immunity. We demonstrated by flow cytom- etry that the majority of the cohort maintained detectable fre- quencies of spike- and RBD-specific MBCs, even when they had lost circulating Abs capable of live virus neutralization. Tetramer staining allowed accurate ex vivo quantification and characteriza- tion of antigen-specific MBCs, revealing that individuals who had lost detectable nAbs had lower frequencies of spike- and RBD-spe- cific MBCs, with a preserved classical memory phenotype but class-switching skewed away from IgG toward IgA. Elderly and younger recovered individuals infected in the same care home outbreaks maintained similar frequencies of spike- and RBD- specific tetramer-staining B cells, with comparable isotypes but an increase in activated RBD-specific MBCs in the elderly. Impor- tantly, functional assessment using ELISPOT assays demonstrat- ed that the persisting spike, and particularly RBD-specific, MBCs had reduced potential for Ab production in the elderly. The success of an infection or vaccine at inducing durable humoral immunity is dependent on the generation of long-lived plasma cells and MBCs (17–19). The longevity of the plasma cell response, capable of sustaining Abs, varies widely following dif- ferent viral infections (5–7). A recent study demonstrated the pres- ence of bone marrow plasma cells secreting IgG against SARS- CoV-2 spike protein in 15 of 19 individuals examined 7 months after infection (44), a finding in line with the durability of some Abs in the first year after mild infection. Nevertheless, many stud- ies have also highlighted the potential for nAbs against SARS- CoV-2 to wane to a point where there is an, as yet ill-defined, risk of reinfection (45–47). Our study deliberately focused on the role of MBCs in individuals with waning or undetectable nAbs against live virus, despite the persistence of binding Abs. MBCs, previ- ously identified in younger COVID-19 cohorts (11, 26, 48, 49), can provide crucial backup by responding quickly to a pathogen The Journal of Clinical Investigation re-encounter or vaccination to form new plasmablasts, producing potent affinity-matured Abs with more flexible recognition of viral variants (9, 20); this is consistent with the enhanced nAb response described following vaccination of health care workers previously infected with SARS-CoV-2 (50). Our demonstration that B cells of relevant specificities can still be detected even when nAb titers are waning or completely undetectable provides some reassur- ance that a memory response remains intact in the elderly. Future large-scale studies are needed to assess whether B cell memory serves as an independent correlate of protection, or whether reli- ance on MBCs to mount a new response in the absence of existing Abs provides a critical window of opportunity for a virus that repli- cates as rapidly as SARS-CoV-2. One strategy to combat Abs that are waning or unable to cross-recognize emerging variants is the use of booster vaccines. Our finding that the elderly have impaired differentiation of per- sistent spike- and RBD-specific MBCs into Ab-producing cells, as determined by ELISPOT assays, provides a biological rationale for the potential need for more frequent booster vaccinations in this high-risk group. The frequency and class-switching responses of antigen-specific B cells did not reveal obvious changes in the elder- ly group that would account for this functional defect, but pheno- typic analysis of MBCs did reveal an increase in the CD27+CD21– subset. The activated CD27+CD21– subset of MBCs has recently been noted to remain expanded in some resolved COVID-19 patients (51), consistent with emerging literature supporting the possibility of prolonged antigen persistence, exemplified by a recent study detecting SARS-CoV-2 in the small bowel 4 months after asymptomatic infection (9). Our finding of an expanded population of CD27+CD21– MBCs selectively within RBD-specific (and not global) responses in the older age group raises the pos- sibility there is more prolonged antigen persistence and resultant B cell activation following SARS-CoV-2 infection in the elderly. The aging immune system is characterized by a tendency toward low-level chronic inflammation (52, 53), which could also contrib- ute to prolonged activation; however, this might be expected to affect all MBCs irrespective of antigen specificity. Analogous to our findings in elderly care home residents, both older individuals and those with HIV have been found to have persistent circulating MBCs but defective plasmablast formation, resulting in reduced influenza vaccine–induced Abs (54, 55). Such age-related defects in B cell responses to vaccination have been attributed to a com- bination of B cell–intrinsic senescence and defective Tfh cells in germinal centers (56–58). One limitation of our study is that the majority of care home residents and staff were female, implying that results cannot necessarily be generalized to males. Residents and staff were matched by the fact they were infected during the same care home outbreaks, therefore likely with the same SARS-CoV-2 strain and time frame. A potential confounder may be that more residents than staff members were symptomatic, although all symptomat- ic individuals in both groups were mild, with no requirement for oxygen or hospital attendance. Data on other cofactors that could have influenced immunity in addition to age were not available; larger cohorts would be needed to assess these. Another caveat to our study is that we were only able to study circulating B cells, whereas additional recall responses may be compartmentalized 8 J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 RESEARCH ARTICLEThe Journal of Clinical Investigation within the mucosa. A recent study suggested that mild infection can stimulate mucosal SARS-CoV-2–specific IgA secretion in the absence of circulating Abs (59). The bias toward the reten- tion of IgA+ spike- and RBD-specific MBCs in those who had lost all detectable serum nAbs against live virus could therefore be reflective of a stronger mucosal response in these individuals. An increase in mucosal-homing IgA responses has been described as a feature of the aging immune response (60), consistent with the older composition of our cohort. Alternatively, the relative pres- ervation of IgA rather than IgG spike- and RBD-specific MBCs in those with the fastest waning nAbs may simply reflect the recent observations that IgA dominates the early nAb response to SARS- CoV-2 infection and may not decline as fast as the IgG response (9, 61). In vitro ELISPOT assays may underestimate the full extent of residual SARS-CoV-2–specific responses able to mount function- al memory in vivo. In addition, several studies have shown that the magnitude of the MBC response to SARS-CoV-2 continues to increase beyond 6 months (9, 23, 51, 62), again implying that we may have underestimated the extent of recall potential in our cohort at 5 months. Future studies should also examine the pres- ervation of non-spike-specific MBCs with the potential to produce Abs mediating antiviral effects beyond neutralization, since other viral proteins (ORF3a, membrane and nucleocapsid) can play a dominant role in triggering Ab-dependent NK cell activation (63). In conclusion, by focusing on an elderly cohort with a high pro- portion of nAb loss, we demonstrated that this waning in the first line of humoral defense could be compensated by the presence of a reserve of adaptive B cell memory in the majority of cases. Our findings highlight the importance of including measures of B cell memory in larger studies of natural infection and vaccination to determine their role as additional correlates of protection. Our data underscore the idea that identifying antigen-specific B cells by tetramer antigen staining is useful for quantitation and thor- ough ex vivo characterization, but may not necessarily equate with the preservation of a functional response, in line with discrepan- cies between the frequency and function of MBCs described in chronic viral infection (43, 64). The relative preservation of IgA antigen–specific MBCs in those with waned serum nAb raises the possibility that mucosal sequestered immunity may outlast that which is detectable in the circulation. Increased expansion of activated MBCs in the elderly highlights the need to investigate whether these cells are more prone to prolonged stimulation from persistent reservoirs of SARS-CoV-2 antigen. A finding of con- cern was the lack of detectable functional recall to RBD in elderly donors who had lost nAbs; given that RBD is the dominant site for nAbs, this observation supports the need for additional monitor- ing and/or booster vaccines to maintain sufficient Abs to neutral- ize emerging variants in this highly vulnerable group. Methods Participants. SARS-CoV-2 antigen–specific MBC responses were com- pared between elderly care home residents and younger staff counter- parts exposed to the virus in the same environment. Individuals from 6 care homes that reported SARS-CoV-2 outbreaks to Public Health England (PHE) were recruited for longitudinal SARS-CoV-2 reverse transcription PCR (RT-PCR) and serological follow-up in April 2020 (T0; refs. 1, 21). The serostatus of the individuals in these care homes at 1 month and 5 months after the outbreaks was established using binding and functional assays as previously described (21, 22). Briefly, seropositivity was assessed with a native virus lysate ELISA assay, RBD binding, and virus neutralization using the England 2 SARS CoV-2 prototype virus (21, 22). A total of 42 individuals with mild or asymptomatic SARS-CoV-2 infection (n = 32 elderly residents; n = 10 staff members), all of whom were seropositive according to at least 1 of the binding assays described above at both sampling time points (month 1: May 2020, month 5: Sep- tember 2020; Supplemental Table 2), were recruited along with 11 con- trol SARS-CoV-2–seronegative individuals from 3 of the care homes. Participants donated 30 mL blood to be processed for PBMCs and serum 5 months after the initial outbreaks (month 5). Stored pre-pan- demic samples from 7 healthy individuals were used as controls. Sample processing and data collection. Venepuncture blood sam- ples collected in lithium heparin–coated tubes, and serum separa- tion tubes were used for isolation of PBMCs and serum, respectively. PBMCs were isolated by density centrifugation using Pancoll human (PAN-Biotech). Isolated PBMCs were frozen in FBS supplemented with 10% DMSO (MilliporeSigma). Prior to use, samples were thawed and washed in PBS. Serum was collected following centrifugation and stored at –80°C prior to use. Clinical and laboratory data including age, sex, symptoms, and SARS-CoV-2 RT-PCR status at the time of the initial outbreak were available for participants (Supplemental Table 1 and ref. 1). Protein expression and purification. Recombinant spike and spike RBD proteins of SARS-CoV-2 for antigen-specific B cell flow cytome- try and ELISPOT were expressed and purified as previously described (15). Briefly, spike glycoprotein trimer (uncleaved spike stabilized in the prefusion conformation (GGGG substitution at furin cleavage site and 2P mutation; ref. 65) and RBD protein (12) were cloned into a pHLsec vector containing Avi and 6xHis tags. Biotinylated spike and RBD were expressed in Expi293F cells (Thermo Fisher Scientific). Supernatants were harvested after 7 days and purified. For the production of bioti- nylated protein, spike- and RBD-encoding plasmids were cotransfect- ed with BirA and PEI-Max in the presence of 200 μM biotin. Recombinant S1 protein constructs spanning SARS-CoV-2 res- idues 1–530 for ELISPOT were produced as previously described (28, 30). Briefly, codon-optimized DNA fragments were cloned into mammalian expression vector pQ-3C-2xStrep to create plasmids, which were then transfected into Expi293F cells growing at 37°C in a 5% CO2 atmosphere using ExpiFectamine reagent (Thermo Fisher Scientific). Proteins were purified by strep-tag affinity followed by size exclusion chromatography. Flow cytometry. High-dimensional, multiparameter flow cytome- try was performed for ex vivo identification of spike- and RBD-spe- cific B cells. Two panels (surface and intranuclear) of mAbs were used to phenotype global and antigen-specific subsets (Supplemental Table 3). Biotinylated tetrameric spike (1 μg) and RBD (0.5 μg) were fluorochrome linked for flow cytometry by incubation with streptavi- din-conjugated allophycocyanin (APC) (ProZyme) and phycoerythrin (PE) (ProZyme) for spike and with BV421 (BioLegend) for RBD, for 30 minutes in the dark on ice. PBMCs were thawed and incubated with Live/Dead fixable dead cell stain (UV, Thermo Fisher Scientific) and saturating concentra- tions of phenotyping mAbs (Supplemental Table 2) diluted in 50% 1× PBS 50% Brilliant Violet Buffer (BD Biosciences). For identification J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 9 RESEARCH ARTICLEof SARS-CoV-2 antigen–specific B cells, 1 μg per 500 μL stain each of tetrameric spike-APC and spike-PE and 0.5 μg per 500 μL stain of tetrameric RBD-BV421 were added to the cell preparation. Parallel samples stained with an identical panel of mAbs, but excluding the SARS-CoV-2 proteins (fluorescence minus one [FMO] controls), were used as controls for nonspecific binding. Cells were incubated in the staining solution for 30 minutes at room temperature, washed with PBS, and subsequently fixed with either fixa- tion and permeabilization solution (BD Biosciences) or a FOXP3 Buffer Set (BD Biosciences) according to the manufacturer’s instructions for surface and intranuclear staining, respectively. Saturating concentra- tions of mAbs diluted in 1× PBS were added following permeabilization for the detection of intranuclear proteins. All samples were acquired on a Fortessa-X20 (BD Biosciences) and analyzed using FlowJo (TreeStar). B cell subsets were defined as CD19+CD20+ MBCs, excluding the IgD+, CD38hi, and CD21+CD27– naive fractions (see gating strategy in Supplemental Figure 1A) and CD19+CD20+CD38+/–CD21–CD27– CD11chiCXCR5lo DN2 cells. For analysis of RBD-costaining cells, a sufficient magnitude of spike-specific MBCs (≥20 dual spike+ cells) was required. For phenotypic analysis of spike- and RBD-specific cells, a sufficient magnitude of responses (≥50 cells in the relevant parent gate) was required. MBC recall response to SARS-CoV-2 by ELISPOT. To activate MBC differentiation, 1 × 106 PBMCs were stimulated with 1 μg/mL R848 (TLR7/8 agonist; resiquimod, InvivoGen) diluted in complete RPMI (cRPMI) (RPMI supplemented with 10% FBS plus recombinant human IL-2; 20 IU/mL; Peprotech), as previously described (66, 67). Activated cells were incubated for 6 days with a media change on day 3. ELISPOT plates (Mabtech) were precoated with recombinant SARS-CoV-2 trimeric spike (1 μg/mL), S1 (1 μg/mL), and RBD (10 μg/ mL) and anti–human IgG (1 μg/mL, Jackson ImmunoResearch) over- night at 4°C. Coated plates were blocked with cRPMI with 10% FBS prior to the addition of cells. Cultured PBMCs were added at varying concentrations depending on SARS-CoV-2 antigen and incubated at 37°C in 5% CO2 for 18 hours: 50,000 cells/ well to detect spike- specific, IgG-secreting cells; 100,000 cells/well to detect S1 and RBD IgG-secreting cells; and 1000 cells/well to detect total IgG-secret- ing cells. To control for nonspecific binding, uncoated control wells were incubated with 100,000 prestimulated cells. The following day, ELISPOT plates were washed in filtered PBS supplemented with 0.5% Tween-20 (Merck) and incubated for 4 hours in the dark at room tem- perature with 1 μg/mL goat anti–human IgG HRB Ab (Jackson Immu- noResearch). Cells were again washed 3 times with PBS–Tween-20 (0.5%) and 3 times with PBS, and then developed with 3-amino-9- ethylcarbazole (AEC) substrate (BD Biosciences) according to the manufacturer’s instructions. ELISPOT plates were washed with ddH20 before analysis using ViruSpot (Autoimmun Diagnostika). All conditions were performed in duplicate and the responses averaged. For the detection of IgM- and IgA-secreting cells, PBMCs were stimulated as before for 5 days. Cultured PBMCs were added to coat- ed and blocked plates, as described above, and incubated at 37°C in 5% CO2 for 6 hours. ELISPOT plates were then washed in filtered PBS supplemented with 0.5% Tween-20 (Merck) and incubated overnight in the dark at 4°C with 1 μg/mL goat anti–human IgM- or IgA-HRP Ab (Jackson ImmunoResearch). Procedures under all conditions were performed in duplicate and the responses averaged. A control well coated with anti-IgM or anti-IgA was used for the detection of total The Journal of Clinical Investigation IgM- or IgA-secreting cells. Data are presented minus the background, calculated from an uncoated well. Statistics. Data were analyzed using GraphPad Prism (GraphPad Software). Descriptive statistical analyses were performed. Contin- uous data that did not follow a normal distribution were described as medians with IQRs, and differences were compared using the Mann-Whitney U test (2-tailed), Wilcoxon’s paired t test (2-tailed), or the Kruskal-Wallis test with Dunn’s post hoc test for pairwise multiple comparisons as appropriate. Contingency table analyses were con- ducted using Fisher’s exact test. Correlations for nonparametric data were assessed using Spearman’s rank correlation with a 95% CI. A P value of less than 0.05 was considered significant. Study approval. The study protocol was reviewed and approved by the PHE Research Ethics and Governance Group (REGG reference NR0204). Written information regarding the study was provided to all participants. Verbal informed consent for testing was obtained by care home managers from staff members and residents or their next of kin as appropriate. Stored pre-pandemic samples from 7 healthy individuals were used as controls, recruited under ethics number 11/ LO/0421 and approved by the South East Coast – Brighton and Sussex Research Ethics Committee. Author contributions AJS, MZ, LEM, and MKM conceptualized the study. AJS, ARB, LM, KJD, SNL, LEM, and MKM designed the methodology. AJS, ARB, SL, JD, MP, RG, CRS, and LEM performed experiments. AJS, FA, SNL, and JYC collected samples and clinical data. AJS, ARB, SL, MP, RG, LEM, and MKM analyzed the data. Funding acquisition: SNL, JYC, MZ, and MKM acquired funding. SNL, MZ, LEM, and MKM supervised the study. AJS and MKM wrote the original draft of the manuscript. All authors reviewed and edited the manuscript. Acknowledgments The authors are very grateful to the care home managers, staff members, and residents; without their support and engagement, this study would not have been possible. The authors would also like to thank the staff in the Immunisation and Countermeasures Department, in particular Maria Zavala; the Virus Reference Department; PHE Operations; the London Coronavirus Response Cell, in particular Nalini Iyanger and Jonathan Fok; PHE Field ser- vices; and members of the Maini laboratory for their help coordi- nating this study. We thank Peter Cherepanov of the Francis Crick Institute for supplying recombinant S1 antigen. This work was sup- ported by PHE and by a Medical College of St. Bartholomew’s Hos- pital Trustees Clinical Research Fellowship (to AJS), and National Institute for Health Research (NIHR) Efficacy and Mechanism Evaluation (EME), EU Horizon 2020, and UK Research and Inno- vation (UKRI), NIHR UK-CIC grants (to MKM). LEM is supported by a Medical Research Council Career Development Award (MR/ R008698/1). Additional support was provided by the UCL Coro- navirus Response Fund made possible through generous dona- tions from UCL’s supporters, alumni, and friends (to LEM). KJD is supported by the King’s Together Rapid COVID-19 Call. Address correspondence to: Mala K. Maini, UCL Institute of Immunity and Transplantation, The Pears Building, Pond Street, London, NW3 2PP, United Kingdom. Email: m.maini@ucl.ac.uk. 1 0 J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 RESEARCH ARTICLEThe Journal of Clinical Investigation 1. Ladhani SN, et al. Investigation of SARS-CoV-2 outbreaks in six care homes in London, April 2020. EClinicalMedicine. 2020;26:100533. 2. McMichael TM, et al. Epidemiology of Covid-19 in a long-term care facility in King County, Wash- ington. N Engl J Med. 2020;382(21):2008–2011. 3. Edridge AWD, et al. Seasonal coronavirus protective immunity is short-lasting. Nat Med. 2020;26(11):1691–1693. 4. Siggins MK, et al. Durability of immunity to SARS-CoV-2 and other respiratory viruses. Trends Microbiol. 2021;29(7):648–662. 5. Amanna IJ, et al. Duration of humoral immunity to common viral and vaccine antigens. N Engl J Med. 2007;357(19):1903–1915. 6. Amanna IJ, Slifka MK. Mechanisms that determine plasma cell lifespan and the dura- tion of humoral immunity. Immunol Rev. 2010;236(1):125–138. 7. Hammarlund E, et al. Plasma cell survival in the absence of B cell memory. Nat Commun. 2017;8(1):1–11. 8. Arkhipova-Jenkins I, et al. Antibody response after SARS-CoV-2 infection and implications for immunity. Ann Intern Med. 2021;174(6):811–821. 9. Gaebler C, et al. Evolution of antibody immunity to SARS-CoV-2. Nature. 2021;591(7851):639. 10. Harris RJ, et al. Serological surveillance of SARS- CoV-2: Six-month trends and antibody response in a cohort of public health workers. J Infect. 2021;11(5):162–169. 11. Hartley GE, et al. Rapid generation of durable B cell memory to SARS-CoV-2 spike and nucleo- capsid proteins in COVID-19 and convalescence. Sci Immunol. 2020;5(54):eabf8891. 12. Seow J, et al. Longitudinal observation and decline of neutralizing antibody responses in the three months following SARS-CoV-2 infection in humans. Nat Microbiol. 2020;5(12):1598–1607. 13. Röltgen K, et al. Defining the features and duration of antibody responses to SARS-CoV-2 infection associated with disease severity and outcome. Sci Immunol. 2021;5(54):eabe0240. 14. Rees-Spear C, et al. The effect of spike muta- tions on SARS-CoV-2 neutralization. Cell Rep. 2021;34(12):108890. 15. Graham C, et al. Neutralization potency of monoclonal antibodies recognizing dominant and subdominant epitopes on SARS-CoV-2 Spike is impacted by the B.1.1.7 variant. Immunity. 2021;54(6):1276–1289. COVID-19: Prospective cohort study, England. EClinicalMedicine. 2020;28:100597. 22. Jeffery-Smith A, et al. Infection and transmission of SARS-CoV-2 in London care homes reporting no cases or outbreaks of COVID-19: prospective observational cohort study, England 2020. Lan- cet Reg Health Eur. 2021;3:100038. 23. Dan JM, et al. Immunological memory to SARS- CoV-2 assessed for up to 8 months after infec- tion. Science. 2021;371(6529):eabf4063. 24. Amanna IJ, Slifka MK. Quantitation of rare mem- ory B cell populations by two independent and complementary approaches. J Immunol Methods. 2006;317(1–2):175–185. 25. Townsend SE, et al. Single epitope multiple stain- ing to detect ultralow frequency B cells. J Immu- nol Methods. 2001;249(1–2):137–146. 26. Rodda LB, et al. Functional SARS-CoV-2-specific immune memory persists after mild COVID-19. Cell. 2021;184(1):169–183. 27. Liu L, et al. Potent neutralizing antibodies against multiple epitopes on SARS-CoV-2 spike. Nature. 2020;584(7821):450–456. 28. Ng KW, et al. Preexisting and de novo humoral immunity to SARS-CoV-2 in humans. Science. 2020;370(6522):1339–1343. 29. Brouwer PJM, et al. Potent neutralizing antibodies from COVID-19 patients define multiple targets of vulnerability. Science. 2020;369(6504):643–650. 30. Rosa A, et al. SARS-CoV-2 recruits a haem metab- olite to evade antibody immunity [preprint]. https://doi.org/10.1101/2021.01.21.21249203. Posted on medRxiv January 26, 2021. 31. McCallum M, et al. N-terminal domain antigenic mapping reveals a site of vulnerability for SARS- CoV-2. Cell. 2021;184(9):2332–2347. 32. Valiathan R, et al. Effects of ageing on the immune system: infants to elderly. Scand J Immu- nol. 2016;83(4):255–266. 33. Ma S, et al. B cell dysfunction associated with aging and autoimmune diseases. Front Immunol. 2019;10:318. 34. Phalke S, Marrack P. Age (autoimmunity) associ- ated B cells (ABCs) and their relatives. Curr Opin Immunol. 2018;55:75–80. 35. Jenks SA, et al. Distinct effector B cells induced by unregulated toll-like receptor 7 contribute to pathogenic responses in systemic lupus erythe- matosus. Immunity. 2018;49(4):725–739. 36. Woodruff MC, et al. Extrafollicular B cell 16. Planas D, et al. Sensitivity of infectious SARS- CoV-2 B.1.1.7 and B.1.351 variants to neutralizing antibodies. Nat Med. 2021;27(5):917–924. responses correlate with neutralizing antibod- ies and morbidity in COVID-19. Nat Immunol. 2020;21(12):1506–1516. 17. Inoue T, et al. Generation of memory B 37. Moir S, et al. B cells in early and chronic HIV cells and their reactivation. Immunol Rev. 2018;283(1):138–149. 18. Quast I, Tarlinton D. B cell memory: understand- infection: evidence for preservation of immune function associated with early initiation of antiret- roviral therapy. Blood. 2010;116(25):5571–5579. ing COVID-19. Immunity. 2021;54(2):205–210. 38. Davis CW, et al. Longitudinal analysis of the 19. Akkaya M, et al. B cell memory: building two walls of protection against pathogens. Nat Rev Immunol. 2020;20(4):229–238. 20. Purtha WE, et al. Memory B cells, but not long-lived plasma cells, possess antigen spec- ificities for viral escape mutants. J Exp Med. 2011;208(13):2599–2606. 21. Ladhani SN, et al. High prevalence of SARS- CoV-2 antibodies in care homes affected by human B cell response to ebola virus infection in brief. Cell. 2019;177(6):1566–1582. 39. Lau D, et al. Low CD21 expression defines a population of recent germinal center graduates primed for plasma cell differentiation. Sci Immu- nol. 2017;2(7):eaai8153. 40. Knox JJ, et al. T-bet-expressing B cells during HIV and HCV infections. Cell Immunol. 2017;321:26–34. 41. Knox JJ, et al. T-bet+ B cells are induced by human viral infections and dominate the HIV gp140 response. JCI Insight. 2017;2(8):e92943. 42. Barnett BE, et al. Cutting edge: B cell-intrinsic T-bet expression is required to control chronic viral infection. J Immunol. 2016;197(4):1017–1022. 43. Burton AR, et al. Circulating and intrahepatic antiviral B cells are defective in hepatitis B. J Clin Invest. 2018;128(10):4588–4603. 44. Turner JS, et al. SARS-CoV-2 infection induces long-lived bone marrow plasma cells in humans. Nature. 2021;595(7867):421–425. 45. Lumley SF, et al. Antibody status and incidence of SARS-CoV-2 infection in health care workers. N Engl J Med. 2021;384(6):533–540. 46. Overbaugh J. Understanding protection from SARS-CoV-2 by studying reinfection. Nat Med. 2020;26(11):1680–1681. 47. Hall V, et al. Do antibody positive healthcare workers have lower SARS-CoV-2 infection rates than antibody negative healthcare workers? Large multi-centre prospective cohort study (the SIREN study), England: June to November 2020 [preprint]. https://doi.org/10.1101/2021.01.13.21 249642. Posted on medRxiv January 15, 2021. 48. Juno JA, et al. Humoral and circulating follicular helper T cell responses in recovered patients with COVID-19. Nat Med. 2020;26(9):1428–1434. 49. Ogega CO, et al. Durable SARS-CoV-2 B cell immunity after mild or severe disease. J Clin Invest. 2021;131(7):145516. 50. Reynolds CJ, et al. Prior SARS-CoV-2 infection res- cues B and T cell responses to variants after first vaccine dose [published online April 30, 2021]. Science. https://doi.org/10.1126/science.abh1282. 51. Sokal A, et al. Maturation and persistence of the anti-SARS-CoV-2 memory B cell response. Cell. 2021;184(5):1201–1213. 52. Mueller AL, et al. Why does COVID-19 dispro- portionately affect older people? Aging (Albany NY). 2020;12(10):9959–9981. 53. Franceschi C, et al. Inflamm-aging. An evolution- ary perspective on immunosenescence. Ann N Y Acad Sci. 2000;908:244–254. 54. Frasca D, et al. The generation of memory B cells is maintained, but the antibody response is not, in the elderly after repeated influenza immuniza- tions. Vaccine. 2016;34(25):2834–2840. 55. Rinaldi S, et al. Antibody but not memory B-cell responses are tuned-down in vertically HIV-1 infected children and young individuals being vaccinated yearly against influenza. Vaccine. 2014;32(6):657–663. 56. Frasca D, et al. Age-related factors that affect B cell responses to vaccination in mice and humans. Immunol Rev. 2020;296(1):142–154. 57. Linterman MA. How T follicular helper cells and the germinal centre response change with age. Immunol Cell Biol. 2014;92(1):72–79. 58. Silva-Cayetano A, et al. A booster dose enhances immunogenicity of the COVID-19 vaccine candi- date ChAdOx1 nCoV-19 in aged mice. Med (N Y). 2021;2(3):243–262. 59. Cervia C, et al. Systemic and mucosal antibody responses specific to SARS-CoV-2 during mild versus severe COVID-19. J Allergy Clin Immunol. 2021;147(2):545–557. 60. Arranz E, et al. Immunosenescence and muco- J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 1 1 RESEARCH ARTICLEThe Journal of Clinical Investigation sal immunity: significant effects of old age on secretory IgA concentrations and intraepithelial lymphocyte counts. Gut. 1992;33(7):882–886. 61. Sterlin D, et al. IgA dominates the early neutraliz- ing antibody response to SARS-CoV-2. Sci Transl Med. 2021;13(577):2223. 62. Sherina N, et al. Persistence of SARS-CoV-2 specific B- and T-cell responses in convalescent COVID-19 patients 6-8 months after the infec- tion. Med. 2021;2(3):281–295. 63. Fielding CA1, et al. ADNKA overcomes SARS-CoV2-mediated NK cell inhibition through non-spike antibodies [preprint]. https://doi. org/10.1101/2021.04.06.438630. Posted on bioRxiv April 6, 2021. 64. Salimzadeh L, et al. PD-1 blockade partially recovers dysfunctional virus-specific B cells in chronic hepatitis B infection. J Clin Invest. 2018;128(10):4573–4587. 65. Wrapp D, et al. Cryo-EM structure of the 2019- NCoV spike in the prefusion conformation. Sci- ence. 2020;367(6483):1260–1263. 66. Jahnmatz M, et al. Optimization of a human IgG B-cell ELISpot assay for the analysis of vac- cine-induced B-cell responses. J Immunol Meth- ods. 2013;391(1–2):50–59. 67. Crotty S, et al. Tracking human antigen-specific memory B cells: a sensitive and generalized ELISPOT system. J Immunol Methods. 2004;286(1–2):111–122. 1 2 J Clin Invest. 2022;132(2):e152042 https://doi.org/10.1172/JCI152042 RESEARCH ARTICLE
10.1186_s12903-020-1021-0	Zhao et al. BMC Oral Health (2020) 20:34 https://doi.org/10.1186/s12903-020-1021-0 R E S E A R C H A R T I C L E Open Access Adjunctive subgingival application of Chlorhexidine gel in nonsurgical periodontal treatment for chronic periodontitis: a systematic review and meta-analysis Han Zhao1,2†, Jingchao Hu2,3† and Li Zhao4,5,6* Abstract Background: Subgingival applications of chlorhexidine (CHX) gel are commonly used as an adjunct in nonsurgical periodontal treatment (NSPT) for chronic periodontitis (CP). However, there is lack of systematic review and meta- analysis justifying the effects of adjunctive CHX gel on clinical outcomes. The objective of this meta-analysis was to evaluate the efficacy of adjunctive subgingival administration of CHX gel in NSPT compared to NSPT alone for CP. Methods: An electronic search of four databases and a manual search of four journals were conducted up to August 2019. Only randomized controlled trials reporting on the clinical outcomes of subgingival use of CHX gel adjunct to scaling and root planing (SRP), as compared to SRP alone or with placebo, for at least 3 months were included. Primary outcomes were probing pocket depth (PPD) reduction and clinical attachment level (CAL) gain at 3 and 6 months, when data on at least three studies were obtained. Results: Seventeen studies were included for qualitative analysis and seven studies for quantitative analysis (four studies for the application of CHX gel adjunct to SRP at selected sites with at least pocket depth ≥ 4 mm and three studies for comparison of full-mouth disinfection (FMD) with subgingival use of CHX gel and full-mouth scaling and root planing (FMSRP). For subgroups, the clinical outcomes between adjunctive use of Xanthan-based CHX gel (XAN-CHX gel) and CHX gel were analyzed. Results indicated a significant improvement of PPD reduction following local adjunctive administration of XAN-CHX gel for SRP at selected sites (MD: 0.15 mm). However, no difference was found in CAL gain. Moreover, no significant difference was observed in PPD and CAL at both 3 and 6 months post- treatment between FMD and FMSRP. Conclusion: Adjunctive subgingival administration of XAN-CHX gel at individual selected sites in NSPT appears to provide slight benefits in PPD reduction compared to NSPT alone for CP. Due to the lack of high-quality studies, further studies with larger sample sizes and strict standards are needed to confirm the conclusions. Keywords: Chronic periodontitis, Chlorhexidine, Subgingival irrigation, Root planing; meta-analysis * Correspondence: lizhao2010@hospital.cqmu.edu.cn †Han Zhao and Jingchao Hu contributed equally to this work. 4Department of Prosthodontics, Stomatological Hospital of Chongqing Medical University, Chongqing 400015, China 5Chongqing Key Laboratory of Oral Diseases and Biomedical Sciences, Chongqing 400015, China Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Zhao et al. BMC Oral Health (2020) 20:34 Page 2 of 12 Background Chronic periodontitis (CP) is characterized as a complex progressive chronic inflammatory process, which leads to the destruction of periodontal supportive tissue and a further loss of teeth. CP occurs when the magnitude ef- fects of the pathogenic microbial load in the periodontal pocket are larger than that of the hosts immune re- sponse [1, 2]. The basis of periodontal treatment is elim- ination or suppression of periodontal pathogens. The golden standard of which is mechanical debridement by scaling and root planing (SRP). However, large limita- tions of physical treatment have been observed due to the difficulty of accessing deep periodontal defects, which compromises the effectiveness of biofilm removal. The persistence of periodontal pathogens, such as Aggre- gatibacteractinomycetemcomitans and Porphyromonas gingivalis (P.g), were often found following SRP and can result in microbial re-colonization and the consequent destruction of periodontal tissue [3–6]. In regards to this issue, adjunctive systemic and localized antibiotics have been applied to compensate for the limitation of mech- anical therapy. Despite the rapid development of a var- iety of adjunctive local periodontal treatments in recent years, such as metformin, antioxidants, photodynamic treatment and so on [7–9], chlorhexidine (CHX) re- mains one of the most effective local antimicrobial agents, and is widely used for the local treatment of peri- odontitis [10–13]. Through the rapid attraction of the negatively charged bacterial cell surface to the cationic CHX molecule, CHX shows strong antibacterial activity in the periodontal pocket, along with a lack of toxicity, incompliance from patients and an emergence of resist- ance microorganisms. However, the high clearance of CHX from the periodontal pocket leads to subtherapeu- tic CHX concentrations in the local environment after only a short time of subgingival CHX application [14], which results in an insufficient treatment effectiveness [1, 15]. Given this limitation, CHX Gel with CHX con- centration up to 15 times than liquid carriers was devel- oped for periodontal years, numerous of studies have reported the effectiveness of adjunctive CHX to nonsurgical periodontal treatment (NSPT). However, contrary results were presented [10– 13], there is still no consensus on this issue. So far, only one systematic review without quantitative analysis indi- cated that the positive effect of local subgingival applica- tion of CHX Gel adjunctive to NSPT could be not justified as compared to NSPT alone [16]. Therefore, there is lack of strong evidence for support the beneficial effect of subgingival use CHX as adjunct to NSPT. In recent treatment. Full-mouth disinfection (FMD) was proposed by Quir- ynen in 1995, with the aim of eradicating periodontal pathogens in a short time from all the oropharyngeal habitats (mucous membranes, tongue, tonsils and saliva) [17]. CHX gel as an adjunct was used in the FMD proto- col, which was described as full-mouth scaling and root planing (FMSRP) in 1–2 sessions within 24 h combined with full-mouth subgingival irrigation with CHX gel, as well as a tongue brush and mouthwash by means of CHX [2, 17–20]. However, whether the use of antiseptics played a role in FMD is still unclear. The aim of this systematic review and meta-analysis was to evaluate the benefits of a subgingival administra- tion of CHX gel as an adjunct to NSPT for the treatment of CP. Method Focus questions Whether subgingival chlorhexidine gel application as an adjunct to nonsurgical periodontal treatment provides additional benefit in chronic periodontitis? to clinical outcomes Search strategy The review and meta-analysis were based on the Pre- ferred Reporting Items for Meta-Analysis (PRISMA) statement [21]. Three reviewers (HZ, JCH and LZ) con- ducted an independent search of three databases, includ- ing PubMed, EMBASE and the Cochrane Collaboration Library on the 20 August 2019 for articles addressing the focused question. Furthermore, a search of the Open Grey database was performed, and a hand search was conducted of following journals: Journal of Dental Re- search, Journal of Periodontology, Journal of Clinical Periodontology and Journal of Periodontal Research from 2000 until 2019. Study selection Titles and abstracts were reviewed for eligibility by two independent reviewers (HZ, JCH) according to the inclu- sion criteria. Studies that met all inclusion criteria or met some of inclusion criteria but did not meet any of the exclusion criteria were admitted for full-text review. In this phase, full-text papers were assessed in line with the exclusion criteria. And the reasons for exclusion were recorded (Additional file 1: Table S1). Any dis- agreements were resolved on discussion between the three reviewers and a consensus was reached through voting. The agreement value between the reviewers was calculated using Kappa statistics, which is used to meas- ure inter-rater reliability. The classification of Kappa Value was suggested: ≤ 0 as indicating no agreement, 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial, and 0.81–1.00 as perfect agreement [22]. The search strategy for PubMed (adapted to the other databases) is listed below: Zhao et al. BMC Oral Health (2020) 20:34 Page 3 of 12 (periodontitis OR periodontal disease) AND ((((chlor- hexidine, OR chlorhexidine gluconate, OR xanthan OR xanthan chlorhexidine) AND gel) AND (subgingival, OR subgingival curettage, OR dental scaling, OR root plan- ing OR dental prophylaxis)) OR full mouth disinfection) Primary and secondary outcomes The primary outcomes were probing pocket depth (PPD) reduction and the clinical attachment level (CAL) gain at 3 and 6 months post-therapy. The secondary out- come was adverse events. Eligibility criteria The inclusion criteria for the studies were: 1) random- ized controlled trials (RCTs); 2) comparison of SRP alone/placebo and CHX gel adjunct to SRP; or compari- son of FMSRP alone/placebo and FMD, including sub- gingival use of CHX gel; 3) follow-up of at least 3 months; 4) reported data on clinical parameters (CAL and PPD) and 5) publication in English only. The exclusion criteria were: 1) not RCTs; 2) duplicate publications; 3) Inadequate treatment strategy: CHX was used as mono-therapy or CHX was used as adjunct to surgery or other treatment; 4) follow-up less than 3 months; 5) reported only microbiological findings with no reference to clinical result; 6) not in English. Quality assessment The methodology quality of the included articles was evaluated independently by two reviewers (HZ and JCH) based on recommendations from the CONSORT state- ment [23]. Quality assessments of the included studies were conducted using the revised risk of bias assessment tool from the Cochrane Collaboration’s handbook ver- sion 5.2.0 [24], which includes seven criteria: random se- quence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assess- ment; incomplete outcome data; selective reporting and other sources of bias. Each category was estimated on whether it could impact the overall results and was fur- ther qualified as either low, high or unclear. Overall, each article was judged as (i) low risk of bias, (ii) unclear risk of bias or (iii) high risk of bias. When a trial did not meet all four criteria for randomization and blinding methods, it was excluded from quantitative analysis, as its low quality and high bias may have subverted the val- idity of the results and conclusions. Data collection process/data items Data of each included study were recorded using a stan- including study design, dardized data extraction form, number of patients, demographics, inclusion criteria, types of CHX gel, timing and frequency of CHX gel application, number of adverse events and length of follow-up. Data synthesis The meta-analysis was performed using RevMan version 5.3 (2014). Mean differences (MD) with 95% confidence intervals (95% CI) were used for continuous data. The I2 value was used to access the statistical heterogeneity of the studies. If the heterogeneity was evaluated as I2 ≤ 50%, a fixed effects model was applied. When the het- erogeneity was assessed as I2 > 50%, a random effects model was used. The inverse-variance method per- formed, and the overall effect was defined as statistically significant if the p value < 0.05.The I2 value was classi- fied into four levels: i) no heterogeneity, between 0 and 25%; ii) low heterogeneity, 25–50%; iii) moderate hetero- geneity, 50–75% and iiii) high heterogeneity, 75–100% [25]. Results Study selection In the initial search, a total of 487 studies were identi- fied; 171 PubMed, 166 Embase and 139 from the Cochrane Library database. Three papers were found through the Open Grey search and eight papers were se- lected following a manual search. After removal of dupli- (n = 221), 266 papers were included in the cates selection phase of titles and abstracts. A total of 240 arti- cles were excluded, and 27 papers were selected for full- text reading. In this phase, 10 studies were further ex- cluded (Additional file 1: Table S1) and 17 papers were finally included in the qualitative analysis [2, 10–12, 17– 19, 26–35]., The kappa value for inter-reviewer agree- ment was 0.92 indicating high degree of inter-rater reli- ability. Figure 1 shows the study identification flowchart based PRISMA19 with the reasons for exclusion. Description of the included studies Seventeen articles met the criteria and were included for qualitative analysis. Thirteen studies reported subgingi- val application of CHX gel adjunct to SRP at selected sites with a moderate to deep probing depth (at least 4 mm in all studies) [8–10, 25–34]. Nine studies were split-mouth RCTs [10, 12, 27–29, 32–35], four were par- allel RCTs [11, 26, 30, 31] and three studies used a pla- cebo in the control [12, 31, 33]. From the 13 papers, 10 showed the clinical outcomes of adjunctive subgingival delivered Xanthan-based CHX gel (XAN-CHX gel) in SRP and SRP alone [26–35]; CHX concentration in the XAN-CHX gel was 1.5% in nine studies and 2.5% in one study [34], and another three studies reported the use of gels containing 0.5, 1 and 2% CHX without Xanthan gum [10–12]. Patient samples ranged from five to 98. One included study compared the clinical outcomes Zhao et al. BMC Oral Health (2020) 20:34 Page 4 of 12 Fig. 1 Flow chart of the study identification based PRISMA19 with the reasons for exclusion between SRP plus XAN-CHX gel and SRP alone for pa- tients with diabetes mellitus type 2 [27]. The timing and frequency of CHX Gel application varied between the trials. In all 13 studies but four [10, 12, 27, 32], the CHX gel was applied once at baseline after SRP. In the other four studies, the application of CHX gel was described as three times at baseline, 10 day and 20 day follow-ups [27], once at 1 month after treatment [31] and three times within 10 min at baseline [10, 12]. The follow-ups ranged from 1 month to 6 months after SRP. An additional arm of the four studies evaluated the results between FMD and FMSRP [2, 17–19]. All studies were RCTs, and one used a placebo gel and solution in the FMSRP group. The number of partici- pants ranged from 18 to 38. Follow-ups varied from 1 month to 12 months. One study included patients with diabetes mellitus type 2 [18]. A 1% CHX gel was used in all of the trials. The timing and frequency variations for the CHX gel ranged from once at base- line [17, 19] and three times in 10 min at baseline [18] to three times within 10 min at first session, sec- ond session of FMSRP and at 1 week of follow-up, re- spectively [2]. Table 1 shows the summary of the characteristics of the included studies. Risk of bias assessment All studies were RCTs. Seven studies did not report on their randomization and allocation methods in detail [11, 12, 27, 28, 30, 32, 34, 35], and from these, six studies also did not describe the blinding methods of participants and personnel as well as their assessment [11, 27, 28, 30, 32, 35], continued CHX rinsing stains the tooth and tongue surfaces, examiners could deduce which subjects were re- ceiving CHX though these changes, and all the four stud- ies in the analysis for comparison between FMD and FMSRP were considered at most to be single-blinded2,1719. Given examiner blinding was performed strictly in three studies, the detection bias for the three articles was quali- fied as ‘unclear’ [2, 17, 19]. Overall, for all 17studies, six were assessed to have a low risk of bias [10, 17, 18, 26, 29, 31], two were judged as an unclear risk of bias [12, 33], nine were considered to have a high risk of bias [2, 11, 19, 27, 28, 30, 32, 34, 35], and six were excluded from the quantitative analysis [11, 27, 28, 30, 32, 35]. . The sum- mary of quality assessment is showed in Table 2. Synthesis of results All 17 studies reported on clinical outcomes with the use of adjunctive CHX Gel. The clinical results of these Zhao et al. BMC Oral Health (2020) 20:34 Page 5 of 12 Table 1 Characteristics of included studies. Variables were listed in this systematic review (including:study design, patient demographics, methodology, number of adverse events and length of follow-up). Outcome difference is reported only between adjunctive CHX gel to SRP and SRP alone Administration Study Design Participants Methodology AE Follow- up (m) Inclusion criteria SD age Description of Gel CHX Gel Application N (C/ T) 68 (34/ 34) 30 (30/ 30) 30 (30/ 30) 20 (20/ 20) 40 (20/ 20) 22 (12/ 10) 46 (46/ 46) 10 (10/ 10) 98 (98/ 98) 30 (30/ 30) 5 (5/ 5) 15 (8/ 7) 10 (10/ 10) 30 (15/ 15) 37 (18/ 19) 18 (9/ 9) Application at selected sites Full-mouth application Faramarzi M et al. (2017) [26] Phogat M et al. (2014) [27] Jain M et al. (2013) [28] Chitsazi MT et al. (2013) [29] Chauhan AS et al. (2013) [30] Matesanz P et al. (2013) [31] Verma A et al. (2012) [32] Kranti K et al. (2010) [33] Paolantonio M (2009) [34] Gupta R et al. (2008) [35] lecic J et al. (2016) [10] unsal E et al. (1994) 11 Oosterwaal PJM et al. (1991) [12] Fonseca DC et al. (2015) [17] Santos VR et al. (2013) [18] P S S S P P (placebo) S S (placebo) S S S P S (placebo) P P (placebo) swierkot et al. (2009) [19] Quirynen M et al. (2006) P P at least eight teeth with PD 4-8 mm 2- at least 3 nonadjacent interproximal sites with PD 4-8 mm DM no 2 sites located on the same side PD between 5 to 7 mm no 30– 60 years 30– 50 years 30– 60 years one site per quadrant with PD ≥4 mm and BOP (+) no mean at least 8 teeth with PD 4-8 mm no 46.5 years 30– 65 years at least 16 teeth and at least 3 teeth per quadrant,4–10 pockets with PPD > 4 mm and BOP(+), or at a programmed supportive visit at least two non-adjacent interproxi- mal sites with PD5-8 mm and BOP(+) at least 4 periodontal pockets with PPD 5-8 mm at least two teeth with PD ≥5 mm and BOP (+) at least three teeth, (at least one tooth apart), with PPD 5-8 mm and BOP (+) at least two bilateral PPD ≥ 5 mm at least 3 teeth in each quadrant with 2 sites with PPD ≥ 4 mm and BOP(+) at least 4 interdental PPD 7-9 mm in single rooted teeth and BOP(+) mild to moderate chronic periodontitis, at least 18 natural teeth at least 15 teeth, 30% of the sites with concomitant PD and CAL > 4 mm no elder than 30 years 30– 65 years yes 25–65 no ears no no no no 24– 58 years 25– 75 years 21– 52 years 30– 57 years 33– 62 years yes 35– 60 years 37– 75 years 2- DM no at least 20 teeth with at least six sites PPD ≥5 mm and BOP(+) no 28 (14/ at least 18 teeth, at least 2 multi- rooted and/or 2 single-rooted teeth no 28– 63 years 30– 75 XAN-CHX 1.5% CHX gel XAN-CHX 1.5% CHX gel XAN-CHX 1.5% CHX gel XAN-CHX 1.5% CHX gel XAN-CHX 1.5% CHX gel XAN-CHX 1.5% CHX gel XAN-CHX 1.5% CHX gel XAN-CHX 1.5% CHX gel XAN-CHX 2.5% CHX gel XAN-CHX 1.5% CHX gel 0.5% CHX gel one time after 2nd SRP (baseline, 2 week after 1st SRP) one time at baseline, 10 days and 20 days / / 3,6 1,3 one time at baseline / 1.5,3,6 one time at baseline 0 1,3 one time at baseline / 1,3 one time at baseline 0 1,3,6 one time 1 month after SRP / 1,3 one time at baseline / 3,6 one time at baseline / 3,6 one time at baseline / 1,3 three times within 10 min / 1,3 1% CHX gel One time at baseline / 3 2% CHX gel 3 times within 10 / 1,3,6 min at baseline 1% CHX gel one time at baseline / 3,6 1% CHX gel 3 times within 10 min at baseline 3,6,12 T: 17 C: 12 1% CHX gel one time at baseline 0 1,2,4,8 1% CHX gel three times within 10 min at first / 2,4,8 Zhao et al. BMC Oral Health (2020) 20:34 Page 6 of 12 Table 1 Characteristics of included studies. Variables were listed in this systematic review (including:study design, patient demographics, methodology, number of adverse events and length of follow-up). Outcome difference is reported only between adjunctive CHX gel to SRP and SRP alone (Continued) Administration Study Design Participants Methodology AE Follow- up (m) N (C/ T) 14) [2] Inclusion criteria SD age Description of Gel CHX Gel Application in the first quadrant, at least 6 sites PPD 6 mm, radiographic bone loss≥25% years sessecion, second session, and 1-week follow-up Studies varied according to the design type of studies, the inclusion or exclusion of patients with systemic disease, different concentration and composition of chlorhexidine gel and different timing and frequency of CHX gel application. Adverse events and follow-up period were recorded P Intersubject parallel study, S Intrasubject split-mouth study, N Number, T Test group, C Control group, SD Systemic disease, 2-DM Diabetes mellitus type 2, XAN Xanthan gum, CHX Chlorhexidine, XAN-CHX Xanthan-based chlorhexidine, min Minutes, AE Adverse events, m Month/months Table 2 Risk of bias assessment Author (year) Faramarzi M et al. (2017) [26] Phogat M et al. (2014) [27] Jain M et al. (2013) [28] Chitsazi MT et al. (2013) [29] Chauhan AS et al. (2013) [30] Verma A et al. (2012) [32] Matesanz P et al. (2013) [31] Kranti K et al. (2010) [33] Paolantonio M. (2009) [34] Gupta R et al. (2008) [35] Fonseca DC 2015 [17] santos VR 2013 [18] Quirynen M 2006 [2] Swierkot K 2009 [19] Random sequence generation ○ ? ? ○ ? ? ○ ○ ○ ? ○ ○ ○ ○ ? unsal E 1994 [11] lecic J 2016 [10] ○ ○ Oosterwaal PJM 1991 [12] Allocation concealment Blinding of participants and personnel ○ ? ? ○ ? ? ○ ○ ? ? ○ ○ ○ ○ ? ○ ? ? X X ? X X ○ ○ X X ○ ○ x ? x ? ○ Blinding of outcome assessment ○ x x ○ x x ○ ○ ○ x ? ? ? x x ○ ○ Incomplete outcome data Selective reporting Other bias Risk of bias ○ ? ○ ○ ○ ? ○ ? ○ ? ○ ○ ○ ○ ○ ○ ? ○ ? ? ○ ? ? ○ ? ○ ? ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ ○ low high exclusion high exclusion low high exclusion high exclusion low unclear high high exclusion low low high high high exclusion low unclear ○: low risk of bias;?: unclear risk of bias; x: high risk of bias; Exclusion: when a trial did not meet all four criteria for randomization and blinding methods, it was excluded from quantitative analysis, as its low quality and high bias may have subverted the validity of the results and conclusions Zhao et al. BMC Oral Health (2020) 20:34 Page 7 of 12 studies are summarized in Additional file 2: Table S2. There was no consensus on the clinical efficacy of ad- junctive CHX gel to SRP at selected sites. A significant improvement in PPD and/or CAL was reported in a number of studies using XAN-CHX gel [27, 28, 30, 32– 35]. Whereas, several studies showed no additional bene- fit in clinical outcomes with the adjunctive use of CHX gel [10–12, 26, 29, 31]. In addition, all three studies using CHX gels that did not contain Xanthan gum re- ported no clinical benefits in the test group [10–12]. For comparing FMD and FMSRP, one study showed a sig- nificant improvement of PPD at 6 months [17]. In the other three studies, no sufficient evidence supported that FMD provided any significant improved clinical out- comes in terms of PPD and CAL [2, 18, 19]. Quantitative analysis was performed when data on at least three studies at 3 and/or 6 months follow-up (± 2 months) was obtained. Six trials were excluded because of an unreached methodological quality for the require- ment of this meta-analysis. Four trials were not included in the quantitative synthesis due to a lack of clinical out- comes in terms of PPD and CAL at follow-up [2, 12, 33, 34]. Finally, four studies were included for the quantita- tive analysis of subgingival application of CHX gel at se- in terms of PPD reduction and CAL lected sites gain10,26,29,31, three studies were included for analysis of full-mouth subgingival application of CHX gel in terms of the mean PPD and mean CAL at 3–4 and 6–8 months [17–19]. Four trials reported the adverse events after treatment [18, 19, 29, 32]. Changes in PPD and CAL at selected sites 6 months after CHX gel administration and the mean bleeding of probing (BOP) value at follow- ups after treatment were not conducted due to a lack of data available in the meta-analysis. Pooled outcomes For the adjunctive application of CHX gel to SRP com- pared to SRP alone at selected sites, the meta-analysis showed a significant improvement in PPD reduction, with a mean MD of 0.15 mm (MD: 0.15 [95% CI: 0.04– 0.25]; p = 0.005), no heterogeneity was observed among the studies (I2 = 0%) (Fig. 2a); No significant differences were found on the CAL gain between the groups (MD: 0.03 [95% CI: − 0.09–0.15]; p = 0.09) and moderate het- erogeneity was indicated (I2 = 54%) (Fig. 2b). For sub- group analysis, adjunctive XAN-CHX gel provided a significant PPD reduction, with a MD of 0.15 mm with no heterogeneity (MD: 0.15 [95% CI: 0.04–0.25]; p = 0.005, I2 = 11%) (Fig. 3a). Whereas, no additional benefit for CAL gain was showed in the XAN-CHX group with a low heterogeneity among the studies (MD: 0.05 [95% CI: − 0.05–0.15]; p = 0.33, I2 = 50%) (Fig. 3b). For full-mouth use of CHX gel, both the mean PPD and CAL showed no significant differences at 3–4 and 6–8 months. The overall effect size for PPD was − 0.18 mm at 3–4 months and − 0.12 mm at 6–8 months, and a high heterogeneity was observed among the studies [3– 4 months (MD: -0.43 [95% CI: − 0.63–0.27]; p = 0.43, I2 = 76%) (Fig. 4a), 6–8 months (− 0.12 [95% CI: − 0.58– 0.35]; p = 0.62, I2 = 78%) (Fig. 4b)]. CAL was 0.09 mm at Fig. 2 Forest plots comparing the adjunctive use of chlorhexidine (CHX) gel to scaling and root planing (SRP) and SRP alone at selected sites at 3 months: a probing pocket depth (PPD) reduction; b clinical attachment level (CAL) gain Zhao et al. BMC Oral Health (2020) 20:34 Page 8 of 12 Fig. 3 Forest plots for subgroup analysis of PPD reduction and the CAL gained between the adjunctive use of CHX gel to SRP and SRP alone at selected sites at 3 months: a PPD reduction; b CAL gain 3–4 months and 0.05 mm at 6–8 months with no hetero- geneity [3–4 months (MD: 0.09 [95% CI: − 0.27–0.46]; p = 0.61, I2 = 0%) (Fig. 5a), 6–8 months (MD: 0.05 [95% CI: − 0.29–0.39]; p = 0.78, I2 = 0%) (Fig. 5b)]. Adverse events Four studies reported adverse effects after treatment [17, 18, 28, 31]. Only one study comparing FMD and FMSRP reported that 17 subjects in the FMD and 12 in the FMSRP groups had one or two adverse events following mouth rinses, including changes in taste perception, dry mouth and staining [17]. Discussion Four trials comparing adjunctive CHX gel and SRP with SRP alone at selected sites were included for quantitative analysis. The results showed that adjunctive administra- tion of CHX gel provided a significant improvement in PPD reduction with a small overall effect size of 0.15 mm and no benefit to CAL. For subgroup analysis, ad- junctive subgingival administration of XAN-CHX gel containing 1.5% CHX provided also a slightly greater im- provement of PPD reduction of MD 0.15 mm. In qualitative analysis, CHX gel without Xanthan gum was applied as adjunct to SRP at selected sites in three studies, and showed no beneficial clinical outcomes [10– 12]. The results were consistent with various studies reporting minimal benefits in the local use of traditional CHX gel as a monotherapy [36, 37] or as an adjunct to SRP [10–13]. The effect of locally delivered antimicrobial drugs depends on its concentration and contact time in subgingival the environment local [12]. For Zhao et al. BMC Oral Health (2020) 20:34 Page 9 of 12 Fig. 4 Forest plots of the mean PPD at 3 and 6 months comparing full-mouth disinfection (FMD) and full-mouth scaling and root planing (FMSRP): a at 3–4 months, b at 6–8 months administration of drugs, the outflow of crevicular fluid may play an important role. Evidence indicated that the outflow of crevicular fluid is about 20 ml/hour, which might be the main cause of the short-term half-life of the gel within the periodontal pocket [14, 38]. Ooster- waal et al. applied fluorescein gel in four pockets of 10 patients, samples were taken from 1 of the 4 pockets at 5, 10, 20 and 40 min. The results showed that the most locally delivered gels in the pocket disappeared within 5 min after application, which might be due to the elasti- after drug tissue, city of pocket bleeding soft administration and spreading of the gel. And then the gel was washed out linearly and gradually by crevicular fluid flow and released from the adherent surface of the periodontal pocket [14]. Given the high clearance of CHX within the pockets, CHX gel seemed to not be an effective adjuvant to SRP. XAN-CHX gel has been ap- plied for local periodontal treatment within the recent 10 years, which contains a mixture of CHX digluconate and CHX dihydrochloride, incorporated in a Xanthan gum. XAN-CHX gel demonstrated a greater capacity to increase viscosity of the carrier (CHX) and maintained Fig. 5 Forest plots of the mean CAL at 3 and 6 months comparing full-mouth disinfection (FMD) and full-mouth scaling and root planing (FMSRP): a at 3–4 months, b at 6–8 months Zhao et al. BMC Oral Health (2020) 20:34 Page 10 of 12 the bacteriostatic and bactericidal concentrations for at least 2 weeks inside the periodontal pocket [35], which could further promote its pharmacotherapeutic effects. Based on this evidence, XAN-CHX gel may overcome the limitations of the previously used CHX gel. Considering the results of this meta-analysis for subgroups, XAN-CHX gel provided only a minor additional improvement with mean MD of 0.15 mm of PPD reduction, and no benefit of CAL gain. So far, no sufficient data have supported the clinical efficacy of adjunctive subgingival applications of XAN-CHX gel according to existing research. Evidence has reported that CHX has a high affinity for salivary or serum proteins and blood, which might lead to its rapid concentration decrease in the subgingival environment [39–41]. Furthermore, the behaviour of pathogenic bac- teria in periodontal pocket may also resist the effectiveness of local CHX gel. Evidence supported that P.g releases ves- icles capable of inactivating the CHX molecule, thereby protecting themselves and other bacteria from the bacteri- cidal activity [15]. These features may markedly negatively regulate the effects of subgingival administrations of a XAN-CHX gel. Microbiological outcomes of various stud- ies have confirmed the minimal efficacy of locally deliv- ered XAN-CHX gel as an adjunct to SRP, which showed minor bacterial count reductions in an adjunctive XAN- CHX gel group as compared to control [29, 31, 34, 42]. Due to a lack of microbiological data from consistent test- ing methods and standards, microbiological outcomes were not analysed in the review. In recent years, CHX gel has been commonly used for FMD protocol in the treatment of periodontal disease. Considering the clinical benefit of FMD in varying de- grees, FMD protocol has been conducted in a large number of studies [43–46]. In addition to full-mouth subgingival applications of CHX gel, tongue brush with CHX gel and mouthwash with CHX solution were also performed with the aim of maximum elimination of periodontal pathogens in the mouth. Despite these, no additional benefits for the adjunctive use of CHX in FMSRP were shown in this meta-analysis. This result is consistent with other studies and reviews, which indi- cated that the benefits of FMD probably resulted from the short-term full-mouth mechanical debridement, ra- ther than the beneficial effects of CHX [15, 47]. A high heterogeneity was detected for analysis of the mean PPD between the FMD and FMSRP groups at 3 months (I2 = 76%) and 6 months (I2 = 78%). Regarding the small num- ber of included studies and limited data available, there were variable factors impacting on the results, such as the general health of the included patients, the initial disease severity of the chronic periodontitis, the fre- quency of CHX gel application and the influence of other adjunctive means of CHX included in FMD and its period and frequency. Noticeably, side effects and adverse events related to the use of the local administration of CHX in the treat- ment of periodontitis should be taken into account and be weighed against the potential benefits. Although the local application of antiseptics or antibiotics overcomes uncertainties in the systemic use of antibiotics, adverse events, such as changes in taste perception, dry mouth, erythema, oral ulceration, gingival tingling, periodontal abscesses, root sensitivity and staining of tongue or teeth, were reported [48]. For FMD, due to long-term mouthwash using CHX solution, staining could occur in most patients [18]. This fact should remind clinicians that the balance between the small effect size of clinical benefit and high possibility of tooth staining should be taken in consideration when developing a treatment plan for periodontal patients. results with previous publications. Limitations To the best of our knowledge, this is the first systematic review focusing on the effects of adjunctive subgingival application of CHX gel to SRP. Therefore, we cannot compare our In addition, because of the lack of RCTs with high quality, only seven studies with a small number of participants were included for quantitative meta-analysis, and long- term clinical results comparing the test and control groups were not calculated. More RCTs with more par- ticipants and long-term follow-up are needed in the future. Conclusion Based on the results of this meta-analysis, adjunctive subgingival administration of XAN-CHX gel at individ- ual selected sites with PPD at least 4 mm promotes a slight additional benefit in PPD reduction. Adjunctive antiseptics of CHX gel at specific sites might be advis- able, but SRP always plays the dominate role in the treatment of chronic periodontitis. Due to a lack of high-quality studies, more RCTs with larger sample sizes and strict standards are needed to confirm these conclusions. Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s12903-020-1021-0. Additional file 1: Table S1. Reasons for exclusion of studies. Additional file 2: Table S2. Summary of clinical outcomes. Abbreviations CAL: Clinical attachment levels; CHX: Chlorhexidine; CP: Chronic periodontitis; FMD: Full-mouth disinfection; FMSRP: Full-mouth scaling and root planning; NSPT: Nonsurgical periodontal treatment; PPD: Probing pocket depth; SRP: Scaling and root planing; XAN-CHX: Xanthan-based chlorhexidine Zhao et al. BMC Oral Health (2020) 20:34 Page 11 of 12 Acknowledgements The authors would like to thank the support of China Scholarship Council and Beijing Stomatological Hospital, School of Stomatology, Capital Medical University for this study. Authors’ contributions HZ and JCH contributed to study conception and design. HZ collected the data and informations, and drafted the manuscript; JCH critically revised the manuscript; LZ performed the statistical analysis. All authors gave final approval and agree to be accountable for all aspects of the work. Funding JCH was supported by the China Scholarship Council for research abroad in University Witten/Herdecke. HZ was supported by Beijing Stomatological Hospital, School of Stomatology, Capital Medical University. Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1Multi-disciplinary Treatment Center, Beijing Stomatological Hospital, School of Stomatology, Capital Medical University, Tian Tan Xi Li Number.4, Beijing 100050, China. 2Department of Periodontology, Witten/Herdecke, University, Alfred-Herrhausen-Str. 45, 58445 Witten, Germany. 3Department of Periodontics, Beijing Stomatological Hospital, School of Stomatology, Capital Medical University, Tian Tan Xi Li Number 4, Beijing 100050, China. 4Department of Prosthodontics, Stomatological Hospital of Chongqing Medical University, Chongqing 400015, China. 5Chongqing Key Laboratory of Oral Diseases and Biomedical Sciences, Chongqing 400015, China. 6Chongqing Municipal Key Laboratory of Oral Biomedical Engineering of Higher Education, Chongqing 400015, China. Received: 28 October 2019 Accepted: 23 January 2020 References 1. Quirynen M, Teughels W, De Soete M, van Steenberghe D. Topical antiseptics and antibiotics in the initial therapy of chronic adult periodontitis: microbiological aspects. Periodontol 2000. 2002;28:72–90. Quirynen M, De Soete M, Boschmans G, Pauwels M, Coucke W, Teughels W, et al. Benefit of “one-stage full-mouth disinfection” is explained by disinfection and root planing within 24 hours: a randomized controlled trial. J Clin Periodontol. 2006;33:639–47. Renvert S, Wikström M, Dahle’n G, Slots J, Egelberg J. Effect of root debridement on the elimination of Actinobacillus actinomycetemcomitans and Bacteroides gingivalis from periodontal pockets. J Clin Periodontol. 1990; 17:345–50. Renvert S, Dahle’n G, Wikström M. The clinical and microbiological effects of non-surgical periodontal therapy in smokers and non-smokers. J Clin Periodontol. 1998;25:153–7. 2. 3. 4. 5. Mombelli A, Gmür R, Gobbi C, Lang NP. Actinobacillus actinomycetemcomitans in adult periodontitis. II. Characterization of isolated strains and effect of mechanical periodontal treatment. J Periodontol. 1994;65:827–34. 6. Mombelli A, Schmid B, Rutar A, Lang NP. Persistence patterns of Porphyromonas gingivalis, Prevotella intermedia/nigrescens, and Actinobacillus actinomyetemcomitans after mechanical therapy of periodontal disease. J Periodontol. 2000;71:14–21. Akram Z, Vohra F, Javed F. Locally delivered metformin as adjunct to scaling and root planing in the treatment of periodontal defects: a systematic review and meta-analysis. J Periodontal Res. 2018;53:941–9. 7. 8. 9. Castro MML, Duarte NN, Nascimento PC, Magno MB, Fagundes NCF, Flores- Mir C, et al. Antioxidants as adjuvants in periodontitis treatment: a systematic review and meta-analysis. Oxidative Med Cell Longev. 2019;22: 9187978 eCollection 2019. Azaripour A, Dittrich S, Van Noorden CJF, Willershausen B. Efficacy of photodynamic therapy as adjunct treatment of chronic periodontitis: a systematic review and meta-analysis. Lasers Med Sci. 2018;33:407–23. Lecic J, Cakic S, Janjic Pavlovic O, Cicmil A, Vukotic O, Petrovic V, et al. Different methods for subgingival application of chlorhexidine in the treatment of patients with chronic periodontitis. Acta Odontol Scand. 2016;74:502–7. 11. Unsal E, Akkaya M, Walsh TF. Influence of a single application of subgingival 10. chlorhexidine gel or tetra-cycline paste on the clinical parameters of adult periodontitis patients. J Clin Periodontol. 1994;21:351–5. 12. Oosterwaal PJ, Mikx FH, van’t Hof MA, Renggli HH. Comparison of the antimicrobial effect of the application of chlorhexidine gel, amine fluoride gel and stannous fluoride gel in debrided periodontal pockets. J Clin Periodontol. 1991;18:245–51. 13. Vinholis AH, Figueiredo LC, Marcantonio Júnior E, Marcantonio RA, Salvador SL, Goissis G. Subgingival utilization of a 1% chlorhexidine collagen gel for the treatment of periodontal pockets. A clinical and microbiological study. Braz Dent J. 2001;12:209–13. 14. Oosterwaal PJ, Mikx FH, Renggli HH. Clearance of a topically applied fluorescein gel from periodontal pockets. J Clin Periodontol. 1990;17:613–5. 15. Cosyn J, Sabzevar MM. A systematic review on the effects of subgingival chlorhexidine gel administration in the treatment of chronic periodontitis. J Periodontol. 2005;76:1805–13. 18. 17. 16. Quirynen M, Bollen CML, Vandekerckhove BNA, Dekeyser C, Papaioannou W, Eyssen H. Full- vs. partial-mouth disinfection in the treatment of periodontal infections: short-term clinical and microbiological observations. J Dent Res. 1995;74:1459–67. Fonseca DC, Cortelli JR, Cortelli SC, Miranda Cota LO, Machado Costa LC, Moreira Castro MV, et al. Clinical and microbiologic evaluation of scaling and root Planing per quadrant and one-stage full-mouth disinfection associated with azithromycin or Chlorhexidine: a clinical randomized controlled trial. J Periodontol. 2015;86:1340–51. Santos VR, Lima JA, Miranda TS, Gonçalves TE, Figueiredo LC, Faveri M, et al. Full-mouth disinfection as a therapeutic protocol for type-2 diabetic subjects with chronic periodontitis: twelve-month clinical outcomes: a randomized controlled clinical trial. J Clin Periodontol. 2013;40:155–62. Swierkot K, Nonnenmacher CI, Mutters R, Flores-de-Jacoby L, Mengel R. One-stage full-mouth disinfection versus quadrant and full-mouth root planing. J Clin Periodontol. 2009;36:240–9. Lang NP, Tan WC, Krähenmann MA, Zwahlen M. A systematic review of the effects of full-mouth debridement with and without antiseptics in patients with chronic periodontitis. J Clin Periodontol. 2008;35(Suppl 8):8–21. 21. Moher D, Altman DG, Liberati A, Tetzlaff J. PRISMA statement. Epidemiology. 20. 19. 2011;22:128. 22. Mchugh ML. Interrater reliability: the kappa statistic. Biochem Med (Zagreb). 2012;22(3):276–82. 23. Moher D, Schulz KF, Altman DG, CONSORT GROUP (consolidated standards of reporting trials). The CONSORT statement: Revised recommendations for improving the quality of reports of parallel-group randomized trials. Ann Intern Med. 2001;134:657–62. 24. Higgins JPT, Altman DG, JAC S. Chapter 8: Assessing risk of bias in included studies. In: JPT H, Churchill R, Chandler J, Cumpston MS, editors. Cochrane Handbook for Systematic Reviews of Interventions version 5.2.0 (updated June 2017); 2017. p. 1–73. 25. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency 26. in meta-analyses. BMJ. 2003;327:557–60. Faramarzi M, Shirmohammadi A, Chitsazi M, Shamami MS, Ghanitab S. Clinical and metabolic effects of subgingival application of xanthan-based chlorhexidine gel in type 2 diabetic patients with chronic periodontitis. Dent Res J (Isfahan). 2017;14:299–305. 27. Phogat M, Rana T, Prasad N, Baiju CS. Comparative evaluation of 28. subgingivally delivered xanthan-based chlorhexidine gel and herbal extract gel in the treatment of chronic periodontitis. J Indian Soc Periodontol. 2014; 18:172–7. Jain M, Dave D, Jain P, Manohar B, Yadav B, Shetty N. Efficacy of xanthan based chlorhexidine gel as an adjunct to scaling and root planing in treatment of the chronic periodontitis. J Indian Soc Periodontol. 2013;17: 439–43. Zhao et al. BMC Oral Health (2020) 20:34 Page 12 of 12 29. Chitsazi MT, Kashefimehr A, Pourabbas R, Shirmohammadi A, Ghasemi Barghi V, Daghigh AB. Efficacy of subgingival application of xanthan-based Chlorhexidine gel adjunctive to full-mouth root Planing assessed by real- time PCR: a microbiologic and clinical study. J Dent Res Dent Clin Dent Prospects. 2013;7:95–101. 30. Chauhan AS, Bains VK, Gupta V, Singh GP, Patil SS. Comparative analysis of hyaluronan gel and xanthan-based chlorhexidine gel, as adjunct to scaling and root planing with scaling and root planing alone in the treatment of chronic periodontitis: a preliminary study. Contemp Clin Dent. 2013;4:54–61. 31. Matesanz P, Herrera D, Echeverría A, O’Connor A, Gon-zález I, Sanz M. A randomized clinical trial on the clinical and microbiological efficacy of a xanthan gel with chlor-hexidine for subgingival use. Clin Oral Invest. 2013;17:55–66. 32. Verma A, Sanghi S, Grover D, Aggarwal S, Gupta R, Pandit N. Effect of insertion 33. of xanthan-based chlorhexidine gel in the maintenance phase following the treatment of chronic periodontitis. J Indian Soc Periodontol. 2012;16:381–5. Kranti K, Seshan H, Sameer Z. Clinical evaluation of topical subgingival application of biodegradable xanthan based 1.5% chlorhexidine gel for treatment on periodontal pockets. J Adv Dent Res. 2010;1:47–54. 34. Paolantonio M, D'Ercole S, Pilloni A, D'Archivio D, Lisanti L, Graziani F, et al. Clinical, microbiologic, and biochemical effects of subgingival administration of a xanthan-based chlorhexidine gel in the treatment of periodontitis: a randomized multicenter trial. J Periodontol. 2009;80:1479–92. 35. Gupta R, Pandit N, Aggarwal S, Verma A. Comparative evaluation of subgingivally delivered 10% doxycycline hyclate and xanthan-based chlorhexidine gels in the treatment of chronic periodontitis. J Contemp Dent Pract. 2008;9:25–32. 36. Oosterwaal PJ, Mikx FH, van’t Hof MA, Renggli HH. Short-term bactericidal activity of chlorhexidine gel, stannous fluoride gel and amine fluoride gel tested in periodontal pockets. J Clin Periodontol. 1991;18:97–100. 37. Perinetti G, Paolantonio M, Cordella C, D’Ercole S, Serra E, Piccolomini R. Clinical and microbiological effects of subgingival administration of two active gels on persistent pockets of chronic periodontitis patients. J Clin Periodontol. 2004;31:273–81. 38. Goodson J. Pharmacokinetic principles controlling efficacy of oral therapy. J Dent Res. 1989;68:1625–32. 39. Wade WG, Addy M. In vitro activity of a chlorhexidine containing mouthwash against subgingival bacteria. J Periodontol. 1989;60:521–5. 40. Rabe LK, Hillier SL. Effect of chlorhexidine on genital microflora, Neisseria 41. gonorrhoeae, and Trichomonas vaginalis in vitro. Sex Transm Dis. 2000;27:74–8. Slots J. Selection of antimicrobial agents in periodontal therapy. J Periodontal Res. 2002;37:389–98. 42. Manthena S, Ramesh A, Srikanth A, Ramoji Rao MV, Preethi PL, Samatha YP. Comparative evaluation of subgingivally delivered chlorhexidine varnish and chlorhexidine gel in reducing microbial count after mechanical periodontal therapy. J Basic Clin Pharm. 2014;6:24–8. 43. Boia S, Stratul Ş-I, Boariu M, Ursoniu S, Goţia SL, Boia ER, et al. Evaluation of antioxidant capacity and clinical assessment of patients with chronic periodontitis treated with non-surgical periodontal therapy and adjunctive systemic antibiotherapy. Romanian J Morphol Embryol. 2018;59:1107–13. 44. Almeida ML, Duarte PM, Figueira EA, et al. Effects of a full-mouth disinfection protocol on the treatment of type-2 diabetic and non-diabetic subjects with mild-to-moderate periodontitis: one-year clinical outcomes. Clin Oral Investig. 2019. https://doi.org/10.1007/s00784-019-02927-8. 45. Pockpa AD, Soueidan A, Louis P, Coulibaly NT, Badran Z, Struillou X. Twenty Years of Full-Mouth Disinfection: The Past, the Present and the Future. Open Dent J. 2018;12:435–42. Santuchi CC, Cortelli JR, Cortelli SC, Cota LO, Fonseca DC, Alencar CO, et al. Scaling and root Planing per quadrant versus one-stage full-mouth disinfection: assessment of the impact of chronic periodontitis treatment on quality of life--a clinical randomized, controlled trial. J Periodontol. 2016;87:114–23. 46. 47. Quirynen M, Mongardini C, De Soete M, Pauwels M, Coucke W, van Eldere J, et al. The role of chlorhexidine in the one-stage full-mouth disinfection treatment of patients with advanced adult periodontitis. Long-term clinical and microbiological observations. J Clin Periodontol. 2000;27:578–89. 48. Matesanz-Pérez P, García-Gargallo M, Figuero E, Bascones-Martínez A, Sanz M, Herrera D. A systematic review on the effects of local antimicrobials as adjuncts to subgingival debridement, compared with subgingival debridement alone, in the treatment of chronic periodontitis. J Clin Periodontol. 2013;40:227–41. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s13018-023-03907-1	Li et al. Journal of Orthopaedic Surgery and Research (2023) 18:424 https://doi.org/10.1186/s13018-023-03907-1 RESEARCH ARTICLE Journal of Orthopaedic Surgery and Research Open Access Study on the effect of internet plus continuous nursing on functional recovery and medication compliance of patients with knee joint replacement Yan Li1, Zongyun Gu1, Rende Ning1 and Hao Yin1* Abstract Objective To evaluate the effect of "Internet + " continuity of care on postoperative functional recovery and medica- tion compliance in patients with knee arthroplasty. Methods In this retrospective study, 100 patients who underwent knee replacement in our hospital between Janu- ary 2021 and December 2022 were recruited and assigned to receive routine care (routine group) or "Internet + " continuity of care (continuity group), with 50 patients in each group. Outcome measures included knee function, sleep quality, emotional state, medication compliance, and self-care ability. Results Patients in the continuity group showed better knee function after discharge and during follow-up versus those in the routine group (P < 0.05). Continuity care resulted in significantly lower Pittsburgh Sleep Quality Index (PSQI), Self-Rating Anxiety Scale (SAS), and Self-Rating Depression Scale (SDS) scores versus routine care (P < 0.05). Patients in the continuity group showed higher treatment compliance, ability of daily living (ADL) scores, and nursing satisfaction than those in the routine group (P < 0.05). Conclusion The "Internet + " continuity of care is highly feasible and can effectively promote the postoperative func- tional recovery of knee replacement patients, improve patients’ medication compliance, sleep quality, and self-care ability, mitigate negative emotions, and provide enhanced home care. Keywords "Internet + " continuity of care, Knee arthroplasty, Functional recovery, Medication compliance Introduction The knee is the most complex structure in the human body and a joint with high demands on motor func- tion. Knee injuries will result in severely compromised self-care ability, motor functions, and quality of life of patients [1]. Total knee arthroplasty (TKA) [2] provides Correspondence: Hao Yin yingjia59349642@163.com 1 Department of Joint Surgery, Hefei First People’s Hospital, Hefei, Anhui, China pain relief, improves knee function and quality of life, and is currently the main clinical management for knee disorders [3]. However, treatment modalities based on knee replacement alone may be insufficient to fulfill the needs of all patients. A growing body of the literature suggests that the majority of patients after knee replace- ment have distinct psychological profiles characterized by catastrophic thoughts, dysfunctional disease percep- tion, poor mental health, anxiety, and depression [4]. These factors may hinder physical recovery and lead to poor postoperative outcomes. Previous studies suggest that rehabilitation outcomes after knee arthroplasty are © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Li et al. Journal of Orthopaedic Surgery and Research (2023) 18:424 Page 2 of 8 closely associated with nursing interventions and patient self-care [5, 6]. Therefore, individualized treatment based on the biopsychosocial health model is indicated for patients to improve the continuity of postoperative care and the effectiveness of rehabilitation exercises. In recent years, with the rapid development of the mobile inter- net, the application of mobile care in healthcare allows effective real-time information exchange and knowledge transfer, which overcomes the limitations of physical space and space [7], such as online platforms that allow patients to receive health education remotely [8, 9]. The present study was conducted to evaluate the effect of "Internet + " continuity of care on postoperative func- tional recovery and medication compliance in patients with knee arthroplasty. Materials and methods Participants In this retrospective study, 100 patients who under- went knee replacement in our hospital between January 2021 and December 2022 were recruited and assigned to receive routine care (routine group) or "Internet + " continuity of care (continuity group), with 50 patients in each group. The study was approved by the ethics com- mittee of our hospital, and all patients provided under- signed informed consent. Inclusion and exclusion criteria Inclusion criteria (1) all patients were aged over 18 years with normal cog- nitive ability and understanding; (2) all patients, regard- less of gender, were scheduled for knee arthroplasty; (3) patients or family members in the two groups were able to use the Internet proficiently; (4) who had not partici- pated in similar studies. Exclusion criteria (1) with heart, liver, or kidney or other serious uncontrol- lable diseases; (2) with intolerance to surgery or contrain- dications to relevant treatment; (3) with postoperative infection or serious complications; (4) with old or patho- logical fractures; (5) with other malignancies. Treatment methods Total knee arthroplasty Preoperative preparation was routinely performed to exclude contraindications to surgery and to improve relevant preoperative examinations. With the patient in the supine position, combined spinal and epidural anes- thesia or general anesthesia was performed, followed by routine sterilization and draping. A 10–15 cm longi- tudinal incision was made in the anterior mid-knee to expose the joint by incising the medial support band and joint capsule along the medial aspect of the patella. The anterior–posterior and bicondylar lines were marked in preparation for osteotomy, the patella was turned later- ally, and the hyperplasia, ligaments, and meniscus were removed. The force lines of the lower extremity and the balance of the extended and flexed knee gaps were assessed, and trial molds of the prosthesis were fitted. A suitable prosthesis was selected and fixed with bone cement, followed by hemostasis, irrigation of the wound, suturing layer-by-layer, and placement of a drainage tube [10, 11]. Routine nursing The patients were regularly provided with health edu- cation to improve their disease awareness. The patients were assisted in various examinations, treatments, and rehabilitation exercises and were explained in detail about postoperative rehabilitation exercises. They were educated on the correct use of medications and provided with nutritional advice, instructed to observe and per- form rehabilitation exercises before discharge, and were provided with regular hospital checkups and regular vis- its to inquire about rehabilitation conditions [12]. "Internet + " continuity of care (1) Patient files were created, including general patient information, disease assessment, postoperative pathol- ogy results, key points of care and follow-up plans. (2) A public WeChat account and TKA patient WeChat group were created to share postoperative care meas- ures and knowledge. (3) A continuity of care team net- work platform was established to train nursing staff on post-knee replacement rehabilitation nursing knowledge and network platform skills through an online learning platform. (4) Internet continuity of care programs was performed. After discharge, the patients were regularly provided with pain relief methods, complication preven- tion and response methods (skin, body temperature, cir- culation), use of antithrombotic drugs), prosthetic care methods (artificial knee joint materials, duration of use, daily maintenance, prosthetic prolapse treatment meth- ods), rehabilitation training methods (walking, swim- ming, bicycling, etc.). Video assessment was conducted to timely observe and evaluate the recovery of patients, including knee joint activities, rehabilitation exercise methods, compliance with medical advice, pain, self-care, quality of life, mental and psychological disorders, and psychological status. Rehabilitation-related daily ques- tions were addressed by the responsible medical staff or physicians, and general questions were summarized and answered collectively. The patients were instructed to perform gentle knee flexion and extension, sit-to-stand transition, hip extension, walking, walking up and down Li et al. Journal of Orthopaedic Surgery and Research (2023) 18:424 Page 3 of 8 stairs and other active exercises, 10–15 min/time, 2–4 times/day. Muscle loading and weight-bearing walks were started in the 3rd month after discharge. Patients were encouraged to do housework and perform light exer- cises such as cycling and swimming. The psychological status of patients was regularly assessed, and appropri- ate psychological counseling was provided. The patients were reminded to visit the hospital regularly for review to understand their recovery. 5) The patients and their fam- ilies were encouraged to communicate with each other to alleviate negative emotions, improve the self-service ability of patients and their families, help create a positive and active health mindset, and increase the compliance and initiative of patients and their families. Outcome measures Knee function score Knee function was assessed using the Hospital for Special Surgery Knee-Rating Scale (HSS), which includes pain (30 points), function (22 points), mobility (18 points), muscle strength (10 points), flexion deformity (10 points), and stability (10 points). The scores were rated at different time points after discharge (at discharge, one week after discharge, two weeks after discharge, one month after discharge, two months after discharge, and six months after discharge), and the scores were propor- tional to the patient’s knee function, i.e., the higher the score the better the knee function. Sleep quality and emotion The Pittsburgh Sleep Quality Index (PSQI) was used to assess the patient’s sleep quality, and the Self-Rating Anx- iety Scale (SAS) and Self-Rating Depression Scale (SDS) were used to assess the patient’s anxiety and depression. PSQI scores were inversely proportional to sleep status, i.e., higher scores indicated poorer sleep quality, and SAS and SDS scores were positively proportional to psycho- logical status, i.e., higher scores indicated higher levels of anxiety and depression. Medication compliance The medication compliance questionnaire (Cronbach’s alpha coefficient 0.847, retest validity 0.858), which was self-administered by our hospital, was used for assess- ment. The total score was 100, with above 80 for com- plete compliance, 60–80 for basic compliance, and below 60 for non-compliance. Total compliance rate = (number of complete compliance cases + number of basic compli- ance cases)/total number of cases. Ability of daily living The ability of daily living (ADL) was used to assess the patient’s self-care ability. The total score was 100, and the score was proportional to the patient’s self-care abil- ity, i.e., the higher the score, the better the ability of daily living. Patient satisfaction Home-made Nursing Satisfaction Questionnaires (including attitude of nursing staff, efficiency of nurs- ing staff, and explanation of diseases by nursing staff ) (divided into highly satisfied, satisfied, less satisfied, and unsatisfied) were used to understand the satisfaction of patients in both groups. Total satisfaction = (number of highly satisfied cases + number of satisfied cases)/total number of cases. Statistical analysis GraphPad Prism 8 was used to process the images, and SPSS 26.0 software was used to organize and statistically analyze the data. The measurement data were expressed as mean ± standard deviation (mean ± s) and analyzed using the t test. The count data were expressed as rate(%) and processed using the chi-square test. P < 0.05 was used as a cut-off for statistical significance. Results Patient characteristics There were 50 patients in the routine group, including 12 males, 38 females, aged 50–83 (64.16 ± 10.02) years, and 50 patients in the continuity group, including 13 males, 37 females, aged 50–81 (66.30 ± 7.74) years. The patient characteristics between the two groups were comparable (P > 0.05). (Table 1). Knee function scores Patients in the continuity group had knee function scores at discharge (50.48 ± 3.45), one week after discharge (52.88 ± 2.89), two weeks after discharge (60.51 ± 5.15), one month after discharge (64.56 ± 4.22), two months after discharge (69.14 ± 3.41), and six months after dis- charge (74.16 ± 3.84). Patients in the continuity group had knee function scores at discharge (50.65 ± 3.17), one week after discharge (55.54 ± 3.15), two weeks after discharge (64.54 ± 2.63), one month after discharge (70.54 ± 3.15), two months after discharge (82.14 ± 3.15), and six months after discharge (89.94 ± 3.01). Patients in the continuity group showed better knee function after discharge and during follow-up versus those in the rou- tine group (P < 0.05). (Fig. 1). Quality of care and emotions There was no statistically significant difference between the two groups in the sleep quality and emotional status before the intervention (P > 0. 05). After the interven- tion, PSQI, SAS, and SDS scores in the continuity group Li et al. Journal of Orthopaedic Surgery and Research (2023) 18:424 Page 4 of 8 Table 1 Patient characteristics Routine (n = 50) Continuity (n = 50) Sex Male Female Age (years) – Mean BMI (kg/m2) – Mean Reason for surgery Knee Osteoarthritis Rheumatoid arthritis Surgical site Left knee Right knee Duration of knee lesion (yrs) – Mean Time-lapses after discharge (d) – Mean Duration of education (years) – Mean BMI: body mass index 12 (24.00) 38 (76.00) 50–83 64.16 ± 10.02 19–26 22.84 ± 0.59 24 (48.00) 26 (52.00) 25 (50.00) 25 (50.00) 1–9 4.88 ± 1.21 1–45 35.34 ± 1.65 6–21 12.27 ± 3.31 13 (26.00) 37 (74.00) 50–81 66.30 ± 7.74 19–26 22.94 ± 0.68 25 (50.00) 25 (50.00) 28 (56.00) 22 (44.00) 1–9 5.02 ± 1.03 1–45 35.08 ± 1.77 6–21 12.01 ± 3.45 t – – – 1.195 – 1.111 – – – – – 0.881 – 1.074 – 0.544 P – – – 0.235 – 0.268 – – – – – 0.379 – 0.284 – 0.587 t a e r o c s n o i t c n u f t n o i j e e n K * Regular group Continuation group e g r a h c s i d r e t f a e m i t t n e r e f f i d 100 80 60 40 * * * 2 month after discharge 1 month after discharge half a year after discharge 2 week after discharge 1 week after discharge On discharge Fig. 1 Knee function scores. Note: * indicates statistical differences (P < 0.05) when compared with the routine group Li et al. Journal of Orthopaedic Surgery and Research (2023) 18:424 Page 5 of 8 (7.01 ± 0.66, 34.16 ± 3.15, 40.14 ± 2.44) were signifi- cantly lower than those in the routine group (9.98 ± 0.75, 48.45 ± 3.14, 50.15 ± 3.14) (P < 0.05). (Table 2). Medication compliance In the routine group, there were 16 cases of complete compliance, 23 cases of basic compliance and 11 cases of non-compliance, while in the continuity group, there were 21 cases of complete compliance, 27 cases of basic compliance and 2 cases of non-compliance. Patients in the continuity group showed higher treatment compli- ance than those in the routine group (P < 0.05) (Table 3). Self‑care ability No statistically significant differences were found in the sleep quality and emotional status of the two groups before the intervention (P > 0.05). After the interven- tion, the ADL score of the continuity group (83.45 ± 7.22) was significantly higher than that of the routine group (75.15 ± 8.15) (P < 0.05) (Table 4). Patient satisfaction In the routine group, 15 patients were highly satisfied, 25 were satisfied, 6 were less satisfied, and 4 were dissatis- fied. In the continuity group, 20 patients were highly sat- isfied, 27 were satisfied, 2 were less satisfied, and 1 were dissatisfied. Total patient satisfaction was significantly Table 2 PSQI, SAS, and SDS scores Routine (n = 50) Continuity (n = 50) t P Before interven- tion PSQI SAS SDS After intervention PSQI SAS SDS 13.54 ± 0.88 57.45 ± 2.56 59.45 ± 3.14 13.49 ± 0.91 57.89 ± 2.14 59.21 ± 3.42 0.395 1.319 0.517 0.693 0.189 0.606 9.98 ± 0.75* 48.45 ± 3.14* 50.15 ± 3.14* 7.01 ± 0.66* 34.16 ± 3.15* 40.14 ± 2.44* 29.728 < 0.001 32.129 < 0.001 25.172 < 0.001 Indicates statistical differences between the pre- and post-treatment data; PSQI: Pittsburgh Sleep Quality Index; SAS: Self-Rating Anxiety Scale; SDS: Self- Rating Depression Scale Table 4 Self-care ability Routine (n = 50) Continuity (n = 50) t P 61.15 ± 10.51 Before interven- tion After intervention 75.15 ± 8.15 61.35 ± 9.94 0.138 0.890 83.45 ± 7.22 7.623 < 0.001 Indicates statistical differences between the pre- and post-treatment data higher in the continuity group (94.00%) than in the rou- tine group (80.00%) (P < 0.05) (Fig. 2). Discussion TKA provides joint repair and reconstruction and improves joint function for patients with knee joint dis- ease, and its clinical effectiveness has been well-estab- lished by the medical community. However, due to the lack of knowledge about knee replacement rehabilitation, most patients are prone to anxiety and panic after dis- charge, and the interruption of professional rehabilitation care after discharge results in difficulties in carrying out standardized and effective rehabilitation education, caus- ing poor motor performance and compromised recov- ery [13]. Therefore, scientific and effective rehabilitation care contributes to the recovery of knee function and improves the mobility of the knee joint after surgery. The results of the present study showed that patients in the continuity group showed better knee function after discharge and during follow-up versus those in the rou- tine group. Continuity care resulted in significantly lower PSQI, SAS, and SDS scores versus routine care. Patients in the continuity group showed higher treatment com- pliance, ADL scores, and nursing satisfaction than those in the routine group. These results suggested that "Inter- net + " continuity care provides bidirectional communi- cation between nursing and patients, promotes patients’ recovery, and improves the quality of life. Continued care services were provided to patients after discharge through WeChat groups or WeChat public accounts, including methods of pain relief after knee replacement, methods of complication prevention and response, meth- ods of prosthesis maintenance, and methods of reha- bilitation training to improve patients’ compliance with Table 3 Medication compliance Routine (n = 50) Continuity (n = 50) Complete compliance Basic compliance Non-compliance Total compliance rate 16 (32.00) 23 (45.00) 11 (22.00) 39 (78.00) 21 (42.00) 27 (54.00) 2 (4.00) 48 (96.00) x2 – – – P – – – 7.162 < 0.001 Li et al. Journal of Orthopaedic Surgery and Research (2023) 18:424 Page 6 of 8 * 29.00% Very satisfied 50.00% Satisfied 13.00% Not very satisfied 8.00% Dissatisfied 40.00% Very satisfied 53.00% satisfied 5.00% Not very satisfied 2.00% dissatisfied Regular group patient satisfaction Continuation group patient satisfaction Fig. 2 Patient satisfaction. Note: * indicates statistical differences between the pre- and post-treatment data rehabilitation education. Also, patients were instructed to perform rehabilitation exercises, such as knee flexion, muscle strength, flexibility, and weight-bearing capacity, to facilitate the recovery of knee function. Patients regu- larly reported feedback to medical staff on the effects of rehabilitation exercises and obtained professional and personalized advice and suggestions to improve patient compliance and satisfaction [14]. Patients were also pro- vided with comprehensive nursing advice such as nutri- tional advice and rehabilitation exercises to provide strong support for post-discharge care [15, 16]. Com- munication and encouragement enhance the patient’s initiative to promote self-rehabilitation management and recovery of joint function, thereby improving their ADL. Continuous care is an extension of inpatient care, providing uninterrupted medical services for dis- charged patients, and enables nursing staff to under- stand patient recovery and issues with patient behavior and care processes after discharge. However, in previ- ous related studies, telephone follow-up was conducted for patients with total knee arthroplasty, and the results showed that it could improve patients’ rehabilitation after discharge, but there was little improvement in postoperative rehabilitation exercises, joint rehabilita- tion activities, and living conditions. Mere reliance on telephone follow-up is clearly insufficient for exhaustive and intuitive follow-up and is of limited use for home rehabilitation. Web-based continuity of care, how- ever, allows caregivers to communicate with patients and provide advice on home rehabilitation, reduces psychological, physical, and social barriers, enhances rehabilitation outcomes, and also prevents complica- tions [17, 18]. In this study, Internet-based continuity of care was applied to the care of patients with TKA, and the intervention was extended from the hospital to the home through video assessments, demonstrations, training, and question-and-answer sessions to provide interventions for patients. comprehensive nursing The results showed that "Internet + " continuity of care could effectively improve knee function, reduce patients’ negative emotions, and significantly improve patients’ home rehabilitation, which is consistent with the results of prior studies [19, 20]. Notwithstanding the strength of "Internet + " con- tinuity of care, several feasible protocols concerning functional rehabilitation have been reported before. Stefani L et al. [21] pointed out that individually pre- scribed home-based exercise programs were cost effec- tive and safe and resulted in modest improvements in body composition, strength, and total body water distribution with little to no adverse effect on cardiac function. Paravlic AH et al. [22] reported that motor imagery (MI) practice plus physical therapy improves both objective and subjective measures of patients’ physical function after TKA and facilitates transfer of MI strength task on functional mobility. In addi- tion, supervised training and home exercises improved long-term outcome in patients with ankylosing spon- dylitis versus educational-behavioral program or no intervention [23]. Masiero S et al. revealed that follow- ing knee surgery, rehabilitation can dramatically affect the postoperative course and the final outcomes of the procedure. And the authors systematically reviewed the current literature comparing clinical outcomes of home-based and outpatient supervised rehabilitation protocols following knee surgery and concluded that supervised and home-based protocols did not show an overall significant difference in the outcomes achieved within the studies reviewed. Rompe JD et al. [24] indi- cated that both corticosteroid injection and home training were significantly less successful than shock wave therapy at 4-month follow-up. Corticosteroid injection was significantly less successful than home training or shock wave therapy at 15-month follow-up. Li et al. Journal of Orthopaedic Surgery and Research (2023) 18:424 Page 7 of 8 Speculatively, the above-mentioned programs might result in favorable outcomes combined with "Inter- net + " continuity of care in the current study. Conclusion The "Internet + " continuity of care is highly feasible and can effectively promote the postoperative functional recovery of knee replacement patients, improve patients’ medication compliance, sleep quality and self-care ability, mitigate negative emotions, and provide enhanced home care. Our study has the strength of a systematic and com- prehensive assessment of "Internet + " continuity of care in postoperative functional recovery of knee replace- ment patients; however, important limitations should be outlined. The sample studied comprised a select and relatively small number of participants, facts which affect the generalizability of the findings. Due to the design characteristics, causality was not pursued. Hence, a pro- spective and controlled trial in a larger sample is needed to strengthen our hypothesis. In addition, due to multi- ple factors, we didn’t reveal comparison of major post- OP medications like analgesics and psychologic drugs as anti-anxiety ones between both groups, which would possibly bias our results toward the null. Therefore, ongo- ing studies with comparison of major post-OP medica- tions are warranted. Prospects In recent years, with the development and progress of sci- ence and technology, the Internet is the inevitable trend of the future. The "Internet + " continuity of care breaks through the time and space limitations of traditional extended care, with the advantages of economy, practi- cality and rapidity, and facilitates patient recovery, which demonstrates great potential for clinical promotion. Acknowledgements Not applicable. Authors’ contributions Zongyun Gu and Rende Ning wrote the main manuscript text. Yan Li and Hao Yin prepared figures and tables. All authors reviewed the manuscript. All authors have read and approved the manuscript. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Availability of data and materials The datasets used and analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This study has been approved by the ethics committee of Hefei First People’s Hospital. Patients and their families were informed of the research content and voluntarily signed the informed consent. All the methods were carried out in accordance with the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Received: 9 March 2023 Accepted: 4 June 2023 References 1. Hummer E, et al. Knee joint biomechanics of patients with unilateral total knee arthroplasty during stationary cycling. J Biomech. 2021;115: 110111. 2. Canovas F, Dagneaux L. Quality of life after total knee arthroplasty. Orthop Traumatol Surg Res. 2018;104(1s):S41-s46. 3. Rossi R, et al. Total knee arthroplasty in the varus knee: tips and tricks. Int Orthop. 2019;43(1):151–8. 4. Wainwright TW, et al. Consensus statement for perioperative care in total hip replacement and total knee replacement surgery: Enhanced Recovery After Surgery (ERAS(®)) Society recommendations. Acta Orthop. 2020;91(1):3–19. 5. Russell TG, et al. Rehabilitation after total knee replacement via low- bandwidth telemedicine: the patient and therapist experience. J Telemed Telecare. 2004;10(Suppl 1):85–7. 6. Chen H, et al. Is it necessary to perform prehabilitation exercise for patients undergoing total knee arthroplasty: meta-analysis of rand- omized controlled trials. Phys Sportsmed. 2018;46(1):36–43. Shukla H, Nair SR, Thakker D. Role of telerehabilitation in patients follow- ing total knee arthroplasty: evidence from a systematic literature review and meta-analysis. J Telemed Telecare. 2017;23(2):339–46. 7. 8. Rognsvåg T, et al. Development of an internet-delivered cognitive behavioral therapy program for use in combination with exercise therapy and education by patients at increased risk of chronic pain following total knee arthroplasty. BMC Health Serv Res. 2021;21(1):1151. 9. Wang Q, et al. The effectiveness of internet-based telerehabilitation among patients after total joint arthroplasty: an integrative review. Int J Nurs Stud. 2021;115: 103845. 10. Kopp SL, et al. Anesthesia and analgesia practice pathway options for total knee arthroplasty: an evidence-based review by the American and European societies of regional anesthesia and pain medicine. Reg Anesth Pain Med. 2017;42(6):683–97. 11. Chen C, Li R. Cementless versus cemented total knee arthroplasty in young patients: a meta-analysis of randomized controlled trials. J Orthop Surg Res. 2019;14(1):262. 12. Lim JA, Thahir A. Perioperative management of elderly patients with osteoarthritis requiring total knee arthroplasty. J Perioper Pract. 2021;31(6):209–14. 13. King PJ, et al. The fate of patients not returning for follow-up five years after total knee arthroplasty. J Bone Joint Surg Am. 2004;86(5):897–901. 14. Greene KA, Harwin SF. Maximizing patient satisfaction and functional results after total knee arthroplasty. J Knee Surg. 2011;24(1):19–24. 15. Zhang X, et al. A social media-promoted educational community of joint replacement patients using the Wechat app: survey study. JMIR Mhealth Uhealth. 2021;9(3): e18763. 16. Lindberg MF, et al. The effectiveness of exercise therapy and education plus cognitive behavioral therapy, alone or in combination with total knee arthroplasty in patients with knee osteoarthritis—study protocol for the MultiKnee trial. BMC Musculoskelet Disord. 2021;22(1):1054. 17. Russell TG, et al. Internet-based outpatient telerehabilitation for patients following total knee arthroplasty: a randomized controlled trial. J Bone Joint Surg Am. 2011;93(2):113–20. 18. Zachwieja E, et al. Web-based self-directed exercise program is cost- effective compared to formal physical therapy after primary total knee arthroplasty. J Arthroplasty. 2020;35(9):2335–41. Li et al. Journal of Orthopaedic Surgery and Research (2023) 18:424 Page 8 of 8 19. Klement MR, et al. Web-based, self-directed physical therapy after total knee arthroplasty is safe and effective for most, but not all. Patients J Arthroplasty. 2019;34(7s):S178-s182. 20. Kenanidis E. Letter to the editor on “web-based, self-directed physical therapy after total knee arthroplasty is safe and effective for most, but not all, patients.” J Arthroplasty. 2019;34(8):1858–9. 21. Stefani L, Klika R, Mascherini G, Mazzoni F, Lunghi A, Petri C, Petreni P, Di Costanzo F, Maffulli N, Galanti G. Effects of a home-based exercise rehabilitation program for cancer survivors. J Sports Med Phys Fitness. 2019;59(5):846–52. https:// doi. org/ 10. 23736/ S0022- 4707. 18. 08908-9. 22. Paravlic AH, Maffulli N, Kovač S, Pisot R. Home-based motor imagery intervention improves functional performance following total knee arthroplasty in the short term: a randomized controlled trial. J Orthop Surg Res. 2020;15(1):451. https:// doi. org/ 10. 1186/ s13018- 020- 01964-4. 23. Masiero S, Poli P, Bonaldo L, Pigatto M, Ramonda R, Lubrano E, Punzi L, Maffulli N. Supervised training and home-based rehabilitation in patients with stabilized ankylosing spondylitis on TNF inhibitor treat- ment: a controlled clinical trial with a 12-month follow-up. Clin Rehabil. 2014;28(6):562–72. https:// doi. org/ 10. 1177/ 02692 15513 512214. 24. Papalia R, Vasta S, Tecame A, D’Adamio S, Maffulli N, Denaro V. Home- based vs supervised rehabilitation programs following knee surgery: a systematic review. Br Med Bull. 2013;108:55–72. https:// doi. org/ 10. 1093/ bmb/ ldt014. 25. Rompe JD, Segal NA, Cacchio A, Furia JP, Morral A, Maffulli N. Home train- ing, local corticosteroid injection, or radial shock wave therapy for greater trochanter pain syndrome. Am J Sports Med. 2009;37(10):1981–90. https:// doi. org/ 10. 1177/ 03635 46509 334374. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12889-021-11008-z	Rooij et al. BMC Public Health (2021) 21:1013 https://doi.org/10.1186/s12889-021-11008-z R E S E A R C H A R T I C L E Open Access Assessing training needs in infectious disease management at major ports, airports and ground-crossings in Europe Doret de Rooij1,2* Aura Timen1,2 and for the EU HEALTHY GATEWAYS Joint Action consortium , Evelien Belfroid1, Christos Hadjichristodoulou3, Varvara A. Mouchtouri3, Jörg Raab4, Abstract Background: The implementation of core capacities as stated in the International Health Regulations (IHR) is far from complete, and, as the COVID-19 pandemic shows, the spreading of infectious diseases through points of entry (POEs) is a serious problem. To guide training and exercises, we performed a training needs assessment on infectious disease management among professionals at European POE. Methods: We disseminated a digital questionnaire to representatives of designated airports, ports, and ground- crossings in Europe. Topics were derived from the IHR core capacities for POEs. Based on the importance (4-point Likert scale) and training needs (4-point Likert scale), we identified the topics with the highest priority for training. These results were put in further perspective using prior experience (training < 3 year, exercise < 5 years, events < 5 years). Also, preferences for training methodologies were assessed. Results: Fifty questionnaires were included in the analyses, representing 50 POEs from 19 European countries. Importance is high for 26/30 topics, although scores widely vary among respondents. Topics with a high training need (16/30) are amongst others the handling of ill travelers; using and composing the public health emergency contingency plan, and public health measures. Respondents from ports and airports attribute equal importance to most topics, but respondents from ports showed higher training needs on 75% of the topics. POEs are unevenly and generally little experienced. The most preferred training methods were presentations. Simulation is the preferred methodology for training the handling of ill or exposed travelers. Conclusions: The European workforce at designated ports, airports and ground-crossings has a different level of experience and perceives varying importance of the topics assessed in our study. We identified the topics on which training is required. We call for European collaboration between POEs to agree upon the importance of infectious disease management, and to jointly build a trained and prepared workforce that is ready to face the next crisis. Keywords: Points of entry, Port, Airport, Ground-crossing, Infectious disease control, Training, Exercise, Education, Training needs assessment * Correspondence: doret.de.rooij@rivm.nl 1Centre for Infectious Disease Control, National Institute for Public Health and the Environment, Bilthoven, The Netherlands 2Athena Institute, Free University, Amsterdam, The Netherlands Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Rooij et al. BMC Public Health (2021) 21:1013 Page 2 of 11 Background In the globalized world, infectious diseases spread easily from one country to another via travelers and goods [1–5]. The intercontinental spread of SARS (2003), A/H1N1v (2009), Ebola Virus Disease (2014–2015) and COVID-19 (2019–20) are major examples from recent history. These examples show that threats easily spread through air-, mari- time- and land travel and that countries need to be pre- pared to respond to outbreaks but also to prevent them through measures at their points of entry. To safeguard a collaborative effort in preventing the cross-border spread- ing of disease, the International Health Regulations (IHR) are being abided globally, and the Decision 1082 of the European Commission in Europe [6, 7]. According to IHR, countries are committed to designate at least one airport and one port, and may designated a ground-crossing, where the core capacities should be in place to respond effectively to infectious disease threats [6]. It is hard to say how prepared Europe’s designated ports, airports and ground-crossings – together called points of entry (POEs) – are to respond to infectious disease threats, but there are serious indications that they are not prepared enough yet. The IHR yearly self- assessment in 2019 showed that less than 60% of the core capacities have been implemented in Europe [8, 9]. And this is not even the full equation since many coun- tries do not report their preparedness status at all, and those who do, have self-assessed their status. The World Health Organization (WHO) Joint External Eval- uations (JEE) (2016–2018), which until now have been performed in only 14 of 49 countries in the WHO EURO region, show a wide variety of capacity imple- mentation among countries both in daily situations as well as for a public health emergency of international concern (PHEIC) [10]. Worryingly, other studies show a general lack of sustainable training programs and a infectious disease pre- general paredness among POE professionals [11, 12]. lack of awareness of Because of the large number of continental flights and the way travel restrictions among European countries have decreased over the past decades, European POEs are largely dealing with the same body of travelers. Their interdependency urges a joint level of preparedness at European POEs, as is currently supported by the Euro- pean Union (EU) Joint Action Healthy Gateways (2018– 2021) [13]. However, complete contact networks among POEs are lacking and a study among the European workforce as a whole has not been performed. In this way, it is unclear, from a workforce perspective, which issues have the highest priority and how training should be performed. For example, training needs could vary from performing a risk assessment, the implementation of measures, or routine inspections. And training designs interactive might differ between a series of lectures, discussion of case studies, or a full-scale simulation exercise. Therefore, in this study, we reached out to Europe’s designated ports, airports and ground-crossings and per- formed the first European wide training needs assess- ment from a POE perspective. To guide future training efforts, we aimed to identify training priorities, and cor- responding training methods. Methods This study was conducted between August 2018 and August 2019. Digital questionnaires for airports, ports and ground-crossings were developed seperately and dis- seminated to designated ports, airports and ground- crossings in Europe with the aim to assess training needs from a workforce perspective. We collected data on the importance of different topics, the training needs for these topics, and prior, experience with infectious dis- ease management (preparedness and response) at POEs. Study population We invited professionals involved in infectious disease preparedness and response at European designated POEs to complete our questionnaire. Because no validated and complete list of European designated POEs exists, we used an indirect sampling method. First, the question- naire was disseminated via a link in an email by the co- ordinator of the EU Joint Action Healthy Gateways to the national representatives of all 26 participating coun- tries in this EU Joint Action. Then, these national part- ners were asked to forward the link to a selected professional per designated POE who was involved in in- fectious disease preparedness and response and could represent this designated POE regarding their training needs. A single respondent represented a POE but was encouraged to consult colleagues during completion. The questionnaire The digital questionnaire was built in Formdesk [14]. We used the IHR Annex 1B ‘core capacity requirements for designated airports, ports and ground crossings' to extract the different issues for POEs (topic A – H) which were further divided into corresponding subtopics (A1 – H3) [6]. In line with IHR Annex 1B, we made a distinc- tion between topics that are required ‘at all times’ (topics A, B, C, D) and those that are required during a PHEIC (topics E, F, G, H). For airports and ground-crossings, 29 subtopics were extracted. The questionnaire for ports had one extra subtopic on ballast water management (subtopic C5). Terms specific for the type of POE were adjusted accordingly (e.g. aircraft; ship; vessel). This resulted in similar questionnaires for ports, airports and ground-crossings, file 1, Additional file 2, and Additional file 3 respectively. The shown in Additional as Rooij et al. BMC Public Health (2021) 21:1013 Page 3 of 11 questionnaire was pilot-tested by two professionals in- volved with hygiene and infectious disease management at POEs. Based on their feedback, we made textual changes and included the option to consult colleagues during the completion of the questionnaire. job title, The first part of the questionnaires captured demo- graphic characteristics, such as their country of origin, involvement in the designated PoE, gender, managerial and/or operational tasks, working level (local, regional, national), sector (private, public), and years of working experience in the current job. Then, respon- dents scored importance and training needs using 4- point Likert scales (3 = high, 2 = moderate, 1 = low, 0 = no, or I don’t know). Lastly, respondents rated per topic their previous education or training in the last 3 years (yes, no, I don’t know), and their training methods of ‘discussions’, preference (‘presentations’, ‘e-modules’, ‘other, ‘simulations’, namely … ’; multiple answers were allowed). Respon- dents were invited to fill out their e-mail address for follow-up questions. All other questions in the question- naire were mandatory. The questionnaires were distrib- uted between 23 October and 19 November 2018, and 7 February and 10 March 2019. Several reminders were sent. ‘case-studies’, ‘no preference’, or Additional data collection We sent a questionnaire to respondents that had pro- vided their e-mail address for follow up questions on training and practical experience in the past 5 years (Additional file 4). We asked whether and how often multi-disciplinary exercises were scheduled and how often they had managed an event at their point of entry in the last year (2018) and in the last 5 years. Data preparation Data were imported in Microsoft Excel and verified for correct storage and missing data, before analysis in IBM SPSS Statistics 24_0_0_1 (IBM Corp., Armonk, NY, USA) [15]. Questionnaires without any data other then demographic characteristics of the respondent were ex- cluded, as were double submissions (the first submis- sion was kept) or those not agreeing with the privacy statement. We calculated the response rates for coun- tries. A non-response analysis (one sample t-test) on country-level was performed using capacity scores from the IHR self-assessment of 2018 [9]. For the follow-up questionnaire, a non-response analysis was performed based on working level and average training needs (one sample t-test). We translated nominal variables into bin- ary variables for each answer option. Experience with real events was dichotomized into any experience (value = 1) vs. no experience (value = 0) in order to analyze training needs among these groups. We analysed the consistency of importance and training needs and developed constructs of scores per main topic if Cron- bach’s alpha > 0,7 [16]. Data analysis Baseline characteristics were calculated. The previous experience, importance, training needs, and preference for methodologies were analysed primarily by calculating the modus. Regarding importance and training needs, the frequencies on ‘low’ and ‘no’ were added up and in- dicated as ‘low/no’. Also means and standard deviations were calculated to enable further analyses. To assess to what extent importance and training need were valued differently, statistical differences between importance and training needs were analysed using the Wilcoxon- signed rank test. The same analyses were performed for all POEs together and for ports, airports and ground- crossings separately. Furthermore, the relations between experience from prior training and importance, and experience from prior training and training needs were assessed using a Spearman rank-order correlation. A t-test was per- formed to compare the experience with real cases (any real cases < 5 year; none) and training needs. A limit of α ≤ .05 was used for statistical significance. Results Respondent characteristics The questionnaire was completed 58 times, representing POEs in nineteen countries. Of these 58 questionnaires, eight questionnaires were excluded from the main ana- lyses because they were a double submission on the same working level (n = 4), or from a non-designated POE (n = 4). The port of Finland was non-designated, but its representatives’ results could be included after confirmation that designation is in process and the port already functions as such. This led to a response rate on country level of 73%. The non-response analysis showed that non-responding countries score slightly lower on the IHR core capacities for PoE than countries that com- pleted the questionnaire (mean 53.3% vs. 55.6% respect- performed ively; indirectly, a wide variety of professionals responded to the questionnaire. Fourteen respondents described their job as environmental or health inspector; eight as an en- vironmental, health or public health officer; six as a chief; head or director position; seven as a medical doc- tor; and several others had single job titles. Further char- acteristics of the 50 included respondents are shown in Table 1. The second questionnaire to collect additional data was filled out by thirteen of the 50 respondents (26%). A non-response analysis for this set did not show signficiant difference (responders’ Training Needs = 2.07 vs. non-responders’ Training Needs = 1.83; p = .320). sampling was p = <.001). Since Rooij et al. BMC Public Health (2021) 21:1013 Page 4 of 11 Table 1 Baseline characteristics of respondents for all POEs together and per POE type Variable Number (n) Countries Response Rate for countries Gender Male Female Working level National level Regional level Port level Sector Public sector Private sector Highest completed education Secondary education Vocational education Higher professional education University All POE (n, (%)) 50 (100) 19 (100) 19/26 (73) 26 (52) 24 (48) 20 (40) 16 (32) 14 (28) 48 (95) 2 (5) 2 (4) 1 (2) 5 (10) 42 (84) Ports (n, (%)) 27 (54.0)* 17 (89)* 17/26 (65) 14 (52) 13 (48) 7 (26) 10 (37) 10 (37) 26 (96) 1 (4) 1 (4) 1 (4) 3 (11) 22 (81) Airports (n, (%)) 18 (36)* 17 (89)* 17/26 (65) 9 (50) 9 (50) 10 (56) 4 (22) 4 (22) 17 (94) 1 (6) 0 (0) 0 (0) 2 (11) 16 (89) Ground-crossings (n, (%)) 5 (10)* 4 (21)* 4/26 (15) 3 (60) 2 (40) 3 (60) 2 (40) 0 (0) 5 (100) 0 (0) 1 (20) 0 (0) 0 (0) 4 (80) Years experience in the current job 14.8 (8.2)^ 14.9 (7.7)^ 14.5 (9.6)^ 15.6 (7.2)^ *)% of All PoE; ^Mean (Standard Deviation); POE point of entry, n number Experience Respondents have generally little experience with prior education or training, multi-disciplinary exercises and real events. Education and training received in the last 3 years ranged between 8 and 44% of respondents (Table 2). Lowest percentages are identified for H. Af- fected animals and F. Recommended measures in case of a PHEIC; highest for topic C. routine inspections and topic A. Different health risks. Experience from prior training or education did weakly correlate (coeff. <.31) with importance and training needs. Experience with real events ranged between 0 and 20 events in the last 5 years, with a median of 1. Multi- disciplinary exercises were regularly performed by five of 13 POEs, with a range between 0 and 10 and a median of 1. Real events and multidisciplinary exercises were equally divided between airports and ports; results for ground-crossings were insufficient. Respondents with experience with participating in exercises or real events showed higher training needs, although results are not significant (p = .229;.685 resp.). Importance Twenty-four of the 30 topics scored high on importance, as shown in Table 2. Among these topics, the highest mean scores were identified for the use of personal pro- tective equipment both in routine (D2) and response situations (G2), composing and updating the Public health emergency contingency (PHEC) plan (E1), hy- gienic public health measures (F1), and the handling of ill travelers (G1–3). Incongruent scores among respon- dents were identified for for ballast water management (C5) and infection control on animals (H2), with an equal frequency for high and low/no importance. Training needs Sixteen of the 30 topics were scored high on training needs. Among these, the handling of ill travelers rou- tinely (D) and during a PHEIC (G), preventive measures (F), Standard operating procedures (C1), and adequate and timely usage of the PHEC plan (E2) had the hightest mean score. Food- and water safety (B3, C3) and re- quired health documents (C4) had an equally high fre- quency for high and low/no scores. All topics with a high training need, based on the modus, are also regarded highly important, as indicated with the ‘!’ in Table 2. The scores on importance and training needs signifi- cantly differed for all topics except for six, being: chem- ical agents (A2), radiological agents (A3), ballast water management (C5), treating containers or vessels (F3), ar- rangements with local veterinary services (H1), and care or treatment of animals (H3). Of these topics, ballast water management (C5) was only assessed by ports and Rooij et al. BMC Public Health (2021) 21:1013 Page 5 of 11 Table 2 Results of importance and training needs per topic for all respondents (ports, airports and groundcrossings) Topics Training in last 3 years n (%) Importance Training needs Difference High score Modus (H; M; L/ N) mean (SD) (0 = N; 1 = L; 2 = M; 3 = H) Modus (H; M; L/ N) mean (SD) (0 = N; 1 = L; 2 = M; 3 = H) Δ (Z-score (p-value))c A – Knowledge of public health risks 20 (40) 15 (30) A1 – biological agents A2 – chemical agents A3 - radiological agents B – Safe environment B1 - Inspection programs B2 - Vector control at/near the PoE B3 - Food- and water safety B4 - Public washrooms; waste management B5 - Air quality C – Routine vessel inspections 22 (44) C1 – standard operating procedures; C2 – Sewage; solid- and medical wastes; C3 –Food- and water safety; C4 – Assessment of required health documents; C5 – Ballast water managementa. D – Ill travelers D1 – Use of protective equipment; D2 – Safe removal of travellers for assessment, care, quarantine or isolation; D3 – Triage; 16 (32) D4 – Approaching diagnostic facilities and medical services. E – PHEC plan 17 (34) E1 – Composing and updating; E2 – Adequate and timely usage; E3 – Arrangements with local medical services. F – Prevention measures 10 (20) M H M M H H H H M 2.09 (.70) (α.842)b 2.44 (.76) 2.00 (.80) 1.77 (.83) 2.16 (.72) (α.859)b M M M M L/N 1.81 (.69) (α.857)b 1.94 (.79) 1.81 (.79) 1.66 (.82) 1.75 (.86) (α.911)b 2.39 (.81) H; M 1.94 (1.00) 2.27 (.93) H; M 1.98 (.98) 2.38 (.92) 2.06 (.87) L/N 1.71 (1.01) L/N L/N L/N 1.84 (1.06) 1.54 (1.03) 1.46 (.98) H H H H H H; L/N H H H H H H H H H H 2.23 (.79) (α.926)b H; L/N 1.87 (.92) (α.971)b 2.47 (.79) H 2.04 (.99) 2.06 (.99) L/N 1.73 (.97) 2.29 (.96) H; L/N 1.88 (1.05) 2.31 (.96) 1.92 (1.10) 2.41 (.86) (α.952)b 2.50 (.85) 2.42 (.90) 2.44 (.97) 2.29 (.94) 2.49 (.72) (α.836)b 2.56 (.76) 2.48 (.79) 2.43 (.89) 2.38 (.66) (α.836)b H H H H H H H M M M M H 1.90 (1.10) 1.92 (1.02) 2.15 (.90) (α.948)b 2.24 (.95) 2.20 (.96) 2.17 (1.02) 2.00 (.97) 1.92 (.93) (α.942)b 1.94 (1.00) 2.02 (.98) 1.87 (1.00) 2.08 (.88) (α.908)b −1.885 (.059) −3.570 (<.001)a −1.153 (.249) −.590 (.555) −3.847 (<.001)a −3.598 (<.001)a −2.083 (.037)a −3.842 (<.001)a −3.701 (<.001)a −1.989 (.047)a −3.178 (.001)a −3.631 (<.001)a −2.724 (.006)a −3.021 (.003)a −2.634 (.008)a −.104 (.917) −2.374 (.018)a −2.368 (.018)a −1.969 (.049)a −2.156 (.031)a −2.442 (.015)a −4.001 (<.001)a −3.665 (<.001)a −3.535 (<.001)a −3.945 (<.001)a −2.560 (.010)a ! ! ! ! ! ! ! ! ! ! ! ! ! Rooij et al. BMC Public Health (2021) 21:1013 Page 6 of 11 Table 2 Results of importance and training needs per topic for all respondents (ports, airports and groundcrossings) (Continued) Topics Training needs Importance Difference High score Training in last 3 years n (%) Modus (H; M; L/ N) mean (SD) (0 = N; 1 = L; 2 = M; 3 = H) Modus (H; M; L/ N) mean (SD) (0 = N; 1 = L; 2 = M; 3 = H) Δ (Z-score (p-value))c F1 – disinsection, deratting, disinfection, decontamination; F2 – Treating goods, baggage, cargo or postal parcels; F3 – Treating containers or vessels. G – Ill and exposed travelers 15 (30) G1 – Use of protective equipment; G2 – Use of space for assessment, care, quarantine or isolation; G3 – Interview and triage; G4 – Safe transfer of suspected travelers H – Affected animals 4 (8) H1 – Arrangements with local veterinary services; H2 – Infection control; H3 – Care or treatment H H H H H H H H L/N L/N 2.60 (.64) 2.23 (.88) 2.31 (.75) 2.53 (.68) (α.923)b 2.67 (.66) 2.55 (.74) 2.50 (.80) H H M H H H H 2.17 (.91) 2.00 (1.03) 2.08 (.92) 2.13 (.87) (α.939)b 2.28 (.91) 2.17 (.90) 2.04 (1.01) 2.42 (.82) H/M 2.02 (.95) 1.71 (1.04) (α.950)b L/N 1.41 (1.08) (α.979)b 1.72 (1.07) H; L/N 1.82 (1.13) L/N 1.62 (1.07) L/N L/N L/N 1.45 (1.06) 1.48 (1.13) 1.31 (1.12) ! ! ! ! ! ! ! −3.377 (.001)a −2.000 (.046)a −1.932 (.053) −3.432 (.001)a −3.260 (.001)a −3.378 (.001)a −3.359 (.001)a −3.285 (.001)a −2.534 (.011)a −1.880 (.060) −2.348 (.019)a −1.949 (.051) aOnly applicable for ports; b Crohnback’s alpha; c Wilcoxon-signed rank test. H high, M moderate, L low, N no, ! a topic with high importance and high training need its analysis is based on a lower number of values. Of these topics, A2, A3, H1, H3 are among the lowest scored topics and regard other aspects than infectious disease control focused on humans. The full dataset of rough and prepared data can be found in Additional file 5. Ports, airports, and ground-crossings Generally, respondents from ports and airports reported higher importance and had higher training needs than respondents from ground-crossings. Due to the low re- sponse rate for ground-crossings, we are not able to pro- vide results for ground-crossings specifically. Specific results for airports and ports are shown in Table 3. Difference among points of entry for different subtopics are shown in Additional file 6. Respondents from ports considered six out of eight topics highly important (B, C, D, E, F, G). Ports expressed high training needs for routine vessel inspec- tions (C), the handling of ill travelers (D), prevention measures (F) and ill travelers during PHEIC situations (G). Highest discongurence among respondents for ports was identified for training needs regarding handling of affected animals (H). The discongruence is shown by the modus being low/no importance (n = 11), but also six re- spondents checked ‘I don’t know’ for this topic, and six scored it as having a high training need. Airports also considered six of the eight topics highly important. A safe environment (B) and public health risks (A) were considered moderately important. Re- spondents from airports had a high training need for the PHEC plan (E). For all others topics, a moderate training need is identified. Preference for training methodologies The training methodology of overall preference was the use of presentations (22%), followed by discussions (19%) and e-modules (19%). Simulation exercises were most preferred for training handling of ill travelers roun- tinely (21%) and during a PHEIC (20%). Respondents se- lected on average 2.56 prefered methodologies per topic. Least preference for a methodology was identified for af- fected animals (H) for which 14 respondents selected no methodology of preference. In the open answers, sugges- tions for other methodologies were made by four re- spondents. They suggested for different topics practical Rooij et al. BMC Public Health (2021) 21:1013 Page 7 of 11 Table 3 Importance and training needs for ports and airports Topic Ports (n = 27) Airports (n = 18) Importance (mean (SD) – modus (n)) Training needs (mean (SD) – modus (n)) Importance (mean (SD) – modus (n)) Training needs (mean (SD) – modus (n)) A –public health risks 2.02 (.83)a B – Safe environment 2.38 (.70) High (16) C – Routine vessel inspections 2.33 (.84) High (15) D – Ill travelers E – PHEC plan F – Prevention measures G – Ill and exposed travelers H – Affected animals 2.38 (.90) High (15) 2.60 (.67) High (18) 2.44 (.70) High (14) 2.58 (.67) High (16) 1.78 (1.02) Low/No (9) 1.78 (.77) Mod (16) 1.94 (.91) Mod (12) 2.03 (1.00) High (12) 2.19 (.95) High (14) 1.96 (1.00) Mod (10) 2.19 (.90) High (12) 2.17 (.97) High (13) 2.17 (.49) Mod (13) 1.96 (.69) Mod (9) 2.10 (.77) High (8) 2.60 (.70) High (12) 2.57 (.56) High (11) 2.41 (.57) High (8) 2.61 (.64) High (12) ! ! ! ! 1.82 (.62) Mod (13) 1.54 (.76) Mod (9) 1.66 (.77) Mod (9) 2.24 (.79) Mod (9) 1.94 (.72) High (12) 2.08 (.82) Mod (9) 2.15 (.76) Mod (9) 1.56 (1.11) Low/No (11) 1.84 (.1.05) High(6);Mod (6) 1.46 (1.07) Low/No (8) ! N number, SD standard Deviation, mod moderate; aequal scores for high, moderate and low/no;! a topic with high importance and high training need training for the use of personal protective equipment (for topic D and G), face-to-face training (for all topics), simulation MARSEC (for topic D), and on-site training (for topic A and B). Detailed results on training method- ology preference can be found in Additional file 7. Discussion In this study, we performed the first European wide train- ing needs assessment from a POE perspective. This train- ing needs assessment aimed to gain insight into the training needs on infectious disease management among dedicated staff at designated airports, ports, and ground- crossings in Europe. Handling ill travelers, public health measures at PoE, and routine inspections have the highest priority for training among ports, airports and ground- crossings together. Combining the moderate to high train- ing needs, the low percentage of respondents that received recent training, and the few real events that were experi- enced, we call for additional training efforts to enhance the workforce preparedness at European POEs. Interpretation of the results Our results are univocal regarding issues that are both important and have high training needs, such as the handling of ill travelers and several public health mea- sures. Here, according to our sample, additional training efforts should be made. However, less clear is the con- clusion for issues that are considered little or not im- portant, such as handling animals, or health risks from a chemical or radiological essence. History can mark sev- eral chemical events affecting international travel and trade that require effective response by points of entry [17, 18]. Also, several cross-border public health events can be pointed outin recent history involving zoonoses and the transport of animals [19], such as tularemia [20], bovine spongiform encephalopathy or BSE [21], or avian influenza [22]. It is therefore of no surprise that these are stated in the core capacity list in the IHR and recent landmark guidelines. Several interpretations are possible with regard to the fact that these animal and chemical threats seem to re- ceive little attention. First, they might be scored as less important relative to the other topics instead of not be- ing important at all in an absolute sense. Or second, re- these issues being spondents may be unaware of important or do not consider it a POE problem. The lit- tle experience from prior training, exercise and real cases that we identified prudently supports this second interpretation. Because if the workforce is not trained to focus on an issue, and no direct consequences follow from a lack of attention, one can concede to attribute lit- tle importance to it. Chemical events have shown to be disruptive. And as the number of zoonotic (re) emerging diseases increases [19, 23], from a one health perspective animal handling does indeed require attention in the training of POE personnel. Also, the diverging training needs among respondents needs further attention. For many topics, we saw high training needs as well as low training needs. This incon- gruence indicates varying perspectives on infectious dis- ease preparedness among POEs, which is not explained by differentiating between ports, airports and ground- crossings. Again, this is in line with the generally low and widely varying level of preparedness among POEs, as shown by the results of our study, the IHR self- assessment [9], and the Joint External Evaluations of Rooij et al. BMC Public Health (2021) 21:1013 Page 8 of 11 IHR core capacities. In the light of increasing travel with and within Europe [24] and the experiences of the current COVID-19 pandemic, it is of utmost importance that the awareness for the role POEs in cross-border disease preparedness and control, and subsequently, the development being implemented. prepared workforce of is a We identified a very low number of infectious disease events at POEs, with few exceptions reporting several events a year. This finding correlates with the results of a literature review conducted in 2013 in which less than 70 events were identified between 1990 and 2013 in the categories ‘European ports’, ‘the Mediterranean Sea’, or ‘worldwide’ [25]. The latter is named here because European crew and ships might be involved. The com- bination of the workforce’s little experience with events in practice, and the little attention of the topic in educa- tion, training and exercises in the last 5 years raises the question to what extent the training needs merely indi- cate a gut feeling, or are a reliable estimation related to real practice. Also, the low and highly varying number of events may suggest uneven chances for events to occur at different points of entry. Another possibility is that it signals an incomplete identification and notification of infectious disease at several POEs. What the right inter- pretation is needs to be studied by determining the dif- ferences in risk for events to occur at different per points of entry. Airports, ports and ground-crossings perceive slightly different needs. Airports and ports have the highest training needs in PHEIC situations. This difference can be explained by several events. First, the large Ebola out- break of 2014–2015 led to enhanced screening at ports and airports worldwide [26, 27]. In addition, in the meantime, major EU Joint Actions AIRSAN and SHIP SAN supported countries extensively with the develop- ment of effective infectious disease control at ports and airports [28, 29]. Ground-crossings, however, have re- ceived less attention since they only had a minor role in the spreading of Ebola and have not had a EU Joint Ac- tion aimed at enhancing their preparedness. However, a recent report shows a substantial and growing number of travelers that enter Europe via train and roadways and a suboptimal prepared workforce for dealing with infectious disease threats [30]. In combination with the current COVID-19 pandemic, we expect more attention for and awareness at ground-crossings on infectious dis- ease management shortly. The high preference for presentations as training methodology is hard to resolve with currently leading educational theories, such as the Adult Learning Theory. This theory promotes interactive, problem-based learn- ing in real environments to be most effective [31]. Fu- ture organizers of training programs should note this review on effective discrepancy in preferences between our respondents and leading theories, and consider consulting didactic profes- sionals during training development. Even more, because the literature on training in infectious disease control mostly leaves us here, as is shown in a recent but still unpublished literature training methods performed by this study’s authors. However, very promising tools have been developed and tested to enhance active learning and interaction during presenta- tions, such as the use of audience response systems [32] and online methodologies. (Online) E-modules are the other preferred methodology and already better suited for problem-based learning and limited interaction among learners. Since there is a need for European-wide training at POEs, this might be a very suitable method to reach this geographically spread target group. Locally at POEs, however, our respondents simulation exercises for practical skills such as the use of personal protective equipment and the handling of ill persons. The Covid-19 pandemic Between the data collection and reporting of this study, the COVID-19 pandemic has confronted many European POEs with the response to infectious disease threats on an unpre- cedented scale. News reports, first scientific publications and the authors’ experiences indicate that indeed several POEs perceived enormous challenges to handle cruise ships with cases on board [33, 34], implement public health measures at airports and on land-borders [35]. These events again emphasize, how important the capacity of personnel and organization and the necessary training at PoEs are. The re- sults of this study outline the starting point from which, in January and February 2020, Europe’s POEs started the re- quired ad hoc preparations for the COVID-19 crisis. This crisis, however, inevitable has lead to new insights on what kind of training is required for effective infectious disease management in the near and more distant future. It is too early to capture these new insights since these will keep on changing till the entire international community has recali- brated its position towards the prevention of international spread of infectious diseases in the light of a highly globalized world, and the subsequent roles for POEs. What remains, however, is the need for a well pre- pared workforce at POEs both on the individual and the collective level to face COVID-19, other conventional in- fectious diseases such as influenza, tuberculosis, and measles, as well as any new Disease X. This geographic- ally spread and divided workforce with widely varying needs will be one of the key players in restarting inter- national travel and trade again. Focus on essential roles and tasks and a collaborative policy will be of utmost importance in the coming months of European COVID- 19 recovery. Our findings draw attention to this crucial resource and provides a starting point for this collective Rooij et al. BMC Public Health (2021) 21:1013 Page 9 of 11 approach, which, however, needs to be combined with the emerging needs of our fast-changing international society. for developing training goals, Strengths and limitations This study had several challenges that might have influ- enced our results. We assessed training needs on core capacity level based on the IHR [6]. Our results point out the capacities that require further attention. How- ever, these should be translated into trainable knowledge, skills, and attitudes, or so-called competencies [36, 37]. Second, we had to apply indirect sampling methods for the questionnaire since there is currently no contact network of European PoE. In this way, we had only to some degree an idea who represented the POEs in our sample and to what extent they were representative for the POE situation. Still, the ones who selected the respondents for us were partners in a European joint action program. In this way, they are reliable in selecting representatives of points of the low response rate from ground- entry. Third, crossings also is a limitation. That is why we included their response in the general results, but were not able to specifically report on ground-crossings. Overall, a better insight into the designated POEs and the public health networks at designated POEs should remain the focus of future research to reach this group of professionals for the dissemination of new informa- tion, the invitation for education or training, and for fu- ture assessments of training needs. A more broadly distributed survey to numerous workers at each PoE would contribute to richer data than the representation of the entire POE as is the case in this study. However, this would require that personnel at POEs can be dir- ectly contacted. Last but not least, the current COVID- 19 crisis has put major focus on infectious disease man- gement at POEs. Our results are from a pre-COVID-19 status, in which the respondents had not perceived the crisis that they have now, but indicate with which train- ing needs they entered the COVID-19 crisis. Conclusions To our knowledge, this study is the first in Europe to assess the collective training needs of POEs from a POE perspective. In light of the current COVID outbreak, it is shown how important a prepared workforce at POE is. This study can be used during the development of the training agenda for training and exercises in the near future. We showed the issues requiring highest attention according to our sample from 50 different POEs, but above all, identified that preparedness at POEs requires a major place in European capacity building to be collab- oratively ready to deal with the current and the next crisis. Abbreviations EU: European Union; IHR: International Health Regulations; JEE: Joint external evaluation; PHEC: Public health emergency contingency; PHEIC: Public health emergency of international concern; POE: Point of entry; WHO: World Health Organization Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12889-021-11008-z. Additional file 1. Questionnaire Ports.pdf. The questionnaire for participants representing ports. Additional file 2. Questionaire Airports.pdf. The questionnaire for participants representing airports. Additional file 3. Questionnaire Ground-crossings.pdf. The questionnaire for participants representing ground-crossings. Additional file 4. Additional questions on experience.pdf. Additional questions disseminated in an additional data collection round. Additional file 5. Rough and prepared data.xls. In this file, the rough data file and prepared data set is presented. The file contains several sheets: Introduction: here the other sheets are introduced again. Rough File: The data as exported from Formdesk, with the headings used in Formdesk. In this sheet are highlighted: the unfinished questionnaires (orange), and questionnaires without consent for processing the data (red). Prepared: In this slide the unfinished questionnaires and those without formal consent have been excluded. Headings were added, and nominal data were splitted in colomns with binary data. Later on, the additional information from the follow-up questionnaire was added in column GL-GP. Prepared-final: Here, we excluded double submissions and non-designated points of entry. This is the database as it was uploaded into SPSS for further analysis. Additional file 6. Importance & Training needs per POE type.pdf. Results for subtopics presented for ports, airports and ground-crossings sepe- rately, including p-values as a result of a one-way anova analysis. Additional file 7. Preferred training methodologies.pdf. A table with frequencies of preferred training methodologies presented per topic (A- H), presented for all POEs together and for respondents for ports, airports and ground-crossings seperately. Acknowledgements The authors are thankful to all respondents for their participation in this study. Authors’ contributions All authors were involved in designing the study. DdR extracted data from the literature. CH and VM sent out the digital questionnaire and several reminders. DdR and EB collected and analysed the data and drafted the first manuscript together with multiple feedback round swith JR and AT. All authors critically reviewed and approved the final versions of the manuscript. All authors read and approved the final manuscript. Funding This publication has been produced with the support of the European Commission’s Consumers, Health, Agriculture and Food Executive Agency (CHAFEA) for the Healthy Gateways Joint Action (grant agreement no. 801493) and support from the Dutch Ministry of Health, Welfare and Sport. For both financial sources, this study was included among more general formulated goals. None of the sources were used or granted for specific parts of the study. The content of this publication represents the views of the authors only and is their sole responsibility; it cannot be considered to reflect the views of the European Commission and/or the Consumers, Health, Agriculture and Food Executive Agency (CHAFEA) or any other body of the European Union. The European Commission and the Agency do not accept any responsibility for use that may be made of the information it contains. Rooij et al. BMC Public Health (2021) 21:1013 Page 10 of 11 Availability of data and materials All data generated or analysed during this study are included in this published article and the Additional files. Declarations Ethics approval and consent to participate The study protocol (LCI-386) was reviewed by the Clinical Expertise Centre of the National Institute for Public Health and the Environment. Based on the study protocol, the Clinical Expertise Centre concluded that this research lays outside the scope of the Medical Research Involving Human Subjects Act (WMO). This study was conducted in the framework of the EU HEALTHY GATEWYAS Joint Action (Grant Agreement number: 801493). The contract described the methodology of the study conducted and was approved by the governmental authorities designated by the ministries of health of the countries that participated in the study and this Joint Action. All precautions were taken to well inform participants on the study’s scope and aims, and to protect their anonymity and confidentiality. All participants were offered the possibility to ask questions before participating in the questionnaire. Participants were asked to sign if they agreed and gave consent for processing the personal information, including according to the privacy and personal data protection statement. They were told their right to access, correct, delete or restrict their input. When respondents did not sign for informed consent, their personal information was not included in the questionnaire. Consent for publication All respondents included in the study approved publication of the data by signing consent to the following sentence in the privacy and personal data protection statement: “The data can be used for (scientific) publications.” The privacy statement (true/false) for all participants can be found in Additional file 5, in tab ‘Prepared’ in column E named ‘privacy statement’. Competing interests The authors declare that they have no competing interests. Author details 1Centre for Infectious Disease Control, National Institute for Public Health and the Environment, Bilthoven, The Netherlands. 2Athena Institute, Free University, Amsterdam, The Netherlands. 3Department of Hygiene and Epidemiology, University of Thessaly, Thessaly, Greece. 4Department of Organization Studies, School of Social and Behavioral Sciences, Tilburg University, Tilburg, The Netherlands. Received: 27 May 2020 Accepted: 7 May 2021 References 1. Wilder-Smith A, Gubler DJ. Geographic expansion of dengue: the impact of international travel. Med Clin North Am. 2008;92(6):1377–90. https://doi. org/10.1016/j.mcna.2008.07.002. 2. Mohr O, Askar M, Schink M, Eckmanns T, Krause G, Poggensee G. Evidence for airborne infectious disease transmission in public ground transport – a literature review. Euro Surveill. 2012;17(35):pii=20255. European Center for Disease Control. Guidelines for the surveillance of invasive mosquitoes in Europe. Stockholm: European Center for Disease Control; 2012. 3. 4. Wilson ME, Chen LH. Travellers give wings to novel coronavirus (2019- nCoV). J Travel Med. 2020;27(2): doi: https://doi.org/10.1093/jtm/taaa015. 5. Mizumoto K, Kagaya K, Zarebski A, Chowell G. Estimating the asymptomatic proportion of coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship, Yokohama, Japan, 2020. Euro Surveill. 2020; 25(10):pii=2000180. 6. World Health Organization. International health regulations (2005). World 7. Health Organization, 2016. Available from https://www.who.int/ihr/en. [cited 9 July 2018] European commission Eurostat. Decision No 1082/2013/EU of the European Parliament and of the Council of 22 October 2013 on serious cross-border threats to health and repealing Decision No 2119/98/EC (Text with EEA relevance). Official journal of the European Union 5.11.2013:L293. Available from http://data.europa.eu/eli/dec/2013/1082/2013-11-05. [cited 9 July 2018] 8. Isla N. [title unknown]. Results presented at: EU joint action general Assemblee; 2019 ;Hamburg. 9. World Healthy Organization. Electronic State Parties Self-assessment Annual Reporting Tool (e-SPAR). World Health Organization 2019. Available from https://extranet.who.int/e-spar. [cited 4 Sept 2019] 10. World Health Organization. WHO European Region: JEE mission reports. Available from https://www.who.int/ihr/procedures/mission-reports-europe/ en/. [cited 31 March 2020] 11. Rosales CB, Nuno T, Dieke A, Galvez FN, Dutton RJ, Guerrero R, et al. US- Mexico cross-border workforce training needs: survey implementation. J Inj Violence Res. 2011;3(1):1–11. https://doi.org/10.5249/jivr.v3i1.55. 14. 13. 12. Yeskey K, Hughes J, Galluzzo B, Jaitly N, Remington J, Weinstock D, et al. Ebola virus training: a needs assessment and gap analysis. Emerg Pathog Health Secur. 2017;15(3):225–9. https://doi.org/10.1089/hs.2016.0116. EU Healthy Gateways. EU Healthy Gateways Joint Action: preparedness and action at points of entry (ports, airports, ground crossings). Available from https://www.healthygateways.eu/. [cited 31 March 2020] Innovero Software Solutions BV. Formdesk [Internet]. Wassenaar: 2019. Available from: https://nl.formdesk.com. IBM Corp. Released 2016. IBM SPSS statistics for windows, version 24. Armonk: IBM Corp. Lance CE, Butts MM, Michels LC. The sources of four commonly reported cutoff criteria: what did they really say? Organ Res Methods 2006;9:202–2 doi: https://doi.org/10.1177/1094428105284919. 15. 16. 17. Dutch Safety Board. Oil Spill Port or Rotterdam. The Hague: The Dutch Safety Board. Available from: https://www.onderzoeksraad.nl/en/page/12207/ oil-spill-port-of-rotterdam. [updated 12 March 2020; cited 4 February 2021] 18. Rocha V. Port of L.A. fire spews toxic funes, prompting air-quality concerns. Los Angeles Times: California 2014. Available from: https://www.latimes. com/local/lanow/la-me-ln-port-los-angeles-fire-toxic-air-20140923-story.html. [cited 4 Feb 2021] Simons RR, Horigan V, Gale P, Kosmider RD, Breed AC, Snary EL. A generic quantitative risk assessment framework for the entry of bat-borne zoonotic viruses into the European Union. Plos One. 2016;11(10):e0165383. https:// doi.org/10.1371/journal.pone.0165383. 19. 21. 20. Avashia SB, Petersen JM, Lindley CM, Schriefer ME, Gage KL, Cetron M, et al. First reported prairie dog-to-human tularemia transmission, Texas, 2002. Emerg Infect Dis. 2004;10(3):483–6. https://doi.org/10.3201/eid1003.030695. Salman M, Silano V, Heim D, Kreysa J. Geographical BSE risk assessment and its impact on disease detection and dissemination. Prev Vet Med. 2012; 105(4):255–64. https://doi.org/10.1016/j.prevetmed.2012.01.006. 22. Radin JM, Shaffer RA, Lindsay SP, Araneta MR, Raman R, Fowler JH. International chicken trade and increased risk for introducing or reintroducing highly pathogenic avian influenza a (H5N1) to uninfected countries. Infect Dis Model. 2017;2(4):412–8. https://doi.org/10.1016/j.idm.2017.09.001. 26. 24. 25. 23. White RJ, Razgour O. Emerging zoonotic diseases originating in mammals: a systematic review of effects of anthropogenic land-use change. Mammal Rev. 2020:40(4). https://doi.org/10.1111/mam.12201. International Air Transport Association. Press Release No:62: IATA Forecast predicts 8.2 billion air travelers in 2037. Available from https://www.iata.org/ en/pressroom/pr/2018-10-24-02/. [cited 29 January 2020] EU SHIPSAN ACT Joint Action. European manual for hygiene standards and communicable disease surveillance on passesenger ships. EU SHIPSAN. 2016. Available from http://www.shipsan.eu/Portals/0/docs/EU_Manual_ Second_Edition.pdf. [cited 6 June 2018] Kesten JM, Audrey S, Holding M, Coope C, Young N, Brown CS, et al. Qualitative study of Ebola screening at ports of entry to the UK. BMJ Glob Health. 2018;3(3):e000788. https://doi.org/10.1136/bmjgh-2018-000788. 27. Cohen NJ, Brown CM, Alvarado-Ramy F, Bair-Brake H, Benenson GA, Chen TH, et al. Travel and border health measures to prevent the international spread of Ebola. MMWR Suppl. 2016;65(3):57–67. https://doi.org/10.15585/ mmwr.su6503a9. EU AIRSAN. AIRSAN: Coordinated action in the aviation sector to control public health threats. Available from http://www.airsan.eu/. [cited 23 January 2020] EU SHIPSAN. SHIPSAN ACT: the impact on maritime transport of health threats due to biological, chemical and radiological agents, including communicable diseases. Available from http://www.shipsan.eu/. [cited 23 January 2020] EU Healthy Gateways Joint Actions. State of the art Report: Ground- crossings of the EU Healthy Gateways Joint Actions. Version 2. 2020. [cited on 25 January 2021]. 28. 30. 29. Rooij et al. BMC Public Health (2021) 21:1013 Page 11 of 11 31. Parmelee D, Michaelsen LK, Cook S, Hudes PD. Team-based learning: a practical guide: AMEE guide no. 65. Med Teach. 2012;34(5):e275–87. https:// doi.org/10.3109/0142159X.2012.651179. 32. Waltz EC, Maniccia DM, Bryde RL, Murphy K, Harris BR, Waldenmaier MN. Training the public health workforce from Albany to Zambia: technology lessons learned along the way. Public Health Rep. 2010;125(5):61–9. https:// doi.org/10.1177/00333549101250S509. 33. Nakazawa E, Ino H, Akabayashi A. Chronology of COVID-19 cases on the diamond princess cruise ship and ethical considerations: a report from Japan. Disaster Med Public Health Prep. 2020;(4):1–27. https://doi.org/10.101 7/dmp.2020.50. Fernandes EG, Santos JD, Sato HK. Outbreak investigation in cargo ship in times of COVID-19 crisis, Port of Santos, Brazil. Rev Saude Publica. 2020;54: 34. https://doi.org/10.11606/s1518-8787.2020054002461 eCollection 2020. Shengen Vis Info. 2020. Available from https://www.schengenvisainfo.com/ news/21-26-schengen-countries-have-already-closed-borders-while-ec-still- mulls-schengen-suspension-idea/. [cited 13 April 2020] Stoto MA, Nelson C, Savoia E, Ljungqvist I, Ciotti M. A public health preparedness logic model: assessing preparedness for cross-border threats in the European region. Health Secur. 2017;15(5):473–482. Foldspang A, Birt CA, Otok R. ASPHER’s European list of Core competences for the public health professional. Assoc Sch Public Health Eur Reg. 2018;46: 1–52. 34. 35. 36. 37. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12879-023-08392-9	Keenan et al. BMC Infectious Diseases (2023) 23:414 https://doi.org/10.1186/s12879-023-08392-9 BMC Infectious Diseases RESEARCH Open Access Unravelling patient pathways in the context of antibacterial resistance in East Africa Katherine Keenan1, Kathryn J. Fredricks1, Mary Abed Al Ahad1, Stella Neema2, Joseph R. Mwanga3, Mike Kesby1, Martha F. Mushi3, Annette Aduda4, Dominique L. Green1, Andy G. Lynch1, Sarah I. Huque1, Blandina T. Mmbaga5, Hannah Worthington1, Catherine Kansiime2, Emmanuel Olamijuwon1, Nyanda E. Ntinginya6, Olga Loza1, Joel Bazira7, Antonio Maldonado‑Barragán1, VAnne Smith1, Arun Gonzales Decano1, John Mwaniki Njeru4, Alison Sandeman1, John Stelling8, Alison Elliott9,10, David Aanensen11, Stephen H. Gillespie1, Gibson Kibiki12, Wilber Sabiiti1, Derek J. Sloan1, Benon B. Asiimwe2, John Kiiru4, Stephen E. Mshana3, Matthew T. G. Holden1 and HATUA Consortium Abstract Background A key factor driving the development and maintenance of antibacterial resistance (ABR) is individuals’ use of antibiotics (ABs) to treat illness. To better understand motivations and context for antibiotic use we use the concept of a patient treatment‑seeking pathway: a treatment journey encompassing where patients go when they are unwell, what motivates their choices, and how they obtain antibiotics. This paper investigates patterns and deter‑ minants of patient treatment‑seeking pathways, and how they intersect with AB use in East Africa, a region where ABR‑attributable deaths are exceptionally high. Methods The Holistic Approach to Unravelling Antibacterial Resistance (HATUA) Consortium collected quantita‑ tive data from 6,827 adult outpatients presenting with urinary tract infection (UTI) symptoms in Kenya, Tanzania, and Uganda between February 2019‑ September 2020, and conducted qualitative in‑depth patient interviews with a subset (n = 116). We described patterns of treatment‑seeking visually using Sankey plots and explored explanations and motivations using mixed‑methods. Using Bayesian hierarchical regression modelling, we investigated the associa‑ tions between socio‑demographic, economic, healthcare, and attitudinal factors and three factors related to ABR: self‑treatment as a first step, having a multi‑step treatment pathway, and consuming ABs. Results Although most patients (86%) sought help from medical facilities in the first instance, many (56%) described multi‑step, repetitive treatment‑seeking pathways, which further increased the likelihood of consuming ABs. Higher socio‑economic status patients were more likely to consume ABs and have multi‑step pathways. Reasons for choos‑ ing providers (e.g., cost, location, time) were conditioned by wider structural factors such as hybrid healthcare systems and AB availability. Conclusion There is likely to be a reinforcing cycle between complex, repetitive treatment pathways, AB consump‑ tion and ABR. A focus on individual antibiotic use as the key intervention point in this cycle ignores the contextual challenges patients face when treatment seeking, which include inadequate access to diagnostics, perceived inef‑ ficient public healthcare and ease of purchasing antibiotics without prescription. Pluralistic healthcare landscapes Correspondence: Katherine Keenan katherine.keenan@st‑andrews.ac.uk Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 2 of 12 may promote more complex treatment seeking and therefore inappropriate AB use. We recommend further attention to healthcare system factors, focussing on medical facilities (e.g., accessible diagnostics, patient‑doctor interactions, information flows), and community AB access points (e.g., drug sellers). Keywords Antibacterial Resistance, Africa, Antibiotics, Treatment seeking, Healthcare system, Urinary tract Infection, Mixed methods, Patient pathways Background Antibacterial resistance (ABR) is a significant global health threat which compromises the treatment of infec- tions with antibiotics (ABs). ABR was associated with an estimated 4·95 million deaths globally in 2019 [1]. Such deaths are highest in the Sub-Saharan African region, [1] where there is high burden of infectious diseases and fewer resources to tackle ABR. The process of ABR is influenced by the way individuals use ABs, which is, in turn, impacted by social, political, and economic sys- tems operating on a variety of scales [2–5]. It is there- fore crucial to understand treatment-seeking behaviours, or ‘patient pathways’ people take when they are unwell, how these relate to AB use, and the potential presence of ABR [6–8]. This study takes as its central focus individual patient pathways, as told by patients in Kenya, Tanzania, and Uganda, and interrogates these behaviours within their social, economic, and political contexts. Self-treatment with ABs is one example of ‘inappropri- ate’ individual treatment-seeking behaviour that is pos- ited to contribute to ABR, [9] but which is influenced by a plethora of social and structural factors. A recent mixed-methods paper from six low- and middle-income countries (LMICs) showed that AB self-treatment was common: reported by 55% of respondents surveyed in Vietnam, 46% in Bangladesh, and 36% in Ghana [10]. Propensity for AB self-medication, while influenced by individual characteristics such as socioeconomic sta- tus is affected by structural determinants, including AB dispensing regulations, availability of ABs through alter- native providers (e.g., drug sellers), and public trust in different types of healthcare providers [11]. In LMICs, some patients choose to self-treat with ABs rather than access them via prescription because of reduced access and perceived deficiencies in the formal healthcare sys- tem, such as long queues, short consultation times, and high out-of-pocket expenses [12, 13]. Other influencing factors include recurrent episodes of infections, cultural beliefs and practices, and symptoms stigma [11]. The structure of healthcare systems is vital for under- standing patient pathways. Healthcare and treatment landscapes in LMICs are sometimes described as ‘plu- ralistic’ [14], which typically means there are multiple sources of public and private clinics, alongside pharma- cists, drug sellers, and complementary and traditional/ herbal medicine providers. The East African countries studied here are no exception, but with some notable differences between them. Kenya has higher per capita health care expenditure than Tanzania or Uganda, and while out-of-pocket costs are reducing year on year, they still make up a substantial share of spending in all three countries [15]. Health insurance coverage is far from universal [16], but is highest in Kenya, lower in Tanza- nia [17], and lowest in Uganda, where development of a national health insurance scheme is ongoing [18]. Finally, while drug sellers and pharmacies play a vital role in pro- viding access to medication in all three countries [19], access to ABs without prescription through these outlets is relatively common [10, 18, 20–22]. In this paper, we focus on urinary tract infections (UTIs) as a lens for understanding treatment-seeking and ABR more generally. UTIs are bacterial illnesses that are usually treated with ABs, [23] but which can be stig- matizing and reduce quality of life [24]. Globally, over 150 million individuals are diagnosed with UTIs each year, most of whom are women [25]. Empiric use of ABs to treat UTIs contributes to ABR of the uropathogens responsible (typically Escherichia coli and other Entero- bacteriacae), and presents a growing global challenge [26]. In East Africa, high community prevalence of UTIs, combined with high levels of AB self-medication, may further exacerbate ABR, [23] particularly considering self-management for UTI symptoms is extremely com- mon [27, 28]. Research questions The study aims to assess the socioeconomic, attitudinal, and contextual factors associated with patients’ treat- ment-seeking pathways for UTI-like symptoms in Kenya, Tanzania and Uganda and explore how key pathway points intersect with AB use. We focus on three aspects theoretically related to the ABR: patients self-treating rather than seeking help at a medical facility, having multi-step pathways (e.g., multiple treatment attempts), and AB consumption during the pathway. We use mixed- methods data to explore the lived experience of treat- ment-seeking and shed light on situational barriers and facilitators of such behaviours. Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 3 of 12 Fig. 1 Flow of questions used to collect quantitative data on the patient treatment‑seeking pathway for UTI‑like symptoms Methods Context and theoretical approach This study is part of a multi-country interdisciplinary consortium “Holistic Approach to Unravel Antibacterial Resistance in East Africa (HATUA)” [29]. We conceptu- alise ABR as an assemblage of interconnected, multi-sca- lar social, political, and biological influences (see Fig. 1 in [29]). At the heart of this complexity is the ‘patient path- way’, nested in a pluralistic healthcare landscape com- prising various formal and informal healthcare providers [30]. Pathway analysis has been used to investigate care for tuberculosis, [6] abortion, [31] and cancer [32] but has rarely been used to understand AB use [8]. Incorporating medical syncretism, the patient pathway may contain any number of steps, and include delays in seeking treatment, different choice of provider, self-treatment and self-med- ication, and differences in regimen adherence. Rather than a linear sequence of actions, [7] the patient pathway can be iterative and repetitive. To unravel this complex- ity, we analyse qualitative and quantitative data about AB use pathways in parallel. We avoid the assumption that human behaviour is entirely driven by the individual decisions [33] by evaluating patient pathways within their social, economic, and political contexts. Sample The sample consists of 6,827 adult outpatients, aged 18 and over (or pregnant and < 18) who were recruited from healthcare facilities in Kenya, Tanzania, and Uganda, within three sites per country (Kenya: Nairobi, Nany- uki, and Makueni; Tanzania: Mwanza, Kilimanjaro, and Mbeya; Uganda: Mbarara, Nakapiripirit, and Nakason- gola), between February 2019- September 2020. Full study details, including sample size considerations, inclu- sion and exclusion criteria are published elsewhere [29]. We recruited patients at primary and secondary facili- ties (levels 2–5) which were predominantly government- funded (Table S1). Clinicians identified patients with symptomatic and probable UTI for inclusion. In all sites, less than 1% of those approached declined to partici- pate. Patients provided a mid-stream urine sample and answered a questionnaire with trained fieldworkers on treatment-seeking for UTI symptoms, AB use practices and attitudes, and sociodemographic characteristics. We excluded 219 patients who came to the recruitment clinic for non-UTI symptoms and had not attempted to treat their symptoms, leaving 6,608 patients for analysis: 3,190 (48·3%) from Tanzania, 1,757 (26·6%) from Uganda, and 1,661 (25·1%) from Kenya. The patient urine samples underwent microbiological culture, and UTI (defined by the presence of > 104 colony-forming units per millilitre (CFU/mL) of one or two uropathogens) was present in 2,264 (24%) of patients. In-depth interviews (IDIs) were conducted 1–2 weeks after the clinic visit with a purposively selected subset of patients (n = 116) who had microbiologically confirmed UTI, reported longer treatment pathways or were diag- nosed with a multi-drug resistant UTI pathogen. IDIs Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 4 of 12 were conducted in person, at the respondents’ homes and in their primary language, using standardised topic guides, which were subsequently translated into Eng- lish by fieldworkers. IDIs focused on mapping individual pathways based on the patient’s account of recent treat- ment-seeking action, AB use, knowledge and attitudes, and motivations for behaviours. Patient qualitative and quantitative data were linked using numeric identifiers, rather than personal details, to allow us cross-reference between types of data to understand biomedical, social, economic, and attitudinal characteristics of patients. Par- ticipants gave written informed consent. Patients were not involved in the design, or conduct, or reporting, or dissemination plans of our research. Ethical approval was obtained from National and Institutional Research Ethics Committees (see protocol) [29]. primary, secondary, and tertiary), employment status (formal employment, informal employment, homemaker, not working), self-reported difficulty in meeting health- care costs (easy, difficult, very difficult), and a within- country asset index grouped into quintiles (Table S2 for details). Healthcare factors included the level of medical facility the patient was recruited from (lower-level com- munity health centres (levels 2–3) vs. high-level clinics or referral hospitals (levels 4 +), and whether they had previous experience of UTI (did not have UTI before, had UTI before, or did not know what UTI was). We include a binary variable indicated whether the patient had any type of health insurance. We also included indi- cators which measured whether patients felt that UTI symptoms were stigmatised (yes/no), which may impact treatment-seeking. Study variables Treatment seeking behaviours Figure 1 illustrates the structured questionnaire used to collect patient pathway information, which identified the types of providers consulted, treatments taken, and rea- sons for these choices. We study three binary outcomes with a theoretical or empirical link to ABR risk identified in previous studies: [11, 34, 35] 1. Self-treatment as a first step (defined as going to a drug shop/pharmacy, seeking advice from friends or family, or using traditional or home remedies) vs seeking help at a medical facility; 2. Multi-step pathway: Having two or more steps in the pathway prior to coming to the recruitment centre vs. having fewer steps; 3. Taking ABs from any source to treat their symptoms vs not taking any ABs prior to coming to the recruit- ment clinic. The last variable was derived from patients’ self-reports of the names of medicines taken during the pathway. During the interview, respondents were prompted using a drug bag or drug card developed specifically for each site [36]. Statistical methods We used Sankey plots to visualise quantitative data on patient pathways, showing counts and percentages of types of providers chosen and type of treatment obtained (if any) at each step. We excluded patients without valid data on the steps considered (n = 230), leaving a sample of n = 6,378. Subsequently, we used Bayesian hierarchical logistic regression models to assess socioeconomic and attitudinal factors associ- ated with three binary outcomes outlined above (full model specification in Supplementary Sect. 3). The models were estimated in R using the Nimble package [37]. Approximately 8% of our sample had missing val- ues on the outcomes or covariates, which we account for within a Bayesian modelling framework. Regres- sion models had four levels: patients were nested in 25 clinics, clinics in nine sites, and sites within three countries. Results were reported in terms of odds ratios (ORs) and 95% highest posterior density intervals (HPDI) due to the skewed posterior distribution of all independent variables. We conducted a sensitivity anal- ysis, with the same models restricted to the patients with microbiologically confirmed UTI (reported in Supplementary material S13). Other variables Qualitative data analysis We included self-reported variables that might impact treatment-seeking patterns [11]. Sociodemographic fac- tors included gender (male/female), age (categorised into < 25; 25–34; 35–44; 45–54; 55–64; 65 + years), mari- tal status (married, never married, and other, which included cohabiting, widowed, divorced), and household size (1–2 people, 3–6 people, 7 + people). Socioeconomic status was measured by education level (no formal, interview Translated English-language transcripts were linked to quantitative data using patients IDs, and analysed using NVivo software [38]. We used iterative thematic content analysis, beginning with first-round coding based on interview questions, such as how patients sought treatment and obstacles to treatment- seeking. Subsequent rounds of in-depth coding were undertaken to identify differences and similarities in Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 5 of 12 treatment-seeking pathways between patients, as well as potential contributing factors to decision-making around treatment seeking. Results Sample characteristics Most patients (79%) were females of reproductive age, and the majority were married (Table S3). In Uganda and Tanzania, most had little or no formal education, whereas in Kenya 86% had secondary or higher educa- tion. In the pooled data, most were either in informal employment or homemakers (41% and 25% respectively), and 60% lived in households of 3–6 people. Ugandan respondents were least likely to report it was ‘easy’ to meet healthcare costs, compared to Tanzania and Kenya. Around half of respondents from Tanzania and Kenya reported they had been previously diagnosed with UTI, compared with one quarter in Uganda. Possibly related is that symptoms stigma was higher among Ugandan participants (39%) compared to those in Tanzania and Kenya (16% and 23%, respectively). Patients in Tanzania and Uganda were mostly recruited at lower-level facili- ties (level 2–3), whereas in Kenya most were recruited at higher-level facilities or referral hospitals (levels 4 +). Sociodemographic characteristics for the qualitative sample reflected those in the quantitative (Table S4). Overview of patient pathways Figure 2 illustrates the choices of provider and whether ABs were obtained for each pathway step pooling data across 3 countries; analogous country-specific plots are shown in Supplementary Figure S7. The path- ways as shown are composed of a maximum of three steps. At step 1 or 2, patients are classified as going to either a government clinic, private clinic, drug shop, self-treating, or going directly to a recruitment clinic. At the third step, participants reported either trying a third option or attending the recruitment clinic. Across all countries, most patients (86%) went to medical facilities as their first step in treating UTI-like symp- toms (Fig. 2). The largest group of patients (45%) went directly to a HATUA recruitment clinic, 26% visited another government-funded facility, 15% visited a pri- vate facility, 7% visited drug shops/pharmacies and 7% self-treated either with home remedies, herbs, or drugs provided by friends/relatives. Kenyan patients had the simplest pathways and were most likely to go to a recruitment centre as their first step. Visiting private clinics as a first step was more common in Tanzania and Uganda than in Kenya, whereas visits to drug shops and pharmacies were more common in Kenya than the other countries. Fig. 2 Sankey plot describing patient treatment seeking pathways for UTI‑like symptoms for pooled analysis (N = 6,378). Figure notes: n = 230 were excluded from the analysis sample due to missing data on the relevant variables. Percentages might not add up to 100% due to rounding Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 6 of 12 Fig. 3 Descriptive statistics on reasons for patient’s first choice of treatment (n = 3,546). Notes: Includes patients who tried to treat their UTI symptoms before going to the recruitment clinic (n = 3,546); patients could choose multiple reasons Reasons for first choice of treatment The reported reasons for first choice of treatment are shown in Fig. 3 (country-specific results in Fig. S8). Those who chose a government or private facility were most often motivated by location, but cost was also cited as a reason for choosing government clinics, drug sellers, and self-treatment (less so private facilities). Time was a com- mon reason for choosing drug shops or self-treatment. Qualitative data shows that cost was often related to both location (due to transportation expenses) and time (due to loss of work), which can influence both the choice of medical facility and the decision to self-treat at step one: “[I live] far from the hospital or health centre, if I feel pain the first thing is to go [to the] pharmacy, there they will assist me with a drug that will ease pain at that moment. But if pain persists, I will have to go to the hospital for further treatment and advice from the doctor.” (male patient, Mbeya, Tanzania) ‘Trust’ was the most cited reason among those using self-treatments. Qualitative accounts indicated this can be related to trusting one’s own knowledge to treat illness: “Normally, I first observe the condition as I take the concoction for ginger, garlic, lemon and honey. That mixture heals everything, I don’t know why it does not heal HIV/AIDS.” (male patient, Nanyuki, Kenya) In some cases, ‘trust’ in formal medical facilities is eroded when patients’ expectations are not met, and symptoms remain unresolved, prompting patients to favour other treatment sources. Qualitative data also highlighted that drug shops were seen as convenient alternatives to medical facilities, and that drug sellers were sometimes viewed as part of the trusted cadre of healthcare professionals. For example, one patient explained: “I trust only medical help from doctors and nurses, so the health centres and drug shops is where I only go” (female patient, Nakapiripirit, Uganda). Results from multivariable regression suggest the importance of perceived costs on decision to self- treat. After adjustment for socio-demographic vari- ables, self-treating from a drug shop or with home remedies at step 1, versus choosing a medical facility, was more likely among patients who find it ‘a little difficult’ (OR = 1·29; 95%HPDI = 1·06, 1·57) or ‘very difficult’ (OR = 1·71; 95%HPDI = 1·35, 2·16) to meet healthcare costs, relative to those who found it ‘easy’ (Fig. 4, panel 1). However, this association was strong- est in Kenya, relative to the other countries (see coun- try specific plots, Figure S9). Having health insurance was associated with 20% reduced odds of self-treating (OR = 0.80, 95% HPDI 0.65–0.98), and this association was strongest in Tanzania (OR = 0.50, 95%HPDI 0.37, 0.71) (Figure S9). Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 7 of 12 Self−treating in step 1 Having 2 or more steps Antibiotic consumption at step 1 or 2 Gender: Male Age: 25−34 Age: 35−44 Age: 45−54 Age: 55−64 Age: 65+ Marital status: Never−Married Marital status: Other Education: Secondary/High school Education: Further/higher/tertiary Wealth: 2nd asset quintile Wealth: 3rd asset quintile Wealth: 4th asset quintile Wealth: 5th asset quintile Working: informal employment Not working: homemaker Not working: other Household size: 3−6 Household size: 7+ Meet healthcare cost: little difficult Meet healthcare cost: very difficult Having Insurance: Yes Feeling symptoms stigma: Yes Had a previous UTI episode Don't know what is UTI Higher hospital level: 4−5/6 0.0 0.5 1.0 1.5 OR [95% HPDI] 2.0 2.5 3.0 0.0 0.5 1.0 1.5 OR [95% HPDI] 2.0 2.5 3.0 0.0 0.5 1.0 1.5 OR [95% HPDI] 2.0 2.5 3.0 Fig. 4 Odds ratios and 95% HPDI from adjusted logistic regression models for outcomes of self‑treating in step 1, having 2 + steps in the pathway and taking ABs in the pathway (N = 6,608). Notes: Antibiotic consumption at step 1 or 2 outcome (n = 3,546) excludes patients going to the recruitment clinic as their 1st step. Reference categories: Feeling symptoms stigma (‘No’) Meeting healthcare costs (‘Easy’); Had previous UTI episode (‘No’); Age (< 25 years); Education: No quals/ primary; Marital status (‘married’); Wealth quintile (1.st‑ poorest); Working (‘formal employment’); Household size (1–2 people); Hospital level (2–3) Attempted self-treatments consuming herbal remedies and ABs obtained in drug shops with- out a prescription, as shown in this pathway from a female patient in Nakapiripirit, Uganda. included Step 0, Delay: “When I started feeling this pain I first ignored [it]” Step 1, Traditional Medicine: “Later I went to [a traditional healer] where I was told that I was bewitched … The herbalist gave me some leaves for bathing … then for drinking I was given moringa, he told me to take for one month ... I felt fine for that one month, but it didn’t help me.” Step 2, Drug seller: “I began buying the medicine in the drug shop. I was given amoxicillin capsule and metronidazole. I bought [a] full dose that I took for 3 weeks I was told to take 23 a day, which I did. In addition, I stayed for some months.” Step 3, Drug Seller: “After some time, it started again, and I decided to go and buy more medicine which was not full dose because I didn’t have enough money to buy the medicine.” Step 4, Recruitment Facility: “Then in March I met sister who… [advised me] to come for check-up and I was given medication for one month from [this] heath centre.” Here, initial self-treatment behaviour was driven by confusion: “I just don’t know what to do” (female patient, Nakapiripirit, Uganda). After unsuccessful treatment attempts with the traditional medicine and drug sell- ers, the patient reflected that for future ailments: “I will be coming to get medication at the hospital not somewhere else since they prescribe well and give the right medication.” Understanding multi‑step pathways Overall, 56% of patients tried some other form of treat- ment before arriving at the recruitment facility; we label these as having ‘multi-step pathways’. Of these patients (n = 3,546), 53% took one step before the recruitment clinic (30% of the full sample), 29% had two additional steps (16% of the full sample), and 18% had three or more steps (10% of the full sample) (Fig. 5). Multi-step path- ways were more common in Uganda and Tanzania than Kenya. While some multi-step pathways did include self- treatment or drug seller visits, most began with patients seeking treatment from medical facilities, after which pathways became convoluted, often involving repeated visits to medical facilities, causing great frustration, as evident in the pathway description by a male patient from Mbarara, Uganda: Step 1, Facility 1: “When I got infected, I went to [facility 1] … He injected me and gave me drugs to take and told me that I had recovered. I … took the medicine but I did not recover. I remained sick.” Step 2, Facility 1: “I went back and … I asked him, haven’t I healed now, he wrote for me more medi- cine, I went to the pharmacy and bought the drugs ... I took the medicine and I seemed to recover but pain reduced.” Step 3, Drug Seller: “When I finished [the prescrip- tion], I thought I would recover but I went back to the same state. So, I looked at the place I went to buy Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 8 of 12 Fig. 5 Number of pathway steps as percent of the total within each country (N = 6,608) the medicine, I went there, and they wrote for me this one.” Step 4, ‘Leave it’: “[The treatment] has not worked. So, I decided to leave it for a while.” Step 5, Facility 2: “When the sickness continued, I [visited another] doctor … He tested me, gave me medicine, and injected me six times ... Then he took my urine samples ... He told me that the medicine he was injecting couldn’t cure it. He then brought another type of medicine … [but] when the dose was over … I started feeling like the disease had come back. Then I wondered how I was, if I had gone everywhere and the disease was failing, now where was I to go!” Step 6, Recruitment facility: “The doctor … had called a specialist who would test my urine … and give me proper treatment … I decided to go there, and they got my urine samples.” We conducted analysis on a subset of data collected after January with data on diagnostic use in the path- way. This showed that 73% had had their urine sample tested (although it is unknown what kind of test); and that urine testing was more common at private than government facilities (Fig. S12). Having a multi-step pathway was more common in patients older than 35 years, in middle asset quin- tiles, who had health insurance, and those with a previously diagnosed UTI (Fig. 4, panel 2). Stigma around UTI symptoms was associated with longer pathways, dependant on context. In Uganda, where 39% of patients felt stigma, it was associated with higher odds of having a multi-step pathway (OR = 1·43; 95%HPDI = 1·06, 1·90) (Figure S10). By contrast, in Tanzania, where only 16% of patients reported UTI symptoms stigma, this was associated with simpler pathways. IDIs also suggested that in some contexts, stigma drove treatment choices. For example, Kenyan patients discussed choosing private or distant medical facilities to avoid being recognized by members of the community. Private clinics were also favoured because treatment was faster, meaning patients could avoid hav- ing to explain absences from work: “If I went to a public hospital, I could have taken a long time, and my friends at the market could have wanted to know what my prob- lem was and this way I tried my level best to hide it” (female patient, Makueni, Kenya). Antibiotic consumption during the pathway Increasing the number of pathway steps provided more opportunities for AB consumption. Among patients with multi-step pathways, nearly half (48%) reported taking ABs at step one, and 42% of those with a second step took ABs (Fig. 2). As Fig. 2 shows, most ABs were consumed after visits to medical facilities. The most common ABs taken were amoxicillin and ciprofloxacin Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 9 of 12 (see Table S5), and some ABs taken were not recom- mended for treating UTI symptoms (e.g., doxycycline). AB consumption was higher among those with UTI symptoms stigma (OR = 1·34; 95%HPDI = 1·09, 1·62), those who had been previously diagnosed with UTI (OR = 1·33; 95%HPDI = 1·07, 1·63), and those who had (OR = 1·43; higher-level educational qualifications 95%HPDI = 1·07, 1·89) (Fig. 4, panel 3); AB consump- tion was lower among those aged 65 + (OR = 0·63; 95%HPDI = 0·44, 0.88), those who don’t know what a UTI is (OR = 0.77, 95%HPDI 0.61,0.96), and among homemakers, relative to those in formal employment (OR = 0.77, 95% HPDI 0.60,0.98). Qualitative data enriched the picture with stories of repeat prescriptions for the same drugs procured from multiple visits to medical centres and drug sellers, evi- dent in the pathway of a female patient from Moshi, Tanzania: Step 1, Private Facility: "They started with amoxi- cillin. Then they gave me ampiclox. Then they gave me amoxiclav. But symptoms persisted." Step 2, Government Facility 1: "Then I … went to Kilimanjaro hospital, and there I still had UTI … They gave me amoxiclav." Step 3, Government Facility 2: “Then I went to [another] hospital, they also gave me amoxiclav… and they said that was a strong drug." Step 4, Recruitment Facility: "I went there because I didn’t get a cure. And I had two problems. The biggest one was ulcers, and I still had UTI symp- toms… [the doctor] suggested the tests I needed to take. I agreed and they tested me … [then] he told me that he is giving me a seven-day dose [of amoxiclav].” This patient further expressed concern that her insurance coverage impacted the drugs she was given: “Sometimes a doctor may prescribe same drugs that you have used before and when you ask him he says it is the insurance” (female patient, Moshi, Tanzania). Quanti- tative data corroborated this picture. In country-spe- cific quantitative models (Figure S11), having health insurance was associated with higher chance of antibi- otic use (OR = 1.64, 95%HPDI 1.20, 2.20) in Tanzania, but not the other countries. In these settings, visits to medical facilities and drug sellers often go hand in hand, due to stockouts in public health facilities: “You find yourself not having money, [and at the hospital] you are just given a prescription and told to buy drugs. If you don’t have money you won’t buy, until you get money that is when you will buy” (female patient, Mwanza, Tanzania). This contributed to decisions to go directly to drug sellers to save the cost of medical consultation. Discussion Using mixed-methods data, we show that patient path- ways in East Africa for a common infection (UTI) are often convoluted, involving reiterative steps and differ- ent healthcare providers. Such complexity was not driven by individual choice; patients were struggling to get care that worked in a confusing, hybrid healthcare landscape riddled with structural constraints. Our findings sup- port others which stress the importance of location, cost, and time in treatment decision making, [11, 39], but we show these are contextualised by community factors such as stigma, illness behaviour, and local understandings of illness which are in turn conditioned by wider socioeco- nomic, geographic, and healthcare structures. These pat- terns also need to be placed in the context of struggles to access formal healthcare alongside relative ease of access to antibiotics at pharmacies and drug shops in the region [39, 40]. Even though patients predominantly chose what might be regarded as clinically ‘optimal’ paths, i.e., attending medical facilities, this often did not resolve their symp- toms. The causes of these treatment failures in healthcare settings deserve further investigation. Over three quar- ters of the patients in the study did not have a microbio- logically confirmed UTI, so empirical treatment for this infection was not likely to resolve their symptoms. Some of those with microbiologically confirmed UTI will have forms of drug resistance which make their UTI more difficult to treat. Improvements in access to diagnostic capabilities in LMICs [41] could detect forms of infec- tion and drug resistance and guide appropriate treat- ments at earlier stages in the treatment journey. One area to address could be better ways to record medical his- tory so that clinicians understand what drugs have been taken and why. Another issue may be equitable access to appropriate treatments: around two thirds of those attending clinics ended up not taking any medicines, and our data suggest this may be due to stockouts at clinics and prohibitive costs of medicine [11, 13]. Future studies should address clinic-related factors in depth, including the patient-doctor consultation, medical records keeping and information flows, and other barriers that prevent medical staff and patients from following best practice advice around ABs. Given that most patients sought care at public health facilities, this underlines the importance of understanding limitations in healthcare systems and infrastructure to address the threat of AMR. Patients’ treatment choices were often motivated by time and financial constraints, which can lead to Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 10 of 12 systematically different pathways among the poorest and the richest patients. More educated and wealthier patients were both more likely to consume ABs and have multi-step pathways, corroborating findings from other studies [10, 42]. As described elsewhere, [13] wealthier subgroups may be also more likely to choose private facil- ities and buy medicines from drug sellers to save time. On the other hand, poorer subgroups or those strug- gling to afford healthcare are more likely to self-treat and struggle to afford appropriate AB treatment. This study has some limitations. The linked quan- titative–qualitative sample is representative of the patient population attending public facility outpatient services for UTI symptoms at study sites (future con- sortium papers will address community members not attending clinics). We repeated the analysis restricted to patients with microbiologically confirmed UTI and the same pattern of effects was seen (Figure S13). We would also recommend repeating the study with a focus on other symptoms for other common condi- tions that prompt antibiotic use (e.g. upper respiratory tract infections), because they potentially have differ- ent levels of stigma and are confused with other con- ditions such as COVID-19 [21]. As mentioned above, further emphasis on clinic-related factors, such as diagnostics, medicine availability, prescribing patterns is warranted. Given the self-reported nature of treat- ment-seeking behaviour and its predictors, we cannot rule out reporting bias. Conclusion This study has taken a patient-centric perspective, but our results suggest that treatment-seeking is never an individuated behaviour; actions are influenced by situ- ational constraints and are contextually dependent. Thus, AMR should be considered a system rather than a set of individual actions. Complex treatment pathways are likely related to various individual and structural factors, but another important driver is likely to be ABR itself. As ABR continues to evolve, the cyclical treatment attempts we observed here for UTI-like illness will become more common, reflecting the vicious socio-biological cycle of ABR. Drug resistance means treatment attempts are more likely to fail, thus fuelling ABR by necessitat- ing further AB treatments. In many LMICs, there are key structural weaknesses which facilitate this vicious cycle: including (but not limited to) under-resourced public healthcare, insufficient diagnostic capacity, and ample opportunities to purchase ABs without prescrip- tion. Thus, we advocate that attention be paid towards addressing these upstream factors which drive both ABR and complex patient pathways. Abbreviations ABR ABs HATUA UTI LMIC IDIs OR HDPI Antibacterial resistance Antibiotics The Holistic Approach to Unravelling Antibacterial Resistance Consortium Urinary tract infection Low‑ and middle‑income country In‑depth interviews Odds Ratio Highest posterior density interval Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12879‑ 023‑ 08392‑9. Additional file 1. Acknowledgements HATUA Consortium includes Kathryn J. Fredricks, PhD1, Mary Abed Al Ahad, MSc1, Stella Neema, PhD2, Joseph R. Mwanga, PhD3, Mike Kesby, PhD1, Martha F. Mushi, PhD3, Annette Aduda, PhD4, Dominique L. Green, PhD1, Andy G. Lynch, PhD1, Sarah I. Huque, PhD1, Blandina T. Mmbaga, PhD5, Hannah Worthington, PhD1, Catherine Kansiime, PhD2 , Emmanuel Olamijuwon, PhD1, Nyanda E. Ntinginya, PhD6, Olga Loza, PhD1, Joel Bazira, PhD7, Antonio Maldonado‑Barragán, PhD1, V Anne Smith, PhD1, Arun Gonzales Decano, PhD1, John Njeru Mwaniki, PhD4 , Alison Sandeman, PhD1, John Stelling, MD8, Alison Elliott, PhD9,10, David Aanensen, PhD11, Stephen H. Gillespie, DSc1, Gibson Kibiki, MD12, Wilber Sabiiti, PhD1, Derek J. Sloan, MD1, Benon B. Asiimwe, PhD2, John Kiiru, PhD4, Stephen E. Mshana, PhD3, Benjamin Sunday, PhD Pendo Ndaki, MSc3 Fernando Benitez‑Paez, PhD1 Madeleine Clarkson, MSc1 Xuejia Ke, MSc1 Eveline T. Konje, PhD3 Matthew T. G. Holden, PhD1 Katherine Keenan, PhD1, 1University of St Andrews, UK 2Makerere University, Uganda 3Catholic University of Health and Allied Sciences, Tanzania 4Kenya Medical Research Institute, Kenya 5Kilimanjaro Clinical Research Institute and Kilimanjaro Christian Medical University College, Tanzania 6NIMR‑Mbeya Medical Research Centre, Tanzania 7Mbarara University, Uganda 8Brigham and Women’s Hospital, United States Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 11 of 12 9London School of Hygiene & Tropical Medicine, UK 10Medical Research Council/Uganda Virus Research Institute and London School of Hygiene & Tropical Medicine Uganda Research Institute, Uganda 11University of Oxford, UK 12Africa Research Excellence Fund, UK Authors’ contributions KK co‑designed the study, conceptualised the paper, supervised analysis of quantitative data, and wrote the manuscript. KJF conceptualised the paper, analysed qualitative data, and wrote the manuscript. MAAA conceptualised the paper, assisted with data preparation, analysed the quantitative data, and edited the manuscript. SN co‑designed the study, supervised data collec‑ tion, and reviewed the manuscript. JRM co‑designed the study, supervised data collection, and reviewed the manuscript. MK co‑designed the study, conceptualised the paper, supervised qualitative data analysis, and edited the manuscript. MFM supervised data collection and reviewed the manuscript. AA supervised data collection and reviewed the manuscript. DLG assisted with data preparation and reviewed the manuscript. AGL co‑designed the study, supervised quantitative data analysis, and edited the manuscript. SIH analysed qualitative data and edited the manuscript. BTM co‑designed the study, supervised data collection, and reviewed the manuscript. HW provided statistical advice and edited the manuscript. CK supervised data collection and reviewed the manuscript. EO assisted with quantitative data analysis and reviewed the manuscript. NEN supervised data collection and reviewed the manuscript. OL reviewed the manuscript. JB supervised data collection and reviewed the manuscript. AM‑B reviewed the manuscript. VAS co‑designed the study and reviewed the manuscript. AGD reviewed the manuscript. JNM supervised data collection and reviewed the manuscript. AS coordinated the study and reviewed the manuscript. JS co‑designed the study and reviewed the manuscript. AE co‑designed the study and reviewed the manuscript. DA co‑designed the study and reviewed the manuscript. SHG co‑designed the study and reviewed the manuscript. GK co‑designed the study and reviewed the manuscript.WS co‑designed the study and reviewed the manuscript. DJS co‑designed the study and reviewed the manuscript. BBA co‑designed the study, supervised data collection, and reviewed the manuscript. JK co‑ designed the study, supervised data collection, and reviewed the manuscript. SEM co‑designed the study, supervised data collection, and reviewed the manuscript. MTGH led the design of the project, was the guarantor of the work, and reviewed the manuscript. Funding The Holistic Approach to Unravel Antibacterial Resistance in East Africa is a Global Context Consortia Award (MR/S004785/1) funded by the National Insti‑ tute for Health Research, Medical Research Council, and the Department of Health and Social Care. The award is also part of the EDCTP2 programme sup‑ ported by the European Union. This work is supported in part by the Makerere University‑Uganda Virus Research Institute Centre of Excellence for Infection and Immunity Research and Training (MUII). MUII is supported through the DELTAS Africa Initiative (grant number 107743). The DELTAS Africa Initiative is an independent funding scheme of the African Academy of Sciences and Alli‑ ance for Accelerating Excellence in Science in Africa and is supported by the New Partnership for Africa’s Development Planning and Coordinating Agency with funding from the Wellcome Trust (grant number 107743) and the UK Government. This paper was funded in part by a grant from the National Insti‑ tutes of Health (grant number U01CA207167), and a Scottish Funding Council GCRF Consolidator Award. The funders had no role in study design, data col‑ lection and analysis, decision to publish or preparation of the manuscript. Availability of data and materials The datasets generated and/or analysed during the current study are not publicly available due to ethical and data access agreements with individual country ethical boards but are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The study received ethical approval from the University of St Andrews, UK (number MD14548, 10/09/19); National Institute for Medical Research, Tanzania (number 2831, updated 26/07/19); CUHAS/BMC Research Ethics and Review Committee (number CREC /266/2018, updated on 02/2019); Mbeya Medical Research and Ethics Committee (number SZEC‑2439/R.A/V.1/303030); Kilimanjaro Christian Medical College, Tanzania (number 2293, updated 14/08/19); Uganda National Council for Science and Technology (number HS2406, 18/06/18); Makerere University, Uganda (number 514, 25/04/18); and Kenya Medical Research Institute (04/06/19, Scientific and Ethics Review Committee (SERU) number KEMRI/SERU/CMR/P00112/3865 V.1.2). For Uganda, administrative letters of support were obtained from the district health officers to allow the research to be conducted in the respective hospitals and health centres. All participants provided written informed consent to partici‑ pate. All methods were carried out in accordance with relevant guidelines and regulations. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Author details 1 School of Geography and Sustainable Development, University of St. Andrews, St Andrews KY16 9AL, UK. 2 Makerere University, Kampala, Uganda. 3 Catholic University of Health and Allied Sciences, Mwanza, Tanzania. 4 Kenya Medical Research Institute, Nairobi, Kenya. 5 Kilimanjaro Clinical Research Institute and Kilimanjaro Christian Medical University College, Moshi, Tanzania. 6 NIMR‑Mbeya Medical Research Centre, Mbeya, Tanzania. 7 Mbarara University, Mbarara, Uganda. 8 Brigham and Women’s Hospital, Boston, USA. 9 London School of Hygiene & Tropical Medicine, London, UK. 10 Medical Research Coun‑ cil/Uganda Virus Research Institute and London School of Hygiene & Tropical Medicine Uganda Research Institute, Entebbe, Uganda. 11 University of Oxford, Oxford, UK. 12 Africa Research Excellence Fund, London, UK. Received: 5 August 2022 Accepted: 9 June 2023 References 1. Murray CJ, Ikuta KS, Sharara F, Swetschinski L, Robles Aguilar G, Gray A, et al. Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis. Lancet. 2022;399:629–55. https:// doi. org/ 10. 1016/ S0140‑ 6736(21) 02724‑0. 2. Davis A, Sharp J. Rethinking one health: emergent human, animal and environmental assemblages. Soc Sci Med. 2020;258:113093. https:// doi. org/ 10. 1016/j. socsc imed. 2020. 113093. 3. Chatterjee A, Modarai M, Naylor NR, Boyd SE, Atun R, Barlow J, et al. Quantifying drivers of antibiotic resistance in humans: a systematic review. Lancet Infect Dis. 2018;18:e368–78. https:// doi. org/ 10. 1016/ S1473‑ 3099(18) 30296‑2. 4. Ahmad M, Khan AU. Global economic impact of antibiotic resistance: A 5. review. J Glob Antimicrob Resist. 2019;19:313–6. https:// doi. org/ 10. 1016/j. jgar. 2019. 05. 024. Tompson AC, Manderson L, Chandler CIR. Understanding antibi‑ otic use: practices, structures and networks. JAC‑Antimicrob Resist. 2021;3(4):dlab150. https:// doi. org/ 10. 1093/ jacamr/ dlab1 50. 6. Hanson CL, Osberg M, Brown J, Durham G, Chin DP. Conducting patient‑ pathway analysis to inform programming of tuberculosis services: methods. J Infect Dis. 2017;216 suppl_7:S679‑85. https:// doi. org/ 10. 1093/ infdis/ jix387. 7. Haenssgen MJ, Ariana P. Healthcare access: a sequence‑sensitive 8. approach. SSM ‑ Population Health. 2017;3:37–47. https:// doi. org/ 10. 1016/j. ssmph. 2016. 11. 008. Lucas PJ, Uddin MR, Khisa N, Salim Akter SM, Unicomb L, Nahar P, et al. Pathways to antibiotics in Bangladesh: a qualitative study investigating how and when households access medicine includ‑ ing antibiotics for humans or animals when they are ill. PLoS ONE. 2019;14:e0225270. https:// doi. org/ 10. 1371/ journ al. pone. 02252 70. 9. Ocan M, Obuku EA, Bwanga F, Akena D, Richard S, Ogwal‑Okeng J, et al. Household antimicrobial self‑medication: A systematic review and meta‑analysis of the burden, risk factors and outcomes in developing Keenan et al. BMC Infectious Diseases (2023) 23:414 Page 12 of 12 countries. BMC Public Health. 2015;15:742. https:// doi. org/ 10. 1186/ s12889‑ 015‑ 2109‑3. 10. Do NTT, Vu HTL, Nguyen CTK, Punpuing S, Khan WA, Gyapong M, et al. Community‑based antibiotic access and use in six low‑income and middle‑income countries: a mixed‑method approach. Lancet Glob Health. 2021. https:// doi. org/ 10. 1016/ S2214‑ 109X(21) 00024‑3. 11. Torres NF, Chibi B, Middleton LE, Solomon VP, Mashamba‑Thompson TP. Evidence of factors influencing self‑medication with antibiotics in low and middle‑income countries: a systematic scoping review. Public Health. 2019;168:92–101. https:// doi. org/ 10. 1016/j. puhe. 2018. 11. 018. 12. Larson PS. Access to health care in Sub‑Saharan Africa: challenges in a changing health landscape in a context of development. Health Ecol Perspect Anthropocene. 2019;95–106. https:// link. sprin ger. com/ chapt er/ 10. 1007/ 978‑ 981‑ 13‑ 2526‑7_8. 13. Green DL, Keenan K, Fredricks KJ, et al. The Role of Multidimensional Pov‑ erty in Antibiotic Misuse: A Study of Self‑Medication and Non‑Adherence in Kenya, Tanzania, and Uganda. Lancet Global Health. 2023;1(1):e59–68. https:// doi. org/ 10. 1016/ S2214‑ 109X(22) 00423‑5. 14. Charani E, McKee M, Ahmad R, Balasegaram M, Bonaconsa C, Merrett GB, et al. Optimising antimicrobial use in humans – review of current evidence and an interdisciplinary consensus on key priorities for research. Lancet Reg Health Eur. 2021;7:100161. https:// doi. org/ 10. 1016/j. lanepe. 2021. 100161. 15. World Health Organization. Global Health Expenditure Database. https:// apps. who. int/ nha/ datab ase. Accessed 15 Jul 2022. 16. Umeh CA. Challenges toward achieving universal health coverage in Ghana, Kenya, Nigeria, and Tanzania. Int J Health Plann Manage. 2018;33:794–805. https:// doi. org/ 10. 1002/ hpm. 2610. 17. Amu H, Dickson KS, Kumi‑Kyereme A, Darteh EKM. Understanding variations in health insurance coverage in Ghana, Kenya, Nigeria, and Tanzania: Evidence from demographic and health surveys. PLoS ONE. 2018;13:e0201833. https:// doi. org/ 10. 1371/ journ al. pone. 02018 33. 18. Namyalo PK, Mutatina B, Byakika S, Walimbwa A, Kato R, Basaza RK. The feasibility analysis of integrating community‑based health insurance schemes into the national health insurance scheme in Uganda. PLoS ONE. 2023;18:e0284246.https:// doi. org/ 10. 1371/ journ al. pone. 02842 46. 19. Goel PK, Ross‑Degnan D, Berman P, Soumerai S. Retail pharmacies in developing countries: A behavior and intervention framework. Soc Sci Med. 1996. 1155–61. https:// doi. org/ 10. 1016/ 0277‑ 9536(95) 00388‑6. 20. Ocan M, Bwanga F, Bbosa GS, Bagenda D, Waako P, Ogwal‑Okeng J, et al. Patterns and predictors of self‑medication in Northern Uganda. PLoS ONE. 2014;9:e92323. https:// doi. org/ 10. 1371/ journ al. pone. 00923 23. 21. Olamijuwon E, Konje E, Kansiime C, Kesby M, Keenan K, Neema S, et al. Antibiotic dispensing practices during COVID‑19 and implications for antimicrobial resistance (AMR): parallel mystery client studies in Uganda and Tanzania. Antimicrob Resist Infect Control. 2023;12:10. https:// doi. org/ 10. 1186/ s13756‑ 022‑ 01199‑4. 22. Muloi D, Fèvre EM, Bettridge J, Rono R, Ong’are D, Hassell JM, et al. A cross‑sectional survey of practices and knowledge among antibiotic retailers in Nairobi, Kenya. J Glob Health. 2019;9:010412. https:// doi. org/ 10. 7189/ jogh. 09. 020412. 23. Flores‑Mireles AL, Walker JN, Caparon M, Hultgren SJ. Urinary tract infec‑ tions: epidemiology, mechanisms of infection and treatment options. Nat Rev Microbiol. 2015;13:5 2015;13:269–84. https:// doi. org/ 10. 1038/ nrmic ro3432. Izett‑Kay M, Barker KL, McNiven A, Toye F. Experiences of urinary tract infection: A systematic review and meta‑ethnography. Neurourology and Urodynamics. 2022. https:// doi. org/ 10. 1002/ NAU. 24884. 24. 25. Stamm WE, Norrby SR. Urinary tract infections: disease panorama and challenges. J Infect Dis. 2001;183 Supplement_1:S1‑4. https:// doi. org/ 10. 1086/ 318850. 26. Mazzariol A, Bazaj A, Cornaglia G. Multi‑drug‑resistant Gram‑negative bacteria causing urinary tract infections: a review. J Chemother. 2017;29:2–9. https:// doi. org/ 10. 1080/ 11200 09X. 2017. 13803 95. 27. Smith CM, Shallcross LJ, Dutey‑Magni P, Conolly A, Fuller C, Hill S, et al. Incidence, healthcare‑seeking behaviours, antibiotic use and natural history of common infection syndromes in England: results from the Bug Watch community cohort study. BMC Infect Dis. 2021;21:1–11. https:// doi. org/ 10. 1186/ s12879‑ 021‑ 05811‑7. 28. der Lelie‑vanZande R, Koster ES, Teichert M, Bouvy ML. Womens’ self‑ management skills for prevention and treatment of recurring urinary tract infection. Int J Clin Pract. 2021;75:e14289. https:// doi. org/ 10. 1111/ ijcp. 14289. 29. Asiimwe BB, Kiiru J, Mshana SE, Neema S, Keenan K, Kesby M, et al. Protocol for an interdisciplinary cross‑sectional study investigating the social, biological and community‑level drivers of antimicrobial resistance (AMR): Holistic Approach to Unravel Antibacterial Resistance in East Africa (HATUA). BMJ Open. 2021;11:e041418. https:// doi. org/ 10. 1136/ bmjop en‑ 2020‑ 041418. 30. Moshabela M, Pronyk P, Williams N, Schneider H, Lurie M. Patterns and implications of medical pluralism among HIV/AIDS patients in rural South Africa. AIDS Behav. 2011;15:842–52. https:// doi. org/ 10. 1007/ s10461‑ 010‑ 9747‑3. 31. Coast E, Norris AH, Moore AM, Freeman E. Trajectories of women’s abor‑ tion‑related care: a conceptual framework. Soc Sci Med. 2018;200:199– 210. https:// doi. org/ 10. 1016/j. socsc imed. 2018. 01. 035. 32. Zøylner IA, Lomborg K, Christiansen PM, Kirkegaard P. Surgical breast cancer patient pathway: Experiences of patients and relatives and their unmet needs. Health Expect. 2019;22:262–72. https:// doi. org/ 10. 1111/ hex. 12869. 33. Chandler CIR. Current accounts of antimicrobial resistance: stabilisation, individualisation and antibiotics as infrastructure. Palgrave Communica‑ tions. 2019;5:15–7. https:// doi. org/ 10. 1057/ s41599‑ 019‑ 0263‑4. 34. Ayukekbong JA, Ntemgwa M, Atabe AN. The threat of antimicrobial resistance in developing countries: Causes and control strategies. Antimicrob Resist Infect Control. 2017;6:47. https:// doi. org/ 10. 1186/ s13756‑ 017‑ 0208‑x. 35. Lu J, Sheldenkar A, Lwin MO. A decade of antimicrobial resistance research in social science fields: a scientometric review. Antimi‑ crob Resist Infect Control. 2020;9(1):178. https:// doi. org/ 10. 1186/ s13756‑ 020‑ 00834‑2. 36. Dixon J, MacPherson E, Manyau S, Nayiga S, Khine Zaw Y, Kayendeke M, et al. The ‘Drug Bag’ method: lessons from anthropological stud‑ ies of antibiotic use in Africa and South‑East Asia. Glob Health Action. 2019;12:1639388. 37. de Valpine P, Turek D, Paciorek C, Anderson‑Bergman C, Temple Lang D, Bodik R. Programming with models: writing statistical algorithms for gen‑ eral model structures with NIMBLE. J Comput Graph Stat. 2017;26:403– 413. https:// doi. org/ 10. 1080/ 10618 600. 2016. 11724 87. 38. QSR International Pty Ltd. NVivo (released in March 2020). 2020. https:// www. qsrin terna tional. com/ nvivo‑ quali tative‑ data‑ analy sis‑ softw are/ home. 39. Dixon J, Macpherson EE, Nayiga S, Manyau S, Nabirye C, Kayendeke M, et al. Antibiotic stories: a mixed‑methods, multi‑country analysis of household antibiotic use in Malawi, Uganda and Zimbabwe. BMJ Global Health. 2021;6:e006920. https:// doi. org/ 10. 1080/ 16549 716. 2019. 16393 88. 40. Ndaki PM, Mushi MF, Mwanga JR, Konje ET, Ntinginya NE, Mmbaga BT, et al. Dispensing antibiotics without prescription at community pharma‑ cies and accredited drug dispensing outlets in Tanzania: a cross‑sectional study. Antibiotics. 2021;10:1025. https:// doi. org/ 10. 3390/ antib iotic s1008 1025. 41. Yadav H, Shah D, Sayed S, Horton S, Schroeder LF. Availability of essential diagnostics in ten low‑income and middle‑income countries: results from national health facility surveys. Lancet Glob Health. 2021;9:e1553– 60. https:// doi. org/ 10. 1016/ S2214‑ 109X(21) 00442‑3. 42. McKinn S, Trinh DH, Drabarek D, Trieu TT, Nguyen PTL, Cao TH, et al. Driv‑ ers of antibiotic use in Vietnam: implications for designing community interventions. BMJ Glob Health. 2021;6:e005875. https:// doi. org/ 10. 1136/ bmjgh‑ 2021‑ 005875. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub‑ lished maps and institutional affiliations.
10.1186_s13048-020-00624-9	Farkas et al. Journal of Ovarian Research (2020) 13:25 https://doi.org/10.1186/s13048-020-00624-9 R E S E A R C H Open Access Comparative analysis of abdominal fluid cytokine levels in ovarian hyperstimulation syndrome (OHSS) Balint Farkas1,2* Jozsef Bodis1,2 and Miklos Koppan1 , Ferenc Boldizsar3, Noemi Bohonyi1, Nelli Farkas4, Saska Marczi5,6, Gabor L. Kovacs7,8,2, Abstract Background: Ovarian hyperstimulation syndrome (OHSS) is a rare, yet severe, iatrogenic complication of ovulation induction therapy during assisted reproductive procedures. Our group previously detected atypical cells in the ascitic fluid of OHSS patients, although no malignancy developed during follow up. Here, the aim was to perform a comparative analysis of the cytokines present in the abdominal fluid of patients affected by OHSS versus patients with advanced ovarian cancer, a benign adnexal mass, or ovarian endometriosis. Methods: This prospective, non-randomized study was conducted at the Clinical Center of the University of Pecs Department of Obstetrics and Gynecology/Reproductive Center between October 2016 and March 2018. Abdominal fluid samples were obtained from 76 patients and subjected to Luminex analysis. The samples were collected from patients with OHSS (OHSS; n = 16), advanced ovarian cancer (OC; n = 22), a benign adnexal mass (BAM; n = 21), or ovarian endometriosis (EM; n = 17). Data were subjected to the non-parametric Kruskal-Wallis test and Spearman’s rank correlation coefficient to identify statistical differences between the four study groups. Results: Leukocytosis and hemoconcentration were detected in the peripheral blood of OHSS patients. Abdominal fluid analysis further revealed significantly higher levels of interleukin (IL)-6, IL-8, IL-10, and transforming growth factor (TGF)-β in both the OHSS and OC groups compared to the BAM and EM groups. The highest concentration of vascular endothelial growth factor (VEGF) was detected in the OC group, while a significantly lower level was detected in the OHSS group. Moreover, VEGF levels in OC and OHSS groups were significantly elevated compared to the levels in the BAM and EM groups. Conclusions: Vasoactive and hematogenic cytokines were present at higher levels in both the OHSS and OC abdominal fluid samples compared to the fluid samples obtained from the peritoneal cavity of the BAM patients. It is possible that these cytokines play an important role in the formation of ascites. Keywords: Ovarian hyperstimulation syndrome, Ovulation induction therapy, Ovarian cancer, Ovarian endometriosis, Benign pelvic mass * Correspondence: dr.balint.farkas@gmail.com 1Department of Obstetrics and Gynecology, University of Pecs, School of Medicine, 17 Edesanyak Str., Pecs, Hungary 2Member of the HAS-UP Human Reproduction Scientific Research Group, Hungarian Academy of Sciences (HAS), Pecs, Hungary Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Farkas et al. Journal of Ovarian Research (2020) 13:25 Page 2 of 8 Background Infertility is defined as an individual’s inability to repro- duce through a natural process. Currently, infertility represents a major healthcare issue in the twenty-first century and it can be the result of male, female, or combined infertility issues. Worldwide, an estimated 48 million women and approximately 7% of men suf- fer from infertility [1, 2]. Thus, a growing need for assisted reproduction techniques, particularly in vitro fertilization (IVF) procedures, exists to facilitate con- ception. However, ovarian hyperstimulation syndrome (OHSS) is a rare, yet potentially life threatening, iatrogenic complication of ovarian induction therapy (OIT) during IVF procedures. OHSS is associated with abdominal pain and/or bloating, nausea, vomiting, and in severe cases, shortness of breath and chest pain. A diagnosis of OHSS is confirmed with laboratory findings of hemoconcentra- tion and ultrasound imaging [3]. Manifestations of the disease can vary from mild to moderate to severe. OHSS often develops after the administration of gonadotropins which are needed to facilitate oocyte maturation and release during IVF procedures. The pathophysiology of OHSS is characterized by the appearance of multiple large luteinized cysts in the ovaries. These cysts are accompan- ied by a simultaneous increase in vascular permeability which leads to a shift in fluids from the intravascular system to the abdominal and pleural cavity [4]. Despite greater insights into the etiology of OHSS, the It has been exact pathomechanism remains unclear. hypothesized that local vasoactive mediators, such as vascular endothelial growth factor (VEGF), substances belonging to the renin-angiotensin system, and cytokines such as interleukin (IL)-6 and IL-8, play major roles in disease pathogenesis [5–9]. These factors can potentially induce fluid redistribution and massive extravasation, thereby resulting in a state of hypovolemic hyponatre- mia with hemoconcentration, as well as hypercoagula- bility [8, 9]. A growing concern among public opinion is a potential link between IVF procedures and malignant disease, and this issue may challenge the safety of assisted reproduction. To date, there is no clear evidence which demonstrates a causative role for IVF procedures in breast cancer [10, 11] or ovarian cancer [12]. However, we previously detected atypical cells in the ascitic fluid of women with severe OHSS [13]; although, no correlation between the presence of these cells and subsequent malignancy was observed [13]. Therefore, the aim of the current study was to analyze and compare the levels of potentially key mediators of OHSS in the ascitic fluid of women with OHSS, advanced ovarian cancer (OC), or ovarian endometriosis, and in the abdominal fluid of women with benign adnexal masses (BAMs). We hypothesize that these results will provide a better understanding of the pathomechanism of OHSS. Results Demographic characteristics The mean age of our study groups were: 34 ± 5 years (range: 26–44) for the OHSS group; 64 ± 13 years (range: 30–84) for the OC group; 51 ± 15 years (range: 24–78) for the BAM group; and 34 ± 8 years (range: 18–47) for the ovarian endometriosis (EM) group. Peripheral blood serum analysis The mean serum levels of Na+ and K+, as well as activity levels for – aspartate transaminase (ASAT), alanine ami- (ALAT), and lactate dehydrogenase notransferase - (LDH), are presented in Fig. 1. In addition, white and red blood cell counts (WBC and RBC, respectively), thrombocyte (TCT) count, blood hemoglobin concentra- tion (Hgb), and hematocrit level (Htc) are also presented in Fig. 1. Application of the non-parametric Kruskal- Wallis test to these data revealed significant differences between the distribution of several values among the study groups (Fig. 1). Abdominal fluid cytokine level analysis From the OHSS patients, an average of 1.1 l of ascites were withdrawn. In a Luminex assay, levels of six cyto- kines were investigated: IL-6, IL-8, IL-10, tumor necrosis factor (TNF)-α, VEGF, and transforming growth factor (TGF)-β. The mean concentration values for IL-6, IL-8, and TGF-β were significantly higher in both the OHSS and OC groups compared to the BAM and EM groups (Fig. 2). The level of VEGF was only significantly higher in the OC group. No statistically significant differences in TNF-α concentrations were observed among the four study groups (Fig. 2). With Spearman’s correlation analysis various significant positive correlations were ob- served in the OHSS group between the WBC count and IL-6 level (r = 0.640; p < 0.01), between the IL-6 and IL- 10 levels (r = 0.677; p < 0.01), between the IL-6 and VEGF levels (r = 0.652; p < 0.01), and between the IL-10 and VEGF levels (r = 0.615; p < 0.01). In contrast, a sig- nificant negative correlation was observed between the serum CA-125 level and VEGF concentration in ascites (r = − 0.584; p < 0.01). Meanwhile, a significant positive correlation was observed between serum CA-125 level and abdominal fluid VEGF concentration in the EM group (r = 0.564; p = 0.02). In the OC group, a significant positive correlation between peripheral blood TCT level (r = 0.568; p < 0.01), and and serum CA-125 level between TCT and VEGF concentration (r = 0.624; p < 0.01), were observed. In the BAM group, the level of IL- 6 in abdominal fluid exhibited a significant positive correlation with the levels of IL-8, IL-10, VEGF, and TGF-β. Similarly, IL-8 levels exhibited a significant posi- tive correlation with the levels of IL-10, VEGF, and Farkas et al. Journal of Ovarian Research (2020) 13:25 Page 3 of 8 Fig. 1 Levels of LDH and Hgb in peripheral blood serum and hematogram values for the four study groups. Comparison were made with non- parametric Kruskal-Wallis test with Bonferroni post hoc test TGF-β, and also between the IL-10 level and the VEGF and TGF-β levels (See Fig. 3.)). Discussion Cytokines are a group of polypeptides which are unable to penetrate the lipid bilayer of cells. Despite this limita- tion, it is still hypothesized that these peptides play an important role in cell signaling. In OHSS, roles for sev- eral cytokines have been well-established, thereby sug- gesting that interactions take place between the immune system and the ovaries during the development of this disease [14]. In the current pilot study, significant alter- ations in the levels of examined cytokines were observed in the abdominal fluid samples collected from our four study groups. The pro-inflammatory cytokine, IL-6, is produced by various cells, including monocytes, T lymphocytes, endothelial cells, and fibroblasts [15]. It has also been proposed that IL-6 is a major mediator of ascites for- mation based on its involvement in angiogenesis and hyperpermeability [16]. The present results confirm that high levels of IL-6 are present in the peritoneal cavity of patients with severe OHSS and in patients with advanced ovarian cancer. In contrast, ascites was not detected in the BAM and EM patients. Recent studies also propose that increased serum levels and peritoneal cavity levels of both IL-6 and IL-10 are as- sociated with factors of worse prognosis in ovarian cancer patients [6, 7]. Numerous studies have demonstrated that the vaso- active protein, VEGF, has a key role in OHSS. VEGF belongs to a family of heparin-binding proteins and is able to induce angiogenesis and vascular permeability [17, 18]. VEGF is secreted by ovarian granulosa cells and its production is stimulated by human chorionic go- nadotropin hormone [4, 19–21]. Elevated levels of VEGF have been measured in both serum and ascitic fluid in patients with OHSS [21, 22]. These results, and those of the current study, are consistent with previous reports that high levels of VEGF are present in the ascitic fluid Farkas et al. Journal of Ovarian Research (2020) 13:25 Page 4 of 8 Fig. 2 Cytokine levels in abdominal fluid samples obtained from the peritoneal cavity in the four study groups. Statistical analysis included non- parametric Kruskal-Wallis test with Bonferroni post hoc test of patients with ovarian cancer [23] and also in the ab- low dominal fluid of OHSS patients [24]. Meanwhile, levels of VEGF were detected in BAM and EM patients in the present study. We hypothesize that increased pro- duction of VEGF is a major factor in the formation of ascites. the anti-inflammatory protein, Macrophages produce IL-8, an important cytokine in the immune system. This cytokine induces chemotaxis- triggered neutrophil migration toward inflammation sites and then stimulates phagocytosis once the neutro- levels of IL-8 and phils arrive onsite [25]. Previously, levels of IL-10, were found at higher concentrations in the ascitic fluid of OHSS patients [26, 27]. In our current investigation, sig- nificantly higher levels of both IL-8 and IL-10 were de- tected in the ascites of OHSS and OC patients compared to the levels detected in the abdominal lavage fluid of BAM and EM patients. Moreover, the levels were high- est in the ascites of the OC patients. This finding is consistent with other recently published data [7] and with an angiogenetic role for IL-8 in malignancy [28] and a pivotal immunosuppressive role for IL-10 in OC- associated ascites when activation of dendritic cells via toll-like receptors is compromised [29]. To date, available literature does not indicate a con- sensus regarding the role of TNF-α in OHSS. For ex- ample, while no statistically significant difference was previously found in the amount of TNF-α in the ascites of OHSS patients compared to controls [30], others re- ported elevated levels in the same experimental setting [31]. The TNF-α data obtained in the present study sup- port a less important role for TNF-α in OHSS. TGF-β is a multifunctional cytokine. In its activated it binds TGF-β receptors by forming a serine/ form, threonine kinase complex [31]. Subsequent activation of a signaling cascade leads to downstream activation of various substrates and regulatory proteins. In addition, the transcription of various target genes is induced, Farkas et al. Journal of Ovarian Research (2020) 13:25 Page 5 of 8 Fig. 3 Spearman correlation matrix of the cytokine values, CA-125 tumor marker parameter and age. Correlation coefficients are shown, red in case of negative, blue in case of positive correlation. X marks: non significant connection thereby contributing to differentiation, chemotaxis, pro- liferation, and activation of many immune cells [31]. In our study groups, a significant increase in the levels of TGF-β were detected in the OHSS and OC groups rela- tive to the BAM and EM groups. The OC group had the highest concentration of TGF-β. Among the immuno- suppressive cytokines associated with advanced ovarian cancer, it has been proposed that TGF-β contributes to impaired anti-tumor immune function [32]. However, the role of TGF-β in OHSS remains unknown. The novelty of our data is that we managed to reveal similarly increased, with no statistically significant differ- ence, in the peritoneal cavity levels of IL-6, IL-8, IL-10, VEGF and TGF-ß both in OC and OHSS patients, but found statistically significantly lower levels of the same cytokines compared to BAM and EM groups. Despite the mean age alteration between OC and OHSS groups the inflammatory responses might be hard to compare, but the cytokine production trend seem to be similar in these two groups. This might suggest same kind of pathomechanism of the ascites formation both in OHSS and in ovarian malignancy. There are limitations associated with the present study. These include a relatively low number of participants, a lack of serum cytokine concentration measurements, dis- crepancies of age between the compared groups, which can influence the inflammatory profile, and some other, potentially important cytokine concentrations were not in- vestigated as yet. Regarding the latter point, IL-2 would have been another cytokine of interest to investigate con- sidering that it has been found at high levels in the peri- toneal cavity of OHSS patients [33]. Furthermore, we could not isolate and identify the origin of the atypical cells present in the ascitic fluid of OHSS patients which we previously described [13, 34]. However, a strength of the present study is the broad spectrum of samples which were examined, including abdominal fluid from patients with various benign adnexal masses and from patients with ovarian “chocolate cysts” (e.g., endometrioma), which served as valid negative controls. In the future a proposed potential clinical implication of our study would have been to find anti-inflammatory citokine agents to reduce the symptoms of OHSS, and to decrease the severity of the disease. Farkas et al. Journal of Ovarian Research (2020) 13:25 Page 6 of 8 Conclusions Local pro- and anti-inflammatory cytokines, as well as vasoactive components, play important roles in both the formation of free abdominal fluid and in the pathogen- esis of advanced ovarian cancer and OHSS compared with benign ovarian disease and ovarian manifestation of endometriosis. In further studies serum cytokine levels and peritoneal cavity immune cell distribution might worth to investigate, with the aim to reveal which cell population are colonize and produce the described cytokines. Methods Patients and study design This prospective, non-randomized study was approved by the University of Pecs Institutional Ethical Review Board (#5273–2/2012) and was conducted at the Clinical Center of the University of Pecs Department of Obstetrics and Gynecology/Reproductive Center between October 2016 and March 2018. Patient participation was on a voluntary basis and all enrolled participants were older than 18 years of age. Written informed consent was completed if patients had an adnexal mass or if they were diagnosed with OHSS after OIT. Evaluation of abdominal fluid Abdominal fluid samples were obtained during ultrasound- guided culdocentesis of patients with a severe form OHSS (n = 16), who represented the investigated population; intra- operative ascites sampling was performed during laparot- omy of patients with advanced ovarian cancer (OC) (n = 22), who were the malignant disease group; sterile saline was collected after intraoperative pelvic lavage during lap- aroscopic cystectomy of patients with a benign adnexal mass (BAM; n = 21), who acted as negative controls; and in- traoperative sampling of free abdominal fluid was per- formed during operative laparoscopy for patients with ovarian endometriosis (EM; n = 17), used as transient / be- nign disease group. Clinical and histological diagnoses of the participants are summarized in Table 1. Peripheral blood analysis Peripheral blood samples were collected preoperatively including 60 patients from all the study participants, who were admitted for surgery on the day of interven- tion, and samples were also collected on the day of hospitalization for the OHSS patients (n = 16). Serum levels of Na+, K+, LDH, ASAT, ALAT, and CA-125 tumor marker were determined. A hemogram was also performed. Measurement of cytokine levels in abdominal fluid Cytokine levels were measured by using the R&D Systems Human Premixed Multi-Analyte Kit Luminex Assay (Cat. no. LXSAH-6; R&D Systems, Minneapolis, MN, USA) and a Luminex 200 instrument (R&D Systems). Levels of IL-6, IL-8, IL-10, TNF-α, and VEGF were measured according to the manufacturer’s instructions. TGF-β levels were measured with the R&D Systems Magnetic Luminex Performance Assay and MAGPIX MILLIPLEX MAP in- strument (MilliporeSigma, Danvers, MA, USA) according to the manufacturer’s instructions. Samples above the standard curve were considered to be maximum value, Table 1 Clinical and histological diagnoses of the participants in the four study groups Study group OHSS Ovarian Cancer Histologic Diagnosis NA Benign adnexal mass Serous papillary adenocarcinoma Clear cell adenocarcinoma Adenosarcoma Adult granulosa cell ovarian tumor Borderline (atypical proliferation) Ovarian fibroma Follicular cyst Granulosa lutein cyst Adult type teratoma Borderline tumor (no atypical proliferation) Paraovarian cyst Endometriosis Endometrioma of the ovaries FIGO International Federation of Obstetrics and Gynecology FIGO Stage 3b-c NA NA 3b NA Grade High NA Low NA NA Number (n) 16 22 16 2 2 1 1 21 6 3 6 1 3 2 17 Farkas et al. Journal of Ovarian Research (2020) 13:25 Page 7 of 8 and samples under the curve sensitivity were annotated as 0. All data are displayed in pg/ml. Statistical analysis Statistical analyses were performed by using IBM SPSS Statistic 25 software (IBM Corporation) at the University of Pecs, Institute of Bioanalysis (performed by NF). The total sample size (n) was 76. Comparisons were made between serum and cytokine levels detected for the four study groups according to the non-parametric Kruskal- Wallis test with Bonferroni post hoc test. To examine the relationship between cytokine levels and serum pa- rameters, Spearman’s rank correlation coefficient was ap- plied. Mean data are reported ± standard deviation (SD). Statistical significance was set at p < 0.05, or *p < 0.01. Abbreviations ALAT: Alanine Aminotransferase; ASAT: Aspartate transaminase; BAM: Benign adnexal mass; EM: Endometriosis; Hgb: Hemoglobin; Htc: Hematocrit, packed cell volume; IL: Interleukin; IVF: In vitro fertilization; LDH: Lactate dehydrogenase; OC: Ovarian cancer; OHSS: Ovarian hyperstimulation syndrome; OIT: Ovulation induction therapy; RBC: Red blood cell; SD: Standard deviation; TCT: Thrombocyte; TGF-β: Transforming growth factor-beta; TNF-α: Tumor necrosis factor-alpha; VEGF: Vascular endothelial growth factor; WBC: White blood cell Acknowledgements We would like to express our gratitude to the medical staff of the University of Pecs, Department of Obstetrics and Gynecology, and the Reproduction Center. We especially thank Gabriella Boskovits, head of the OR nurses, for allowing us to collect samples and obtain data. We thank Agnes Kemeny PhD (Associate Professor at the Department of Pharmacology and Pharmacotherapy, University of Pecs, Medical School) for her help in the TGFbeta MagPix measurement. We would also like to thank Prof. Dr. Peter M. Gocze for his useful comments regarding our manuscript. Authors’ contributions BF collected samples, set up the study design, and wrote the manuscript; NB collected samples; FB and SM performed the Luminex assay; NF performed the statistical analyses; MK, GLK, and JB provided financial support and edited the manuscript. All of the authors have read and approved the final manuscript. Funding The current study was funded by grant, GINOP-2.3.2-15-2016-00021, „The use of chip-technology in increasing the effectiveness of human in vitro fertilization”. Open access funding provided by University of Pécs (PTE). Availability of data and materials The datasets generated and/or analyzed in the current study are not publicly available in order to prevent compromise of individuals’ privacy. However, the data are available from the corresponding author upon reasonable request. Ethics approval and consent to participate This prospective cohort study was approved by the University of Pecs Institutional Ethical Review Board (#5273–2/2012). All of the participating patients provided written informed consent. Consent for publication The current manuscript does not contain any individual person’s data in any form. Competing interests The corresponding author (BF) and two other authors (JB and GLK) have multiple affiliations and JB has received financial support from the Hungarian Academy of Sciences (HAS; Budapest, Hungary). The remaining authors have no conflicts of interest to report regarding the present study. Author details 1Department of Obstetrics and Gynecology, University of Pecs, School of Medicine, 17 Edesanyak Str., Pecs, Hungary. 2Member of the HAS-UP Human Reproduction Scientific Research Group, Hungarian Academy of Sciences (HAS), Pecs, Hungary. 3Department of Immunology and Biotechnology, University of Pecs, School of Medicine, Pecs, Hungary. 4School of Medicine, Institute of Bioanalysis, University of Pecs, Pecs, Hungary. 5Laboratory of Molecular and HLA Diagnostics, University Hospital Osijek, Clinical Institute of Transfusion Medicine, Osijek, Croatia. 6Department of Medical, Chemistry, Biochemistry and Clinical Chemistry, University of Osijek, Faculty of Medicine, Osijek, Croatia. 7Szentágothai Research Center, University of Pecs, Pecs, Hungary. 8Department of Laboratory Medicine, Faculty of Medicine, University of Pecs, Pecs, Hungary. Received: 11 November 2019 Accepted: 17 February 2020 References 1. Mascarenhas MN, Flaxman SR, Boerma T, Vanderpoel S, Stevens GA. 2. 5. 4. 6. 3. National, regional, and global trends in infertility prevalence since 1990: a systematic analysis of 277 health surveys. PLoS Med. 2012;9:e1001356. Lotti F, Maggi M. "ultrasound of the male genital tract in relation to male reproductive health" (PDF). Hum Reprod Update. 2014;21:56–83. Royal College of Obstetricians and Gynaecologists. The management of ovarian hyperstimulation syndrome. Green-top Guideline. 2006:5 Available at: http://www.rcog.org.uk/resources/Public/pdf/green_top_5_ management_ohss_a.pdf. Accessed January 20 2009. Gomez R, Soares SR, Busso C, Garcia-Velasco JA, Simon C, Pellicer A. Physiology and pathology of ovarian hyperstimulation syndrome. Semin Reprod Med. 2010;28:448–57. Várnagy A, Bódis J, Mánfai Z, Wilhelm F, Busznyák C, Koppán M. Low-dose aspirin therapy to prevent ovarian hyperstimulation syndrome. Fertil Steril. 2010;93:2281–4. Sanguinete MMM, Oliveira PH, Martins-Filho A, Micheli DC, Tavares-Murta BM, Murta EFC, Nomelini RS. Serum IL-6 and IL-8 correlate with prognostic factors in ovarian Cancer. Immunol Investig. 2017;46:677–88. Rodrigues ISS, Martins-Filho A, Micheli DC, Lima CA, Tavares-Murta BM, Murta EFC, Nomelini RS. IL-6 and IL-8 as prognostic factors in peritoneal fluid of ovarian Cancer. Immunol Investig. 2019;22:1–12. Nastri CO, Ferriani RA, Rocha IA, Martins WP. Ovarian hyperstimulation syndrome: pathophysiology and prevention. J Assist Reprod Genet. 2010;27: 121–8. Nelson SM. Prevention and management of ovarian hyperstimulation syndrome. Thromb Res. 2017;151(Suppl 1):S61–4. Sergentanis TN, Diamantaras AA, Perlepe C, Kanavidis P, Skalkidou A, Petridou ET. IVF and breast cancer: a systematic review and meta-analysis. Hum ReprodUpdate. 2014;20:106–23. Taheripanah R, Balash F, Anbiaee R, Mahmoodi M, Akbari Sene A. Breast cancer and ovulation induction treatments. Clin Breast Cancer. 2018;18:395–9. 12. Zhao J, Li Y, Zhang Q, Wang Y. Does ovarian stimulation for IVF increase 10. 11. 9. 7. 8. gynaecological cancer risk? A systematic review and meta-analysis. Reprod Biomed Online. 2015;31:20–9. 13. Hatzipetros I, Gocze PM, Cziraky K, Kovacs K, Kalman E, Farkas B. Assessment of cells in the ascitic fluid of women with ovarian hyperstimulation syndrome: the clinical implications for subsequent ovarian malignancy. Reprod Biol Endocrinol. 2013;11:91–5. 14. Mathur RS, Jenkins JM, Bansal AS. The possible role of the immune system in the aetiopathogenesis of ovarian hyperstimulation syndrome. Hum Reprod. 1997;12:2629–34. 15. Aboulghar MA, Mansour RT, Serour GI, El Helw BA, Shaarawy M. Elevated levels of interleukin-2, soluble interleukin-2 receptor alpha, interleukin-6, soluble interleukin-6 receptor and vascular endothelial growth factor in serum and ascitic fluid of patients with severe ovarian hyperstimulation syndrome. Eur J Obstet Gynecol Reprod Biol. 1999;87:81–5. 16. Andus T, Gross V, Holstege A, Weber M, Ott M, Gerok W, Schölmerich J. Evidence for the production of high amounts of interleukin-6 in the peritoneal cavity of patients with ascites. J Hepatol. 1992;15:378–81. 17. Hoeben A, Landuyt B, Highley MS, Wildiers H, Van Oosterom AT, De Bruijn EA. Vascular endothelial growth factor and angiogenesis. Pharmacol Rev. 2004;56:549–80. Farkas et al. Journal of Ovarian Research (2020) 13:25 Page 8 of 8 18. Bates DO. Vascular endothelial growth factors and vascular permeability. 19. Cardiovasc Res. 2010;87:262–71. Elchalal U, Schenker JG. The pathophysiology of ovarian hyperstimulation syndrome--views and ideas. Hum Reprod. 1997;12:1129–37. 20. Namavar Jahromi B, Parsanezhad ME, Shomali Z, Bakhshai P, Alborzi M, Moin Vaziri N, Anvar Z. Ovarian hyperstimulation syndrome: a narrative review of its pathophysiology, risk factors, prevention, classification, and management. Iran J Med Sci. 2018;43:248–60. Krasnow JJ, Berga SL, Guzick DS, et al. Vascular permeability factor and vascular endothelial growth factor in ovarian hyperstimulation syndrome. Fertil Steril. 1996;65:552–5. 21. 22. Abramov Y, Barak V, Nisman B, Schenker JG. Vascular endothelial growth factor plasma levels correlate to the clinical picture in severe ovarian hyperstimulation syndrome. Fertil Steril. 1997;67:261–5. 23. Dalal V, Kumar R, Kumar S, Sharma A, Kumar L, Sharma JB, Roy KK, Singh N, Vanamail P. Biomarker potential of IL-6 and VEGF-A in ascitic fluid of epithelial ovarian cancer patients. Clin Chim Acta. 2018;482:27–32. 24. Chen CD, Wu MY, Chen HF, Chen SU, Ho HN, Yang YS. Prognostic importance of serial cytokine changes in ascites and pleural effusion in women with severe ovarian hyperstimulation syndrome. Fertil Steril. 1999; 72:286–92. Köhidai L, Csaba G. Chemotaxis and chemotactic selection induced with cytokines (IL-8, RANTES and TNF-alpha) in the unicellular Tetrahymena pyriformis. Cytokine. 1998;10:481–6. 25. 26. Revel A, Barak V, Lavy Y, Anteby E, Abramov Y, Schenker JJ, Amit A, Finci- Yeheskel Z, Mayer M, Simon A, Laufer N, Hurwitz A. Characterization of intraperitoneal cytokines and nitrites in women with severe ovarian hyperstimulation syndrome. Fertil Steril. 1996;66:66–71. 27. Manolopoulos K, Lang U, Gips H, Braems GA. Elevated interleukin-10 and sex steroid levels in peritoneal fluid of patients with ovarian hyperstimulation syndrome. Eur J Obstet Gynecol Reprod Biol. 2001;99:226–31. 28. Brat DJ, Bellail AC, Van Meir EG. The role of interleukin-8 and its receptors in gliomagenesis and tumoral angiogenesis. Neuro-Oncology. 2005;7:122–33. 29. Brencicova E, Jagger AL, Evans HG, Georgouli M, Laios A, Attard Montalto S, Mehra G, Spencer J, Ahmed AA, Raju-Kankipati S, Taams LS, Diebold SS. Interleukin-10 and prostaglandin E2 have complementary but distinct suppressive effects on toll-like receptor-mediated dendritic cell activation in ovarian carcinoma. PLoS One. 2017;12:e0175712. Friedlander MA. Loret de Mola JR, Goldfarb JM: elevated levels of interleukin-6 in ascites and serum from women with ovarian hyperstimulation syndrome. Fertil Steril. 1993;60:826–33. 30. 31. Massagué J. TGFβ signalling in context. Nat Rev Mol Cell Biol. 2012;13:616–30. Santin AD, Bellone S, Ravaggi A, Roman J, Smith CV, Pecorelli S, Cannon MJ, 32. Parham GP. Increased levels of interleukin-10 and transforming growth factor-beta in the plasma and ascitic fluid of patients with advanced ovarian cancer. BJOG. 2001;108:804–8. Lisa C. Grossman, Konstantinos G. Michalakis, hyacinth Browne, mark D. Payson, James H. Segars: the pathophysiology of ovarian hyperstimulation syndrome: an unrecognized compartment syndrome. Fertil Steril. Author manuscript; available in PMC 2011 Jun 27. Fertil Steril. 2010;94:1392–8. 34. Bódis J, Papp S, Vermes I, Sulyok E, Tamás P, Farkas B, Zámbó K, Hatzipetros 33. I, Kovács GL. "Platelet-associated regulatory system (PARS)" with particular reference to female reproduction. J Ovarian Res. 2014;16(7):55–60. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s13054-020-03210-z	Combes et al. Critical Care (2020) 24:490 https://doi.org/10.1186/s13054-020-03210-z R E S E A R C H Open Access ECCO2R therapy in the ICU: consensus of a European round table meeting Alain Combes1,2* Guglielmo Consales9†, Wojciech Dabrowski10†, David De Bels11†, Francisco Javier González de Molina Ortiz12,13†, Antje Gottschalk14†, Matthias P. Hilty15†, David Pestaña16,17†, Eduardo Sousa18†, Redmond Tully19†, Jacques Goldstein20 and Kai Harenski21 , Georg Auzinger3,4†, Gilles Capellier5,6†, Damien du Cheyron7†, Ian Clement8†, Abstract Background: With recent advances in technology, patients with acute respiratory distress syndrome (ARDS) and severe acute exacerbations of chronic obstructive pulmonary disease (ae-COPD) could benefit from extracorporeal CO2 removal (ECCO2R). However, current evidence in these indications is limited. A European ECCO2R Expert Round Table Meeting was convened to further explore the potential for this treatment approach. Methods: A modified Delphi-based method was used to collate European experts’ views to better understand how ECCO2R therapy is applied, identify how patients are selected and how treatment decisions are made, as well as to identify any points of consensus. (Continued on next page) * Correspondence: alain.combes@aphp.fr †Georg Auzinger, Gilles Capellier, Damien du Cheyron, Ian Clement, Guglielmo Consales, Wojciech Dabrowski, David De Bels, Francisco Javier González de Molina Ortiz, Antje Gottschalk, Matthias P. Hilty, David Pestaña, Eduardo Sousa and Redmond Tully contributed equally to this work. 1Sorbonne Université, INSERM, UMRS_1166-ICAN, Institute of Cardiometabolism and Nutrition, 47, Boulevard de l’Hôpital, F-75013 Paris, France 2Service de Médecine Intensive-Réanimation, Institut de Cardiologie, APHP Hôpital Pitié–Salpêtrière, F-75013 Paris, France Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Combes et al. Critical Care (2020) 24:490 Page 2 of 10 (Continued from previous page) Results: Fourteen participants were selected based on known clinical expertise in critical care and in providing respiratory support with ECCO2R or extracorporeal membrane oxygenation. ARDS was considered the primary indication for ECCO2R therapy (n = 7), while 3 participants considered ae-COPD the primary indication. The group agreed that the primary treatment goal of ECCO2R therapy in patients with ARDS was to apply ultra-protective lung ventilation via managing CO2 levels. Driving pressure (≥ 14 cmH2O) followed by plateau pressure (Pplat; ≥ 25 cmH2O) was considered the most important criteria for ECCO2R initiation. Key treatment targets for patients with ARDS undergoing ECCO2R included pH (> 7.30), respiratory rate (< 25 or < 20 breaths/min), driving pressure (< 14 cmH2O) and Pplat (< 25 cmH2O). In ae-COPD, there was consensus that, in patients at risk of non-invasive ventilation (NIV) failure, no decrease in PaCO2 and no decrease in respiratory rate were key criteria for initiating ECCO2R therapy. Key treatment targets in ae-COPD were patient comfort, pH (> 7.30–7.35), respiratory rate (< 20–25 breaths/min), decrease of PaCO2 (by 10–20%), weaning from NIV, decrease in HCO3 haemodynamic stability. Consensus was reached on weaning protocols for both indications. Anticoagulation with intravenous unfractionated heparin was the strategy preferred by the group. Conclusions: Insights from this group of experienced physicians suggest that ECCO2R therapy may be an effective supportive treatment for adults with ARDS or ae-COPD. Further evidence from randomised clinical trials and/or high-quality prospective studies is needed to better guide decision making. − and maintaining Keywords: Acute respiratory distress syndrome, Chronic obstructive pulmonary disease, CO2 removal, Consensus, Driving pressure, ECCO2R, Gas exchange, Lung protective ventilation, Tidal volume, Therapy experience lung ventilation (UPLV) Background Advances in technology to deliver extracorporeal car- bon dioxide removal (ECCO2R) therapy have simpli- fied this approach, making it easier to deploy for the management of adults with both hypoxaemic and hypercapnic acute respiratory failure (ARF) [1–4]. In patients with acute respiratory distress syndrome (ARDS), ECCO2R therapy may be used to allow ultra- reduce protective ventilator-induced lung injury (VILI) by decreasing tidal volume (VT), both plateau (Pplat) and driving pressures and respiratory rate, while also controlling respiratory acidosis [5–14]. In patients with acute ex- acerbations of chronic obstructive pulmonary disease (ae-COPD) with severe respiratory acidosis and hyper- capnic respiratory failure, ECCO2R therapy may be applied to prevent intubation in patients at risk of non-invasive ventilation (NIV) failure [15]. It may also be used to hasten weaning from mechanical ventilation (MV) and early extubation in those who require inva- sive ventilation [10, 15–17]. and However, there is currently limited evidence regarding the use of ECCO2R therapy in these indications, with available data limited to the description of single cases or to case series that include a small number of patients [16, 18–21], as well as a few retrospective matched cohort studies [15, 22]. Additionally, questions remain on how best to implement a therapy that might be associated with serious side-effects [1]. Ongoing and published trials such as VENT-AVOID (NCT03255057), REST (NCT02654327) [2] and SUPERNOVA (NCT0228 2657) [11, 12, 23] are expected to provide valuable evidence to support decision making. Given the potential of ECCO2R therapy to provide effective supportive treatment for a wide range of patient groups, we convened a European ECCO2R therapy Expert Round Table Meeting to better understand how ECCO2R therapy is applied in key diagnostic groups, e.g. patients with ARDS or ae-COPD, identify how patients are selected, understand how treatment decisions are made and delineate areas of consensus in the group. Methods Research questions and objectives The ECCO2R therapy Expert Round Table Meeting was held in Brussels in July 2019 and was attended by 14 clinicians who regularly provide ECCO2R therapy in hospitals across Europe in order to provide a European perspective on ECCO2R therapy. Each attendee was a senior clinician/intensivist invited based on their experi- ence delivering ECCO2R therapy, with and without continuous renal replacement therapy, using different devices. The attendees had direct clinical experience with a wide range of ECCO2R devices, including ALung, iLA, Prismalung and PALP (the later had been removed from the market at the time of the meeting due to loss of the distribution agreement). In addition, several of the attendees are principal investigators in recently com- pleted or ongoing clinical trials, including randomised such as REST and SUPERNOVA. controlled trials Combes et al. Critical Care (2020) 24:490 Page 3 of 10 Conflict of interest declarations for the attendees can be found at the end of the manuscript. The meeting objectives were to better define and understand the application of ECCO2R therapy in key indications (ARDS and ae-COPD), to identify patient selection criteria and when to initiate and stop/wean patients from treatment and to determine points of consensus and differences in clinical practice in those centres represented at the meeting. A non-systematic search of MEDLINE, ClinicalTrials.gov and other sites was performed to identify key studies and trials to sup- port the development of the questions and the content of the meeting. and post-meeting Data collection and analysis A modified Delphi-based method (Fig. 1) was used to collate the clinicians’ views in three rounds of question- ing [24]. The meeting questions as well as the pre- meeting questionnaires were developed by JG and KH before being reviewed and approved by AC. JG and KH were present as Baxter employees and moderators, but were not permitted to provide answers or responses, either to the survey ques- tions or during the meeting. Round 1 data were collected via an interactive PDF questionnaire circulated in advance of the meeting, and results were analysed an- onymously. Round 2 data were collected during the meeting, attendees were divided into 4 subgroups and the questions were presented by an independent facilita- tor. Open questions were used to encourage freedom of response, and the meeting was designed to allow the attendees adequate time to consider and respond to the questions based on their experience. Attendees could respond to the questions either through anonymous electronic voting or by inputting responses into a micro- computer, with responses collected and discussed openly by the group. Round 3 was a second interactive PDF questionnaire, circulated post-meeting, designed to fol- low up on discussion points raised at the face-to-face meeting, with results analysed anonymously. Details on the process for information gathering and the questions are provided in Additional file 1. Target values for ventilation parameters of interest— criteria for initiation of ECCO2R therapy and treatment targets for ECCO2R therapy in both ARDS and ae- COPD—were collected during the three rounds of questioning. These values were subsequently evaluated for consensus. To facilitate the analysis of the responses for certain questions, a scoring system was employed. Participants were asked to score their responses in order of importance, giving them a score (e.g. from 1 to 8, de- pending on the number of variables). Scores were then combined to give a total score for each parameter, with higher scores indicating a higher perceived importance. To determine whether a consensus was reached or not based on participant responses to the questions, a threshold of ≥ 80% of participants in agreement was used to define if consensus was reached, a level that has been used in previous analyses [25]. Majority agreement indicates that ≥ 50% of participants agreed, but consen- sus level was not reached, and no agreement means that < 50% of participants agreed. The report was drafted by an independent medical writing company (SciMentum, Nucleus Global) and paid for by Baxter in line with Good Publication Practice 3. The various drafts were reviewed and approved by AC before being reviewed by the full author team. All authors provided their approval to submit and meet the ICMJE criteria for authorship. Results Attendee clinical experience Twelve clinicians completed the pre-meeting survey: eight worked in Combined Surgical and Medical inten- sive care units (ICUs), while the others were employed Fig. 1 Overview of the five-step Delphi method used in the Round Table Meeting. Each step was a distinct process that was completed before the following step was initiated. Results and discussions from each step were independently analysed and used to inform the direction and content of the following steps, e.g. if the group were split on a topic, then clarifying questions were crafted to guide the discussions in the following step(s) to identify and explore points of consensus or difference. GPP3, Good Publication Practice 3 Combes et al. Critical Care (2020) 24:490 Page 4 of 10 in Medical ICUs (n = 2), Surgical ICUs (n = 2) and Cardiac Surgery ICUs (n = 2); respondents could be employed at more than one type of centre. ICUs had a median of 20 beds/unit and 400–2000 admissions/year. Extracorporeal membrane oxygenation (ECMO) experi- ence of participants ranged from 0 to 80 veno-venous ECMO procedures/year and 0 to 220 veno-arterial ECMO procedures/year. Indications and rationale for ECCO2R based on pre- meeting survey Analysis of the Round 1 pre-meeting survey responses revealed that ARDS was considered the primary indica- tion for ECCO2R therapy by 7 participants, while 3 participants considered ae-COPD to be the primary indi- cation. Severe asthma was also mentioned as another potential ECCO2R indication, although less frequently. The median number of ARDS admissions (as per the Berlin definition [26]) was 60 patients per centre per year, with some centres admitting up to 500 patients per year. While the most common criteria stated in the pre- meeting responses for initiating ECCO2R therapy in patients with ARDS were to manage hypercapnia with acidosis, although specific criteria varied across the ICUs, likewise, weaning criteria shared at Round 1 varied significantly, with no clearly consistent management pattern being identified between centres. However, most participants they would place indicated that patients with ARDS in the prone position when using ECCO2R therapy. The number of ae-COPD admissions ranged from 0 to 250 patients per centre per year (median 50). Participants indicated that ECCO2R therapy was predominantly initiated to prevent intub- ation in patients at risk of NIV failure or to facilitate extubation in patients who had been intubated after NIV failure. (92%) Use of ECCO2R therapy in patients with ARDS During the Expert Round Table Meeting and post- meeting survey (Rounds 2 and 3, respectively), the group considered the ventilation parameters for implementa- tion of a lung protective ventilation (LPV) strategy in all patients with ARDS and agreed upon the following targets: driving pressure, 10–14 cmH2O; positive end- expiratory pressure (PEEP), 10–14 cmH2O; Pplat, 25– 29 cmH2O; and respiratory rate either 20–25 or 25–30 breaths/min, although most of the group would target a respiratory rate of 25 breaths/min. There was some vari- ation in responses among the group when asked about target pH, with half of participants opting for a target pH value of 7.25–7.30, while others indicated the target should be > 7.30 (n = 4), < 7.25–7.30 (n = 2) or < 7.20 (n = 1). Finally, the panellists thought VT should be set at 6.0 mL/kg of predicted body weight (PBW), although 6.1–7.0 or 7.1–8.0 mL/kg PBW were also considered to be reasonable targets. When asked in the post-meeting survey (Round 3) about the preferred ventilation mode used for patients with ARDS undergoing LPV, the group were split with respect to pressure control (pressure assist) (n = 8) and flow control (volume assist) (n = 6) modes of ventilation. These recommendations agreed with the most recent guidelines for the ventilation man- agement of patients with ARDS [27, 28]. that voted in favour]) There was consensus among the group (91% [2 participants were unavailable for this question, 11 of n = 12/14 the primary treatment goal of ECCO2R therapy for patients with ARDS was to apply UPLV via managing CO2 levels. For initiating ECCO2R therapy in patients with ARDS, driving pressure (≥ 14 cmH2O) followed by Pplat (≥ 25 cmH2O) was important criteria, and this was confirmed in the post-meeting survey (Tables 1 and 2). Additional key parameters included pH (< 7.25), reducing VT to < 6 mL/kg PBW, PaCO2 (> 60–80 mmHg), respiratory rate (≥ 25 to > (100–200) and PEEP 30 breaths/min), PaO2/FiO2 (combined findings from Rounds 2 and 3). considered the most i.e. Participants were evenly split during the meeting on the primary rationale for ECCO2R therapy, being rescue therapy in patients with ARDS undergoing injurious MV, those with very high plateau and driving pressures despite reduced VT and PEEP (n = 7), or to facilitate UPLV to prevent the deleterious effects of MV in patients already undergoing LPV (n = 7). Based on the results of the post-meeting survey, a consensus was reached among the group (12/14, 86% of participants) that ECCO2R was a strategy they would consider select- ing for rescue in patients with ARDS. Typical character- initiating ECCO2R in a rescue situation istics obtained as part of the post-meeting survey are summarised in Table 2. A majority (10/14, 71% of participants) indicated that they would select ECCO2R as a means of facilitating UPLV for patients with ARDS, and typical characteristics for selecting patients are summarised in Table 2. for For both potential indications, patients would not be considered suitable for an ECCO2R strategy if they met the indications for ECMO, such as severe or refractory ARDS [29] and presence of severe right heart failure (ECMO may be a more adequate treatment for these pa- tients), in cases where anticoagulation is contraindicated and for those with major comorbidities and/or predicted survival of < 1 year. The group considered treatment targets for their patients with ARDS undergoing ECCO2R. A consensus was reached regarding driving pressure (< 14 cmH2O) and respiratory rate (< 25 or < 20 breaths/min). There was majority agreement with respect to targets for Pplat Combes et al. Critical Care (2020) 24:490 Page 5 of 10 Table 1 ECCO2R treatment criteria for patients with ARDS Parameter Target Initiation criteria Driving pressure Pplat PaCO2 pH Reduce VT to < 6 mL/PBW Respiratory rate PaO2/FiO2 PEEP Treatment targets Driving pressure Pplat Respiratory rate ≥ 14 cmH2O ≥ 25 cmH2O > 60–80 mmHg < 7.25 – ≥ 25 to > 30 100–200 – < 14 cmH2O < 25 cmH2O < 25 or < 20 breaths/min pH VT PaCO2 > 7.30 ≤6 mL/PBW < 50–55 mmHg Score 31 22 21 20 18 14 10 8 66* 57* 44* 39* 39* Consensus Consensus Majority agreement Majority agreement Majority agreement Majority agreement Majority agreement No agreement Consensus Majority agreement† Consensus Majority agreement Majority agreement Majority agreement Criteria for ECCO2R treatment considered to be of importance and selected from the provided list. Target describes any potential target values identified, with ‘–’ indicating that no target parameter was provided or considered relevant. Score indicates the combined total score, with higher scores indicating a higher perceived importance. Consensus means a consensus threshold (≥ 80%) was reached, majority agreement means ≥ 50% agreed but consensus level was not reached, and no agreement means < 50% agreed Based on the post-meeting survey. †Note, for Pplat, a consensus threshold of 80% was not reached in the meeting; in the post-meeting survey, it was rated as the second most important target 30 (< 25 cmH2O), pH (> 7.30 [Rounds 2 and 3]), PaCO2 (< 50 or < 55 mmHg) and VT (≤ 6 mL/kg PBW). Other tar- get parameters were not proposed by the group (Table 1). The expected average length of time patients with ARDS would remain on ECCO2R therapy was sug- gested to be 1–3 days (n = 5) and 4–6 days (n = 9). Following discussion during the meeting on a protocol for weaning from ECCO2R in patients with ARDS, a protocol was proposed and reviewed as part of the post-meeting survey (Table 3). The group voted on each step and reached consensus (92% of participants, n = 13) that this proposal was a suitable weaning strategy. Table 2 Typical characteristics for initiating ECCO2R for rescue therapy and to facilitate ultra-protective ventilation in ARDS Parameter Target for initiation in: Rescue Target for initiation in: Ultra-protective ventilation Driving pressure Pplat PaCO2 pH > 15 to 20 cmH2O > 30 to 35 cmH2O ≥ 60 mmHg < 7.25–7.30 > 13 to 15 cmH2O ≥ 25 cmH2O ≥ 60 mmHg < 7.25–7.30 Respiratory rate > 20 to 30 breaths/min > 20 breaths/min PaO2/FiO2 PEEP < 150 > 8 to 15 < 150 ≥ 8 Responses were captured during the post-meeting survey (Round 3) and general themes were identified Use of ECCO2R therapy in patients with ae-COPD There was consensus during the meeting that patients with ae-COPD who should receive ECCO2R therapy were those at risk of NIV failure, as well as patients re- cently initiated on MV after NIV failure to allow for early extubation within 24 h of initiating ECCO2R therapy. Other patient groups would be considered (e.g. patients on prolonged MV who require weaning from invasive ventilation and patients who are refusing intub- ation), but a consensus was not reached. The group agreed that for patients with ae-COPD at ‘no decrease in PaCO2’ and ‘no risk of NIV failure, Table 3 ECCO2R weaning protocol for patients with ARDS Weaning criteria and steps for weaning for ECCO2R in ARDS ECCO2R will be applied for at least 48 h PaO2/FiO2 > 200 mmHg before testing weaning possibility Set VT at 6 mL/PBW and PEEP 5–10 cmH2O Driving pressure should be < 14 cmH2O Respiratory rate should be 20–30 breaths/min Reduce gas flow to zero, using 2 L/min decremental steps While weaning, pH should remain > 7.30 and respiratory rate < 25 breaths/min Patient will be weaned off ECCO2R therapy after a minimum of 12 h of stability under these settings (including pH > 7.30 and respiratory rate < 25 breaths/min) *A consensus was reached for all of these criteria and steps Combes et al. Critical Care (2020) 24:490 Page 6 of 10 decrease in respiratory rate’ while on NIV were both key initiation criteria for ECCO2R therapy (Table 4). These criteria were considered indicative of NIV failure. Clinical signs of respiratory failure and pH (< 7.25 [n = 5] or 7.25–7.30 [n = 6]) would be considered as initiation criteria by most of the participants. Baseline PaCO2 and respiratory rate as main triggers were favoured by less than half of participants. For patients with ae-COPD who had already been intubated, criteria for initiating ECCO2R therapy varied (Table 4). Factors for excluding patients with ae-COPD from ECCO2R typically included patients with end-stage disease (the group highlighted that markers for this in- clude severe functional limitation and cachexia); contra- indications to anticoagulation; problems with vascular access; patient’s wishes, e.g. refusal to be intubated, ex- cept in cases where ECCO2R therapy represented the last resource accepted by the patient; poor quality of life; and the patient not being a candidate for MV. Treatment targets for patients with ae-COPD receiving ECCO2R therapy were, in order of perceived importance (Table 5), comfortable patient, pH (> 7.35/7.30; no Table 4 ECCO2R treatment initiation criteria for patients with ae-COPD Initiation criteria for patients at risk of NIV failure Parameter No decrease in PaCO2 while on NIV Consensus No decrease in respiratory rate while on NIV Consensus Clinical signs of respiratory failure pH 7.25–7.30 Baseline PaCO2 Baseline respiratory rate Majority agreement Majority agreement No agreement No agreement Initiation criteria for patients who are already intubated - Patients who look like they will not be extubated early without ECCO2R ○ Previous intubation for ae-COPD ○ Has failed a spontaneous breathing trial due to increased dyspnoea ○ Reintubation after first extubation attempt despite NIV ○ Patients with severe bronchospasm who are difficult/impossible to ventilate adequately or otherwise not responding to medical treatment ○ Patients who remain hypercapnic and not improving with MV - No hypoxemia preventing extubation - MV < 72 h - Patients with home NIV and good quality of life Criteria for ECCO2R treatment considered to be of importance and selected from the provided list. Target describes any potential target values identified. Consensus means a consensus threshold (≥ 80%) was reached, majority agreement means ≥ 50% agreed but consensus level was not reached, and no agreement means < 50% agreed Scoring and ranking was not conducted for this section during the meeting consensus on specific pH), respiratory rate (< 20–25 breaths/min), decrease of PaCO2 by 10–20%, weaning from NIV, decrease in HCO3 and maintaining haemo- dynamic stability. Consensus on a weaning protocol for patients with ae-COPD was reached during the meeting (Table 5). − Anticoagulation strategy for patients receiving ECCO2R Responses obtained during Round 1 (pre-meeting sur- vey) showed that heparin was the preferred choice of anticoagulant used during ECCO2R therapy (~ 80% of participants stated that heparin was their anticoagulant of choice). This was confirmed in the post-meeting survey, the anticoagulant of choice for the majority (~ 90% of partic- ipants). The proposed heparin anticoagulation protocol agreed by the group is shown in Table 6. Lastly, argatro- ban was the group’s preferred anticoagulant in case of proven heparin-induced thrombocytopenia (HIT). in which unfractionated heparin was Discussion The responses obtained from the Expert Round Table Meeting and accompanying pre- and post-meeting sur- veys have provided further insights into the use of ECCO2R therapy across Europe. During a typical Delphi process [24], 100% agreement is rare, and any consensus is the result of multiple rounds of voting and discussion that lead to a convergence of opinion. However, in areas where clinical evidence is limited, as is the case for ECCO2R therapy in patients with ARDS and ae-COPD, using a modified Delphi method may offer insight into the current practice of experienced users, which could help inform decision making in local clinical practice. Additionally, the use of the Delphi method to guide these discussions and reach points of consensus will be of potential benefit for the design of future trials. Specif- ically, the discussions provide insight relevant to inclu- sion criteria, guidance on the management of patients while receiving ECCO2R therapy and possible primary and secondary endpoints. Key areas of consensus for the use of ECCO2R therapy in the treatment of patients with ARDS or ae-COPD were identified. There was consensus among the group that the primary treatment goal of ECCO2R therapy for patients with ARDS was to apply UPLV via managing CO2 levels; this is in agreement with the findings of a systematic literature review [30]. The group reached a consensus that, when initiating ECCO2R therapy in pa- tients with ARDS, driving pressure (≥ 14 cmH2O) followed by Pplat (≥ 25 cmH2O) was the most important criteria to consider. Higher PEEP, lower peak and plat- eau pressures and lower respiratory rate have been shown to correlate with improved survival in patients with ARDS [7, 11, 31]. However, only the driving Combes et al. Critical Care (2020) 24:490 Page 7 of 10 Table 5 ECCO2R treatment targets and weaning protocol for patients with ae-COPD Treatment targets for patients with ae-COPD Parameter Comfortable patient pH Respiratory rate Decrease of PaCO2 by 10–20% Weaning from NIV − Decrease in HCO3 Maintaining haemodynamic stability Target – > 7.35/7.30, no consensus on specific pH < 20–25 breaths/min – – – – Score 27 23 19 18 9 9 7 ECCO2R weaning protocol for patients with ae-COPD 1. Patient weaned from NIV for > 6 h a. Excluding patients on home NIV or candidates for long-term NIV 2. Intubated patients weaned from MV for > 6 h 3. SpO2 ≥ 88% with supplemental O2 if needed 4. Reduce sweep gas flow rate by 1–3 L/min; check arterial blood gas after 1 h for: a. pH ≥7.35 with respiratory rate < 25 breaths/min b. PaO2 > 55 mmHg c. SpO2 > 88% d. FiO2 < 40% 5. Repeat sweep gas reduction until zero gas flow reached, while arterial blood gas targets maintained 6. Remove ECCO2R after 6 h of stability of the aforementioned criteria Treatment targets for ECCO2R considered to be of importance and selected from the provided list. Target describes any potential target values identified. Score indicates the combined total score, with higher scores indicating a higher perceived importance. Consensus means a consensus threshold (≥ 80%) was reached, majority agreement means ≥ 50% agreed but consensus level was not reached, and no agreement means < 50% agreed. The ECCO2R weaning protocol for patients with ae-COPD was developed and voted on during the meeting, with all attendees in agreement pressure was associated with increased mortality using a multilevel mediation analysis in a large retrospective co- hort study of patients with ARDS [32]. It is therefore perhaps not surprising that the key treatment targets for ECCO2R in ARDS identified by the group were reduc- tions in driving pressure and respiratory rate. A pH of < 7.25 was also considered by most of the group to be a criterion for initiation of ECCO2R therapy Table 6 Heparin anticoagulation strategy 1. Anticoagulation with intravenous unfractionated heparin, preferably applied to the extracorporeal circuit 2. Monitor aPTT or anti-Xa or both a. To obtain an aPTT of 1.5–2.0 times normal baseline (45–70 s), or anti-Xa activity of 0.3–0.5 UI/mL 3. Initial bolus of heparin a. 40–80 units/kg PBW b. Bolus will not be performed in patients already on full anticoagulation c. Bolus routinely performed when guidewires have been inserted/or after catheter insertion 4. Patients with proven HIT-2 a. Argatroban protocol, e.g. 0.5–2.0 μg/kg/min independently in this patient group. Indeed, a lower pH was recently shown to be associated with ICU mortality in the large prospective LUNG SAFE registry [31]. Most of the group also agreed that ECCO2R should be initiated at PaCO2 levels > 60–80 mmHg. While it was suggested that permissive hypercapnia provided protection against lung injury in terms of lung perme- ability, oxygenation and lung mechanics [33], more recent data have shown a positive correlation between hypercapnic acidosis and mortality [34, 35]. Raising pH (> 7.30 or > 7.25) and decreasing PaCO2 levels were considered important treatment targets, indicating that there is a perception that ECCO2R is an important therapy for the management of respiratory acidosis. The experts were evenly split on the primary rationale for ECCO2R therapy, either as a rescue therapy in patients with ARDS undergoing injurious MV, or to facilitate UPLV to prevent VILI. The results from the post-meeting survey highlighted that the group agreed that they would at least consider selecting ECCO2R as a strategy in both settings. Ongoing (NCT02654327) [11] randomised trials may help clarify the role of ECCO2R, allowing UPLV in patients with acute hypoxemic respiratory failure. Combes et al. Critical Care (2020) 24:490 Page 8 of 10 To the best of our knowledge, this is the first publica- tion of a proposed weaning strategy for ECCO2R in pa- tients with ARDS. The group reached a consensus regarding a strategy for weaning patients from ECCO2R in this setting. It was agreed that ECCO2R therapy should be applied for at least 48 h in patients with ARDS, and that a test for PaO2/FiO2 > 200 mmHg while maintaining a driving pressure < 14 cmH2O should be carried out to determine weaning possibility. It was also agreed that patients should be stable for a minimum of 12 h at the ventilation parameters outlined (see Table 3) before any weaning attempt takes place [11]. In a randomised study exploring the role of helium/oxy- gen in ae-COPD, the rate of patients failing on NIV and requiring MV was 15% [36]. Identifying the subgroup of patients with ae-COPD at high risk of NIV failure is in- deed crucial to improve their outcomes by deploying ef- fective preventive strategies. The panel identified ‘lack of decrease in PaCO2’ and ‘respiratory rate during NIV’ as important indicators of increased risk of NIV failure and an indication for ECCO2R initiation. The group also felt that it was important to allow enough time to show that NIV was ineffective before initiating ECCO2R therapy. Furthermore, there are numerous factors involved in NIV failure, and the benefit of ECCO2R for this patient group is still a matter of debate due to lack of data from rando- mised clinical trials [15, 22]. For patients with ae-COPD who are already intubated, the intended use of ECCO2R therapy is to rapidly allow extubation, to facilitate oral nutrition and early physio- therapy and to prevent muscle deconditioning [3]. Treatment targets identified by the group clearly fit in with the strategy of reducing the duration of MV and are in line with published data and wider views on the use of ECCO2R therapy [1, 19]. The VENT-AVOID trial (NCT03255057) is currently randomising patients to further investigate the benefits of ECCO2R therapy in patients at risk of NIV failure or who already have been intubated after NIV failure. Anticoagulation with intravenous unfractionated hep- arin was the preferred strategy of the group. This reflects recent studies in the literature in which unfractionated heparin appears to be the anticoagulant most frequently used in this setting [10, 11]. The post-meeting survey highlighted that anticoagulant activity should be moni- tored using activated partial thromboplastin time (aPTT) and/or approach remains dependent on local practice. For patients with proven HIT, argatroban was the group’s preferred anticoagulant [37, 38]. the monitoring anti-Xa; Limitations The findings presented here relate to the experiences of a relatively small number of physicians from centres across Europe; evidence from a larger group of intensi- vists from multiple regions of the world may be required to support these observations. Certain topics were not covered due to the scope of the meeting. Firstly, the questions covered current practice and did not explore if practices, e.g. inclusion policies of the respective centres, had changed over time. Secondly, certain rarer indica- lung transplant, were not covered, as the tions, e.g. meeting focussed on the broader population of patients requiring ECCO2R therapy, e.g. patients with ARDS or ae-COPD. These questions could be covered as part of a follow-up meeting. Additionally, while the authors took every opportunity to ensure all relevant major articles were cited, the purpose of the meeting was to under- stand current practice as opposed to conducting a com- prehensive literature analysis. Finally, the experiences outlined are the physicians’ respective personal experi- ences and are not a replacement for formal guidelines. The reader should consider their patients’ needs and local guidelines when performing ECCO2R therapy. Conclusions The insights from this group of experienced physicians suggested that ECCO2R therapy may be a useful and ef- fective supportive treatment for adults in the ICU with both ARDS and ae-COPD. They have however highlighted an urgent need for further evidence in the form of randomised clinical trials and/or high-quality prospective studies to help guide decision making. On- going and published trials such as VENT-AVOID (NCT03255057), REST (NCT02654327) [2] and SUPER- NOVA (NCT02282657) [11, 12, 23] should provide the data to support these guidelines. Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s13054-020-03210-z. Additional file 1: Expanded methods. Details on the process for information gathering and the questions. Abbreviations ae-COPD: Acute exacerbations of chronic obstructive pulmonary disease; aPTT: Activated partial thromboplastin time; ARDS: Acute respiratory distress syndrome; ARF: Acute respiratory failure; ECCO2R: Extracorporeal carbon dioxide removal; ECMO: Extracorporeal membrane oxygenation; FiO2: Fraction of inspired oxygen; HIT: Heparin-induced thrombocytopenia; ICU: Intensive care unit; LPV: Lung protective ventilation; MV: Mechanical ventilation; NIV: Non-invasive ventilation; PaCO2: Partial pressure of carbon dioxide; PBW: Predicted body weight; PEEP: Positive end-expiratory pressure; Pplat: Plateau pressure; SpO2: Oxygen saturation; UPLV: Ultra-protective lung ventilation; VILI: Ventilator-induced lung injury; VT: Tidal volume Acknowledgements Supported by Baxter Inc. Medical writing support was provided by Daniel Johnson and Ruth Brown of SciMentum Inc. (Nucleus Global), funded by Baxter Inc., under the authors’ conceptual direction and based on feedback from the authors. Combes et al. Critical Care (2020) 24:490 Page 9 of 10 Authors’ contributions All authors participated in the discussions in the Round Table Meeting as well as the questionnaire rounds, contributed data, critically reviewed the manuscript providing interpretation of the data and their implications and provided approval for the final version to be published. Authors’ information FJGDM: Member of the Acute Respiratory Failure Working Group and Chair of the Nephrological Intensive Care Working Group of the Spanish Society of Critical and Intensive Care Medicine and Coronary Units (Sociedad Española de Medicina Intensiva, Críticos y Unidades Coronarias [SEMICYUC]). Funding The meeting and associated expenses as well as the publication of this research was supported by Baxter Inc. Availability of data and materials Not applicable. Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests All authors received financial compensation from Baxter for travel and accommodation for attending the meeting. MPH has received financial compensation from Baxter for travel and accommodation in the past. Author details 1Sorbonne Université, INSERM, UMRS_1166-ICAN, Institute of Cardiometabolism and Nutrition, 47, Boulevard de l’Hôpital, F-75013 Paris, France. 2Service de Médecine Intensive-Réanimation, Institut de Cardiologie, APHP Hôpital Pitié–Salpêtrière, F-75013 Paris, France. 3Department of Critical Care, King’s College Hospital, London SE5 9RS, UK. 4Department of Critical Care, Cleveland Clinic, London SW1Y 7AW, UK. 5Service de Médecine Intensive-Réanimation CHRU Besançon, EA 3920 University of Franche Comte, Besançon, France. 6Australian and New Zealand Intensive Care Research Centre, Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia. 7Service de Médecine Intensive-Réanimation, Caen University Hospital, 14000 Caen, France. 8Critical Care Unit, Royal Victoria Infirmary, Newcastle upon Tyne NE1 4LP, UK. 9Department Emergency and Critical Care, Prato Hospital, Azienda Toscana Centro, Prato, Italy. 10Department of Anaesthesiology and Intensive Care, Medical University of Lublin, Jaczewskiego Street 8, 20-954 Lublin, Poland. 11Service des Soins Intensifs Médico-chirurgicaux, CHU Brugmann, 4 Place A Van Gehuchten, 1020 Brussels, Belgium. 12Department of Critical Care, University Hospital Mútua Terrassa, Universitat de Barcelona, Terrassa, Barcelona, Spain. 13Department of Critical Care, University Hospital Quirón Dexeus, Universitat Autònoma de Barcelona, Barcelona, Spain. 14Department of Anaesthesiology, Intensive Care Medicine and Pain Medicine, University Hospital Münster, Münster, Germany. 15Institute of Intensive Care Medicine, University Hospital of Zürich, Rämistrasse 100, 8091 Zürich, Switzerland. 16Department of Anesthesiology and Surgical Critical Care, Hospital Universitario Ramón y Cajal, IRYCIS, Carretera de Colmenar Viejo km 9, 28034 Madrid, Spain. 17Universidad de Alcalá de Henares, Madrid, Spain. 18Serviço de Medicina Intensiva, Centro Hospitalar e Universitário de Coimbra, Praceta Mota Pinto, 3000-075 Coimbra, Portugal. 19Department of Intensive Care, Royal Oldham Hospital, Northern Care Alliance, Oldham OL1 2JH, UK. 20Baxter World Trade SPRL, Acute Therapies Global, Braine-l’Alleud, Belgium. 21Baxter, Baxter Deutschland GmbH, Unterschleissheim, Germany. Received: 1 May 2020 Accepted: 28 July 2020 References 1. Boyle AJ, Sklar MC, McNamee JJ, Brodie D, Slutsky AS, Brochard L, et al. Extracorporeal carbon dioxide removal for lowering the risk of mechanical ventilation: research questions and clinical potential for the future. Lancet Respir Med. 2018;6:874–84. 2. McNamee JJ, Gillies MA, Barrett NA, Agus AM, Beale R, Bentley A, et al. pRotective vEntilation with veno-venouS lung assisT in respiratory failure: a protocol for a multicentre randomised controlled trial of extracorporeal carbon dioxide removal in patients with acute hypoxaemic respiratory failure. J Intensive Care Soc. 2017;18:159–69. Brodie D, Slutsky AS, Combes A. Extracorporeal life support for adults with respiratory failure and related indications: a review. JAMA. 2019;322:557–68. Combes A, Pesenti A, Ranieri VM. Fifty years of research in ARDS. Is extracorporeal circulation the future of acute respiratory distress syndrome management? Am J Respir Crit Care Med. 2017;195:1161–70. Needham DM, Colantuoni E, Mendez-Tellez PA, Dinglas VD, Sevransky JE, Dennison Himmelfarb CR, et al. Lung protective mechanical ventilation and two year survival in patients with acute lung injury: prospective cohort study. BMJ. 2012;344:e2124. Terragni PP, Del Sorbo L, Mascia L, Urbino R, Martin EL, Birocco A, et al. Tidal volume lower than 6 ml/kg enhances lung protection: role of extracorporeal carbon dioxide removal. Anesthesiology. 2009;111:826–35. Bellani G, Laffey JG, Pham T, Fan E, Brochard L, Esteban A, et al. Epidemiology, patterns of care, and mortality for patients with acute respiratory distress syndrome in intensive care units in 50 countries. JAMA. 2016;315:788–800. Fanelli V, Ranieri MV, Mancebo J, Moerer O, Quintel M, Morley S, et al. Feasibility and safety of low-flow extracorporeal carbon dioxide removal to facilitate ultra-protective ventilation in patients with moderate acute respiratory distress sindrome. Crit Care. 2016;20:36. Schmidt M, Jaber S, Zogheib E, Godet T, Capellier G, Combes A. Feasibility and safety of low-flow extracorporeal CO2 removal managed with a renal replacement platform to enhance lung-protective ventilation of patients with mild-to-moderate ARDS. Crit Care. 2018;22:122. 3. 4. 5. 6. 7. 8. 9. 10. Augy JL, Aissaoui N, Richard C, Maury E, Fartoukh M, Mekontso-Dessap A, et al. A 2-year multicenter, observational, prospective, cohort study on extracorporeal CO2 removal in a large metropolis area. J Intensive Care. 2019;7:45. 11. Combes A, Fanelli V, Pham T, Ranieri VM. European Society of Intensive Care Medicine Trials Group and the “Strategy of Ultra-Protective lung ventilation with Extracorporeal CO2 Removal for New-Onset moderate to severe ARDS” (SUPERNOVA) investigators. Feasibility and safety of extracorporeal CO2 removal to enhance protective ventilation in acute respiratory distress syndrome: the SUPERNOVA study. Intensive Care Med. 2019;45:592–600. 12. Goligher EC, Combes A, Brodie D, Ferguson ND, Pesenti AM, Ranieri VM, et al. Determinants of the effect of extracorporeal carbon dioxide removal in the SUPERNOVA trial: implications for trial design. Intensive Care Med. 2019;45:1219–30. 13. Bein T, Weber-Carstens S, Goldmann A, Müller T, Staudinger T, Brederlau J, et al. Lower tidal volume strategy (≈3 ml/kg) combined with extracorporeal CO2 removal versus ‘conventional’ protective ventilation (6 ml/kg) in severe ARDS: the prospective randomized Xtravent-study. Intensive Care Med. 2013;39:847–56. 14. Allardet-Servent J, Castanier M, Signouret T, Soundaravelou R, Lepidi A, Seghboyan JM. Safety and efficacy of combined extracorporeal CO2 removal and renal replacement therapy in patients with acute respiratory distress syndrome and acute kidney injury: the pulmonary and renal support in acute respiratory distress syndrome study. Crit Care Med. 2015; 43:2570–81. 15. Del Sorbo L, Pisani L, Filippini C, Fanelli V, Fasano L, Terragni P, et al. Extracorporeal CO2 removal in hypercapnic patients at risk of noninvasive ventilation failure: a matched cohort study with historical control. Crit Care Med. 2015;43:120–7. 16. Winiszewski H, Aptel F, Belon F, Belin N, Chaignat C, Patry C, et al. Daily use of extracorporeal CO2 removal in a critical care unit: indications and results. J Intensive Care. 2018;6:36. 17. Diehl JL, Piquilloud L, Richard JM, Mancebo J, Mercat A. Effects of extracorporeal carbon dioxide removal on work of breathing in patients with chronic obstructive pulmonary disease. Intensive Care Med. 2016;42: 951–2. 18. Deniau B, Ricard JD, Messika J, Dreyfuss D, Gaudry S. Use of extracorporeal carbon dioxide removal (ECCO2R) in 239 intensive care units: results from a French national survey. Intensive Care Med. 2016;42:624–5. 19. Morelli A, Del Sorbo L, Pesenti A, Ranieri VM, Fan E. Extracorporeal carbon dioxide removal (ECCO2R) in patients with acute respiratory failure. Intensive Care Med. 2017;43:519–30. Combes et al. Critical Care (2020) 24:490 Page 10 of 10 20. Gattinoni L, Pesenti A, Rossi G, Vesconi S, Fox U, Kolobow T, et al. Treatment of acute respiratory failure with low-frequency positive-pressure ventilation and extracorporeal removal of CO2. Lancet. 1980;316:292–4. 21. Hilty MP, Riva T, Cottini SR, Kleinert E-M, Maggiorini A, Maggiorini M. Low flow veno-venous extracorporeal CO2 removal for acute hypercapnic respiratory failure. Minerva Anestesiol. 2017;83:812–23. 22. Braune S, Sieweke A, Brettner F, Staudinger T, Joannidis M, Verbrugge S, et al. The feasibility and safety of extracorporeal carbon dioxide removal to avoid intubation in patients with COPD unresponsive to noninvasive ventilation for acute hypercapnic respiratory failure (ECLAIR study): multicentre case-control study. Intensive Care Med. 2016;42:1437–44. 23. Combes A, Tonetti T, Fanelli V, Pham T, Pesenti A, Mancebo J, et al. Efficacy and safety of lower versus higher CO2 extraction devices to allow ultraprotective ventilation: secondary analysis of the SUPERNOVA study. Thorax. 2019;74:1179–81. 24. Dalkey N, Helmer O. An experimental application of the Delphi method to 25. the use of experts. Manag Sci. 1963;9:458–67. Eubank BH, Mohtadi NG, Lafave MR, Wiley JP, Bois AJ, Boorman RS, et al. Using the modified Delphi method to establish clinical consensus for the diagnosis and treatment of patients with rotator cuff pathology. BMC Med Res Methodol. 2016;16:56. 26. ARDS Definition Task Force, Ranieri VM, Rubenfeld GD, Thompson BT, Ferguson ND, Caldwell E, et al. Acute respiratory distress syndrome: the Berlin Definition. JAMA. 2012;307:2526–2533. 27. Papazian L, Aubron C, Brochard L, Chiche JD, Combes A, Dreyfuss D, et al. 28. Formal guidelines: management of acute respiratory distress syndrome. Ann Intensive Care. 2019;9:69. Fan E, Del Sorbo L, Goligher EC, Hodgson CL, Munshi L, Walkey AJ, et al. An official American Thoracic Society/European Society of Intensive Care Medicine/Society of Critical Care Medicine clinical practice guideline: mechanical ventilation in adult patients with acute respiratory distress syndrome. Am J Respir Crit Care Med. 2017;195:1253–63. 29. Combes A, Hajage D, Capellier G, Demoule A, Lavoué S, Guervilly C, et al. 30. 31. Extracorporeal membrane oxygenation for severe acute respiratory distress syndrome. N Engl J Med. 2018;378:1965–75. Taccone FS, Malfertheiner MV, Ferrari F, Di Nardo M, Swol J, Broman LM, et al. Extracorporeal CO2 removal in critically ill patients: a systematic review. Minerva Anestesiol. 2017;83:762–72. Laffey JG, Bellani G, Pham T, Fan E, Madotto F, Bajwa EK, et al. Potentially modifiable factors contributing to outcome from acute respiratory distress syndrome: the LUNG SAFE study. Intensive Care Med. 2016;42:1865–76. 32. Amato MB, Meade MO, Slutsky AS, Brochard L, Costa EL, Schoenfeld DA, et al. Driving pressure and survival in the acute respiratory distress syndrome. N Engl J Med. 2015;372:747–55. Laffey JG, Jankov RP, Engelberts D, Tanswell AK, Post M, Lindsay T, et al. Effects of therapeutic hypercapnia on mesenteric ischemia-reperfusion injury. Am J Respir Crit Care Med. 2003;168:1383–90. 33. 34. Nin N, Muriel A, Peñuelas O, Brochard L, Lorente JA, Ferguson ND, et al. 35. 36. 37. Severe hypercapnia and outcome of mechanically ventilated patients with moderate or severe acute respiratory distress syndrome. Intensive Care Med. 2017;43:200–8. Tiruvoipati R, Pilcher D, Buscher H, Botha J, Bailey M. Effects of hypercapnia and hypercapnic acidosis on hospital mortality in mechanically ventilated patients. Crit Care Med. 2017;45:e649–e56. Jolliet P, Ouanes-Besbes L, Abroug F, Ben Khelil J, Besbes M, Garnero A, et al. A multicenter randomized trial assessing the efficacy of helium/ oxygen in severe exacerbations of chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2017;195:871–80. Kimmoun A, Oulehri W, Sonneville R, Grisot PH, Zogheib E, Amour J, et al. Prevalence and outcome of heparin-induced thrombocytopenia diagnosed under veno-arterial extracorporeal membrane oxygenation: a retrospective nationwide study. Intensive Care Med. 2018;44:1460–9. 38. Pollak U. Heparin-induced thrombocytopenia complicating extracorporeal membrane oxygenation support: review of the literature and alternative anticoagulants. J Thromb Haemost. 2019;17:1608–22. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12985-019-1252-3	Bruno et al. Virology Journal (2019) 16:148 https://doi.org/10.1186/s12985-019-1252-3 R E S E A R C H Open Access HPV16 persistent infection and recurrent disease after LEEP Maria Teresa Bruno1,2* , Nazzario Cassaro1, Salvatore Garofalo1 and Sara Boemi1 Abstract Background: About 23% of patients develop CIN2+ after LEEP treatment due to residual or recurrent lesions. The majority of patients with HPV infection were HPV negative before treatment, but 16,4% were still HPV 16 positive after treatment, indicating that conization do not necessarily clear HPV infection rapidly. The aim of this retrospective study was to evaluate the possible correlation existing between the appearance of recurring high- grade lesions and the viral genotype 16, and other risk factors such as residual disease. Methods: One hundred eighty-two HPV positive patients underwent LEEP for CIN2+. The follow-up post treatment was carried out every 6 months. Abnormal results during follow-up were confirmed histologically and considered recurrent high-grade intraepithelial cervical lesions (CIN2/CIN3 or CIS). Statistical analysis was performed by using the SPSS software package for Windows (version 15.0, SPSS, Chicago, IL, USA). Descriptive statistics are expressed as frequency, arithmetic mean, standard deviation (S.D.) and percentages. We calculated significance (P < 0.5) with the Easy Fischer Test. We calculated the Odds Ratio (OR) of women with peristent HPV 16 infection and positive margin, to have a recurrence. Results: In our study, the rate of persistent infection from HPV 16, after LEEP, was 15.9% (29/182) with 94% (17/18) of the recurring disease occurring within 18 months of follow up. From this study it was found that the persistence of genotype 16 is associated with a greater rate of relapse post-conization of CIN 2+ lesions, with respect to other genotypes. Our study further supports those studies that demonstrate that the risk for residual disease or relapse is not to be overlooked, also when the margins are negative, but persistent HPV infection is present. In our case study, 40% of relapses were in women with negative margin, but with persistent HPV 16 infection. Even more so, the margins involved in HPV16 positive subjects is another prediction factor for relapse. Conclusions: Our results show the importance of genotyping and that persistent HPV 16 infection should be considered a risk factor for the development of residual/recurrent CIN 2/3. Keywords: Papillomavirus infection, LEEP, CIN2+, Relapse, Recurrent desease, Positive margin Background The treatment of Cervical Intraepithelial Neoplasia grade 2+ (CIN2+) (CIN3-CIS) consists in a conservative surgi- cal approach using loop electrosurgical excision proced- ure (LEEP) radicality and preserves functional integrity of the uterine cervix, con- sidering that young women are the most affected by these pathologies. The treatment for CIN 2/3 or carcin- oma in situ (CIS) with LEEP is efficacious and most that guarantees surgical * Correspondence: mt.bruno@unict.it 1Department of General Surgery and Medical Surgery Specialities, Gynecological Clinic of the University of Catania, Policlinico. Via S. Sofia78, Catania, Italy 2Gynecological Oncology, Humanitas, Catania, Italy patients require no further treatment. However, about 23% of patients develop CIN2+ after conservative treat- ment due to residual or recurrent lesions [1]. It has been seen that the women who do not eliminate the virus have greater recurrence rates. Furthermore, in the ASC-US LSIL triage study (ALTS trial) [2] and in the more recent study of Heymans et al. [3], patients with HPV 16 were identified as being at high risk of recurrence and thus were given a more frequent follow- up. Based on these data, in this study we wanted to inves- tigate if the HPV 16 genotype could be a risk factor with respect to high-grade lesion recurrence after excisional © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Bruno et al. Virology Journal (2019) 16:148 Page 2 of 4 treatment with LEEP and the management of the re- sidual disease in cases of cone with positive margin. Aim of the study The aim of this retrospective study was to evaluate the possible correlation existing between the appearance of recurring high-grade lesions and the viral genotype 16, and other risk factors such as residual disease. Materials and methods The study protocol was approved by the Institutional Review Board of the Department and was conducted. in accordance with the 1975 Declaration of Helsinki. We studied the clinical files of 230 patients who, from April 2015 to April 2017, underwent LEEP for CIN2+ at the Colposcopy Outpatient Service of the Gynaeco- logical/Obstetrics Unit at the Policlinico Universitario, Catania (University of Catania, Italy) and who satisfied the following inclusion criteria: a) Patients with histological diagnosis of CIN2+; b) Patients who had an HPV test before and after treatment; c) Patients who had no anti-HPV vaccination; d) Patients without pathologies of the immune system; e) Patients who had completed at least two-years of follow-up. Moreover, patients were excluded if they were HPV negative or had any suspicion of infiltrating neoplastic pathologie. Only 192 patients satisfied the inclusion criteria; their clinical data was collected, of which: patient’s age, type of pathology, HPV strain, treatment method used, resec- tion margins, follow-up cervical histology, date of follow-up, pretreatment viral genotype, post-treatment HPV genotype, and recurrent cases. All the patients underwent conservative surgical treat- ment of cone excision (conization) electrosurgery with conization (LEEP) of 20 mm width and 12, 15 or 20 mm depth (Utah Medical Products, Midvale, Utah, USA). The histological examination of the cervical cone established a definitive histological diagnosis and evalu- ated the extension of the cone margins, defined as posi- tive if the distance between the CIN lesion and the margin of the resection was less than 1 mm. The follow-up was carried out every 6 months dur- ing the first year after treatment and then once to year (At the moment of excisional treatment (T0) and the successive follow-ups at 6 (T1), 12 (T2), 18 (T3) and 24 (T4) and 30 (T5) months). Abnormal results during follow-up were confirmed histologically and consid- ered recurrent high-grade intraepithelial cervical le- sions (CIN2/CIN3 or CIS). At each follow-up examination the patients were examined by conventional Pap test, HPV test and col- poscopy, moreover, cytological eso-endocervical samples were taken and placed in ThinPrep Solution. The sam- ples were sent to the laboratory for extraction of total DNA for the genotyping of viral DNA by means of gen- etic amplification followed by hybridization with genotype-specific probes able to identify most HPV ge- notypes of the genital region [28 genotypes of high-risk HPV (16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82), low-risk (6, 11, 40, 43, 44, 54, 70) and not defined risk (69,71,74)]. The commercial method used was NucliSenseasy MAG system bio Merieux SA, Marct l’Etoile, France. The HPV genotype was classified as follows: 1) nega- tive, 2) HPV 16, 3) hr HPV (HPV18, 31, 33, 45 and other hr HPV). Colposcopy was carried out using a Zeiss OPM1F colposcope (Carl Zeiss, Jena, Germany) and applying acetic acid and a Lugol iodine solution. Any colposcopic abnormality was classified based on the nomenclature proposed by the International Federation for Colposcopy and Cervical Pathologies (IFCPC) in 3 grades of increas- ing anomalies based on severity: Zone of Abnormal Transformation (AnTZ) grade 1 (AnTZ1) and grade 2 (AnTZ2), or cancer. We evaluated the visibility of the squamo-columnar junction and specific biopsies were taken from the portio and/or endocervical curettage to guide the diagnosis in cases of an abnormal Pap test or suspicion of residual/recurrent disease at colposcopy. Statistical analysis Statistical analysis was performed by using the SPSS software package for Windows (version 15.0, SPSS, Chicago, IL, USA). Descriptive statistics are expressed as frequency, arithmetic mean, standard deviation (S.D.) and percentages. We calculated significance (P < 0.5) with the Easy Fischer Test. We calculated the Odds Ratio (OR) of women with peristent HPV 16 infection and positive margin, to have a recurrence. Results The mean age of the 192 patients was 39.3 ± 8.7 years (range 22 to 73 years). During the follow-up, 18 (9.8%) of the 182 patients relapsed. The average time between LEEP and the diagnosis of recurrent disease was 18 months (range 6 to 30 months). The 192 patients had CIN3-CIS, as determined at bi- opsy or histological examination of the cone. In total, 182 (94.7%) of the 192 patients were positive for the HPV test before surgery, 10 patients that were negative at the HPV test excluded by the study. Bruno et al. Virology Journal (2019) 16:148 Page 3 of 4 Of the 182 patients with a positive HPV test, 104 (57.1%) belonged to genotype 16, 78 (42.8%) to other ge- notypes. (Table 1). Of the 182 patients positive at the HPV test at T0 (be- fore LEEP), 144 (79.1%) were negative at 6 months (T1) (p < 0.001), 29 cases (15.9%) HPV 16 showed virus per- sistence, only 9 case of virus persistence was found in the hr HPV group (Table 2). A negative margin was found in 162/182 (89%) pa- tients, while 20/182 (10.9%) patients had a positive mar- gin. Nine patients with genotype 16 had positive margin, while in the hr HPV group, only 6 patients had positive margin. From the Table 3 we can see how all the patients (9 cases) with positive margin and 16 HPV positive have developed a recurrence (100%), while the hr HPV pa- tients (6 cases) only relapsed in one case. Furthermore, the HPV 16 positive women with a negative margin, de- veloped 8 recurrences (40%). Women with hr HPV but whose margin was negative did not developed any recur- rence. Table 4 show the OR. Discussion The women with CIN2+ with residual or recurring le- sions are at 5 times greater risk of cancer with respect to the general population [4]. Various studies have reported that persistent HPV infection after LEEP is a risk factor for residual/recurrent disease [5]. On the other hand, the viral clearance at follow-up after conization is significatively associated with efficacy of the surgical treatment, as confirmed by Cricca et al. [6]. Authors [7], in particular, reported that the rate of persistence of HPV infection after conization for CIN 3 was approximately 20, and 46% of these patients with persistent HPV infection developed CIN relapse at 4–10 months after treatment. In our study, the rate of persist- ent infection from HPV 16, after LEEP, was 15.9% (29/ 182) and occurring within 18 months of FU. From this study it was found that the persistence of genotype 16 is associated with a greater rate (17/18) (94.4%) (p < .05) of relapse post-conization of CIN 2+ lesions, respect to other genotypes with OR = 11.33 (CL95% = 1.25–102.93). Our study further supports those studies [5] that dem- onstrate that the risk for residual disease or relapse is not to be overlooked, also when the margins are nega- tive, but persistent HPV infection is present. In our case Table 1 patients before treatment Table 2 Patients after treatment HPV Test HPV 16 hrHPV Negative HPV Postoperative patients 29 9 144 15.9% 4.9% 79.1% study, 40% of relapses were in women with negative margin, but with persistent HPV 16 infection. Even more so, the margins involved in HPV 16 posi- tive subjects is another prediction factor for relapse OR = 45 (CL 95% = 2.29–885.65). Viral clearance at follow-up can also occur in cases of positive margins [6], only these cases do not need fur- ther treatment; 6 cases in our study became negative within 24 months of FU. Moreover, in our study, no cases of relapse were found in women who were negative at the HPV test at 6 months, independently from the type of involvement of the margin, indicating a negative predictive value of 100% for relapse during follow-up [8]. Our results show the importance of genotyping [9] and that persistent HPV 16 infection should be consid- ered a risk factor for the development of residual/recur- rent CIN 2/3. Moreover, Authors [10] reported that HPV 16 positivity, 6 months after LEEP, was associated with a 37% increase of the absolute risk at 2 years for CIN 2+, twice that associated with HPV 18 (18.5%), and three times that of other types of oncogenes (10.8%). This result indicates that the HPV genotype should be considered in the policy of post-treatment monitoring, as suggested by many authors [11]. Furthermore, some researchers believe that when the DNA of post-treatment HPV is absent for 3 to 6 months after conization, especially in patients with a negative margin of the cone, the patients can go back to general- population screening [12]. Furthermore, the possibility of re-infection after treat- ment should be considered. Infact, Jung Mi Byun et al. [13] during follow-up, 70.7% of HPV infections were Table 3 Follow-up according to margin status, HPV status and recurrence % No recurrence % Recurrence % Positive Margin 16 HPV Hr HPV negative HPV Negative Margin 9 6 5 45 30 25 0 5 5 0 83.3 100 60 100 100 9 1 0 8 0 0 100 16,6 0 40 0 0 HPV Test HPV 16 hrHPV Preoperative patients 104 78 57.1% 42.8% 16 HPV Hr HPV 20 3 12.3 12 1,8 3 negative HPV 139 85.8 139 Bruno et al. Virology Journal (2019) 16:148 Page 4 of 4 Table 4 Odd Ratio: recurrence after treatment Received: 11 July 2019 Accepted: 11 November 2019 HPV 16 positive margin HPV 16 infection Odds ratio (95% CL) 45 (2.29–885.65) 11.33 (1.25–102.93) P-value .0034 .0126 References 1. Ghaem-Maghami S, Sagi S, Majeed G, et al. Incomplete excision of cervical intraepithelial neoplasia and risk of treatment failure: a meta-analysis. Lancet Oncol. 2007;8:985–93. ASCUS-LSIL Triage Study Group. Results of a randomized trial on the management of cytology interpretations of atypical squamous cells of undetermined significance. Am J Obstet Gynecol. 2003;188:1383–92. Heymans J, Benoy IH, Poppe W, Depuydt CE. Type-specific HPV geno-typing improves detection of recurrent high-grade cervical neoplasia after conisation. Int J Cancer. 2011;129(4):903–9. Soutter WP, Sasieni P, Panoskaltsis T. Long-term risk of invasive cervical cancer after treatment of squamous cervical intraepithelial neoplasia. Int J Cancer. 2006;118:2048–55. Kang WD, Oh MJ, Kim SM, Nam JH, Park CS, Choi HS. Significance of human papillomavirus genotyping with high-grade cervical intraepithelial neoplasia treated by a loop electrosurgical excision procedure. Am J Obstet Gynecol. 2010;203:72.e1–6. Cricca M, Venturoli S, Morselli-Labate AM, Costa S, Santini D, Ambretti S, et al. HPV DNA patterns and disease implications in the follow-up of patients treated for HPV16 high-grade carcinoma in situ. J Med Virol. 2006; 78:494–500. Nagai Y, Maehama T, Asato T, Kanazawa K. Persistence of human papillomavirus infection after therapeutic conization for CIN 3: is it an alarm for disease recurrence? Gynecol Oncol. 2000;79:294–9. Nam K, Chung S, Kim J, Jeon S, Bae D. Factors associated with HPV persistence after conization in patients with negative margins. J Gynecol Oncol. 2009;20:91–5. Bruno MT, Ferrara M, Fava V, Rapisarda A, Coco A. HPV genotype determination and E6/E7 mRNA detection for management of HPV positive women. Virol J. 2018;15(1):52. Kreimer AR, Guido RS, Solomon D, Schiffman M, Wacholder S, Jeronimo J, et al. Human papillomavirus testing following loop electrosurgical excision procedure identifies women at risk for posttreatment cervical intraepithelial neoplasia grade 2 or 3 disease. Cancer Epidemiol Biomark Prev. 2006;15:908–14. 11. Gok M, Coupe VM, Berkhof J, Verheijen RH, Helmerhorst TJ, Hogewoning CJ, et al. HPV16 and increased risk of recurrence after treatment for CIN. Gynecol Oncol. 2007;104:273–5. 12. Verguts J, Bronselaer B, Donders G, Arbyn M, Van Eldere J, Drijkoningen M, et al. Prediction of recurrence after treatment for high-grade cervical intraepithelial neoplasia: the role of human papillomavirus testing and age at conisation. BJOG. 2006;113:1303–7. 13. Byun JM, Jeong DH, Kim YN, Jung EJ, Lee KB, Sung MS, Kim PKT. Persistent HPV-16 infection leads to recurrence of high-grade cervical intraepithelial neoplasia. Medicine (Baltimore). 2018;97(51):e13606. 14. Bruno MT, Ferrara M, Barrasso G, Cutello S, Sapia F, Panella MM. Prevalence genotypes and distribution of human papillomavirus infection in women with abnormal cervical cytology in Catania, Italy. Giornale Italiano di Ostetricia e Ginecologia. 2016;38(5–6):376–80. 15. Bruno MT, Ferrara M, Fava V, Barrasso G, Panella MM. A prospective study of women with ASC-US or LSIL pap smears at baseline and HPV E6/E7 mRNA positive: a 3-year follow-up. Epidemiol Infect. 2018;146(5):612–8. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. new and 29.3% were persistent, indicating a need to pre- vent reinfection after treatment and to regular follow-up for persistent HPV infection. HPV vaccination for HPV 16 type may be useful in preventing recurrence of CIN 2/3 and CIS. The efficacy of vaccination after treatment of CIN should now be investigated. Conclusions The prevalence of the diverse genotypes of HPV seen in our cohort is similar to other Italian and European pop- ulations [14]. HPV 16 has been confirmed as the preva- lent genotype in CIN2+ and is generally, by itself, responsible for the pathology (single genotype) [15]. In the absence of viral infection, the risk of relapse is minimal. On the other hand, the cone with negative margins in the presence of persistent HPV 16 infection has a high incidence of relapse and thus persistent infec- tion with HPV 16 should be considered a risk factor for the development of CIN2+ relapse. 2. 3. 4. 5. 6. 7. 8. 9. HPV vaccination for HPV 16 type may be useful in 10. preventing recurrence of CIN 2/3 and CIS. There are some limitations in our study that include the small sample size, the retrospective design and the limited long-term follow-up. However, follow-up is still underway. Abbreviations CIN2+: All the cases of CIN3, SCC lesion; CIS: Carcinoma In Situ; FU: Follow- Up; HPV: Human Papillomavirus; hr HPV: High risk HPV; SCC: Squamous Cervical Carcinoma Acknowledgments We wish to thank the Scientific Bureau of the University of Catania for language support. Authors’ contributions MTB designed the study; CN and SB collected the data; MTB and SB drafted the manuscript; SG compiled the statistical data. All authors were involved in editing the manuscript. All authors read and approved the final manuscript. Funding No funding was involved in the preparation of this research. Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on request. Ethics approval and consent to participate The study protocol was approved by the Institutional Review Board of the Department and was conducted in accordance with the 1975 Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests.
10.1186_s12889-023-15729-1	Mjöberg et al. BMC Public Health (2023) 23:795 https://doi.org/10.1186/s12889-023-15729-1 RESEARCH BMC Public Health Open Access Supermarket promotions in Western Sweden are incompatible with Nordic dietary recommendations and differ by area-level socioeconomic index Melissa Mjöberg1, Lauren Lissner1 and Monica Hunsberger1 Abstract Background Large supermarket chains produce weekly advertisements to promote foods and influence consumer purchases. The broad consumer reach of these ads presents an opportunity to promote foods that align with dietary recommendations. Thus, the aim of this study was to investigate the health quality of supermarkets’ weekly food pro- motions in a large region of Sweden with attention to more and less advantaged socioeconomic index areas. Methods Analysis of weekly advertisements from 122 individual stores, representing seven chains, was carried out in a large region of Sweden from 2–29 March in 2020. Food promotions were divided into categories according to the Nordic Nutrition Recommendations and World Health Organization Regional Office for Europe’s nutrient profile model, and defined as ‘most healthy’, ‘healthy’, ‘unhealthy’ and ‘most unhealthy’. A mean socioeconomic index was used to classify each store location to determine whether proportions of the ‘most unhealthy’ foods differed between more advantaged and more disadvantaged socioeconomic index areas. Results In total, 29,958 food items were analyzed. Two-thirds of promotions belonged to the food groups consid- ered ‘most unhealthy’ and ‘unhealthy’. In the ‘most unhealthy’ food group ‘sugar-rich beverages and foods’ constituted approximately 23.0% of the promotions. Food promotions had 25% increased odds to be from the ‘most unhealthy’ group (odds ratio 1.25, confidence interval 1.17, 1.33) in more disadvantaged socioeconomic index areas. This associa- tion could be explained by the supermarket chain the stores belonged to. Conclusions Our findings indicate that Swedish supermarkets promote a large proportion of unhealthy foods as classified by the Nordic Nutrition Recommendations. We also observe that certain national supermarket chains tend to locate their stores in more disadvantaged areas and promote a greater proportion of unhealthy foods in their weekly advertisements compared to the more advantaged areas. There is an urgent need for supermarkets to shift promotions toward healthier food items. Keywords Food environment, Food advertising, Nutrition guidelines, Sweden, Supermarket, Healthy diet, Socioeconomic area Correspondence: Melissa Mjöberg melissa.mjoberg@gu.se 1 School of Public Health and Community Medicine, Institute of Medicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden Background The supermarket is a primary venue for grocery shop- ping, where various external factors influence con- sumer buying behavior, e.g. price, promotion, nutritional information, quality, freshness, use of health claims, © The Author(s) 2023, corrected publication 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Mjöberg et al. BMC Public Health (2023) 23:795 Page 2 of 9 placement and labelling [1–5]. Consumers make multi- ple food choices at each visit and when exposed to much information concurrently, choices tend to become less conscious and increasingly vulnerable to marketing influ- ences [1, 6]. In Sweden, both the total energy consumption and the consumption of refined products have increased per cap- ita since 1980 [7]. Above all, the young population con- sumes too little fruit, vegetables and fiber, and excessive amounts of energy-dense foods and beverages high in fat, salt and sugar [8]. The food environment is dominated by formal markets that offer Swedish consumers a variety of food all year round at low prices. The availability of both healthy and unhealthy foods is high, but marketing of unhealthy foods has been reported to be particularly fre- quent [9]. For less affluent households, it is problematic because they spend a larger proportion of the household budget on food [10] and are often extra price sensitive [11, 12]. Extensive marketing of unhealthy foods could increase already existing health gaps since they often cost less per calorie than healthy foods [13, 14]. Supermarkets in Sweden, and many other countries across all continents [15, 16], use weekly advertisement sheets (ad-sheets) to promote what is on weekly sale. These ad-sheets have a wide reach [17] and are often available in both physical and digital form. In addition, they probably reflect in-store sales [18]. As the general Swedish population has high digital competence, well above the European Union average [19], it is important that both the physical and digital consumer food environ- ment is healthy. It may be even more important to study the healthiness of food marketing at a time when Sweden is experiencing high food inflation [20] as they are usually marketed at a discount. The Nordic Nutrition Recommendations (NNR) are produced periodically by the Nordic Council of Ministers [21] and communicated in the Swedish context by the Swedish Food Agency [22]. The latest NNR from 2012 propose an increased intake of vegetables and pulses, fruit and berries as well as dietary fiber and limited con- sumption of red meat, discretionary and refined foods [21]. From a public health perspective, food campaigns by the Swedish retail should be compatible with these recommendations. To the authors’ knowledge, no study in Sweden has evaluated the health quality of supermarkets’ weekly ad-sheet promotions. However, studies documenting unhealthy supermarket food promotions have been con- ducted in United States [17, 23–25], Australia [18], the Netherlands [26, 27], and Brazil [16] together with one international comparison of twelve countries [15]. Thus, the aim of this study was to investigate the health qual- ity of supermarkets’ weekly food promotions in a large region of Sweden with attention to more and less advan- taged socioeconomic index areas. Methods Study design This cross-sectional study is based upon data collected from 2–29 March in year 2020 from weekly online ad- sheets published by seven supermarket chains across a large region in Western Sweden. Data collection The supermarket chains were selected based on leading market share in 2019. All individual supermarkets rep- resenting the chains were identified using Google Maps. From these leading chains, all online versions from 122 individual supermarkets across the region were com- pared to the printed advertisements and found to be identical. The seven chains (coded A-G), number of indi- vidual stores per chain and their promotions are pre- sented in Table 1. Content analysis In order to analyze the content of the ad-sheets, food categories from NNR [21] were used to correctly clas- sify the ads as ‘most healthy’, ‘healthy’, ‘unhealthy’ and ‘most unhealthy’. Table 2 illustrates the content analysis template that is based on NNR version 2012 [21], with slight modifications of some food categories for the purpose of this study. The Swedish Food Agency uses an arrow, which we have reproduced here, to indicated that consumers should aim to exchange items from the ‘unhealthy’ to the ‘healthy’ group. E.g. from refined to wholegrain cereals and from butter-based to vegetable oil-based fat spreads [22]. The national authority’s front of package “keyhole” symbol criteria for foods of high nutritional quality as well as the Swedish Food Agency’s Code of Statutes [32] were used as steering documents to be consist- ent with ambiguous food items. The food classification was performed by a food scientist (MM). Food cat- egories and inclusion conditions used during the data input process are described in Additional file 1. To dif- ferentiate between unprocessed and processed meat, definitions from the Swedish Food Agency were used [33]. World Health Organization Regional Office for Europe’s (WHO Europe) nutrient profile model [34] was used to add food categories that were missing in NNR’s model in order to categorize all food promotions. Bev- erages with any alcohol content are included in the food group ‘most unhealthy’ in this study. According to Swedish law, alcohol products containing 3,5% alcohol Mjöberg et al. BMC Public Health (2023) 23:795 Page 3 of 9 Table 1 Number of promoted items in each chain and individual store Chain Market share of the supermarket conglomerates (%) Individual stores per week (n = 122) A B C D E F G 51.51 17.8a2 16.91 71 4.71 2.11 11 17 39 7 4 28 16 a Chains B and C belong to the same conglomerate and the available data and their market share in 2019 is combined [28, 29] 1 Data from 2019 [29] 2 Data from 2017 [30, 31] Promoted food items over 4 weeks (n = 29,958) 2252 3264 7592 1302 800 7644 7104 Table 2 Content analysis template with arrow referring to a recommended replacement of ‘unhealthy’ to ‘healthy’ foods [21] * This refers to allowable alcohol sales in supermarkets as described in the text, e.g. low-alcohol cider and beer Mjöberg et al. BMC Public Health (2023) 23:795 Page 4 of 9 or less can be sold in supermarkets, all other alcohol must be purchased at the state monopoly and are there- fore not promoted in the ad-sheets [35, 36]. Data input and variables Food items promoted more than one time in the same ad-sheet were recorded individually to represent the number of visual promotions. Non-categorizable promo- tions were allocated to an “excluded” category. Non-food items were not included in this analysis. Area‑level socioeconomic classification of stores Socioeconomic characteristics of stores were classified applying the socioeconomic index (SEI) from Statistics Sweden to different geographical areas within the region of Western Sweden [37]. The SEI ranges from 0–100 per- cent and reflects socioeconomic disadvantage based on three indicators: proportion of inhabitants with 1) low economical standard, 2) basic education, and 3) finan- cial support and/or unemployment. The index calculates the mean of the three proportions to get a percentage for each area, with a higher SEI value indicating more dis- advantaged socioeconomic conditions [38]. The mean SEI was used as a threshold to compare more versus less affluent SEI areas. Method of analyses The proportion of promoted food categories by health quality (four groups ranging from ‘most healthy’ to ‘most unhealthy’) were described in terms of frequency, per- cent, and 99% confidence interval (CI). Comparison between promotion of ‘most unhealthy’ foods in more advantaged versus more disadvantaged neighborhoods were also described. Pearson’s Chi-square was used to test for differences between more and less advantaged areas and promotion of food belonging to the four health groups. Binary logis- tic regression was used to test if there were higher odds of most unhealthy food to be promoted in more disad- vantaged neighborhoods. A multivariable logistic regres- sion analysis was used to test if chain was a confounding factor in the association between SEI area of the store and promotion of most unhealthy food. P < 0.01 was con- sidered as significant. IBM SPSS Statistics 27 was used for all data analyses. A sensitivity analysis was performed to test the robust- ness of the findings by removing the one chain with no stores in more disadvantaged areas. Ethics The study did not require ethical permission because human subjects were not involved. The chains are not identified in this article. Results Promoted foods by healthfulness and food categories After four weeks, 488 ad-sheets containing 42,139 food items were analyzed. After excluding the promo- tions that could not be categorized by health quality, 71.1% (n = 29,958) of the individual food promotions remained for analysis. Of the total promoted foods, 37.4% were categorized as ‘most unhealthy’. See Fig. 1 for proportions of promoted foods across the four health groups. To view all food categories included in each food group we refer to table 2. item was ‘processed meat’ (11.7%), Of the foods promoted, 66.7% belonged to the two unhealthy food groups. The most promoted items in these groups were ‘beverages and foods with added sugar’ (22.8%), ‘high fat dairy’ (9.5%) and ‘red meat’ (7.9%). The least promoted unhealthy ‘alcohol containing products’ (0.4%). In the healthy and most healthy food groups, the most commonly promoted items were ‘vegetable and pulses’ (10.4%), ‘fruit and berries’ (6.8%) and ‘fish and seafood’ (6.6%). The least promoted items in these groups were ‘vegetable oil and oil-based fat spreads’ (1.2%), ‘low fat dairy’ (0.7%), ‘healthier ready-made meals’ (0.7%), ‘nuts and seeds’ (0.3%) and ‘healthier sauces, dips and dressings’ (0.1%). Proportion of promoted unhealthy food in stores located in more disadvantaged and more advantaged areas In this study, the SEI ranged from 2.9% (most advantaged SEI area) to 42.2% (most disadvantaged SEI area). Con- sidering the 122 store areas, the mean SEI and SD were 11.2 (± 6.3). There was a difference in mean proportion of advertised food categories between more advantaged and more disadvantaged neighborhoods, where the unhealth- iest food categories were promoted to a higher extent in the more disadvantaged neighborhoods. The mean proportion of promoted most unhealthy food in stores located in more disadvantaged areas was 40.0%, and in more advantaged areas the mean proportion was 34.9%. The food category that was promoted most frequently was ‘beverages and foods with added sugar’. In the more advantaged SEI areas, the food category had a proportion of 14.7% and in the more disadvantaged SEI areas, the proportion was 17.7%, see Fig. 2. Food promotions in more disadvantaged neigh- borhoods had 25% increased odds to be in the ‘most unhealthy’ group (p < 0.001). However, adjusting for the chain the store belonged to accounted for the association (p = 0.113), see Table 3. Our sensitivity analysis, removing the one chain with no stores in more disadvantaged areas, showed the results to be robust and remained significant. Mjöberg et al. BMC Public Health (2023) 23:795 Page 5 of 9 Fig. 1 Proportion of promoted foods across the four health groups with corresponding recommendations Fig. 2 Promoted food categories shown by more disadvantaged and more advantaged store SEI area Mjöberg et al. BMC Public Health (2023) 23:795 Page 6 of 9 Table 3 Association between store area SEI and promotion of most unhealthy foods Socioeconomic index of the stores More advantaged store areas Less advantaged store areas Total Most unhealthy promotions n (%) 5294 (34.9) 5907 (40.0) 11,201 (37.4) Regression models for association between socioeconomic store index and promotion of most unhealthy foods Model 1 a Model 2 b All other promotions n (%) Total n (%) 9895 (65.1) 8862 (60.0) 18,757 (62.6) OR (99% CI) 1.25 (1.17, 1.33)a 1.00 (1.00 1.01)b 15,189 (50.7) 14,769 (49.3) 29,958 (100.0) P < 0.001a* 0.113b * Significant association. Two-sided level of significance (p < .01) a Value before adjusting for chain (A-G) the stores belong to as a confounding factor b Value after adjusting for chain (A-G) the stores belong to as a confounding factor using multivariable logistic regression analysis Discussion Proportion of promoted healthy and unhealthy food The novelty of this study is that it evaluates the nutri- tional quality of food promotions in Sweden as well as comparing this by area-level SEI of the stores. We found that 66.8% of the promoted foods were from the unhealthy food categories. According to NNR, these items should be exchanged to a healthier alternative or limited in the diet. The most frequently promoted food category was ‘sugary beverages and food’. Our finding that unhealthier food categories were promoted to a larger extent than the healthier catego- ries is consistent with previous research. An interna- tional comparative study found that most countries promoted a high amount of discretionary compared to core food [15]. In a study from Australia, 43.3% con- sisted of discretionary foods, fats and oils and con- cluded that the promotions analyzed were not in line with the Australian dietary recommendations [18]. Another study conducted in Brazil followed the Pan- American Health Organization’s (PAHO) nutritional profile model and reported that all but 3.5% of the food promotions had a less healthy nutrient profile, where ultra-processed foods constituted 66.9% of the pro- motions [16]. Two studies from the Netherlands cat- egorized approximately 70% of the promoted foods as unhealthy [26, 27], and four studies from United States found that most promoted foods were considered as unhealthy [17, 23–25]. Two studies reported that most promotions were for processed foods or simple carbo- hydrates, and few were high in dietary fiber or low in fat and sodium [17, 24]. It has also been observed that many promotions were for meat-based protein foods, where red meat and poultry constituted the majority [23, 25]. This is consistent with our study where pro- cessed meat and red meat constituted the second most promoted food category. This study compares supermarket promotions with the NNR, a guiding framework for both dietary composi- tion and macro- and micronutrient intake. The primary focus of the recommendations is to reduce the risk of diet-related chronic diseases by promoting a nutritious, low-energy diet and a physically active lifestyle to attain energy balance. Although the scientific evidence for total fat intake and health is limited, reducing total fat prevents excessive weight gain, which in turn is associated with health risks. Recommendations on fat intake aim both to reduce the proportion of total fat and to increase the qual- ity of fat, where saturated and trans-fatty acids should be reduced and unsaturated fatty acids increase [21]. Healthiness by store SEI The unhealthiest group which included the food catego- ries ‘processed meat’, ‘sugary beverages and food’, ‘salty food’ and ‘alcohol containing products’ was promoted to a larger extent in ad-sheets from supermarkets geograph- ically located in more disadvantaged SEI areas. However, the association was fully explained by the supermarket chain to which the store belonged. This could indicate that certain chains that promote unhealthier food to a larger extent locate their individual stores in areas which are more socioeconomically disadvantaged. One study conducted in Stockholm, Sweden inves- tigated the difference of outdoor ultra-processed food advertisements in two diverse socioeconomic status (SES) areas [39]. They found a significantly higher pro- portion of ultra-processed food advertised in the less affluent SES area. A study from United States saw a sig- nificant difference in promotions between regions with high and low rates of obesity, where the regions with less obesity prevalence promoted more fruit while the regions with a higher obesity prevalence promoted more sweets [23]. This aligns with our study finding that less affluent Mjöberg et al. BMC Public Health (2023) 23:795 Page 7 of 9 areas are exposed to more unhealthy prompts. However, these other studies did not investigate the influence by chain as we have. The social gradient in healthy dietary patterns is well established in high-income countries [40–44] and can to some extent be explained by factors in the surrounding food environment [1, 2, 45–47]. Promotion of unhealthy foods and soft drinks is contributing to the increase in childhood and adult overweight and obesity, thus the neg- ative raise in non-communicable diseases (NCDs) glob- ally [3, 9]. The results from this study exemplify important aspects of the Swedish consumer nutrition environment that do not support healthy and sustainable consumer choices. These factors might also contribute to the cur- rently stagnant and widening health gaps [48]. To reduce social inequities in diet, healthy and acceptable food choices should be affordable for all consumers [49]. Price incentives in combination with other strategies, e.g. choice architecture and nudging, might be successful to make healthy options more attractive in the store setting [50]. An observation made during data input was that many of the ad-sheet promotions included in the analysis were not price-reduced. To what extent and why certain prod- ucts end up in the ad of weekly offering without being price reduced was not investigated in this study but can be worth investigating in future research. To promote certain products in ad-sheets automatically increases their visual impact, which seems to be a successful nudg- ing strategy to influence consumers’ food choices [2]. The observation that two-thirds of the promotions were clas- sified in this study as unhealthy has been described as a ‘sludging’ as opposed to ‘nudging’ strategy [50]. Public health implications The Public Health Agency of Sweden has recognized the food environment as obesogenic and expressed that more knowledge is needed about the healthiness of the envi- ronment [51]. This study exemplifies weekly ads regularly promote less nutritious alternatives within several food categories. This may create further difficulties for con- sumers to make a healthy choice. Many past health promotion interventions have focused on individual responsibility, although food choices result from an interaction between consumers’ own values and the surrounding food environment [1, 3, 6, 52]. It is there- fore essential that the built food environment is compatible with dietary recommendations that support consumers in making healthy choices. Targeting environmental factors can be an effective upstream strategy because they often have far reaching effects that might contribute to narrow- ing health gaps [3, 10, 53–56]. Strength and limitations A strength with the study is that the food categories and four food healthiness groups align with NNR and WHO Europe’s nutrient profile model which are both reliable sources, established by a substantial amount of previous research. The use of these food categories also facilitates the process of making international compari- sons of food marketing in future research projects [34]. A weakness is that ad-sheets from only one Swed- ish region were analyzed for the study and the results can therefore not be generalized to the whole of Swe- den. However, the large region of western Sweden was covered, and two of the seven chains seem to be using the same ad-sheet version across all Swedish stores. Another limitation is the fact that data were collected from 2–29 March in 2020 so possible seasonal variation of the ads could not be taken into consideration. Conclusions Our findings indicate that Swedish supermarkets pro- mote a large proportion of unhealthy foods as classi- fied by the Nordic Nutrition Recommendations. We also observe that certain national supermarket chains tend to locate their stores in more disadvantaged areas and promote a greater proportion of unhealthy foods in their weekly advertisements compared to the more advantaged areas. There is an urgent need for super- markets to shift promotions toward healthier food items. Abbreviations SEI NNR WHO Europe Socioeconomic index Nordic Nutrition Recommendations WHO Regional Office for Europe Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12889- 023- 15729-1. Additional file 1. Acknowledgements Valter Sundh and Kirsten Mehlig offered statistical consulting and advised MM during data analysis and interpretation. Authors’ contributions MM planned the study, formulated the aim, conducted data collection, performed all data analysis and interpretations, and wrote the manuscript. LL and MH advised and supported the whole study process, from planning the study to writing and editing the manuscript. All authors read and approved the final manuscript. Funding Open access funding provided by University of Gothenburg. None to report. This study was done as part of a master’s thesis. Mjöberg et al. BMC Public Health (2023) 23:795 Page 8 of 9 Availability of data and materials The datasets used during the current study are available from the correspond- ing author on reasonable request. 16. Mendes C, Miranda L, Claro R, Horta P. Food marketing in supermarket circulars in Brazil: an obstacle to healthy eating. Preventive Medicine Reports. 2021;21: 101304. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Received: 20 February 2023 Accepted: 22 April 2023 References 1. Cohen DA, Babey SH. Contextual influences on eating behaviours: heuris- tic processing and dietary choices. Obes Rev. 2012;13(9):766–79. 2. Gidlöf K, Anikin A, Lingonblad M, Wallin A. Looking is buying. how visual attention and choice are affected by consumer preferences and proper- ties of the supermarket shelf. Appetite. 2017;116:29–38. 3. Glanz K, Sallis JF, Saelens BE, Frank LD. Healthy nutrition environments: concepts and measures. Am J Health Promot. 2005;19(5):330–3. Farley T, Rice J, Bodor J, Cohen D, Bluthenthal R, Rose D. Measuring the food environment: shelf space of fruits, vegetables, and snack foods in stores. J Urban Health. 2009;86(5):672–82. 4. 5. Vandevijvere S, Mackenzie T. Indicators of the relative availability of 6. 7. 8. 9. healthy versus unhealthy foods in supermarkets: a validation study. Int J Behav Nutr Phys Act. 2017;14(1):53. Sobal J, Bisogni CA, Jastran M. Food choice is multifaceted, contex- tual, dynamic, multilevel, integrated, and diverse. Mind Brain Educ. 2014;8(1):6–12. The Swedish Board of Agriculture. Livsmedelskonsumtion och näringsin- nehåll. Uppgifter till och med 2019. Accessed 4 April 2023. Swedish Food Agency. Riksmaten vuxna - 2010–11. Livsmedels- och näringsintag bland vuxna i Sverige. Uppsala: Swedish Food Agency; 2012. Available from:https://www.livsmedelsverket.se/globalassets/pub- likationsdatabas/rapporter/2011/riksmaten_2010_20111.pdf [cited date 4 April 2023]. Spires M, Berggreen-Clausen A, Kasujja FX, Delobelle P, Puoane T, Sanders D, et al. Snapshots of urban and rural food environments: EPOCH-based mapping in a high-, middle-, and low-income country from a non-com- municable disease perspective. Nutrients. 2020;12(2):484. 10. Willett W, Rockström J, Loken B, Springmann M, Lang T, Vermeulen S, et al. Food in the Anthropocene: the EAT–Lancet Commission on healthy diets from sustainable food systems. The Lancet. 2019;393(10170):447–92. 11. Puddephatt J-A, Keenan GS, Fielden A, Reaves DL, Halford JCG, Hardman CA. ‘Eating to survive’: a qualitative analysis of factors influencing food choice and eating behaviour in a food-insecure population. Appetite. 2020;147: 104547. 12. Dhuria P, Lawrence W, Crozier S, Cooper C, Baird J, Vogel C. Women’s perceptions of factors influencing their food shopping choices and how supermarkets can support them to make healthier choices. BMC Public Health. 2021;21(1):1070. 13. Darmon N, Drewnowski A. Contribution of food prices and diet cost to socioeconomic disparities in diet quality and health: a systematic review and analysis. Nutr Rev. 2015;73(10):643–60. 14. UNICEF. Food Systems: What they are, why they matter. https:// www. unicef. org/ stori es/ food- syste ms. Accessed 8 Oct 2022. 15. Charlton EL, Kähkönen LA, Sacks G, Cameron AJ. Supermarkets and unhealthy food marketing: an international comparison of the content of supermarket catalogues/circulars. Prev Med. 2015;81:168–73. 17. Ethan D, Basch CH, Rajan S, Samuel L, Hammond RN. A comparison of the nutritional quality of food products advertised in grocery store circulars of high- versus low-income New York City zip codes. Int J Environ Res Public Health. 2013;11(1):537–47. 18. Cameron AJ, Sayers SJ, Sacks G, Thornton LE. Do the foods advertised in Australian supermarket catalogues reflect national dietary guidelines? Health Promot Int. 2017;32(1):113–21. 19. European Commission. Digital Economy and Society Index (DESI) 2022 Sweden. European Commission; 2022. Available from: https:// digital-strategy.ec.europa.eu/en/policies/countries-digitisation-perfor- mance [cited date 4 April 2023]. 20. Statistics Sweden. Inflation rate was 8.0 percent in July 2022. Accessed 4 April 2023. 21. Nordic Council of Ministers. Nordic Nutrition Recommendations 2012 - Integrating nutrition and physical activity. Copenhagen: Nordic Council of Ministers; 2014. ISBN; 978–92–893–2670–4. Available from: https:// www.livsmedelsverket.se/globalassets/publikationsdatabas/andra-sprak/ nordic-nutrition-recommendations-2012.pdf [cited date 17 June 2022]. 22. Swedish Food Agency. Nordiska näringsrekommendationer 2012 - rek- ommendationer om näring och fysisk aktivitet. Uppsala: Swedish Food Agency. Available from: https://www.livsmedelsverket.se/globalassets/ publikationsdatabas/broschyrer-foldrar/nordiska-naringsrekommenda- tioner-2012-svenska.pdf [cited date 17 June 2022]. 23. Martin-Biggers J, Yorkin M, Aljallad C, Ciecierski C, Akhabue I, McKinley J, et al. What foods are US supermarkets promoting? a content analysis of supermarket sales circulars. Appetite. 2013;62:160–5. 24. Ethan D, Samuel L, Basch C. An analysis of bronx-based online grocery store circulars for nutritional content of food and beverage products. J Community Health. 2013;38(3):521–8. 25. Jahns L, Payne C, Whigham L, Johnson L, Scheett A, Hoverson B, et al. Foods advertised in US weekly supermarket sales circulars over one year: a content analysis. Nutr J. 2014;13(1):95. 26. Ravensbergen EAH, Waterlander WE, Kroeze W, Steenhuis IHM. Healthy or unhealthy on sale? a cross-sectional study on the proportion of healthy and unhealthy foods promoted through flyer advertising by supermar- kets in the Netherlands. BMC Public Health. 2015;15(1):470. 27. Hendriksen A, Jansen R, Dijkstra SC, Huitink M, Seidell JC, Poelman MP. How healthy and processed are foods and drinks promoted in supermar- ket sales flyers? a cross-sectional study in the Netherlands. Public Health Nutr. 2021;24(10):3000–8. 28. Coop. Nu blir de första Nettobutikerna Coop. https:// press rum. coop. se/ nu- blir- de- forsta- netto butik erna- coop/ (2019). Accessed 17 June 2022. 29. DLF. Dagligvarukartan 2019. Stockholm: DLF; 2019. 30. Hemköp. Det här är Hemköp. https:// www. hemkop. se/ artik el/ om- hem- kop. Accessed 17 June 2022. 31. DLF. Dagligvarukartan 2017. Stockholm: DLF, DELFI, HUI Research; 2017. 32. Livsmedelsverkets föreskrifter om användningen av symbolen Nyckelhålet (LIVSFS 2005:9). Uppsala: Swedish Food Agency Available from:https://www.livsmedelsverket.se/globalassets/om-oss/lagstift- ning/livsmedelsinfo-till-konsum---markning/livsfs-2005-9-kons-2021-1. pdf [cited date 17 June 2022]. 33. Bjerselius R, Konde ÅB, Sanner Färnstrand J. Consumption of red and processed meats in relation to colorectal cancer - Risk and benefit man- agement report. Uppsala: Swedish Food Agency; 2014. Swedish Food Agency’s report series nr 20/2014. 34. World Health Organization. WHO Regional Office for Europe Nutrient Profile Model. Copenhagen: World Health Organization; 2015. Available from:https://www.euro.who.int/__data/assets/pdf_file/0005/270716/ Nutrient-children_web-new.pdf [cited date 17 June 2022]. 35. Systembolaget. Alkoholrapporten 2021. Stockholm: Systembolaget; 2021. Available from:https://www.omsystembolaget.se/globalassets/pdf/ alkoholrapporten-2021-a4.pdf [cited date 18 June 2022] . 36. Public Health Agency of Sweden. Folköl - beskrivning av servering och detaljhandel. (2022). Accessed 18 June 2022. 37. Statistics Sweden. RegSO - Regionala Statistikområden. Accessed 17 June 2022. Mjöberg et al. BMC Public Health (2023) 23:795 Page 9 of 9 38. Statistics Sweden. Socioekonomiskt index efter RegSO. År 2011–2019 2021. #. Accessed 17 June 2022. 39. Fagerberg P, Langlet B, Oravsky A, Sandborg J, Lof M, Ioakimidis I. Ultra- processed food advertisements dominate the food advertising landscape in two Stockholm areas with low vs high socioeconomic status. Is it time for regulatory action? BMC Public Health. 2019;19(1):1717. 40. Power E. Determinants of healthy eating among low-income Canadians. Can J Public Health. 2005;96:37–42. 41. Ahrens W, Bammann K, de Henauw S, Halford J, Palou A, Pigeot I, et al. Understanding and preventing childhood obesity and related disor- ders—IDEFICS: a European multilevel epidemiological approach. Nutr Metab Cardiovasc Dis. 2006;16(4):302–8. 42. Steptoe A, Wardle J. Motivational factors as mediators of socioeconomic variations in dietary intake patterns. Psychol Health. 1999;14(3):391–402. 43. Wardle J, Steptoe A. Socioeconomic differences in attitudes and beliefs about healthy lifestyles. J Epidemiol Community Health. 2003;57(6):440–3. 44. Allen L, Williams J, Townsend N, Mikkelsen B, Roberts N, Foster C, et al. Socioeconomic status and non-communicable disease behavioural risk factors in low-income and lower-middle-income countries: a systematic review. Lancet Glob Health. 2017;5(3):e277–89. 45. Messerli P, Kim EM, Lutz W, Moatti J-P, Richardson K, Saidam M, et al. Expansion of sustainability science needed for the SDGs. Nature Sustain- ability. 2019;2(10):892–4. 46. Rutter H, Savona N, Glonti K, Bibby J, Cummins S, Finegood DT, et al. The need for a complex systems model of evidence for public health. The Lancet. 2017;390(10112):2602–4. 47. World Health Organization. Essential nutrition actions: mainstreaming nutrition through the life-course. Geneva: World Health Organization; 2019. ISBN; 978–92–4–151585–6. Available from: [cited date 17 June 2022]. 48. Public Health Agency of Sweden. Folkhälsans utveckling. Årsrapport 2020. Solna: Public Health Agency of Sweden; 2020. Report No: 20003. Available from: [cited date 17 June 2022]. 49. Mayén A-L, de Mestral C, Zamora G, Paccaud F, Marques-Vidal P, Bovet P, et al. Interventions promoting healthy eating as a tool for reducing social inequalities in diet in low- and middle-income countries: a systematic review. International journal for equity in health. 2016;15(1):205. 50. Thaler RH. Nudge, not sludge. Science. 2018;361(6401):431. 51. Public Health Agency of Sweden. Matmiljöns betydelse för vår hälsa. Solna: Public Health Agency of Sweden; 2021. Report No: 21030. Avail- able from:https://www.folkhalsomyndigheten.se/contentassets/79017b0 52e0741f98193437e8f04560d/matmiljon-betydelse-halsa.pdf [cited date 17 June 2022]. 52. Glanz K, Bishop DB. The role of behavioral science theory in development and implementation of public health interventions. Annu Rev Public Health. 2010;31(1):399–418. 53. Houghtaling B, Serrano EL, Kraak VI, Harden SM, Davis GC, Misyak SA. A systematic review of factors that influence food store owner and man- ager decision making and ability or willingness to use choice architecture and marketing mix strategies to encourage healthy consumer purchases in the United States, 2005–2017. Int J Behav Nutr Phys Act. 2019;16(1):205. 54. Clark MA, Springmann M, Hill J, Tilman D. Multiple health and environ- mental impacts of foods. Proc Natl Acad Sci. 2019;116(46):23357–62. 55. Afshin A, Penalvo J, Del Gobbo L, Kashaf M, Micha R, Morrish K, et al. CVD prevention through policy: a review of mass media, food/menu labeling, taxation/subsidies, built environment, school procurement, worksite wellness, and marketing standards to improve diet. Curr Cardiol Rep. 2015;17(11):98. 56. O’Rourke D, Lollo N. Transforming consumption: from decoupling, to behavior change, to system changes for sustainable consumption. Annu Rev Environ Resour. 2015;40(1):233–59. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12961-020-00673-y	Onoya et al. Health Res Policy Sys (2021) 19:2 https://doi.org/10.1186/s12961-020-00673-y RESEARCH Open Access Health provider perspectives on the implementation of the same-day-ART initiation policy in the Gauteng province of South Africa Dorina Onoya1† Jacob Bor1,3,4, Jonas Langa2 and Matthew P. Fox1,3,4 , Idah Mokhele1†, Tembeka Sineke1, Bulelwa Mngoma2, Aneesa Moolla1, Marnie Vujovic2, Abstract Background: In September 2016, South Africa (SA) began implementing the universal-test-and-treat (UTT) policy in hopes of attaining the UNAIDS 90-90-90 targets by 2020. The SA National Department of Health provided a further directive to initiate antiretroviral therapy (ART) on the day of HIV diagnosis in September 2017. We conducted a qualitative study to determine the progress in implementing UTT and examine health providers’ perspectives on the implementation of the same-day initiation (SDI) policy, six months after the policy change. Methods: We conducted in-depth interviews with three professional nurses, and four HIV lay counsellors of five pri- mary health clinics in the Gauteng province, between October and December 2017. In September 2018, we also con- ducted a focus group discussion with ten professional nurses/clinic managers from ten clinic facilities. The interviews and focus groups covered the adoption and implementation of UTT and SDI policies. Interviews were conducted in English, Sotho or Zulu and audio-recorded with participant consent. Audio-recordings were transcribed verbatim, translated to English and analysed thematically using NVivo 11. Results: The data indicates inconsistencies across facilities and incongruities between counsellor and nursing pro- vider perspectives regarding the SDI policy implementation. While nurses highlighted the clinical benefits of early ART initiation, they expressed concerns that immediate ART may be overwhelming for some patients, who may be unpre- pared and likely to disengage from care soon after the initial acceptance of ART. Accordingly, the SDI implementation was slow due to limited patient demand, provider ambivalence to the policy implementations, as well as challenges with infrastructure and human resources. The process for assessing patient readiness was poorly defined by health providers across facilities, inconsistent and counsellor dependent. Providers were also unclear on how to ensure that patients who defer treatment return for ongoing counselling. Conclusions: Our results highlight important gaps in the drive to achieve the ART initiation target and demonstrate the need for further engagement with health care providers around the implementation of same-day ART initiation, particularly with regards to infrastructural/capacity needs and the management of patient readiness for lifelong ART Correspondence: donoya@heroza.org †Dorina Onoya and Idah Mokhele contributed equally to this work 1 Health Economics and Epidemiology Research Office, Department of Internal Medicine, School of Clinical Medicine, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa Full list of author information is available at the end of the article © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Onoya et al. Health Res Policy Sys (2021) 19:2 Page 2 of 12 on the day of HIV diagnosis. Additionally, there is a need for improved promotion of the SDI provision both in health care settings and in media communications to increase patient demand for early and lifelong ART. Keywords: HIV, ART attrition, Universal-test-and-treat, Same-day ART , Health provider Background Sub-Saharan Africa remains the region worst affected by the HIV epidemic, accounting for more than two- thirds of the global HIV burden [1]. Despite this, the region has seen substantial gains in the fight against HIV in recent years with the expansion of antiretroviral therapy (ART) eligibility, and subsequent adoption of the World Health Organisation (WHO) recommended universal-test-and-treat (UTT) policy [2–4]. However, many health systems across Sub-Saharan Africa remain weak, under-resourced and overburdened [5–7]. Fur- thermore, many countries in the region faced chal- lenges in meeting UNAIDS 90-90-90 HIV targets and fully realising the benefits of the UTT policy due to persistent health system deficiencies [1, 7, 8]. South Africa (SA) bears the largest HIV burden in the region, with nearly eight million individuals living with HIV, and over four and a half million of these receiv- ing ART in 2019 [9, 10]. However, despite considerable efforts to scale-up access to treatment, an additional three million individuals need to start ART to reach 95% of HIV diagnosed patient on ART by 2030 [4, 11, 12]. Evidence of the benefits of ART are well-documented [13–15]. As a result, the SA government adopted the UTT strategy in 2016 and the ART same-day initiation (SDI) policy in 2017 [13, 15–19]. While the UTT policy removes clinical barriers to ART initiation, the SDI pol- icy aims to reduce the time from HIV diagnosis to ART start to one visit. The SDI policy makes ART initiation logistically easier for patients and can further reduce patient losses in the pre-ART phase of care. How- ever, in SA and other low-and-middle-income coun- tries (LMIC), the implementation of the UTT and SDI policies were not accompanied by expanded human resources or infrastructural capacity [5–7]. The result- ing demand for ART can potentially increase the pres- sure on already burdened public health services, and further compromised the quality of care [16, 20, 21]. Therefore, it is, essential to understand how public sector healthcare providers have received these poli- cies and managed their enactment to identify gaps and devise solutions to maximise and sustain benefits. In this study, we aimed to explore progress towards UTT policy assimilation and examine primary health care providers’ perspectives on the implementation of same- day ART initiation after the policy adoption. Methods Study setting and sites The study was conducted at eleven primary health clin- ics in Johannesburg, South Africa between October 2017 and September 2018. Participant characteristics A total of fifteen health providers from eleven clinics par- ticipated in the study (Table 1). Eleven of the health pro- viders were professional nurses, and eight of these were also facility managers. Four lay HIV counsellors were also interviewed from four different clinics. Key informant interview procedures We conducted key informant interviews with three pro- fessional nurses and four lay HIV counselling and testing counsellors at four of the study sites (Table 1). Interviews lasted approximately 45 min and were conducted in a private space within the clinic by a trained interviewer. The interview guide explored processes involved in HIV testing services, ART initiation, patient management and follow-up under the UTT and SDI policies, and also explored providers’ understanding and attitudes towards the policy changes and implementation processes. Inter- views were conducted in English, Sotho or Zulu, and were audio-recorded. Audio recordings were transcribed verbatim and translated to English for analysis. Focus group procedures Additional data was collected via a focus group discus- sion with professional nurses/clinic managers from ten primary health clinic facilities, covering topics related to the management of the UTT and SDI implementation processes. The discussion lasted approximately 60 min Table 1 Overview of study participants Type of provider Data collection activity Total Key informant interviews Focus group discussion Professional nurse (PN) Clinic manager/professional nurse Lay HIV counsellor (LHC) Total – 3 4 7 3 7 – 10 3 10a 4 15a a Two clinic manager/professional nurse took part in both KI interviews and the FGD Onoya et al. Health Res Policy Sys (2021) 19:2 Page 3 of 12 and was conducted in a venue provided by the research- ers, outside the health providers’ workplaces. Discussions were conducted in English and were also audio recorded. All audio recordings were transcribed verbatim. Data analysis All transcripts were analysed thematically using NVIVO software which facilitated data management and coding. Transcripts were read and coded by three research team members individually. Initial themes were drawn from topics covered in the interview guide. Major trends and cross-cutting themes were identified and then refined over several meetings. Any coder variation identified was resolved through discussion and consensus from all research team members. All participants provided written informed consent before all data collection procedures. Confidentiality and anonymity were safeguarded by removing all identifiers, including participants’ names and names of facilities from the data. This study was approved by the Human Research Ethics Committee (Medical) of the Univer- sity of the Witwatersrand (Wits HREC M1704122 and M170579). Results Primary health care providers highlighted several fac- tors affecting policy implementation at the healthcare management level, clinic and provider levels, and patient- level which are summarised below, and in Fig. 1, with supporting quotations presented in Table 2. Universal‑test‑and‑treat and same‑day ART initiation policy knowledge Acknowledged policy value in the context of South Africa All primary care providers were aware of the policy changes and understood the significance of early ART initiation to improve clinical outcomes but also to pre- vent transmission to HIV negative partners. However, they expressed concerns around the implementation of the same-day ART initiation directive, given the prevail- ing clinic resource and their perception of patient psy- chological needs and social circumstances. “…like I say, the aim [of the policy changes] is to put everyone on [treatment], everybody who is positive so that there must not be any transmission”—PN Fig. 1 Summary of barriers and facilitators to the same-day ART policy implementation at primary healthcare facilities in Johannesburg, South Africa Onoya et al. Health Res Policy Sys (2021) 19:2 Page 4 of 12 Table 2 Policy merit and healthcare worker knowledge related facilitators and barriers to UTT policy and same-day ART initiation implementation Barriers and facilitators Illustrative quotes Facilitators Policy value in the context of South Africa "…like I say, the aim [of the policy changes] is to put everyone on [treatment], everybody Knowledge of policy objectives who is positive so that there must not be any transmission”—PN “I like it [UTT] because some people they say we are going to see, like if the CD4 count is less than 500 they start to initiate. But if it’s more than 500 they don’t initiate that person, and that person they get sick. They get sick even though their CD4 count is high. So, that is not right. To start initiate the person who is very sick, because the side effects worse in that person, it is going to be worse for the person. So, it’s better to test and treat while they’re still healthy”—LHC Barriers Exclusion of facility-level stakeholders in policy formulation “For me, it would have worked very better if before a policy is being designed for the Perceived contradictions between ART guidelines and proposed process for ART initiation Lack of clear operational guidance to implementers facility, even before the level of the operations manager, they call their workshop. So, that they’ve got buy-in and understanding. Because sometimes you will try to explain a policy that you, yourself don’t understand…”—PN “I think again patients’ readiness, there’s subjective readiness and there’s objective readi- ness. My being readiness to [initiate] doesn’t necessarily meaning I qualify for you to initiate, we still have those patients that we will need to fast-track, the guidelines remains the same, antenatal, TB patients. I might be ready to take but then we have to exclude other conditions. You find out I’m ready, I’m saying yes, I’m ready, prob- ably because… when you get deeper into literature, you know those patients, there’s certain categories of patients that you know such patients are better ready compared to these patients.:—PN “Like when you attended meetings… they kept saying you can initiate patients immedi- ately. But, you know the clinicians [nurses] were not that confident because they were saying they don’t have anything written in black and white that guides them.”—PN “…what I noticed especially with the nurses at our clinic is that they were also resistant to the very UTT because it was not implemented like when NIMART implemented, people went for training, you know, expensive training. But with UTT, just a memo came and said from now on you can do this, and you don’t have to wait for this, and you don’t have to wait for that.”—PN Lack of clear operational guidance to implementers Primary healthcare providers highlighted the lack of detailed guidance around the new policy as a critical barrier to the smooth transition to the new policies. Facility managers noted that staff nurses were hesitant to implement immediate ART because they perceived contradictions between the previous (detailed) ART guidelines and the proposed process for ART initiation on the day of HIV diagnosis. As a result, most facility managers struggled to motivate their staff to implement SDI when the directive was first received. Many still stressed the need to ensure that patients were clinically ready for ART before initiation. They were concerned that initiating patients on ART without baseline safety laboratory test results was contrary to their training and prevailing HIV treatment guidelines. Some ref- erenced the extensive training that was provided to nurses in preparation for the implementation of the Nurse Initiated Management of ART (NIMART) as a good example of how the SDI directive should have been introduced [22]. Training for the NIMART pro- gram contained detailed process guidelines creating self-efficacy in nurses who were then tasked with ART initiation and management at primary health care clin- ics [22–24]. “Like when you attended meetings… they kept say- ing you can initiate patients immediately. But, you know the clinicians [nurses] were not that confident because they were saying they don’t have anything written in black and white that guides them.”—PN. “…what I noticed especially with the nurses at our clinic is that they were also resistant to the very UTT because it was not implemented like when NIMART implemented, people went for training, you know, expensive training. But with UTT, just a memo came and said from now on you can do this, and you don’t have to wait for this, and you don’t have to wait for that.”—PN. Inclusion of facility‑level stakeholders in policy formulation and implementation planning Participants underlined the necessity of involving facility- based health providers in policy implementation plan- ning. Primary care providers were familiar with clinic Onoya et al. Health Res Policy Sys (2021) 19:2 Page 5 of 12 conditions and primarily responsible for the actual policy implementation. “For me, it would have worked very better if before a policy is being designed for the facility, even before the level of the operations manager, they call their workshop. So, that they’ve got buy-in and under- standing. Because sometimes you will try to explain a policy that you, yourself don’t understand…”—PN. Health system resources and capacity Extreme human resources and infrastructural constraints Primary care clinics have been tasked with HIV testing, ART initiation and HIV patient management since 2010 as part of efforts to expand patients’ access to ART [23, 24]. The same-day ART initiation policy eliminates the pre-ART activities and now compresses all ART prepara- tion sessions into one long initial (diagnosis) visit, limit- ing the total number of patients that could be seen per day. Human health resource shortages were cited as a critical barrier to policy implementation. Providers per- ceived current staffing capacity to be already strained, and that the same-day ART initiation policy would exac- erbate the already overcrowded conditions at the clinics. Clinic managers also noted that the expectation to lead the implementation of the new policies in clinics that were initially designed and resourced to provide mainly primary (non-HIV) care services was challenging. often remain unfulfilled. The resource limitations require some creativity in the use of clinic space (offering their administration offices or, in some cases, the emergency room to be used as consulting rooms), staff and clinic processes to manage high volumes of patients more effectively (Table 3). There was a heavy reliance on support partners to assist with the implementation of new policies. These included non-governmental organisations (NGOs) that provided technical support and sometimes temporary structures to address human resources and consultation space con- straints. Often, these NGO partners are also relied on for training HIV counselling and testing staff. In most cases, NGO partners place additional counsellors or tracing officers in already overstretched clinics with sometimes unclear long-term added benefits. The support of NGO partners is particularly noted in the capturing and man- agement of ART monitoring data. Nonetheless, primary care providers believed that NGO partners’ technical support efforts would have lit- tle impact on their clinics’ ability to sustainably achieve their goals if resource and infrastructure shortfalls are not addressed. The majority of participants expressed an expectation that NGO partners should negotiate on behalf of facilities that they are contracted to support, considering their familiarity with the clinic resources needs and the partners’ apparent access to higher-level health authorities and funding organisations. “… much as it’s [SDI] not a new service that we are implementing, […] even at the level of the clinician that time that the clinician takes to make a proper initiation […] and the fact that the resources were never altered. The very same clinicians that were doing other things is expected to do that. Now you are seated with the clinician and that counsellor who are compromised in terms of the quality and standard of care that they can provide, but expected on the other hand, to be implementing the policy itself ”—PN. “Because our clinic when it was formed, or built or whatever…I don’t know exactly the history but it wasn’t for all the services. Hence…It’s a very small clinic, and the service was just for family planning, EPI [expanded program on immunization] and but then it started increasing service…and then in June 2016 we started for the first to actually give treat- ment to HIV positive clients”—PN. Adaptable healthcare providers and non‑governmental organisations (NGO) partner dependency Facility managers emphasised that their requests for additional human, infrastructural and material resources “So, with space issues, I think there was a time when we used to do that…when we used to use the emer- gency room and then you are busy with a client and then an emergency comes in and you have to go, where do you go?”—PN. “Hence when we started I mentioned the duties of NGOs…how do they come…how do you say I’m going to support this person…how are you going to? Firstly, the infrastructure is absolutely wrong. You don’t even fit in there [physically], human resource- wise you can’t squeeze somebody there but you say that I’m going to support these people.”—PN. Clinical management of HIV positive patients Knowledgeable patients and ART‑ready patients Nurses highlighted that urban patients are increasingly knowledgeable about HIV and ART, thus reducing some of the ART readiness challenges. HIV and ART informa- tion campaigns lighten the health education burden of HIV counselling and testing counsellors who can focus on correcting myths. However, they expressed fears that knowledge about ART as a prevention intervention would reduce the emphasis on other HIV/STI preven- tion messages, particularly the use of barriers protection Onoya et al. Health Res Policy Sys (2021) 19:2 Page 6 of 12 Table 3 Health system resources and capacity related facilitators and barriers to UTT policy and same-day ART initiation implementation Barriers and facilitators Illustrative quotes Facilitators Existing minimum capacity for immediate policy assimilation “…much as [SDI] it’s not a new service that we are implementing, but if you’re going to do that [implementing SDI], even at the level of the clinician, that time that the clini- cian takes to make a proper initiation to the patient and the fact that the resources were never altered.”—PN Flexible healthcare providers conditions Technical support partner organisation Barriers Lack of provider implementation readiness Extreme human resources and infrastructural constraints “So, with space issues, I think there was a time when we used to do that…when we used to use the emergency room and then you are busy with a client and then an emergency comes in and you have to go, where do you go?”—PN “Sometime I would have to vacate my office, and say okay finish whatever and then I’m just…you know…roaming around the clinic”—PN “So, they were like, no we don’t have anything in black and white that covers us. But, fortunately [because of the NGO support partner], we had something. They gave us something and referred us to the UTT policy and then they gave us the flow chart, just to guide us to that if patients do this, at least we’ve got the flow chart…”—PN “…what I noticed especially with the nurses at our clinic is that they were also resistant to the very UTT because it was not implemented like when NIMART implemented, people went for training, you know, expensive training. But with UTT, just a memo came and said from now on you can do this, and you don’t have to wait for this, and you don’t have to wait for that.”—PN “Because our clinic when it was formed, or built or whatever…I don’t know exactly the history but it wasn’t for all the services. Hence…It’s a very small clinic, and the service was just for family planning, EPI and but then it started increasing service…and then in June 2016 we started for the first to actually give treatment to HIV positive clients”—PN “[…] the fact that the resources were never altered, the very same clinicians that were doing other things is expected to do that. Now you are seated with the clinician and that counsellor who are compromised in terms of the quality and standard of care that they can provide, but expected on the other hand, to be implementing the policy itself”—PN “Hence when we started I mentioned the duties of NGOs…how do they come… how do you say I’m going to support this person…how are you going to? Firstly, the infrastructure is absolutely wrong. You don’t even fit in there [physically], human resource-wise you can’t squeeze somebody there but you say that I’m going to sup- port these people.”—PN methods. Health providers also expressed concerns that some prevention messages (such as the availability of pre- exposure prophylaxis) may be suppressed because of cost and logistical considerations. Hence the need for trained counsellors to handle increasingly complex HIV coun- selling and testing requirements to prepare patients for rapid ART initiation and lifelong adherence to treatment. “Today we live with people who are HIV posi- tive. Although we may speak of people who are in a remote place…and you know…this person…you may give us cues relating to that. But for somebody who is living in CBDs and all that urban life, and all that stuff. They should know about [HIV], even at workplace, I think that’s something…on TVs we talk about HIV…you know…or related things.”—PN. “It doesn’t necessarily mean transmission will only be stopped by treatment. I think these days were just talking about treatment, treatment—we don’t talk about condoms anymore, we don’t talk about healthy lifestyle anymore. We just want treatment, treatment…”—PN. Counsellor skill limitations Health providers were nearly unanimous in their call for additional support and training for lay counsellors because of their critical role patients’ engagement in the HIV care cascade. Additional skills training is needed to assist HIV counsellors in convincing and rapidly pre- paring patients to start ART under the same-day ART initiation policy. There was a perception that patients who are uncertain about starting ART may require pro- fessional counselling, perhaps beyond the skill set of current lay counsellors. Adherence counselling was the broad term used by providers for the information ses- sion on ART. However, providers were uncertain of the exact content of the counselling provided, and opinion Onoya et al. Health Res Policy Sys (2021) 19:2 Page 7 of 12 was divided on whether it was effective. Those who thought it was effective highlighted it as an essential facilitator in informing patients on clinic ART initiation procedures, psychological responses to HIV diagnosis ART and adherence expectations. Those in doubt of the effectiveness of adherence counselling referenced the high number of patient attrition and ART defaulters as an indication of its inadequacy (Table 4). When HIV counsellors were asked about what their approach for addressing patient ambivalence with regards to starting ART, most lay HIV counsellors emphasised the dangers of not taking treatment and personal health benefits of taking up early ART. HIV counsellors may overly highlight the dangers of not starting ART, before resolving social barriers to adherence such as disclosure to partners and family. However, HIV counsellors who have had a positive personal experience with ART were Table 4 Clinical management of HIV positive patients related facilitators and barriers to UTT policy and same-day ART initiation implementation Barriers and facilitators Illustrative quotes Facilitators Knowledgeable patients and ART-ready patients “Today we live with people who are HIV positive. Although we may speak of people who are in a remote place…and you know…this person…you may give us cues relating to that. But for some- body who is living in CBDs and all that urban life, and all that stuff. They should know about [HIV], even at workplace, I think that’s something…on TVs we talk about HIV…you know…or related things.”—PN “It doesn’t necessarily mean transmission will only be stopped by treatment. I think these days were just talking about treatment, treatment—we don’t talk about condoms anymore, we don’t talk about healthy lifestyle anymore. We just want treatment, treatment…”- PN Reduced system barriers to ART “… because like I say, the aim is to put everyone on treatment, everybody who is positive so that there must not be any transmission, and that’s how we deal with AIDS…”—PN Barriers Counsellor skill limitations Limited patient ART demand creation efforts Patient and clinical readiness for ART Excess emphasis on ART targets, need to report ART deferral reasons “… we came up with this (policy), and now we are ready for UTT, but did we go back and look at the cadres of counselling that we have? To say, when they need to communicate that to the patients, how much intense can they go in order for the patients to be able to say “okay I can be motivated” [to initiate ART] or “no give me a chance [to think about initiating ART]”.”—PN “… I used to strongly believe that in issues where you do adherence counselling, there’s a certain length that the counsellor can go up to. Beyond that it needs a professional person.”—PN “We have got adherence counselling, but our non- suppressing patients, their levels [viral load] are very high. Which says to you that maybe the content or the counselling that is done is not getting through to the patients.”—PN “…people expected people will come in numbers to say, yah we want ARVs. But it’s not really hap- pening like that because I think people are still exercising their right to choose whatever is that they want. Much as they said UTT…and then they thought the following day people will just come and say yah I want ARVs. It’s not…I don’t see it happening…”—PN “[Uptake of SDI]is very low. I think maybe it’s our mentality because some of the patients have been tested. They know the policy said “you will be initiated after the blood results”, but now all of a sudden something came up.”—PN “No, I don’t think same-day ART initiation will work, the person has to accept that this is what is hap- pening when he/she comes back from the clinic he/she shouldn’t be surprised… we will be able to assist him/her better than initiating him/her whilst still shocked, whilst crying. What is he/she going to do with the treatment? On the other hand, the husband tells her that he doesn’t want someone that takes a treatment. You see those kind of things? So that is no.”—LHC “… I think that it is fine if we allow a person to go and think about taking ARVs then if he/she has processed it and feels that he/she is ready that is when they can come back and say I am ready for starting on ARVs…”—LHC “Ask the patient to give you back the information that you have just told him/her, you will see that okay, and this person has heard what you said or he/she was not listening. That’s how I can tell that this one can initiate.”—LHC “now I’m doing my stats…and if I say my positivity rate…I tested twenty and my initiation rate for the week, this week, was 35%, I will be asked, why is it 35%. What do I say? It’s 35%, I initiated five out of the twenty and there’s nothing that I can do. It’s just a name-and-shame…and whatever reason you can come up with […] what you are saying, me I always write on the Treatment Reten- tion Acceleration Program (TRAP) the reasons, but it doesn’t change the fact that I am at 35%, so I am pulling the region down. It’s not like these things we don’t say, and sometime if you talk, it’s like you are negative, you are not open to UTT, but those are the realities that we deal with at facil- ity level, those are issues that are there at facility level and we need to talk about them”—PN Onoya et al. Health Res Policy Sys (2021) 19:2 Page 8 of 12 more inclined to underscore the benefits of early ART. None of the HIV counsellors who were interviewed men- tioned the benefit of ART as an HIV prevention tool. “…we came up with this [policy], and now we are ready for UTT/SDI, but did we go back and look at the cadres of counselling that we have? To say, when they need to communicate that to the patients, how much intense can they go in order for the patients to be able to say “okay I can be motivated” [to initiate ART] or “no give me a chance [to think about initi- ating ART]”.”—PN. “We have got adherence counselling, but our non- suppressing patients, their levels [viral load] are very high. Which says to you that maybe the content or the counselling that is done is not getting through to the patients.”—PN. Limited patient ART demand creation efforts Providers indicated that there was no systematic messag- ing or marketing regarding the new universal-test-and- treat and same-day ART policies. Thus, clinics reported low demand for same-day ART initiation, which was mainly provider-driven. Providers noted that patients may still remember past ART initiation procedures, including blood collection for baseline laboratory tests, a second visit (a week later) to receive blood test results and ART initiation later in the process. Additionally, pri- mary care providers seemed uncertain about effective ways for engaging patients who were previously ineligi- ble for ART [25], and forced to defer treatment initia- tion. There was mention of booking them for ongoing or adherence counselling. However, there seemed to be no systematic process for tracing previously ineligible HIV infected patients to now initiate them on ART. “…people expected people will come in numbers to say, yah we want ARVs. But it’s not really happen- ing like that because I think people are still exercis- ing their right to choose whatever is that they want. Much as they said UTT…and then they thought the following day people will just come and say yah I want ARVs. It’s not…I don’t see it happening…”—PN. Patient readiness for ART Besides clinical readiness for same-day ART, many health providers were concerned that patients’ social and emotional readiness for ART may be neglected. ART initiation on the day might be overwhelming and too sudden for some patients, implying that only highly motivated patients would take up ART on the day of HIV diagnosis and remain in care as required. There were concerns that rushing patients who need space to process the new diagnosis could result in disconnection from care after the initial acceptance of ART. Therefore, health providers favoured giving patients the necessary time to absorb the diagnosis and deal with the social prerequisites for sustainable ART adherence such as disclosure to a spouse/partner or family members, and arrangement for proper storage of antiretroviral drugs in their homes or workplaces. Health providers also stressed the importance of patients’ right to choose whether/when they take up ART. However, the process for assessing patient readiness was poorly defined among the different types of pro- viders and clinics. Unless the patient verbally indicated that they were not ready to start treatment, assessment of treatment readiness seems to depend on provider observations and attitude to immediate ART. Some counsellors deemed patients to be ready for ART if they demonstrated understanding of the information shared during the counselling session, others mentioned patients’ adherence to the post-HIV diagnosis follow- up visit schedule (i.e. returning a week later to collect baseline blood results) as an indication of readiness and commitment to lifelong ART. “No, I don’t think same-day ART initiation will work, the person has to accept that this is what is happening when he/she comes back from the clinic he/she shouldn’t be surprised… we will be able to assist him/her better than initiating him/her whilst still shocked, whilst crying. What is he/she going to do with the treatment? On the other hand, the husband tells her that he doesn’t want some- one that takes a treatment. You see those kinds of things? So that is no.”—LHC. “Ask the patient to give you back the information that you have just told him/her, you will see that okay, and this person has heard what you said or he/she was not listening. That’s how I can tell that this one can initiate.”—LHC. “I think again patients’ readiness, there’s subjective readiness and there’s objective readiness. My being readiness to [initiate] doesn’t necessarily mean- ing I qualify for you to initiate, we still have those patients that we will need to fast-track, the guide- lines remains the same, antenatal, TB patients. I might be ready to take but then we have to exclude other conditions. You find out I’m ready, I’m say- ing yes, I’m ready, probably because… when you get deeper into literature, you know those patients, there’s certain categories of patients that you know such patients are better ready compared to these patients.:—PN. Onoya et al. Health Res Policy Sys (2021) 19:2 Page 9 of 12 Emphasis on ART targets, need to report ART deferral reasons Healthcare managers expressed concerns that HIV monitoring indicators focused mainly on ART initiation numbers, excluding provider adherence to clinical guide- lines and patient preferences that may affect longer-term patient outcomes. HIV monitoring tools do not include indicators to explain failures in same-day ART initiation implementation. These monitoring challenges created frustration and sometimes undue influence to initiate patients on ART even when they are not ready for it. “now I’m doing my stats…and if I say my positivity rate…I tested twenty and my initiation rate for the week, this week, was 35%, I will be asked, why is it 35%. What do I say? It’s 35%, I initiated five out of the twenty and there’s nothing that I can do. It’s just a name-and-shame…and whatever reason you can come up with […] what you are saying, me I always write on the Treatment Retention Acceleration Pro- gram (TRAP) the reasons, but it doesn’t change the fact that I am at 35%, so I am pulling the region down. It’s not like these things we don’t say, and sometime if you talk, it’s like you are negative, you are not open to UTT, but those are the realities that we deal with at facility level, those are issues that are there at facility level and we need to talk about them”—PN. Discussions South Africa has made a substantial investment in its commitment to universal treatment coverage by adopt- ing the UTT and SDI guidelines and is committing to achieving the UNAIDS 95-95-95 targets by 2030 [2–4]. This is one of the first studies to gauge the perspectives and experiences of primary healthcare providers in the assimilation and implementation of the UTT and SDI policies in South Africa. This information is critical in understanding gaps in policy implementation process to improve the health systems’ performance in South Afri- can and also other many LMICs that have adopted these policies particularly in the Sub-Saharan African context. We found that primary care providers were knowl- edgeable about the UTT and SDI policies, and generally regarded them positively, highlighting the clinical and public health benefits. Similar opinions were shared by health care providers in high-income and other LMIC settings during ART eligibility expansion from a CD4 of 350–500 cells/µl, and after UTT policy adoption, and likewise among those implementing the Option B+ strategy of universal-testing and initiation of lifelong ART among all HIV-positive pregnant and breastfeeding women [26–31]. Study participants identified health system resource and capacity challenges in the implementation of the UTT and SDI policies in South Africa. Key barriers high- lighted by providers were the surge in workload coupled with constrained healthcare infrastructure and human resources for health. Increases in workload were pro- jected in modelling studies from before policy changes [7, 32]. An estimated 10.3 million HIV-positive individu- als became eligible for ART in Sub-Saharan Africa when the UTT policy was adopted [7, 32–34]. In general, the existing infrastructure and health human resource chal- lenges were not substantially addressed in LMICs [31, 35–38]. Adequate health resources are known to be critical components in ensuring quality health services and positive patient outcomes [28, 38–41]. While NGO partners have been instrumental in supporting policy implementation in South Africa and Sub-Saharan Africa [42, 43], they often cannot improve long-term infrastruc- tural needs [42, 44–46]. In South Africa, government-led initiatives such as the Ideal Clinic Realisation and Main- tenance (ICRM) programs and the Integrated Chronic Disease Management (ICDM) model are attempts to cor- rect resource and infrastructure deficiencies and improve the quality of primary health care services [36, 47]. Similar to other studies, our findings also noted healthcare providers’ concerns regarding patients feel- ing overwhelmed at the prospect of initiating lifelong ART immediately after diagnosis [28, 30, 31, 48, 49]. At the same time, study participants were uncertain of ways of assessing ART readiness challenges and were unclear of measures to ensure that patients who choose to defer ART remain engaged with the health system and promptly initiate ART when ready. Stated health-system and patient-related implementation challenges coupled with pressure on providers to meet ART initiation tar- gets may inadvertently compromise considerations for patient-level factors to ART readiness [21, 49]. Patient readiness to start life-long ART is complex and moti- vated by many personal and social factors and has been established as an important determinant of ART uptake and future adherence to ART [18, 25, 50, 51]. Unprepared patients who are compelled to initiate ART may disen- gage from care, ultimately limiting the potential benefits of the SDI policy provision. ART uptake has increased substantially since the adop- tion of the UTT policy in South Africa and other Sub- Saharan African countries [52–54]. However, recent evidence point to declining patient retention rates [55, 56]. Routine clinic data from Johannesburg and Mopani districts in South African districts showed a 45% higher likelihood of disengagement from care after six months from patients initiated on the same day of diagnosis com- pared to patients initiated later [56]. Onoya et al. Health Res Policy Sys (2021) 19:2 Page 10 of 12 Patient counselling is essential to helping patients to accept their HIV positive status and prepare for life-long ART, particularly among patients who are diagnosed at a relatively healthy state and may not perceive the immedi- ate benefits to early ART [28, 57–59]. Study participants noted challenges regarding current counsellors’ capacity to manage patients ambivalence about immediate ART as well as non-compliant patients. The need for improved, quality counselling in the era of same-day ART initiation was noted among study participants, a factor which has been previously highlighted [28, 59–61]. Helpful strategies to overcome some of the policy implementation challenges include differentiated-ser- vice-delivery (DSD) ART models aimed at decongesting primary health clinics facilities and freeing up profes- sional health worker time to focus on more complicated and sick patients [62–64]. Improved counselling strate- gies are needed to address patient ART readiness and improve long-term ART adherence and retention in care. Additionally, community-focused health promo- tion and media campaigns are needed to improve patient understanding of the benefits of early ART and their demand for same-day ART [28, 59, 65]. The findings from our study provide valuable insights from professionals at the forefront of ART policy imple- mentation to further support South Africa’s commitment to expanding access to ART. Policymakers will need to address the identified implementation challenges in col- laboration with frontline implementers to maximise the demonstrated benefits of the UTT and SDI policies. Limitations Limitations of the study include a small sample size. Data presented are from eleven clinics in the Gauteng prov- ince, which may be different from other facilities in the province and the country. Also, findings were based on opinions and perspectives of key informants who were health providers from a small subset of clinics in the Johannesburg metropolitan area, rather than on empiri- cal data from clinics. Lastly, the qualitative design also limits the generalizability of these findings. Conclusions Our results highlight important gaps in the drive to achieve the second UNAIDS 95% (diagnosed patients on ART) target. Specifically, the study demonstrates the need for further engagement with healthcare providers about the SDI policy, particularly infrastructural/capacity needs. Additionally, improved promotion of immediate ART in health care settings and media communication is needed to increase patient demand for early ART. Abbreviations ART : Antiretroviral therapy; CD4: Cluster of differentiation 4; DSD: Differenti- ated service delivery; FGD: Focus group discussion; HIV: Human immunodefi- ciency virus; HREC: Human Research Ethics Committee; KI: Key informant; LHC: Lay HIV counsellor; NGO: Non-governmental organisation; NIMART : Nurse initiated management of antiretroviral therapy; PHC: Primary health care; PN: Professional nurse; UNAIDS: Joint United Nations Programme on HIV/AIDS; WHO: World Health Organisation. Acknowledgements The authors would like to acknowledge facility managers and staff from participating clinics and the Gauteng provincial Department of Health to providing support for the study implementation. Authors’ contributions DO conceptualised and designed the study. DO, IM and TS developed the protocol and obtained relevant approvals. DO, IM, TS and BM were involved data collection and collation. DO, IM, TS and AM analysed the data and inter- preted findings. DO and IM wrote the original draft manuscript. AM and MPF provided feedback on the manuscript. All authors assisted in interpreting the results, critically reviewed and approved the final version of the manuscript. Funding This study has been made possible by the generous support of the American People and the President’s Emergency Plan for AIDS Relief (PEPFAR) through USAID under the terms of Cooperative Agreements AID-674-A-12-00029 and 72067419CA00004 to Health Economics and Epidemiology Research Office and under the terms of Cooperative Agreement 674-A-00-09-00018-00 to Boston University. The contents are the responsibility of the authors and do not necessarily reflect the views of PEPFAR, USAID or the United States Government. The funders had no role in the study design, collection, analysis and interpretation of the data, in manuscript preparation or the decision to publish. Availability of data and materials Qualitative data extracts are presented in the article to support the findings. The original transcripts are not available to the public as they may contain information that could compromise the confidentiality of study participants. Ethics approval and consent to participate The analysis of the anonymised retrospective data for this study was approved by the Human Research Ethics Committee (Medical) of the University of the Witwatersrand (M1704122 and M170579). Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1 Health Economics and Epidemiology Research Office, Department of Internal Medicine, School of Clinical Medicine, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa. 2 Right to Care, Johannesburg, South Africa. 3 Departments of Global Health, Boston University School of Public Health, Boston, MA, United States of America. 4 Departments of Epidemiology, Boston University School of Public Health, Boston, MA, United States of America. Received: 9 March 2020 Accepted: 16 December 2020 References 1. Joint United Nations Programme on HIV/AIDS (UNAIDS). Factsheet: Global AIDS update. 2019. UNAIDS Geneva; 2019. https ://www.unaid s.org/sites /defau lt/files /media _asset /UNAID S_FactS heet_en.pdf. Accessed 26 Nov 2020. 2. World Health Organization (WHO). Guideline on when to start antiretro- viral therapy and on pre-exposure prophylaxis for HIV. 2015. https ://apps. Onoya et al. Health Res Policy Sys (2021) 19:2 Page 11 of 12 who.int/iris/bitst ream/handl e/10665 /18627 5/97892 41509 565_eng.pdf. Accessed 26 Nov 2020. 3. UNAIDS. Understanding fast-track: accelerating action to end the AIDS epidemic by 2030. Geneva; 2015. https ://www.unaid s.org/sites /defau lt/files /media _asset /JC268 6_WAD20 14rep ort_en.pdf. Accessed 26 Nov 2020. 5. 4. UNAIDS. 90-90-90 An ambitious treatment target to help end the AIDS epidemic. 2014. https ://www.unaid s.org/sites /defau lt/files /media _asset /90-90-90_en.pdf. Accessed 26 Nov 2020. Takarinda K, Harries A, Mutasa-Apollo T. Critical considerations for adopt- ing the HIV “treat all” approach in Zimbabwe: is the nation poised? Public Health Action. 2016;6(1):3–7. Lahuerta M, Ue F, Hoffman S, Elul B, Kulkarni SG, Wu Y, et al. The problem of late ART initiation in Sub-Saharan Africa: a transient aspect of scale- up or a long-term phenomenon? J Health Care Poor Underserved. 2013;24(1):359. 6. 7. Dutta A, Barker C, Kallarakal A. The HIV treatment gap: estimates of the financial resources needed versus available for scale-up of antiret- roviral therapy in 97 countries from 2015 to 2020. PLoS medicine. 2015;12(11):e1001907. https ://doi.org/10.1371/journ al.pmed.10019 07. 8. Nansseu JRN, Bigna JJR. Antiretroviral therapy-related adverse effects: can sub-Saharan Africa cope with the new “test and treat” policy of the World Health Organization? Infect Dis Poverty. 2017;6(1):24. https ://doi. org/10.1186/s4024 9-017-0240-3. 9. Human Sciences Research Council (HSRC). The South African National HIV prevalence, incidence, behaviour and communication survey, 2017. HIV Impact assessment summary. Pretoria; 2017. http://www.hsrc.ac.za/ uploa ds/pageC onten t/9234/SABSS MV_Impac t_Asses sment _Summa ry_ZA_ADS_clear ed_PDFA4 .pdf. Accessed 26 Nov 2020. 10. South African National Department of Health (SA-NDoH). Annual report 2018/2019. In: Health Do, editor. Pretoria: NDOH; 2019. http://www.healt h.gov.za/index .php/2014-08-15-12-56-31?downl oad 3688:annua l-repor t-2018-2019. Accessed 26 Nov 2020. = 11. UNAIDS. Country factsheets: South Africa. Geneva, Switzerland; 2016. https ://www.unaid s.org/en/regio nscou ntrie s/count ries/south afric a. Accessed 26 Nov 2020. 12. UNAIDS. UNAIDS data 2017. UNAIDS: Geneva; 2017. https ://www.unaid s.org/en/resou rces/docum ents/2017/2017_data_book. Accessed 26 Nov 2020. 13. Cohen MS, Chen YQ, McCauley M, Gamble T, Hosseinipour MC, Kumar- asamy N, et al. Antiretroviral therapy for the prevention of HIV-1 transmis- sion. N Engl J Med. 2016;375(9):830–9. 14. Memirie S. Clinical outcome of children on HAART at police referral hospi- tal, Addis Ababa, Ethiopia. Ethiop Med J. 2009;47(2):159–64. 15. Sawyer SM, Afifi RA, Bearinger LH, Blakemore S-J, Dick B, Ezeh AC, et al. Adolescence: a foundation for future health. Lancet. 2012;379(9826):1630–40. 16. Boyer S, Iwuji C, Gosset A, Protopopescu C, Okesola N, Plazy M, et al. Factors associated with antiretroviral treatment initiation amongst HIV-positive individuals linked to care within a universal test and treat programme: early findings of the ANRS 12249 TasP trial in rural South Africa. AIDS Care. 2016;28(Suppl 3):39–51. 17. Motsoaledi A. Health budget vote speech by the Minister of Health. Pretoria: SA_NDoH. 2014. https ://www.gov.za/healt h-budge t-vote-speec h-minis ter-healt h-dr-aaron -motso aledi -mp. 18. SA-NDoH. Guidelines for expanding combination prevention and treat- ment options for sex workers: oral pre-exposure prophylaxis (PrEP) and test and treat (T&T). Pretoria. 2016. http://www.wrhi.ac.za/uploa ds/files / NDoH_PrEP_Natio nal_Guide lines _-_8_Febru ary_2017.pdf. Accessed 26 Nov 2020. 19. WHO. Consolidated guidelines on the use of antiretroviral drugs for treat- ing and preventing HIV infection: recommendations for a public health approach. 2nd ed. Geneva; 2016. https ://www.who.int/hiv/pub/arv/arv- 2016/en/. Accessed 26 Nov 2020. 20. Collins S, Geffen N. Community views: balancing the public health benefits of earlier antiretroviral treatment with the implications for indi- vidual patients–perspectives from the community. Curr Opin HIV AIDS. 2014;9(1):4–10. 21. Kulkarni SP, Shah KR, Sarma KV, Mahajan AP. Clinical uncertainties, health service challenges, and ethical complexities of HIV “Test-and-Treat”: a systematic review. Am J Public Health. 2013;103(6):e14–23. 22. Jones M, Cameron D. Evaluating 5 years’ NIMART mentoring in South Africa’s HIV treatment programme: successes, challenges and future needs. S Afr Med J. 2017;107(10):839–42. 23. Cameron D, Gerber A, Mbatha M, Mutyabule J, Swart H. Nurse-initiation and maintenance of patients on antiretroviral therapy: are nurses inpri- mary care clinics initiating ART after attending NIMART training? S Afr Med J. 2012;102(2):98–100. 24. Georgeu D, Colvin CJ, Lewin S, Fairall L, Bachmann MO, Uebel K, et al. Implementing nurse-initiated and managed antiretroviral treatment (NIMART) in South Africa: a qualitative process evaluation of the STRETCH trial. Implement Sci. 2012;7(1):66. 25. SA-NDoH. National consolidated guidelines for the prevention of mother-to-child transmission of HIV (PMTCT) and the management of HIV in children, adolescents and adults. Pretoria; 2015. http://www. healt h.gov.za/index .php/2014-03-17-09-09-38/polic ies-and-guide lines / categ ory/230-2015p ?downl oad 937:natio nal-art-guide lines -2015fi nal. Accessed 26 Nov 2020. = 26. Evans C, Bennett J, Croston M, Brito-Ault N, Bruton J. “In reality, it is com- plex and difficult”: UK nurses’ perspectives on “treatment as prevention” within HIV care. AIDS Care. 2015;27(6):753–7. 27. Kurth AE, Mayer K, Beauchamp G, McKinstry L, Farrior J, Buchacz K, et al. Clinician practices and attitudes regarding early antiretroviral therapy in the US. J Acquir Immune Defic Syndr (1999). 2012;61(5):e65. 28. Plazy M, Perriat D, Gumede D, Boyer S, Pillay D, Dabis F, et al. Implement- ing universal HIV treatment in a high HIV prevalence and rural South African setting—field experiences and recommendations of health care providers. PLoS ONE. 2017;12(11):e0186883. 29. Dovel K, Phiri K, Mphande M, Mindry D, Sanudi E, Bellos M, et al. Optimiz- ing test and treat in Malawi: health care worker perspectives on barriers and facilitators to ART initiation among HIV-infected clients who feel healthy. Glob Health Action. 2020;13(1):1728830. 30. Orange E. Assessing health policy implementation in South Africa: case study of HIV universal test and treat. Washington 2018. https ://digit al.lib. washi ngton .edu/resea rchwo rks/handl e/1773/41715 . 31. Laar AS, Dalinjong PA, Ntim-Adu C, Anaman-Torgbor JA. Understanding health facility challenges in the implementation of Option B in Ghana: the perspectives of health workers. J Hosp Manag Health Policy. 2018;2(6):29–37. + guidelines 32. Bor J, Ahmed S, Fox MP, Rosen S, Meyer-Rath G, Katz IT, et al. Effect of eliminating CD4-count thresholds on HIV treatment initiation in South Africa: an empirical modelling study. PLoS ONE. 2017;12(6):e0178249. 33. Nash D, Yotebieng M, Sohn AH. Treating all people living with HIV in sub-Saharan Africa: a new era calling for new approaches. J Virus Erad. 2018;4(Suppl 2):1. 34. UNAIDS. Fact Sheet November 2016. Geneva. 2016. 35. Mash R, Howe A, Olayemi O, Makwero M, Ray S, Zerihun M, et al. Reflec- tions on family medicine and primary healthcare in sub-Saharan Africa. BMJ Glob Health. 2018;3:e000662. 36. Hunter JR, Chandran TM, Asmall S, Tucker J-M, Ravhengani NM, Mok- galagadi Y. The Ideal Clinic in South Africa: progress and challenges in implementation. S Afr Health Rev. 2017;1:111–23. 37. Smith A, Ranchod S, Strugnell D, Wishnia J. Human resources for health planning and National Health Insurance: the urgency and the opportu- nity. S Afr Health Rev. 2018;1:23–31. 38. Eze BS, Ochonma OG, Miriam A, Onwujekwe O. The challenges of service delivery, health workforce, and infrustructure for the delivery of antiret- roviral therapy for HIV/AIDS infected-persons in south-east Nigeria. J Dis Glob Health. 2016;5(2):84–94. 39. Datar A, Mukherji A, Sood N. Health infrastructure & immunization cover- age in rural India. Indian J Med Res. 2007;125(1):31. 40. Ravallion M. “Achieving child-health-related millennium develop- ment goals: the role of infrastructure”—a comment. World Dev. 2007;35(5):920–8. 41. Schneider H, Blaauw D, Gilson L, Chabikuli N, Goudge J. Health systems and access to antiretroviral drugs for HIV in Southern Africa: service delivery and human resources challenges. Reprod Health Matters. 2006;14(27):12–23. 42. Hanefeld J. The impact of global health initiatives at national and sub- national level–a policy analysis of their role in implementation processes of antiretroviral treatment (ART) roll-out in Zambia and South Africa. AIDS Care. 2010;22(sup1):93–102. Onoya et al. Health Res Policy Sys (2021) 19:2 Page 12 of 12 43. Moon TD, Burlison JR, Sidat M, Pires P, Silva W, Solis M, et al. Lessons learned while implementing an HIV/AIDS care and treatment program in rural Mozambique. Retrovirology. 2010;3:1–14. 44. Grignon JS, Ledikwe JH, Makati D, Nyangah R, Sento BW, Semo B-W. Maximizing the benefit of health workforce secondment in Botswana: an approach for strengthening health systems in resource-limited settings. Risk Manag Healthcare Policy. 2014;7:91. 45. Kavanagh MM, Dubula-Majola V. Policy change and micro-politics in global health aid: HIV in South Africa. Health Policy Plan. 2019;34(1):1–11. 46. Henry J. Kaiser Family Foundation. Financing the response to AIDS in low- and middle-income countries: international assistance from donor governments in 2014. 2014. https ://kaise rfami lyfou ndati on.files .wordp ress.com/2015/06/7347-11-finan cing-the-respo nse-to-aids-in-low-and- middl e-incom e-count ries.pdf. Accessed on 26 Nov 2020. 47. Mahomed OH, Asmall S, Freeman M. An integrated chronic disease man- agement model: a diagonal approach to health system strengthening in South Africa. J Health Care Poor Underserved. 2014;25(4):1723–9. 48. Black S, Zulliger R, Marcus R, Mark D, Myer L, Bekker L-G. Acceptability and challenges of rapid ART initiation among pregnant women in a pilot programme, Cape Town, South Africa. AIDS Care. 2014;26(6):736–41. 49. Persson A. Significant ambivalence: perspectives of Australian HIV service providers on universal treatment-as-prevention (TasP) for serodiscordant couples. Crit Public Health. 2015;25(2):153–64. 50. Enriquez M, Lackey NR, O’connor MC, McKinsey DS. Successful adherence after multiple HIV treatment failures. J Adv Nurs. 2004;45(4):438–46. 51. Gebrekristos HT, Mlisana KP, Karim QA. Patients’ readiness to start highly active antiretroviral treatment for HIV. BMJ. 2005;331(7519):772–5. 52. Onoya D, Sineke T, Hendrickson C, Mokhele I, Maskew M, Long LC, et al. Impact of the test and treat policy on delays in antiretroviral therapy initiation among adult HIV positive patients from six clinics in Johannes- burg, South Africa: results from a prospective cohort study. BMJ Open. 2020;10(3):e030228. 53. Tymejczyk O, Brazier E, Yiannoutsos CT, Vinikoor M, van Lettow M, Nalugoda F, et al. Changes in rapid HIV treatment initiation after national “treat all” policy adoption in 6 sub-Saharan African countries: regression discontinuity analysis. PLoS Med. 2019;16(6):e1002822. 54. Rufu A, Chitimbire V, Nzou C, Timire C, Owiti P, Harries A, et al. Implemen- tation of the ’Test and Treat’policy for newly diagnosed people living with HIV in Zimbabwe in 2017. Public Health Action. 2018;8(3):145–50. 55. Hirasen K, Fox MP, Hendrickson CJ, Sineke T, Onoya D. HIV treatment outcomes among patients initiated on antiretroviral therapy pre and post-universal test and treat guidelines in South Africa. Ther Clin Risk Manag. 2020;16:169. 56. Lilian RR, Rees K, McIntyre JA, Struthers HE, Peters RP. Same-day antiretro- viral therapy initiation for HIV-infected adults in South Africa: analysis of routine data. PLoS ONE. 2020;15(1):e0227572. 57. Temprano ANRS 12136 Study Group. A trial of early antiretrovirals and isoniazid preventive therapy in Africa. N Engl J Med. 2015;373(9):808–22. 58. Pell C, Reis R, Dlamini N, Moyer E, Vernooij E. “Then her neighbour will not know her status”: how health providers advocate antiretroviral therapy under universal test and treat. Int Health. 2018;11(1):36–41. 59. Nhassengo P, Cataldo F, Magaço A, Hoffman RM, Nerua L, Saide M, et al. Barriers and facilitators to the uptake of Test and Treat in Mozambique: a qualitative study on patient and provider perceptions. PLoS ONE. 2018;13(12):e0205919. 60. Ford N, Migone C, Calmy A, Kerschberger B, Kanters S, Nsanzimana S, et al. Benefits and risks of rapid initiation of antiretroviral therapy. AIDS (London, England). 2018;32(1):17. 61. Katirayi L, Namadingo H, Phiri M, Bobrow EA, Ahimbisibwe A, Berhan AY, + et al. HIV-positive pregnant and postpartum women’s perspectives about in Malawi: a qualitative study. J Int AIDS Soc. 2016;19(1):20919. Option B 62. Long L, Brennan A, Fox MP, Ndibongo B, Jaffray I, Sanne I, et al. Treatment outcomes and cost-effectiveness of shifting management of stable ART patients to nurses in South Africa: an observational cohort. PLoS Med. 2011;8(7):e1001055. 63. Duncombe C, Rosenblum S, Hellmann N, Holmes C, Wilkinson L, Biot M, et al. Reframing HIV care: putting people at the centre of antiretroviral delivery. Trop Med Int Health. 2015;20(4):430–47. 64. Grimsrud A, Bygrave H, Doherty M, Ehrenkranz P, Ellman T, Ferris R, et al. Reimagining HIV service delivery: the role of differentiated care from prevention to suppression. J Int AIDS Soc. 2016;19(1):21484. 65. Derksen L, Muula A, Matengeni A, van Lettow M, Sodhi S, van Oosterhout J. Reducing stigma and increasing HIV testing with a health informa- tion intervention, a cluster-randomized trial from Malawi. J Int AIDS Soc. 2015;18:94–5. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12887-021-02768-z	Huang et al. BMC Pediatrics (2021) 21:298 https://doi.org/10.1186/s12887-021-02768-z R E S E A R C H A R T I C L E Open Access Factors affecting the growth of children till the age of three years with overweight whose mothers have diabetes mellitus: A population-based cohort study Yuan-Der Huang1,2, Yun-Ru Luo1, Meng-Chih Lee3,4,5* and Chih-Jung Yeh1* Abstract Background: The prevalence of diabetes mellitus (DM) during pregnancy and childhood obesity is increasing worldwide. Factors affecting the growth of children with overweight whose mothers had DM are complicated and inconclusive. Few longitudinal studies have focused on the growth of infants with macrosomia born to mothers with DM and the factors influencing their overweight. This study explored risk factors for childhood overweight/ obesity (OWOB) among children of mothers with DM. Perinatal, maternal socio-demographic, infant care, and maternal body weight characteristics as well as child growth until age 3 years were analyzed using a longitudinal design. Methods: In total, 24,200 pairs of mothers and their children from the Taiwan Birth Cohort Study were included. Combined Taiwan Children Growth Curve report classifications were analyzed for infant growth at birth and at 6, 12, 18, 24, and 36 months old (m/o). A multiple logistic regression analysis with different model settings was used to assess factors affecting the growth of high birth weight children of mothers with diabetic mellitus (HODM). Results: Children in the HODM group had a higher average body weight than did those in the non-DM group at different age stages. Relative to the non-DM group, weight gain in the HODM group was slower before 18 m/o but faster from 18 to 36 m/o, particularly after 24 m/o. Maternal DM was a major risk factor for childhood OWOB (odds ratio [OR]: 3.25–3.95). After adjustment for related confounders, the OR was 2.19–3.17. Maternal overweight or obesity and higher gestational weight gain were greater risk factors for childhood OWOB at 3 years old after adjusted maternal DM and other selected confounders (OR: 1.45 and 1.23, respectively). Breastfeeding until 6 m/o was a protective factor against childhood OWOB (OR: 0.95). The HODM and non-DM groups did not differ significantly in perinatal, maternal socio-demographic, or infant care characteristics. Conclusions: Maternal DM is a major factor of childhood OWOB. Maternal body weight before and after pregnancy affects childhood OWOB, and this effect increases with the child’s age. Keywords: Birth cohort, Maternal diabetes, Childhood overweight/obesity, Growth * Correspondence: mengchihlee@gmail.com; alexyeh@csmu.edu.tw 3Department of Family Medicine, Taichung Hospital, Ministry of Health and Welfare, Taichung, Taiwan 1Department of Public Health, Chung-Shan Medical University, Taichung, Taiwan Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Huang et al. BMC Pediatrics (2021) 21:298 Page 2 of 9 Background The prevalence of diabetes mellitus (DM) is increasing rapidly worldwide and is the most common complica- tion of pregnancy, affecting up to 10% of expectant mothers [1]. In Taiwan, the prevalence of diagnosed DM in pregnancy increased from 2.0% in 1996–1998 [2] to 3.5% in 2001–2008 [3]. A Taiwanese study demonstrated that morbidity is higher among children whose mothers had DM than among children of mothers without DM, and overweight newborns had an increased risk of hospitalization and repeat admission [2]. Infants exposed to high glucose levels prenatally have an increased risk of long-term adverse outcomes, includ- ing childhood overweight and obesity (OWOB). Many risk factors contribute to childhood overweight. Mater- nal obesity is a known major risk factor for childhood obesity. Many studies have demonstrated that mothers with higher body mass index (BMI) are more likely to have overweight infants [4]. Although high maternal BMI is associated with a substantially increased risk of gestational DM (GDM) [5], the causal associations of childhood overweight with maternal DM and obesity re- main unclear. In addition to genetic and DM factors of child obesity, the effects of family lifestyle [6], breast- feeding, parents’ education level, and even whether mothers who are able to perceive their children’s body weight status accurately have been discussed [7, 8]. To our knowledge, no studies examining the associ- ation between maternal DM and childhood OWOB have considered all potential key confounders, including ma- ternal obesity, gestational weight gain, maternal and in- fant lifestyle, and care factors. Women with preexisting DM have higher risks of preterm labor, congenital anomalies in their children, adverse perinatal complica- tions, and slowed infant growth in the future. Studies that have not excluded infants with low birth weight born to mothers with DM, especially infants born to women with GDM and type 2 DM, have methodological limitations in their analyses of childhood obesity [9]. A study recommended that infants of women with DM be targeted specifically for obesity prevention [10]. By con- trolling for maternal BMI, our study compared high birth weight of diabetic mothers (HODM) with children whose mothers did not have DM to explore risk factors for subsequent childhood OWOB and growth. The re- sults may provide information for mothers with the most severe DM to prevent childhood obesity in their children. Methods Data source and study population Data used in this study were collected from the Taiwan Birth Cohort Study (TBCS). TBCS is a national population-based cohort study in Taiwan to investigate the health determinants of this 2005–2006 birth cohort infant, child, adolescent, and adult at their newborn, stages, first from the life-course perspective. Formal wave (6 m/o) survey was completed in 2005–2006, followed up in 2006–2007 (18 m/o), 2008 (3 y/o), 2010– 2011 (5 y/o), 2013–2014 (8 y/o), and 2017–2018 (12 y/ o). The TBCS cohort was expected to follow up till 21 y/ o (year 2026–2027). Of the 206,932 live births in the Taiwan birth registration database for 2005, 24,200 pairs of mothers and their children were sampled in the study from the TBCS through a multistage stratified system- atic sampling method [11]. This dataset was compiled from the Millennium Cohort Study (MCS) [12] and the National Children’s Study (NCS) [13]. The mothers pro- vided written consent to participate, and this study was approved by the Medical Ethics Committee and Data Protection Board in Taiwan. A total of 21,248 mother– child pairs (87.8%) completed the first wave survey when child’s age is 6 months old. A total of 20,645 mother– child pairs were enrolled into our study. Of the 20,645 pairs who complete the first wave of study, 19,597 (92.22%, 18 months) and 19,344 (91.03%, 36 months) completed the second wave and the third wave, respect- ively. Our study excluded multiple pregnancies (561 cases) and mothers who did not know their diagnosed DM during pregnancy at interview (43 cases). Children with major or minor congenital fetal anomalies (998 cases) were not calculated into multiple logistic regression analysis models. the time of Variable definitions Children were classified into four groups: children born to mothers without DM (non-DM), Offspring of mother with Diabetic Mellitus (ODM) with high birth weight (>4000 g, HODM), ODM with appropriate birth weight (2500–3999 g, AODM), and ODM with low birth weight (<2500 g, LODM). OWOB was defined as body weight above the 85th percentile at 6, 18, 24, or 36 m/o (sepa- rated by sex) according to Taiwan Children Growth Curve report classifications [14]. Mothers were asked questions regarding awareness of diagnosed DM by doctors during pregnancy by trained interviewers (yes or no), maternal BMIs before preg- nancy and 6 months after delivery (calculated based on the data reported from the first wave survey), nationality, educational level, area of residence (urban or rural), fam- ily income, whether they breastfed or staple fed their children at 6 m/o, and main daytime caregiver (mother or others). Maternal age was defined as mother’s age at delivery (older or younger than 35 years). APGAR scores were classified as more or less than 7. Maternal body weight gain during pregnancy was divided into three cat- egories: <10, 10–14, and >14 kg. Prepregnancy and Huang et al. BMC Pediatrics (2021) 21:298 Page 3 of 9 postpartum BMI were divided into three categories: < 18.5, 18.5–25, and >25 kg/m2. Statistical analyses We classified the variables of influential factors into four major categories, namely perinatal condition, maternal socio-demographic characteristics, infant care factors, and maternal body weight characteristics. Categorical variables were analyzed using the chi-square test among the non-DM, HODM, AODM, and LODM groups. Fish- er’s exact test was conducted if necessary (Table 1). Analysis of variance (ANOVA) in combination with the Tukey honestly significant difference post hoc test was conducted on normally distributed variables to com- pare children’s body weight at different time points (birth, 6, 18, 24, and 36 m/o) and perform a sensitive as- sessment of deviations in growth (Table 2). Figure 1 plots the body weights and changes in growth of all chil- dren born in the non-DM and HODM groups from birth to 36 m/o as a result of a Cochran–Armitage test for trend. Separate models for the HODM and non-DM groups were developed to detect influential factors of perinatal conditions, maternal socio-demographic char- acteristics, infant care, and maternal body weight charac- teristics. Multiple logistic regression analyses were used to estimate odds ratios (ORs) and 95% confidence inter- vals for the associations of influential variables between the HODM and non-DM groups at different ages, in- cluding maternal BMI, gestational weight gain, breast- feeding, and perinatal conditions (Table 3). The non- DM group was used as the reference group, and P < 0.05 indicated statistical significance. Statistical analyses were performed using SAS 9.4. (SAS Institute Inc., Carrie, North Carolina, USA). Results A total of 20,645 (85.3%) pairs of mothers and children were enrolled into our study. Among the pairs, 446 (2.16%) children were born to mothers with diagnosed DM during pregnancy. Regardless of birth weight, all se- lected influential categories, including perinatal condi- tions, maternal socio-demographic characteristics, and maternal body weight characteristics, revealed significant differences between the non-DM group and ODM groups, except infant care factors. Compared with mothers without DM, mothers with DM were more likely to have advanced maternal age and live in urban areas. DM was less likely in immigrant women and those with higher education levels (P < 0.05). Differences in family income, main caregiver, and breastfeeding versus staple-food feeding at 6 m/o were non-significant. More adverse birth outcomes, such as preterm labor, lower APGAR scores 1 and 5 min after birth, and increased prevalence of birth defects, were present in infants in the ODM groups. Maternal BMIs before and after preg- nancy were higher in the DM groups than in the non- DM group. In particular, mothers in the HODM group had obviously higher prevalence of high prepregnancy BMI (46.2% vs. 9.7%) and postpartum BMI (100.0% vs. 63.8%), (P < 0.001). Differences in maternal gestational weight gain among the four groups were nonsignificant (P = 0.157) (Table 1). Figure 2 plots the growth of all children in the HODM and non-DM groups. The average birth weight in these two groups was 4294.5 and 3122.7 g, respectively. From birth to 18 m/o, child growth in both the groups HODM had similar velocity and slope in terms of body weight, and average body weight was higher in the non-DM group. However, the velocity of growth in body weight was faster in the HODM group between 18 and 36 m/o. Average body weight at 36 m/o was 16,830.8 and 14, 939.4 g in the HODM and non-DM groups, respectively. Growth trends in both the groups were significantly dif- ferent (P < 0.05). Figure 3 presents the percentages of childhood OWOB in the four groups. Children in the HODM group had higher rates of child OWOB at each age stage, especially at 36 m/o. The disparities in infant body weight and body weight gain between the HODM and non-DM groups at birth, 6, 18, 24, and 36 m/o are shown in Table 2. Children in the HODM group had a higher average body weight than did the non-DM group at different age stages. Body weight gain among infants differed significantly between the two groups. Body weight gain was slower in the HODM group than in the non-DM group before 18 m/o but faster than the non-DM group, from 18 to 36 m/o showed latter catch-up in HODM group, particularly after 24 m/o. adjusted for maternal Table 3 presents the results of the stepwise model ana- lysis by using a multiple logistic regression. In the null the incidence of OWOB in children in the model, HODM group was greater than that in children in the non-DM group from birth to 36 m/o (OR: 3.95–3.74). Model A, which socio- demographic and perinatal conditions, yielded similar results. When infant care factors in model B were con- sidered, breastfeeding to 6 m/o reduced childhood OWOB among children in the HODM group by ap- proximately 4%, however, mothers as the main daytime caregiver or early staple-food feeding did not affect childhood OWOB (P > 0.05). Model C adjusted for the factors included in model A and maternal body weight factors to analyze the effects of maternal body weight on childhood OWOB. Compared with the non-DM group, mothers in the HODM group were more likely to have OWOB children when their gestational weight gain was >14 kg (OR: 1.12–1.25). Obesity (BMI > 25 kg/m2) before and after pregnancy also increased the risk of childhood Huang et al. BMC Pediatrics (2021) 21:298 Page 4 of 9 Table 1 Maternal socio-demographic, perinatal care and children’s characteristics in Taiwan Non-DM1 (n = 20,199) HODM2 (n = 39) AODM3 (n = 382) LODM4 (n = 25) Perinatal conditions Sex Male Female Prematurity (<37 weeks) Yes No Apgar score(1 min) ≥7 <7 Apgar score(5 min) ≥7 <7 Birth defect Yes No 10,566 (52.3%) 9633 (47.7%) 1387 (6.9%) 18,812 (93.1%) 19,818 (98.1%) 381 (1.9%) 19 (48.7%) 20 (51.3%) 5 (12.8%) 34 (87.2%) 37 (94.9%) 2 (5.1%) 223 (58.4%) 159 (41.6%) 39 (10.2%) 343 (89.8%) 375 (98.2%) 7 (1.8%) 20,148 (99.7%) 39 (100.0%) 381 (99.7%) 51 (0.3%) 0 (0.0%) 10 (0.3%) 960 (4.8%) 19,239 (95.2%) 2 (5.1%) 37 (94.9%) 32 (8.4%) 350 (91.6%) Maternal socio-demographic characteristics Maternal age ≤35 y/o >35 y/o Maternal nationality Taiwanese Non-Taiwanese Maternal education Less than high school High school College and above Family income (NTD) <50,000 50,000–150,000 ≥150,000 Living area Urban Rural Perinatal care Breastfeeding at 6 m/o Yes No 18,448 (91.3%) 1751 (8.7%) 17,446 (86.4%) 2753 (13.6%) 3021 (15.0%) 8085 (40.1%) 9061 (44.9%) 8499 (42.2%) 11,020 (54.7%) 614 (3.0%) 9529 (47.2%) 10,670 (52.8%) 16,611 (82.2%) 3588 (17.8%) Non-staple foods feeding before 6 m/o Yes No 18,148 (89.9%) 2043 (10.1%) Main caregiver at daytime (at 1–6 m/o) 28 (71.8%) 11 (28.2%) 35 (89.7%) 4 (10.3%) 3 (7.7%) 16 (41.0%) 20 (51.3%) 14 (37.8%) 22 (59.5%) 1 (2.7%) 23 (59.0%) 16 (41.0%) 33 (84.6%) 6 (15.4%) 37 (94.9%) 2 (5.1%) 304 (79.6%) 78 (20.4%) 365 (95.5%) 17 (4.5%) 36 (9.4%) 139 (36.5%) 206 (54.1%) 121 (31.8%) 246 (64.6%) 14 (3.7%) 222 (58.1%) 160 (41.9%) 307 (80.4%) 75 (19.6%) 347 (90.8%) 35 (9.2%) 11 (44.0%) 14 (56.0%) 19 (76.0%) 6 (24.0%) 16 (64.0%) 9 (36.0%) 24 (96.0%) 1 (4.0%) 4 (16.0%) 21 (84.0%) 22 (88.0%) 3 (12.0%) 24 (96.0%) 1 (4.0%) 2 (8.0%) 10 (40.0%) 13 (52.0%) 11 (44.0%) 14 (56.0%) 0 (0.0%) 14 (56.0%) 11 (44.0%) 25 (100.0%) (0.0%) 22 (88.0%) 3 (12.0%) Mother 9961 (49.3%) 20 (51.3%) 193 (50.5%) 11 (44.0%) p- valuea NS <0.001* <0.001* p- valueb NS NS NS 0.0032 NS <0.001* NS <0.001* <0.001* <0.001* NS 0.0049 NS 0.0065 NS <0.001* NS 0.0912 NS NS NS NS NS Huang et al. BMC Pediatrics (2021) 21:298 Page 5 of 9 Table 1 Maternal socio-demographic, perinatal care and children’s characteristics in Taiwan (Continued) Non-DM1 (n = 20,199) HODM2 (n = 39) AODM3 (n = 382) LODM4 (n = 25) p- valuea p- valueb Others 10,238 (50.7%) 19 (48.7%) 189 (49.5%) 14 (56.0%) Maternal body weight Pre-pregnancy BMI (kg/m2) <18.5 18.5–25 >25 Postpartum BMI (at 6 m/o) <18.5 18.5–25 >25 Gestational weight gain (kg) <10 kg 10–14 kg >14 kg 4236 (21.0%) 14,040 (69.5%) 1923 (9.5%) 158 (0.8%) 7330 (36.3%) 1 (2.6%) 20 (51.3%) 18 (46.2%) 0 (0.0%) 0 (0.0%) 12,711 (62.9%) 39 (100.0%) 3401 (16.8%) 7919 (39.2%) 8879 (44.0%) 6 (15.4%) 13 (33.3%) 20 (51.3%) 41 (10.7%) 249 (65.2%) 92 (24.1%) 5 (1.3%) 83 (21.7%) 294 (77.0%) 85 (22.3%) 146 (38.2%) 151 (39.5%) 3 (12.0%) 15 (60.0%) 7 (28.0%) 0 (0.0%) 7 (28.0%) 18 (72.0%) 6 (4.0%) 8 (32.0%) 11 (44.0%) 1Non-DM: newborn born from non-DM mothers 2HODM: Newborn born from DM mothers and birth weight more 4000 g 3AODM: Newborn born from DM mothers and birth weight within 2500–3999 g 4LODM: Newborn born from DM mothers and birth weight less 2500 g aAnalysis by χ2 test statistics for 4 groups bAnalysis by χ2 test statistics for non-DM and HODM groups; NS: Not Significant; P < 0.001 <0.001 <0.001* <0.001* <0.001* NS NS OWOB (OR: 1.24–1.44, 1.17–1.31, respectively). When we adjusted for all potential determinants in model D, mater- nal DM remained a major risk factor for childhood OWOB at different ages in the HODM group (OR: 2.19–3.17). Ma- ternal body weight before and after pregnancy was a main factor affecting childhood OWOB, and the risk of child- hood OWOB increased with the children’s age. Discussion GDM prevalence in Asia has been reported to range from 0.7 to 51.0% [15]. A meta-analysis yielded a pooled GDM prevalence rate of 11.5% [16]. This vast disparity in prevalence rates may be due to differences in diagnos- tic criteria [17], screening strategies [18], and ethnicity and population characteristics [19]. Our study included 446 mothers with DM who accounted for approximately 2.2% of all included participants. The TBCS study de- fines whether a mother has DM during pregnancy, in addition to the self-recognition from the interviewees, must be judged by trained interviewers (mainly public health nurses) to determine whether there is “diagnosed DM by doctors during pregnancy”, and refer to the Table 2 Infant growth from birth to 36 m/o in the HODM and non-DM groups HODM Non-DM Birth weight (g) Birth 6 months 18 months 24 months 36 months Weight gain (g) 0–6 months 0–18 months 0–24 months 0–36 months aZ value: Z test statistics bRepeated measure ANOVA 4294.5 ± 272.8 8848.6 ± 1120.8 11,948.4 ± 1346.8 13,900.0 ± 2004.1 16,830.8 ± 2500.2 4555.6 ± 1193.0 7683.4 ± 1380.2 9657.9 ± 1984.1 12,536.3 ± 2511.6 3122.7 ± 424.3 8090.5 ± 1021.4 11,062.9 ± 1332.0 12,373.3 ± 1723.2 14,939.4 ± 2190.7 4965.8 ± 944.6 7933.7 ± 1251.8 9247.2 ± 1654.3 11,816.0 ± 2118.1 Z valuea 0.0028 P valueb 0.0020 0.0029 0.0020 Huang et al. BMC Pediatrics (2021) 21:298 Page 6 of 9 Fig. 1 Growth in body weight of children in 4 groups. aRepeated measure ANOVA test for trend among 4 groups; P < 0.0001. bCochran–Armitage test for trend among non-DM and HODM groups; P < 0.05 Fig. 2 Body weight gain of children in 4 groups. aRepeated measure ANOVA test for trend among 4 groups; P < 0.0001. bCochran– Armitage test for trend among non-DM and HODM groups; P < 0.001 Table 3 Multiple logistic regression models of newborn care and maternal body weight factors with childhood overweight Null model HODM vs. non-DM Model A HODM vs. non-DM Model B HODM vs. non-DM Breast feeding at 6 m/o Main caregiver at daytime (mother vs. others) Non-staple foods feeding before 6 m/o Model C HODM vs. non-DM Gestational weight gain (<10 kg vs. 10–14 kg) Gestational weight gain (>14 kg vs. 10–14 kg) Pre-pregnancy BMI (<18.5 vs. 18.5–25 kg/m2) Pre-pregnancy BMI (>25 vs. 18.5–25 kg/m2) Postpartum BMI at 6 m/o (<18.5 vs. 18.5–25 kg/m2) Postpartum BMI at 6 m/o (>25 vs. 18.5–25 kg/m2) Model D HODM vs. non-DM Breast feeding at 6 m/o Main caregiver at daytime (mother vs. others) Non-staple foods feeding before 6 m/o Gestational weight gain (<10 kg vs. 10–14 kg) Gestational weight gain (>14 kg vs. 10–14 kg) Pre-pregnancy BMI (<18.5 vs. 18.5–25 kg/m2) Pre-pregnancy BMI (>25 vs. 18.5–25 kg/m2) Postpartum BMI at 6 m/o (<18.5 vs. 18.5–25 kg/m2) Postpartum BMI at 6 m/o (>25 vs. 18.5–25 kg/m2) 6 m/o OR 3.95 P-value <0.001 18 m/o OR 3.32 P-value <0.001 24 m/o OR 3.25 P-value 0.0013 36 m/o OR 3.74 P-value <0.001 3.70 <0.001 2.98 0.002 3.03 0.004 3.40 <0.001 3.69 <0.001 2.92 0.003 2.96 0.005 3.34 <0.001 0.99 1.07 1.06 3.19 0.84 1.12 0.85 1.26 1.13 1.17 3.17 0.99 1.06 1.06 0.84 1.12 0.85 1.25 0.99 1.17 NS NS NS <0.001 0.003 0.008 0.003 <0.001 NS 0.002 <0.001 NS NS NS 0.002 0.010 0.003 <0.001 NS 0.002 0.96 1.07 1.17 2.27 0.93 1.22 0.77 1.34 0.88 1.28 2.22 0.96 1.06 1.16 0.93 1.22 0.76 1.34 0.96 1.27 <0.001 NS NS 0.026 NS <0.001 <0.001 <0.001 NS <0.001 0.031 <0.001 NS NS NS <0.001 <0.001 <0.001 <0.001 <0.001 0.95 1.35 1.12 2.26 0.89 1.25 0.63 1.24 0.79 1.25 2.19 0.95 1.35 1.09 0.89 1.25 0.62 1.23 0.95 1.26 <0.001 <0.001 NS 0.045 NS <0.001 <0.001 0.011 NS 0.002 0.055a <0.001 <0.001 NS NS <0.001 <0.001 NS <0.001 0.002 0.96 1.02 1.14 2.61 0.83 1.24 0.68 1.44 1.37 1.31 2.56 0.96 1.03 1.11 0.83 1.23 0.67 1.45 0.96 1.31 <0.0001 NS 0.033 0.005 0.002 <0.001 <0.001 <0.001 NS <0.001 0.006 <0.001 NS NS 0.0014 <0.001 <0.001 <0.001 <0.001 <0.001 Model A: adjustment for perinatal conditions (mode of delivery, gestational age, APGAR score at 1 and 5 min), and maternal socio-demographic characteristics; Model B: adjustment for factors included in model A and newborn care factors (only the ORs of newborn factors are shown); Model C: adjustment for factors included in model A and maternal body weight factors (only the ORs of maternal body weight factors are shown); Model D: adjustment for all factors (only the ORs of newborn care and maternal body weight factors are shown) NS no significance aBorderline significance Huang et al. BMC Pediatrics (2021) 21:298 Page 7 of 9 those of the non-DM group. In particular, mothers in the HODM group had higher prepregnancy BMIs. The disparity in maternal age may explain why mothers with DM had higher education levels; moreover, immigrant and indigenous mothers relatively younger. in Taiwan are Because of the potential confounders including genetic factors, mother and infant lifestyle, and maternal obesity factors are difficult to control, the relationships between maternal DM and risk of childhood OWOB are com- plex. The Taiwan’s NHI has a high coverage rate and people face few barriers to medical assessments. We fo- cused on the HODM group because mothers in the LODM group tended to have pregestational DM that in- creased the incidence of congenital anomalies and pre- term labor. Our results indicate that maternal obesity before and after pregnancy increase the risk of childhood OWOB, and this risk increases as children age. These findings suggest that genetics may play a central role. A previous study identified many genetic polymorphisms in genome-wide association studies of adult BMI and the genetic susceptibility to childhood obesity; the asso- ciation was partially explained by appetitive traits in in- fancy followed by an early childhood increase in BMI [23]. infant and maternal Studies of the effects of gestational DM on childhood overweight and obesity have yielded inconclusive results. A meta-analysis reported inconsistent evidence of an as- sociation of GDM with childhood overweight and obes- ity because of methodological limitations in the included studies [9]. However, our study adjusted for prepreg- nancy obesity, socio- care, demographic factors. Maternal DM was an independent determinant of childhood OWOB, but the associations were attenuated after adjustment for prepregnancy and postpartum maternal BMI. Previous studies have not fo- cused on infants with macrosomia, which may have con- founded their results. Because the infants of mothers with DM, particularly type 1 DM, have a higher risk of congenital anomalies, premature birth, and low APGAR scores, their growth is expected to be slower than that of infants of mothers without DM or even mothers with GDM. We conducted a stepwise logistic regression by using five models to eliminate these confounders. Because of mothers with overweight or obesity have a higher prevalence of DM, our study could not identify the prevalence of DM in ODM children. Mothers with obesity may have a genetic predisposition to have chil- dren who become overweight; in addition, maternal life- style, diet, and maternal cognition to infant body weight are factors of childhood overweight [24]. In our study, main daytime caregiver and early staple-food feeding were not influential factors in Taiwanese cohort. How- ever, high calorie intake in infancy and large appetite Fig. 3 Percentage of OWOB among children in 4 groups at different ages. aχ2 test; *P < 0.001 check-up records on the maternal health booklet issued by Taiwan’s National Health Insurance (NHI). The coverage rate of NHI is >99%. Pregnant women can re- ceive at least 10 prenatal health examinations by obste- tricians, and barriers in access to prenatal care are lower in Taiwan compared with those in other countries. This discrepancy in prevalence rates might be related to the design of the TBCS, which was connected as a self- report questionnaire by mothers administered in inter- views. Although the urine glucose level is a routine par- ameter serum tested during prenatal examinations, glucose levels and oral glucose tolerance tests are not routine items designated by Taiwan’s NHI for prenatal examinations, mothers with DM may not be aware of their DM status. Such issue may cause some GDM mothers not being diagnosed and not included in DM group of this study. However, it is reported that nearly 95% of pregnant women received free prenatal care by the NHI in Taiwan. Doctors are required to screen for maternal DM by NHI guidelines, so we believe it has consistent validity. Our study using a dataset from the TBCS is the first to use a multistage stratified systematic sampling design, and the content of the questionnaire covered a wide range of information. TBSC has been widely applied in research to reveal insights into the health profiles of children growing up in Taiwan. Advanced maternal age, family history of DM, ethni- city, overweight or obesity, and smoking are well- documented risk factors for GDM [20]. In addition to these risk factors, data increasingly indicate that diet and lifestyle factors both before and during pregnancy are as- sociated with GDM [21]. A study in Taiwan indicated that advanced maternal age, a family history of DM, a less than high school education, high prepregnancy BMI, and an indigenous ethnicity were risk factors for GDM [22]. In our study, mothers with DM were older, lived in urban areas, were less likely to be immigrants, and had higher education levels (p < 0.05). Their BMIs before pregnancy and postpartum at 6 months were higher than Huang et al. BMC Pediatrics (2021) 21:298 Page 8 of 9 from 1 year were reported to be related to a higher inci- dence of subsequent childhood obesity [23]. Children who were breastfed until 12 m/o were reported to be 2.7 times less likely to develop childhood obesity [24, 25]. We dem- onstrated that breastfeeding until 6 m/o is a protective fac- tor against childhood OWOB (OR: 0.96). Although the benefits of breastfeeding appear to be a key part of the positive health outcomes associated with the parent-child relationship, the effect of the dose is not informed in this study. The odds ratio of HODM group to the non-DM group for childhood OWOB is range from 3.25 to 3.95 at different age stages. Regardless of whether neonatal care factors were adjusted in Model A or Model B, the odds of HODM in the probability of children being OWOB has similar findings. The effects to the odds of HODM be- tween the former two are much lower than those of Model C and Model D. Therefore, we infer that HODM is an independent factor affecting childhood OWOB, and the impact of maternal overweight or obesity is much greater than that of perinatal conditions and acquired care. Control of DM and restricted weight gain during pregnancy are keys to preventing childhood obesity, espe- cially in mothers with DM. The strength of our study is its focus on a particular infant population for comparison with children whose mothers did not have DM and its exclusion of infants with low birth weight and congenital anomalies to elim- inate the effect of congenital or genetic factors that could confound the statistical results. We also used a stepwise logistic regression model to detect confounders and their interactions between DM and diabetes-related perinatal complications. In addition, we analyzed pre- pregnancy and postpartum BMI to compare the effects of maternal BMI and gestational body weight gain with childhood OWOB. Moreover, this cohort study can offer information on the determinants of childhood OWOB at different ages. Our research has some limitations. The design of the TBCS questionnaires did not differentiate between preg- estational DM and GDM, neither lack of medical records about the severity of hyperglycemia and related medical intervention. This limitation makes genetic factors and the effects of DM control difficult to identify. Another limitation is the high rate of cesarean deliveries in Taiwan; some infants with macrosomia can be delivered earlier if their body weight is estimated to be overweight during prenatal examinations. This phenomenon may have reduced the number of infants in the HODM group. The maternal height and weight were self- reported instead of measurements might latent errors. This cohort study followed children to 36 m/o. Future studies may prolong the follow-up period to analyze the long-term relationships between mothers with DM and childhood obesity. Conclusion and suggestions In this population-based cohort study, maternal DM and BMI of mothers more than 25 kg/m2 prepregnancy and postpartum were the significant factors of childhood OWOB, and the maternal BMI effects increased as chil- dren aged. Prolonged breastfeeding to the age of 6 months and weight gain control during pregnancy are protection factors. DM control during pregnancy, re- stricted maternal gestational body weight gain, maternal BMI control, and prolonged breastfeeding are strategies for mothers with DM to prevent childhood overweight. Abbreviations AODM: Appropriate birth weight Of mothers with Diabetic Mellitus; BMI: Body Mass Index; GDM: Gestational Diabetic Mellitus; HODM: High birth weight children Of mothers with Diabetic Mellitus; LODM: Low birth weight Of mothers with Diabetic Mellitus; MCS: Millennium Cohort Study; NCS: National Children’s Study; NHI: National Health Insurance; ODM: Offspring of mother with Diabetic Mellitus; OWOB: Overweight/ Obesity; TBCS: Taiwan Birth Cohort Study Acknowledgements This manuscript was edited by Wallace Academic Editing (O-2020-008161). Authors’ information (optional) Not applicable. Authors’ contributions YDH: Conceived and designed the analysis, data collecting, analyzing, and paper writing. YRL: data analysis. MCL: developed the theoretical formalism. CJY: Designed the model and the computational framework. All authors discussed the results and contributed to the final manuscript. All authors have read and approved the manuscript. Funding This study was supported by a grant from the Bureau of Health Promotion (grant number DOH93-HP-1702). Availability of data and materials Please contact to Meng-Chih Lee, E-Mail: mengchihlee@gmail.com Declarations Ethics approval and consent to participate This study was approved by the Jianan Psychiatric Center, Ministry of Health and Welfare Medical Ethics Committee and Data Protection Board (approval no. 19-044). Written consent was provided by the mothers. Consent for publication Not applicable. Competing interests The authors have no potential conflicts of interest to disclose. Author details 1Department of Public Health, Chung-Shan Medical University, Taichung, Taiwan. 2Department of Obstetrics and Gynecology, Chia-Yi Hospital, Ministry of Health and Welfare, Chia-Yi, Taiwan. 3Department of Family Medicine, Taichung Hospital, Ministry of Health and Welfare, Taichung, Taiwan. 4Institute of Population Health Sciences, National Health Research Institutes, Miaoli, Taiwan. 5College of Management, Chaoyang University of Technology, Taichung, Taiwan. Huang et al. BMC Pediatrics (2021) 21:298 Page 9 of 9 23. de Lauzon-Guillain B, Koudou YA, Botton J, Forhan A, Carles S, Pelloux V, et al. Association between genetic obesity susceptibility and mother- reported eating behaviour in children up to 5 years. Pediatr Obes. 2019;14: e12496. 24. Reifsnider E, McCormick DP, Cullen KW, Todd M, Moramarco MW, Gallagher MR, et al. Randomized controlled trial to prevent infant overweight in a high-risk population. Acad Pediatr. 2018;18:324–33. Sun J, Wu L, Zhang Y, Li C, Wang Y, Mei W, et al. Association of breastfeeding duration, birth weight, and current weight status with the risk of elevated blood pressure in preschoolers. Eur J Clin Nutr. 2020;74:1325–33. 25. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Received: 9 November 2020 Accepted: 8 June 2021 References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. Statistics Department of the Ministry of the Interior. Annual report on internal affairs statistics: fertility rate of women of childbearing age 2018. https://moi.gov.tw/files/site_stuff/321/2/year/year.html. Huang YC, Fang LR, Huang JP, Xu SB, Zhu CH. Investigation of gestational diabetes in Taiwan. Taipei Med J. 2005;2:P132–7. Chou CY, Lin CL, Yang CK, Yang WC, Lee FK, Tsai MS. Pregnancy outcomes of Taiwanese women with gestational diabetes mellitus: a comparison of Carpenter-Coustan and National Diabetes Data Group criteria. J Women’s Health. 2010;19:935–9. Aldana-Parra F, Vega GO, Fewtrell M. Associations between maternal BMI, breastfeeding practices and infant anthropometric status in Colombia; secondary analysis of ENSIN 2010. BMC Public Health. 2020;20:1–15. Chu SY, Callaghan WM, Kim SY, Schmid CH, Lau J, England LJ, et al. Maternal obesity and risk of gestational diabetes mellitus. Diabetes Care. 2007;30:2070–6. Burke V, Beilin LJ, Dunbar D. Family lifestyle and parental body mass index as predictors of body mass index in Australian children: a longitudinal study. Int J Obes. 2001;25:147–57. Cartagena D, McGrath JM, Masho SW. Differences in modifiable feeding factors by overweight status in Latino infants. Appl Nurs Res. 2016;30:210–5. Brown CL, Skinner AC, Yin HS, Rothman RL, Sanders LM, Delamater AM, et al. Parental perceptions of weight during the first year of life. Acad Pediatr. 2016;16:558–64. Kim SY, England JL, Sharma JA, Njoroge T. Gestational diabetes mellitus and risk of childhood overweight and obesity in offspring: a systematic review. Exp Diabetes Res. 2011;2011:1–9. Svensson V, Jacobsson JA, Fredriksson R, Danielsson P, Sobko T, Schiöth HB, et al. Associations between severity of obesity in childhood and adolescence, obesity onset and parental BMI: a longitudinal cohort study. Int J Obes. 2011;35:46–52. 11. Chiang TL, Lin SJ, Chang MC. Taiwan Birth Cohort Study: backgrounds, design and participants. Taiwan Health Promotion Administration, MOHW; 2009 https://www.hpa.gov.tw/Pages/List.aspx?nodeid=110. 12. Chesbrough KB, Ryan MA, Amoroso P, Boyko EJ, Gackstetter GD, Hooper TI, 13. et al. The Millennium Cohort Study: a 21-year prospective cohort study of 140,000 military personnel. Mil Med. 2002;167:483–8. Landrigan PJ, Trasande L, Thorpe LE, Gwynn C, Lioy PJ, D’Alton ME, et al. The national children’s study: a 21-year prospective study of 100 000 American children. AAP. 2006;118:2173–86. 14. Chen W, Chang M-H. New growth charts for Taiwanese children and adolescents based on World Health Organization standards and health- related physical fitness. Pediatr Neonatol. 2010;51:69–79. 15. Nguyen CL, Pham NM, Binns CW, Duong DV, Lee AH. Prevalence of 16. 17. gestational diabetes mellitus in eastern and southeastern Asia: a systematic review and meta-analysis. J Diabetes Res. 2018;2018:1–10. Lee KW, Ching SM, Ramachandran V, Yee A, Hoo FK, Chia YC, et al. Prevalence and risk factors of gestational diabetes mellitus in Asia: a systematic review and meta-analysis. BMC Pregnancy Childbirth. 2018;18: 494. Lauring JR, Kunselman AR, Pauli JM, Repke JT, Ural SH. Comparison of healthcare utilization and outcomes by gestational diabetes diagnostic criteria. J Perinat Med. 2018;46:401–9. 18. Corrado F, Pintaudi B. Diagnosis of gestational diabetes mellitus: Italian perspectives on risk factor-based Screening. Nutr Diet Mater Diabet. 2017;9: 87–97. 19. Alfadhli EM, Osman EN, Basri TH, Mansuri NS, Youssef MH, Assaaedi SA, et al. Gestational diabetes among Saudi women: prevalence, risk factors and pregnancy outcomes. Ann Saudi Med. 2015;35:222–30. 20. Ben-Haroush A, Yogev Y, Hod M. Epidemiology of gestational diabetes mellitus and its association with Type 2 diabetes. Diabet Med. 2004;21:103– 13. 21. Zhang C, Ning Y. Effect of dietary and lifestyle factors on the risk of 22. gestational diabetes: review of epidemiologic evidence. Am J Clin Nutr. 2011;94:1975S–9S. Lin PC, Hung CH, Chan TF, Lin KC, Hsu YY, Tzeng YL. The risk factors for gestational diabetes mellitus: a retrospective study. Midwifery. 2016;42:16– 20.
10.1186_s12891-023-06345-6	Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 https://doi.org/10.1186/s12891‑023‑06345‑6 RESEARCH BMC Musculoskeletal Disorders Open Access Global prevalence of musculoskeletal disorders among physiotherapists: a systematic review and meta‑analysis Philippe Gorce1,2,3 and Julien Jacquier‑Bret1,2,3* Abstract Background Musculoskeletal disorders (MSD) are one of the most important problems among physiotherapists worldwide. However, there is no meta‑analysis of the MSD prevalence in all body areas among physiotherapists. Objectives The purpose was to investigate and estimate the worldwide prevalence of MSD among physiotherapists using a systematic review‑, meta‑analysis and meta‑regression. Methods The systematic review, meta‑analysis and meta‑regression were performed in 2022 using the PRISMA guidelines. Data sources The search was performed on PubMed/Medline, ScienceDirect, Google Scholar, Medeley and Science. gov databases. Study appraisal The quality appraisal of the included articles was assessed using the critical appraisal tool for cross‑ sectional studies AXIS. Results A total of 722 articles were found. After screening and comparison with the inclusion criteria, 26 studies were retained. Based on the random‑effects model, the worldwide MSD prevalence in neck, upper back, mid back, lower back, shoulders, elbows, wrists/hands, thumb, hips/thighs, knees/legs, and ankles/feet was 26.4% (CI 95%: 21.0– 31.9%), 17.7% (CI 95%: 13.2–22.2%), 14.9% (CI 95%: 7.7–22.1%), 40.1% (CI 95%: 32.2–48.0%), 20.8% (CI 95%: 16.5–25.1), 7.0% (CI 95%: 5.2–8.9), 18.1% (CI 95%: 14.7–21.5%), 35.4% (CI 95%: 23.0–47.8), 7.0% (CI 95%: 5.2–8.8), 13.0% (CI 95%: 10.3–15.8), and 5% (CI 95%: 4.0–6.9) respectively. The neck and shoulder prevalence of four continents were close to the world prevalence. No effect of continent was found on MSD prevalence. The heterogeneity of the results obtained in the meta‑analysis and meta‑regression was discussed. Conclusions Based on the random effects model, the results of the worldwide meta‑analysis showed that lower back pain, thumb, neck and shoulder were the area most at risk for MSD and were therefore those to be monitored as a priority. Recommendations were proposed for future reviews and meta‑analyses. Keywords Musculoskeletal disorders, Prevalence, Body area, Physiotherapists, Meta‑analysis, Meta‑regression, Systematic review *Correspondence: Julien Jacquier‑Bret jacquier@univ‑tln.fr Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 2 of 15 Introduction Musculoskeletal disorders (MSD) in the occupational environment are a major public health issue in the world. They are responsible for multiple work stoppages and significant direct and indirect costs [1]. They affect both work habits and quality of work, quality of life and well- being of workers. Health professionals are populations at risk due to their varied interventions requiring significant loads [2–4]. In this context, the practices of physiotherapists (PT) expose them significantly to MSD. Indeed, several risk factors have been identified inducing important physi- cal loads. Bending and twisting the trunk, transferring patients, performing manual therapy, working in awk- ward postures for long periods of time or repeating the same movements over and over are all factors that con- tribute and reinforce the presence of MSD and associated symptoms [5–7]. Many studies have addressed the high prevalence of MSD among PT worldwide. In studies by Khairy et al. [8], Kinaci et al. [9] or Grooten et al. [10], preva- lence rates exceeded 80%. The 90% threshold has been reported in some countries such as Korea, Australia or the USA [2, 5, 11]. Several studies have investigated the prevalence of MSD by body area. Some work focused on certain areas such as the thumb [12, 13] or the upper limb [14, 15]. Other work extended the analysis to the lower limbs by including MSD hazards in the hip, knee and ankle [6, 11, 16]. Vieira et al. [17] provided an over- view of MSD among physical therapists by summarizing the prevalence by body area during the career of PT. Ten areas were reported among the 32 included studies. The authors reported that lower back was the body area most commonly affected by MSD. However, two limitations can be addressed. The first concerns the heterogeneity of the reported results. The prevalence rates could be based on two different sam- ples. Some studies presented WMSD prevalence in rela- tion to the whole sample tested [18], whereas other works reported prevalence rates in relation to participants who mentioned the presence of WMSD [9]. In this case, the rates were increased since people without WMSD were not considered. It therefore appears important to nor- malize these data to have an accurate estimation of the MSD prevalence in PT. The objective of this study was to perform a systematic review, meta-analysis and meta- regression of the MSD risk in PT normalized to the tested samples. The results would provide an assessment of MSD risk by body area by considering the prevalence of the different works conducted worldwide. The effect of the continent on the prevalence of MSD was also tested. Methods The present systematic review and meta-analysis were conducted on the prevalence of MSD among physiother- apists in 2022. The study was reported according to Pre- ferred Reporting Systematic Reviews and Meta-Analyses (PRISMA) guidelines [19]. The protocol for this review was registered at PROSPERO (CRD42023343473). Search strategy and eligibility criteria Five international databases namely PubMed/Medline, ScienceDirect, Google Scholar, Medeley and Science. gov were explored in March 2022. The following key- words were used: (“Physiotherapy” OR “Physiothera- pist” OR “Physical therapy” OR “Physical therapist”) AND “Musculoskeletal disorders” AND “Prevalence” AND “Body area”. The search focused exclusively on English language peer-reviewed works that quantified the MSD prevalence by body area among physiothera- pists. Reviews, systematic reviews, commentaries, case studies and case series were not retained. Studies were excluded if: not published in English, not among physiotherapists, no sufficient data about sampling, the number of body parts is too low, mixed healthcare pro- fessions without the possibility of distinguishing them, or insufficient MSD prevalence details. Results were imported from the five databases and compiled to remove duplicates. Titles and abstracts of unique records were separately screened by two reviewers (P.G. and J.J.B.) for eligibility. The full text of each article selected from its title/abstract was evalu- ated on the basis of the inclusion criteria separately by two reviewers. Studies that did not meet the criteria were excluded. All discrepancies were resolved by con- sensus and re-review of the articles. The search process is shown in Fig. 1. Methodological quality assessment and risk of bias The critical appraisal tool was used to assess the qual- ity of cross-sectional studies (AXIS) included in the review [20]. Each of the criteria was evaluated on its presence (“Yes”) or absence (“No”). The percentage of items pre- sent is then calculated. The quality appraisal was obtained using McFarland et al. [21] classification and the AXIS repartition: 0–50% has high risk of bias, 50–80% has medium risk of bias and 80–100% has low risk of bias. Two reviewers (P.G. and J.J.B.) performed the quality assessment separately. The discrepancies have been dis- cussed for the final evaluation, involving a third reviewer where necessary. Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 3 of 15 Fig. 1 PRISMA Flow Chart Data extraction The included articles were used to extract the following data: number of respondents, response rate, male and female repartition, age, country, and MSD prevalence by body area. When prevalence rates were calculated from the subsample of physiotherapists with MSD, they were reported to the total sample tested to perform the meta- analysis with homogeneous data. Statistical analysis The meta-analysis was performed based on the work of Neyeloff et al. [22]. Heterogeneity of the stud- ies was assessed using Cochran’s Q test (significance level < 10%) and I2 statistic (significance level > 50%). In case of heterogeneity, random effects model with inverse-variance approach was employed. Otherwise, fixed effects model was applied. A Kurskal-Wallis test was used to compare the prevalence of each body area on the five continents (significance level set at 5%). A meta-regression was performed to analyze the trend in MSD prevalence as a function of the average age of the participants, the year of publication of the included studies, and the Gross Domestic Product (GDP) of the country in which the study was conducted. Analyses were achieved using Statistica (Version 7.1, Statsoft, Tulsa, OK, USA). Results Search results The exploration of the various databases identified 722 articles. Of the 649 unique articles, 87 articles were selected on the basis of their title/abstract and were fully evaluated. After comparison with the inclusion/exclu- sion criteria, 61 were excluded because either the data were mixed and did not meet the objective or the param- eters studied were insufficient. Finally, 26 articles were retained and included in the analysis. Quality appraisal The quality appraisal of the 26 included articles revealed that 18 studies had low risk of bias whereas 8 had medium risk of bias (Table 1.). Study characteristics All studies included in the review were surveys of MSD risk by body area among physiotherapists, physical thera- pists, massage therapists and kinesitherapists. The stud- ies were conducted in 17 different countries on the 5 Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 4 of 15 Table 1. Detailed quality appraisal for risk of biaisaccording to the AXIS [20] continents. The sample sizes were very heterogeneous ranging from a subgroup of 37 participants [18] to 2688 [6]. The response rate was also highly variable between studies, ranging from 37% [9, 23] to 91.9% [24]. All par- ticipants were adult men and/or women between 18 and 55 years old (mean were between 24.25 ± 7.27 years [25] and 43.0 ± 12.0 years [26]). Except the study conducted by Grooten et al. [10] that included only women, all other studies were performed with mixed population but with varying proportions. Table 2. summarizes the general population character- istics, i.e. number of participants, response rate, men/ women repartition, mean age, country, and the preva- lence of MSD by body area of the 26 included studies. Eleven areas were assessed. The most studied were neck and lower back mentioned in all the 26 studies. Shoulder and wrist/hand were studied in 24 studies. Upper back, elbow/forearm, hip/thigh, knee/leg, and ankle/foot were addressed in 21 studies. Finally, thumb and mid back were the less studied body areas addressed in 7 and 3 studies respectively. Meta‑analysis results Heterogeneity among studies was assessed using Q and I2 statistics. Results revealed important heterogeneity for all body areas: neck (Q = 1149.4; df = 25; I2 = 97.8%; p < 0.001), upper back (Q = 888.5; df = 20; I2 = 97.7%; p < 0.001), mid back (Q = 8.47;df = 2; I2 = 76.4; p < 0.05), I2 = 98.1%; p < 0.001), lower back (Q = 1299; df = 25; shoulder (Q = 828.3; df = 23; I2 = 97.2%; p < 0.001), elbow/ forearm (Q = 248.4; df = 20; I2 = 91.9%; p < 0.001), wrist/hand/finger (Q = 434.8; df = 23; I2 = 94.7%; p < 0.001), thumb (Q = 349.2; df = 6; I2 = 98.3%; p < 0.001), hip/thigh (Q = 371.2; df = 20; I2 = 94.6%; p < 0.001), knee/leg (Q = 446.3; df = 20; I2 = 95.5%; p < 0.001), ankle/foot. Neck The prevalence of MSD for the neck was presented in all included studies (26 studies) carried out in many countries of the world (Fig. 2). Based on the random effects model, the neck prevalence was 26.4% (CI 95%: 21.0–31.9%). Upper back The upper back MSD prevalence was evaluated in 21 studies around the world. The overall prevalence was 17.7% (CI 95%: 13.2–22.2%) obtained with the random effects model (Fig. 3). Mid back The prevalence of MSD in the mid back has been the least studied, with only three studies reporting results (Fig. 4). These data came from Taiwan, Turkey, and Australia. The random effects model estimates the prevalence of mid back MSD at 14.9% (CI 95%: 7.7–22.1%). Lower back As for the neck, the prevalence of lower back MSD was found in the 26 included studies all over the world (Fig. 5). Based on the randomized design, the overall prevalence for lower back was 40.1% (CI 95%: 32.2–48.0%). Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 5 of 15 Table 2. Objective and characteristics of the 36 includedstudies by healthcare profession. MSD prevalence by body area was reported foreach study (when available) Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 6 of 15 Fig. 2 Prevalence of musculoskeletal disorders in neck amongst studies included Fig. 3 Prevalence of musculoskeletal disorders in upper back amongst studies included Fig. 4 Prevalence of musculoskeletal disorders in mid back amongst studies included Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 7 of 15 Fig. 5 Prevalence of musculoskeletal disorders in lower back amongst studies included Fig. 6 Prevalence of musculoskeletal disorders in shoulder amongst studies included Shoulder According to Fig. 6, the prevalence of shoulder MSD was mentioned in 24 studies. The results of the random effects model showed that the prevalence of this disorder was 20.8% (CI 95%: 16.5–25.1). Elbow/forearm The prevalence of elbow MSD has been presented in Fig. 7. This was assessed in 21 studies performed in many countries. Based on the results of the random effects model, its prevalence was 7.0% (CI 95%: 5.2–8.9). Wrist/hand The wrist/hand MSD prevalence was evaluated in 24 studies spread over all the continents. The overall preva- lence was 18.1% (CI 95%: 14.7–21.5%) obtained with the random effects model (Fig. 8). Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 8 of 15 Fig. 7 Prevalence of musculoskeletal disorders in elbow/forearm amongst studies included Fig. 8 Prevalence of musculoskeletal disorders in wrist/hand amongst studies included Fig. 9 Prevalence of musculoskeletal disorders in thumb amongst studies included Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 9 of 15 Fig. 10 Prevalence of musculoskeletal disorders in hip/thigh amongst studies included Fig. 11 Prevalence of musculoskeletal disorders in knee/leg amongst studies included Thumb The prevalence of MSD for the thumb was addressed in 7 studies conducted in Canada, Korea, Australia, the United Kingdom, the United States, Taiwan, India and Saudi Arabia (Fig. 9). Based on the results of the random effects model, the thumb MSD prevalence was 35.4% (CI 95%: 23.0–47.8). random effects model showed that the prevalence of this disorder was 7.0% (CI 95%: 5.2–8.8) (Fig. 10). Knee/leg The prevalence of knee/leg MSD has been presented in Fig. 11. This was assessed in 21 studies performed in many countries. Based on the results of the random effects model, its prevalence was 13.0% (CI 95%: 10.3–15.8). Hip/thigh Ankle/foot The prevalence of MSD in hip was reported in 21 stud- ies conducted all around the world. The results of the The prevalence of MSD in ankle/foot was also reported in 21 studies conducted all around the world. The results Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 10 of 15 Fig. 12 Prevalence of musculoskeletal disorders in ankle/foot amongst studies included of the random effects model indicated that its prevalence was 5.5% (CI 95%: 4.0–6.9) (Fig. 12). MSD prevalence by continent Table 3 summarizes the sample size weighted mean prev- alence of the eleven body areas by continent. About the lower back, the prevalence observed in Asia and Europe (36.7% and 41.0%) was close to the world value (40.1%). America presented a lower average rate (33.0%) while Africa and Oceania obtained higher prevalence (53.4% and 50.8% respectively). For neck, the world prevalence was close to the value observed in Africa (26.8%) and Europe (27.5%). Values were slightly lower in America and Asia (23.4% and 22.6% respectively). Oceania displayed higher prevalence of 39.6%. For shoulder, America, Europe and Oceania showed prevalence close to the world value (20.8%). The rate was higher in Africa (27.6%) and slightly lower in Asia. Finally, the prevalence of thumb was different, with a smaller number of studies including this area. Oceania reported a prevalence equivalent to the world value of 35.4% but with only one study. The rate is lower in Asia (24.7%) with 4 studies and in Europe (17.8%) with one study. The prevalence was higher in America (83.3%) with only one study but that was mainly targeted to this area. Kruskal–Wallis analysis revealed no difference in prev- alence between continents for all body areas (Table 3). Meta‑regression Figure 13 illustrates the meta-regression performed for the neck, lower back and shoulder which were the most exposed areas to MSD. Whatever the body area, no effect of year of publication, means age of participants and GDP was evidenced (r2 between 0.0002 and 0.1127, p > 0.05). Discussion The aim of this study was to propose a literature review and meta-analysis to investigate the prevalence of MSD among physiotherapists. The objective was to summa- rize the worldwide MSD prevalence by body area. Par- ticular attention was paid to the way the results were presented in each study. In order to provide a synthesis that included comparable data, prevalence was recal- culated when they did not refer to the global sample. The meta-analysis of the 26 included articles showed that the highest prevalence rates were observed for the lower back (40.1%), thumb (35.4%), neck (26.4%) and shoulder (20.8%). This global result was also observed in national studies. For example, for the lower back, the work of Glover et al. [6] in the UK (37.2%), Głowiński et al. [18] in Poland (41.7%) and [9] in Turkey (37.0%) reported equivalent rates. About the neck, several stud- ies reported similar prevalence on all 5 continents: 30.3% in Egypt [8], 25.0% in the US [26], 26.5% in Saudi Ara- bia [7], 27.5% in Poland [18], and 20.0% in Australia [38]. For the shoulder, equivalent rates were observed mainly in Nigeria, Sweden and Australia with respective prevalence of 22.2% [16], 17.6% [10] and 22.9% [11]. The elbow and lower limbs were the areas least at risk for MSDs among physiotherapists (prevalence ranging from 5.5% for the ankle to 13% for the knee). Three studies Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 11 of 15 t e e f / e l k n A e e n K i p H b m u h T s r e g n fi / d n a h / t s i r W w o b E l l r e d u o h S k c a b r e w o L k c a b d M i k c a b r e p p U k c e N % 5 6 . % 6 2 . % 4 3 . % 8 1 1 . % 7 4 . % 0 5 . % 7 6 . % 6 2 1 . % 6 5 . % 6 3 . % 5 5 . 6 3 0 . = p % 1 1 2 . % 6 4 . % 4 4 . % 1 7 1 . % 5 2 1 . % 9 4 . % 4 1 1 . % 0 4 1 . % 8 8 . % 8 5 . % 0 3 1 . 5 1 0 . = p % 3 4 . % 8 1 . % 1 3 . % 1 1 1 . % 5 2 1 . % 7 3 . % 3 8 . % 8 3 1 . % 1 6 . % 0 3 . % 0 7 . 1 3 0 . = p % 3 3 8 . ‑ % 7 4 2 . % 1 9 1 . % 8 7 1 . ‑ % 6 3 3 . ‑ % 4 5 3 . 0 1 . = p ‑ ‑ % 9 0 2 . % 1 0 1 . % 0 6 1 . % 0 5 1 . % 9 6 1 . % 0 0 1 . % 8 5 1 . % 3 2 1 . % 5 9 1 . % 5 5 . % 1 8 1 . 4 8 0 . = p % 4 6 . % 7 0 . % 0 3 . % 4 6 . % 0 9 . % 7 6 . % 8 7 . % 3 4 1 . % 2 0 1 . % 2 7 . % 0 7 . 3 7 0 . = p % 6 7 2 . % 5 5 1 . % 0 0 2 . % 9 6 2 . % 2 6 1 . % 6 3 1 . % 1 8 1 . % 8 4 1 . % 2 9 1 . % 1 9 . % 4 3 5 . % 0 9 2 . % 0 3 3 . % 4 5 2 . % 7 6 3 . % 7 0 1 . % 0 1 4 . % 1 9 1 . % 8 0 5 . % 6 8 2 . % 8 0 2 . 3 8 0 . = p % 1 0 4 . 1 5 0 . = p % 8 0 2 . % 9 0 1 . ‑ ‑ % 0 1 1 . ‑ % 9 4 1 . 0 1 . = p ‑ ‑ ‑ ‑ % 7 7 1 . 4 5 0 . = p % 4 6 2 . 3 9 0 . = p % 0 0 2 . % 5 1 1 . % 1 5 1 . % 4 4 1 . % 5 9 . % 5 6 . % 8 0 2 . % 1 0 2 . % 0 1 4 . ‑ % 8 6 2 . % 9 5 1 . % 4 3 2 . % 8 8 2 . % 6 2 2 . % 9 9 . % 5 7 2 . % 4 7 1 . % 6 9 3 . % 5 9 1 . n a e M D S n a e M D S n a e M D S n a e M D S n a e M D S i d e d u t s s a e r a y d o b 1 1 e h t f o h c a e r o f t n e n i t n o c y b e c n e a v e r p D S M l f o ) i n o i t a v e d d r a d n a t s ( e g a r e v a d e t h g e w e z i s e p m a S l i 3 e l b a T a c i r e m A a c i r f A e p o r u E a i s A i a n a e c O l l a r e v O ) s i l l l a W – a k s u r K ( p Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 12 of 15 Fig. 13 Meta‑regression to evaluate the trends in the prevalence of neck, lower back and shoulder musculoskeletal disorders in relation to year of publication, age of participants and Gross Domestic Product (GDP) presented prevalence 2 to 6 times higher than the others for the lower limb [17, 24, 29]. These results are consistent with the results proposed by Vieira in his review [2] or more recently in a study involving several health professionals including physi- otherapists [39]. For all body areas, a large heterogene- ity was observed between studies (I2 comprised between 76.4% and 98.3%). These levels are comparable to those observed in other meta-analyses carried out on other health professionals such as nurses (between 76.4% and 99.1% [40]) or surgeons (between 86.8% and 96.8% [4]) Whatever the solutions, i.e. material, societal, related to working conditions or the work environment, it appears important to consider them to reduce MSD risks. An interesting contribution of this meta-analysis was the consideration of MSD prevalence by continent. The aim was to find out whether there were differences between countries among physiotherapists. The analysis has shown that several results by continent were close to the world average (Table 3). For neck, four continents had prevalence differences of less than 4% with the global value of 26.4%. About the lower back, Asia and Europe showed prevalence close to the global values (40.1%) with respective rates of 36.7% and 41.0%. The other three con- tinents had prevalence higher or lower by about 10%. For the shoulder, values similar to the world average (< 5% difference) were found for four continents. Only Africa had a higher rate of 7% (27.6%). Concerning the thumb, a great disparity of results was observed between conti- nents. Only Oceania, through the study of Cromie et al. [11], presented values close to the world value (33.6% vs 35.4%). This can be explained by the few number of stud- ies that specifically distinguished the MSD risks of the thumb from the rest of the hand, which was studied by the majority of studies (24/26). Despite the differences in prevalence observed, no significant difference was found between the continents. This could be explained by the high standard deviations obtained for each of the conti- nents and more generally by the important heterogeneity of the data reported in the different studies included. Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 13 of 15 In order to have a world assessment of the MSD preva- lence among physiotherapists, the search presented in this review was initiated with a reduced number of key- words. This led to the integration of works with very dif- ferent methodological characteristics. Indeed, as shown in Table 2., the age of the participants varied between 20 and 65 years, which leads to differences in years of experience (0.5 to 40 years) and therefore in exposure to MSDs [10, 24, 25]. As shown by Molumphy et al. [41] for work-related lowed back pain in physical therapists, the age of the populations tested and consequently their pro- fessional experience may affect the responses to the ques- tionnaires on the presence of MSD. Other studies, such as Grooten et al. [10] included only women, while other studies included a mixed population. In many studies, the work environments were not explicitly detailed. How- ever, working in a public or private hospital can influence working conditions such as the total number of working hours, number of hours in direct contact with patients, equipment used, etc. These differences may affect the risk of MSDs or their perception. All these parameters characteristic of the samples tested could explain in part the high variability of MSD risks worldwide. The lack of covariance shown by the meta-regression between the parameters studied and the prevalence of MSDs explains reinforces this idea. It would therefore be relevant to take these different parameters into account as inclusion cri- teria in order to try to reduce the heterogeneity between studies and therefore the results. The question would then arise as to the relevance of a meta-analysis since there would probably be fewer studies included per cri- terion studied. Limitations As mentions above, the main limitation was the het- erogeneity. It was also observed into the questionnaires used to investigate MSD. Their nature could influence the results collected, particularly the fact that certain body area prevalence were not reported in several studies. On the other hand, some areas were presented differently in relation to the method used. For example the back could be divided into two (upper and lower back) or three (upper, mid and lower back) parts. Similarly, the thumb may or may not be included in the MSD assessment of the fingers of the hand. This heterogeneity also affects the weight of the differ- ent studies in the meta-analysis. Indeed, for equivalent sample sizes, the method used increases the weight for studies with lower prevalence. However, a greater weight is given to studies with larger sample sizes. To overcome these problems, it would be recom- mended to set up a more standardized protocol allow- ing all the information to be filled in homogeneously. Another alternative would be to pay more particular attention to the experimental conditions and character- istics of the populations in future reviews. This would allow considering only studies whose parameters would be quantified in a same experimental context and thus reduce heterogeneity. Indeed, different workplaces (pub- lic vs. private), gender, experience, age, etc.… are all fac- tors that can affect the occurrence of MSDs. This would provide a more accurate and complete summary of the MSD prevalence among physiotherapists in the context of their professional activity. Psychosocial and societal factors that affect the occurrence of MSD should also be taken into account to complete the assessment. It would also be interesting to have a more precise idea of the practices carried out, particularly in terms of physi- cal demands on the musculoskeletal system, to estimate the impact of the main activity on MSD risks. The work presented shows that increasing the number of stud- ies would likely help reduce heterogeneity in future meta-analyses. Conclusion The literature review, meta-analysis and meta-regression showed the presence of MSD with the highest worldwide prevalence located in the lower back, neck, shoulders and extremities of the hand independently of the continent. Methodological recommendations have been proposed to reduce the heterogeneity observed for future reviews and meta-analyses. Abbreviations MSD PT PRISMA CI Musculoskeletal Disorders Physiotherapist Preferred Reporting Systematic Reviews and Meta‑Analyses Confidence Intervals Acknowledgements None Authors’ contributions JJB and PG separately scanned the databases to extract relevant articles and performed quality appraisal. The two authors have jointly performed data extraction, data analysis, statistics, data formatting, drafting and proofreading the entire manuscript. All author(s) read and approved the final manuscript. Funding None. Availability of data and materials We propose a literature review and a meta‑analysis. All data are from the articles listed in the summary tables included in the manuscript. Declarations Ethics approval and consent to participate The review was approved by the local ethics committee. Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 14 of 15 Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1 International Institute of Biomechanics and Occupational Ergonomics, Toulon, France. 2 Université de Toulon, CS60584‑83041 ‑ TOULON CEDEX 9, Toulon, France. 3 Hôpital Léon Bérard, Avenue du Docteur Marcel Armanet, Hyères 83418, France. Received: 30 June 2022 Accepted: 20 March 2023 References 1. EU‑OSHA, Les troubles musculo‑squelettiques d’origine professionnelle: faits et chiffres — Rapport de synthèse compilé à partir de 10 rapports d’États membres de l’UE. https:// osha. europa. eu/ fr/ publi catio ns/ work‑ relat ed‑ muscu loske letal‑ disor ders‑ facts‑ and‑ figur es‑ synth esis‑ report‑ 10‑ eu‑ member/ view, 2020 (accessed). 2. Vieira ER, Schneider P, Guidera C, Gadotti IC, Brunt D. Work‑related mus‑ culoskeletal disorders among physical therapists: A systematic review. J Back Musculoskelet Rehabil. 2016;29(3):417–28. 3. Davis KG, Kotowski SE. Prevalence of Musculoskeletal Disorders for Nurses in Hospitals, Long‑Term Care Facilities, and Home Health Care: A Compre‑ hensive Review. Hum Factors. 2015;57(5):754–92. Epstein S, Sparer‑Fine E, Tran B, Ruan Q, Dennerlein J, Singhal D, et al. Prevalence of Work‑Related Musculoskeletal Disorders Among Surgeons and Interventionalists: A Systematic Review and Meta‑analysis. JAMA surg. 2017;153:e174947. 4. 5. Chung YC, Hung CT, Li SF, Lee HM, Wang SG, Chang SC, et al. Risk of musculoskeletal disorder among Taiwanese nurses cohort: a nationwide population‑based study. BMC Musculoskelet Disord. 2013;14:144. 6. Glover W, McGregor A, Sullivan C, Hague J. Work‑related musculoskeletal disorders affecting members of the Chartered Society of Physiotherapy. Physiotherapy. 2005;91(3):138–47. 7. Muaidi Q, Shanb A. Prevalence causes and impact of work related mus‑ culoskeletal disorders among physical therapists. J Back Musculoskelet Rehabil. 2016;29(4):763–9. 8. Khairy WA, Bekhet AH, Sayed B, Elmetwally SE, Elsayed AM, Jahan AM. Prevalence, Profile, and Response to Work‑Related Musculoskeletal Dis‑ orders among Egyptian Physiotherapists. Open Access Maced J Med Sci. 2019;7:1692–9. 9. Kinaci E, AtaoĞLu S. Work Related Musculoskeletal Disorders Among the Physiotherapists: Sample of a Region in Turkey. Turk Klin J Med Sci. 2020;5:495–502. 10. Grooten WJA, Wernstedt P, Campo M. Work‑related musculoskeletal disor‑ ders in female Swedish physical therapists with more than 15 years of job experience: Prevalence and associations with work exposures. Physiother Theory Pract. 2011;27:213–22. 11. Cromie JE, Robertson VJ, Best MO. Work‑related musculoskeletal disorders in physical therapists: prevalence, severity, risks, and responses. Phys Ther. 2000;80(4):336–51. 12. Snodgrass SJ, Rivett DA. Thumb Pain in Physiotherapists: Potential Risk Factors and Proposed Prevention Strategies. J Man Manip Ther. 2002;10(4):206–17. 13. McMahon M, Stiller K, Trott P. The prevalence of thumb problems in Aus‑ tralian physiotherapists is high: an observational study. Aust J Physiother. 2006;52(4):287–92. 14. Darragh AR, Huddleston WE, King PM. Work‑related musculoskeletal injuries and disorders among occupational and physical therapists. Am J Occup Ther. 2009;63(3):351–62. 15. Caragianis S. The prevalence of occupational injuries among hand thera‑ pists in Australia and New Zealand. Hand Ther. 2002;15(3):234–41. 16. Adegoke BOA, Akodu AK, Oyeyemi AL. Work‑related musculoskeletal disorders among Nigerian Physiotherapists. BMC Musculoskelet Disord. 2008;9(1):112. 17. Vieira ER, Svoboda S, Belniak A, Brunt D, Rose‑St Prix C, Roberts L, et al. Work‑related musculoskeletal disorders among physical therapists: an online survey. Disabil Rehabil. 2016;38(6):552–7. 18. Głowiński S, Bryndal A, Grochulska A. Prevalence and risk of spinal pain among physiotherapists in Poland. PeerJ. 2021;9:e11715. 19. Harris JD, Quatman CE, Manring MM, Siston RA, Flanigan DC. How to Write a Systematic Review. Am J Sports Med. 2013;42(11):2761–8. 20. Downes MJ, Brennan ML, Williams HC, Dean RS. Development of a critical appraisal tool to assess the quality of cross‑sectional studies (AXIS). BMJ Open. 2016;6(12): e011458. 21. McFarland T, Fischer S. Considerations for Industrial Use: A Systematic Review of the Impact of Active and Passive Upper Limb Exoskeletons on Physical Exposures. IISE Transactions on Occupational Ergonomics and Human Factors. 2019;7(3–4):322–47. 22. Neyeloff JL, Fuchs SC, Moreira LB. Meta‑analyses and Forest plots using a microsoft excel spreadsheet: step‑by‑step guide focusing on descriptive data analysis. BMC Res Notes. 2012;5(1):52. 23. Mahajan R, Singh M, Fahim T, Singh A. Thumb Pain in Physiotherapists Practicing Manual Therapy: Prevalence and Consequences. International Journal of Health Sciences and Research. 2020;10(6):194–200. 24. Meh J, Bizovičar N, Kos N, Jakovljević M. Work‑related musculoskeletal dis‑ orders among Slovenian physiotherapists. J Health Sci. 2020;10(2):115–24. 25. Alnaser MZ, Aljadi SH. Physical therapists with work‑related musculoskel‑ etal disorders in the State of Kuwait: A comparison across countries and health care professions. Work. 2019;63(2):261–8. 26. Bork BE, Cook TM, Rosecrance JC, Engelhardt KA, Thomason ME, Wauford IJ, et al. Work‑related musculoskeletal disorders among physical thera‑ pists. Phys Ther. 1996;76(8):827–35. 27. Albert WJ, Currie‑Jackson N, Duncan CA. A survey of musculoskeletal injuries amongst Canadian massage therapists. J Bodyw Mov Ther. 2008;12(1):86–93. 28. Al‑Eisa E, Buragadda S, Shaheen A, Ibrahim A, Melam G. Work related mus‑ culoskeletal disorders: Causes, prevalence and response among egyptian and saudi physical therapists. Middle East J Sci Res. 2012;12(4):523–9. 29. Anyfantis ID, Biska A. Musculoskeletal Disorders Among Greek Physi‑ otherapists: Traditional and Emerging Risk Factors. Saf Health Work. 2018;9(3):314–8. 30. Alrowayeh HN, Alshatti T, Aljadi S, Fares M, Alshamire M, Alwazan S. Prevalence, characteristics, and impacts of work‑related musculoskeletal disorders: A survey among physical therapists in the State of Kuwait. BMC Musculoskelet Disord. 2010;11:116. 31. Campo M, Weiser S, Koenig KL, Nordin M. Work‑Related Musculoskeletal Disorders in Physical Therapists: A Prospective Cohort Study With 1‑Year Follow‑up. Phys Ther. 2008;88(5):608–19. 32. Holder N, Clark HA, DiBlasio JM, Hughes C, Scherpf JW, Harding L, et al. Cause, prevalence, and response to occupational musculoskeletal injuries reported by physical therapists and physical therapist assistants. Phys Ther. 1999;79(7):642–52. 33. Jang Y, Chi C‑F, Tsauo J‑Y, Wang J‑D. Prevalence and Risk Factors of Work‑ Related Musculoskeletal Disorders in Massage Practitioners. J Occup Rehabil. 2006;16:425–638. 34. Mbada C, Adejuyigbe O, Omole J, Idowu O, Okafor C, Adekanla B, et al. Assessment of knowledge, attitude and perception of Nigerian physi‑ otherapists on manual handling techniques. J Environ occup Health. 2015;4(4):203–9. 35. Nordin NAM, Leonard JH, Thye NC. Work‑related injuries among physi‑ otherapists in public hospitals—a Southeast Asian picture. Clinics. 2011;66:373–8. 36. Rugelj D. Low back pain and other work‑related musculoskeletal prob‑ lems among physiotherapists. Appl Ergon. 2003;34(6):635–9. 37. Salik Y, Ozcan A. Work‑related musculoskeletal disorders : A survey of physical therapists in Izmir‑Turkey. BMC Musculoskelet Disord. 2004;5:27. 38. West DJ, Gardner D. Occupational injuries of physiotherapists in North and Central Queensland. Aust J Physiother. 2001;47(3):179–86. 39. Jacquier‑Bret J, Gorce P. Prevalence of Body Area Work‑Related Musculo‑ skeletal Disorders among Healthcare Professionals: A Systematic Review. Int J Environ Res Public Health. 2023;20(1):841. 40. Tavakkol R, Karimi A, Hassanipour S, Gharahzadeh A, Fayzi R. A Multidis‑ ciplinary Focus Review of Musculoskeletal Disorders Among Operating Room Personnel. J Multidiscip Healthc. 2020;13:735–41. Gorce and Jacquier‑Bret BMC Musculoskeletal Disorders (2023) 24:265 Page 15 of 15 41. Molumphy M, Unger B, Jensen GM, Lopopolo RB. Incidence of Work‑ Related Low Back Pain in Physical Therapists. Phys Ther. 1985;65(4):482–6. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub‑ lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12939-020-01197-1	Celhay et al. International Journal for Equity in Health (2020) 19:122 https://doi.org/10.1186/s12939-020-01197-1 R E S E A R C H Open Access Measuring socioeconomic gaps in nutrition and early child development in Bolivia Pablo Celhay1* , Sebastian Martinez2 and Cecilia Vidal2 Abstract Background: A large body of evidence shows that socioeconomic status (SES) is strongly associated to children’s early development, health and nutrition. Few studies have looked at within sample differences across multiple measures of child nutrition and development. This paper examines SES gaps in child nutritional status and development in Bolivia using a representative sample of children 0–59 months old and a rich set of outcomes, including micronutrient deficiencies, anthropometic measures, and gross motor and communicative development. Methods: We construct direct and proxy measures of living standards based on household expenditures and on ownership of assets combined with access to services and dwelling characteristics. The data for this study come from a nationally representative household survey in Bolivia that contains information on health, nutrition, and child development tests. We used a regression framework to assess the adjusted associations between child development outcomes and socioeconomic status, after controlling for other demographic factors that might affect child’s development. The SES gap in child development was estimated by OLS. To explore when the development gaps between children in different socioeconomic groups start and how they change for children at different ages, we analyze the differences in outcomes between the poorest (Q1) and richest (Q5) quintiles by child’s age by estimating kernel weighted local polynomial regressions of standardized scores for all child development indicators. Results: There are large and statistically significant differences in all anthropometrics z-scores between children in Q5 and children in Q1: height for age (0.95 SD), weight for age (0.70 SD), and weight for height (0.21 SD). When we divide the sample into children at the bottom and top consumption quintiles the results show that 68.6% of children in the poorest quintile are anemic. While this percentage falls to 40.9% for children in the richest quintile, it remains high compared to other countries in the region. The prevalence of vitamin A deficiency is 29.9% for children in the richest quintile and almost 10 percentage points higher for those at the bottom quintile (39.0%); the prevalence of Iron deficiency for children in the top and bottom quintiles is 16.4% and 23.8%, respectively. Compared to the most deprived quintile, children in the wealthiest quintile are less likely to have iron deficiency, anemia, to be stunted, and to have a risk of delays in gross motor and communicative development. At age three, most of these gaps have increased substantially. Our findings are robust to the choice of socioeconomic measurement and highlight the need for targeted policies to reduce developmental gaps. (Continued on next page) * Correspondence: pacelhay@uc.cl 1School of Government, Pontificia Universidad Católica de Chile and Millennium Nuclei for the Study of the Life Course and Vulnerability, Avda. Vicuña Mackenna 4860 – Macul, Santiago, Chile Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 2 of 25 (Continued from previous page) Conclusion: These findings highlight the need for targeted public policies that invest in multiple dimensions of child development as early as possible, including health, nutrition and cognitive and verbal stimulation. From a policy perspective, the large socioeconomic gaps in nutrition outcomes documented here reinforce the need to strengthen efforts that tackle the multiple causes of malnutrition for the poorest. Keywords: Child nutrition, Early childhood development, Socioeconomic gap, Bolivia gradients Introduction In 2010, an estimated 249.4 million children worldwide risked failing to reach their developmental potential be- cause of stunting and poverty [1]1. Studies analyzing so- cioeconomic in child health and early development show that poorer children are less likely to be healthy and well-nourished, achieve optimal cognitive abilities and adequately communicate with others [3–6]. These early disparities are carried through life and have implications for educational attainment, income gener- ation, adult health, risky behaviors2 and other dimen- sions of individual and social wellbeing [7–9]. Losses from poor cognitive and educational performance in poor children account for subsequent lower employabil- ity and earnings [10], high fertility and poorer parenting practices for their children in the future [2]. There are several mediating factors related to poverty that prevent children’s optimal development3. Black et al. [12] find that undernutrition and unstimulating household environments contribute to deficits in chil- dren’s health and development that affect adulthood economic outcomes. Micronutrient deficiencies such as iron and vitamin A have also been found to affect cogni- tive, motor, and socioemotional development [11, 13]. Vitamin A deficiency in children is responsible for over one million child deaths annually, and is the leading cause of preventable child blindness in developing coun- tries [14]. Likewise, Iron deficiency has been associated with poorer child cognitive, motor, and social-emotional functions [11, 13]. Other social determinants of child de- velopment include access to clean water and sanitation, parents education, and access to quality early childhood development services [7]. In this paper, we explore socioeconomic gradients in nutrition and final child development outcomes in 1According to Grantham-McGregor et al. [2], the degree of loss of de- velopmental potential is the discrepancy between children’s develop- mental levels and what they would have achieved in a more nurturing environment with adequate stimulation and nutrition. 2Such as involvement in violent behavior (physical fight, gun use, gang membership). 3Following the framework described in Walker et al. [11], these factors can be grouped in biological risks (such as undernutrition, micronutrient deficiencies and disease) and psychosocial risks (parenting and contextual factors) that affect children’s development through changes in brain function and behavior. Bolivia using a large nationally representative sample of children under five years of age. Although we analyze gaps for all outcomes individually, we also test the asso- ciation between SES and final child development after accounting for nutritional status as a mediating factor. Various elements make Bolivia an interesting case study. While the country has experienced important improve- ments in child nutrition and poverty indicators in the last decade, disparities between socioeconomic groups remain high. In 2016, one of every ten children living in low poverty municipalities was stunted, while this rate almost tripled for high poverty municipalities. At the same time, Bolivia has the second highest prevalence rate of anemia in the Latin American region, with 67% of children under five in the poorest municipalities pre- senting some level of anemia [15]4. Also, Bolivia’s three ecological regions (highlands, valleys and lowlands) differ in multiple environmental and socio-cultural factors, providing an opportunity to explore whether these re- gional differences might contribute or modify SES gaps in child development. Our study contributes to the growing literature of in- equalities in child development in several aspects. The richness of our data allows us to look at child develop- ment outcomes, including gross motor and communica- tion development, as well as nutrition risk factors, such as undernutrition, anemia and micronutrient deficien- cies, within the same representative sample of children. Another contribution of this paper is the use of novel data collected from dried blood samples to measure vita- min A and iron deficiencies. In addition, we measure SES through direct and proxy measures of living stan- dards based on household expenditures and on owner- ship of assets combined with dwelling characteristics. The availability of detailed information to obtain mul- tiple measures of SES is unusual for health surveys and in our study, it permits testing the robustness of our re- sults. Thus, our study complements the existing body of literature that focuses on particular subpopulations of disadvantaged children [4] or analyzes SES gaps in child 4The WHO Global Health Observatory country estimate for 2016 is 46.9% [23.3 – 70.0], the second highest rate in the region, only after Haiti and more than double the regional average of 22.7% (https:// apps.who.int/gho/data/node.main.ANEMIACHILDREN?lang=en). Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 3 of 25 development across countries using a single SES indica- tor and measure of child development [5]. We additionally explore the heterogeneity of our re- sults by looking at changes in SES gaps across age groups and ecological regions. These analyses are par- ticularly relevant as they provide information for the de- sign and targeting of evidence-informed early child development interventions that can benefit the poorest and most disadvantaged children in Bolivia. The remainder of the paper is organized as follows. Section 2 describes the data, explains the construction of SES and child nutrition and development outcomes, and describes the strategy to measure SES gradients. Section 3 presents SES gaps by developmental domain and ex- plores how SES gaps differ across ages. Section 4 pre- sents robustness checks of our results to the choice of the socioeconomic indicator. Section 5 concludes. Methods Data sources The data for this study come from the Health and Nutri- tion Evaluation Survey (ESNUT). The ESNUT is a one- round large household survey jointly implemented be- tween April and December 2012 by the Ministry of Health and the Ministry of Development Planning5. Its sample consists of a multistage cluster design that pro- vides representative statistics for households with chil- dren under five years old at the national and regional levels, and allows also urban-rural disaggregation within regions6. The ESNUT collected data using three main question- naires: i) a household questionnaire used to collect basic sociodemographic characteristics of household members (family composition, education, labor participation and income), physical housing quality, access to basic ser- vices, household asset ownership, and household ex- penditure; ii) a woman’s questionnaire, administered to all women aged 14 to 49 years, used to collect birth his- tories for the five year period preceding the survey (2007-2012) with detailed information on prenatal, birth and postnatal care; and iii) a child’s questionnaire used to gather data from all children under five, including health and weight, hemoglobin levels, immunizations, nutritional practices, and history of visits to health providers (reported and re- corded in health cards). To assess developmental pro- gress in children younger than 3 years, the ESNUT included the gross motor and communicative modules status, measures height of 5The survey was conducted to evaluate the effect of two national social programs: the Bono Juana Azurduy and the Zero Malnutrition Program. 6In a first stage, 424 clusters were selected (300 rural and 124 urban); in a second stage, 20 eligible households were randomly selected within each cluster. of a child development screening questionnaire. In a random subsample of children between 6 and 23 months, the survey additionally collected blood samples to measure vitamin A and iron levels in blood [16]. The full sample of analysis contains 11,358 children from 0 to 59 months in 8,433 households (2,456 urban and 5,977 rural). The subsamples of analysis have 5,763 children from 0 to 36 months with available gross motor and communicative information, and 1,610 children be- tween 6 and 23 months with information for the analysis of vitamin A and iron deficiencies. Assessment of nutritional status and early child development To measure children’s nutritional status, the survey in- cluded anthropometric indicators and hemoglobin level to identify the presence of anemia, and blood concentra- tion of vitamin A and iron to identify micronutrient de- ficiencies. Child growth indicators and standardized scores (z-scores) were constructed following the World for the Health Organization (WHO) guidelines [17] three most common anthropometric indices: height for age, weight for age, and weight for height. The z-score system expresses the anthropometric value as the num- ber of standard deviations from the median of the WHO reference population7,8. In addition, we computed indi- cators of prevalence of chronic malnutrition (stunting), underweight, and overweight, based on height for age and weight for age standard cutoff values of below or above two standard deviations from the reference median. To measure anemia the ESNUT obtained levels of hemoglobin in blood for each child between 3 and 59 months using a HemoCue® test for the photometric de- tection of hemoglobin. This method has been used ex- tensively in household surveys in developing countries, including the Demographic and Health Surveys. Pres- ence of mild, moderate or severe anemia was then deter- mined based on altitude-adjusted hemoglobin levels and standard cutoff values9. One of ESNUT’s specific objectives was to assess defi- cits of key micronutrients in small children; specifically, vitamin A deficiency (VAD) and iron deficiency (ID). To measure VAD and ID, blood samples were obtained from a subsample of 2,000 children ages 6 to 23 months 7Z-score = (observed value - median value of the reference population) / standard deviation value of reference population. 8The WHO reference growth standards were established in 2006 and represent the growth patterns of healthy children from different regions of the world. 9For children between 6 and 59 months of age we used standard WHO cutoff values: mild anemia, Hb 10.0 - 10.9 g/dL; moderate anemia, Hb 7.0 - 9.9 g/dL; severe anemia, Hb <7.0 g/dL. Because there is no standard cutoff to define anemia among infants 3 to 5 months, we used a cutoff of 10.5 g/dl following Chandyo et.al. (2016). Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 4 of 25 using the Dried Blood Spots (DBS) method. This method collects a few blood drops from a heel or finger prick that are then impregnated in filter paper and let to dry10. All DBS were rehydrated and analyzed in a labora- tory using the Enzyme-Linked Immunosorbent Assay method (ELISA)11 [18]. The indicators to estimate con- centration of Vitamin A and iron were the Retinol Bind- ing Protein (RBP) and the Free Transferrin Receptor (sTfR), respectively. VAD was defined as RBP below 0.7μmol/l [19] and ID as sTfR above 8.3 mg/l. Excluding samples with low quality (damaged or small blood spots) and with indication of inflammation, the total sample size for micronutrient analysis was 1,65512. Early child development was assessed through mea- sures of gross motor and communicative development. The ESNUT used 11 age-specific survey questionnaires for children between 3.5 and 36.5 months13 which were based on the second edition of the Ages and Stages (ASQ-2)14. The questionnaires con- Questionnaires® tained items about tasks that the child is (or is not) able to perform according to his or her age. Most items were reported by the child’s caregiver, while some specific items were based on direct observation of the child. To increase scores’ variability, the survey added items of de- creasing and increasing difficulty. Similar adaptations [20–22]. The have been used in other studies 10DBS were collected at the end of the interview after caregiver’s consent. Samples were identified and stored in hermetic plastic containers with drying agents to accelerate the dehydration process and to control humidity during transportation. Once dry they were packed using Ziploc bags. The DBS method had several advantages in field work. First, its simplicity and micro volume of blood required makes it appropriate for its use in small children, with very low health and biological risks. Second, the technique does not require specialized health personnel, intensive training or equipment. Third, DBS are stable to normal outside temperature for relatively long periods, which allows sufficient time to transport samples from rural areas and then to laboratories. In addition, it does not require special transportation conditions (such as cold chain), except humidity control and excess heath, which can both damage samples. Blood samples for DBS were collected separately from samples to determine hemoglobin concentrations. 11For analysis, all samples were flown to a specialized laboratory at the University of Giessen, Germany. 12From the subsample of 2,000 children 6-23 months selected for the DBS study, the ESNUT was able to collect 1,701 valid blood samples (the rest had either quality problems or were not available for analysis). In addition, 46 samples had to be discarded due to identification issues (incorrect household or child id). 13Age groups were: 3.6-6.5 months; 6.6 to 9.5 months; 9.6-11.5 months; 11.6-13.5 months; 13.6-15.5 months; 15.6-17.5 months; 17.6- 19.5 months; 19.6-22.5 months; 22.6-24.5 months; 24.6-30.5 months; 30.6-36.5 months. 14The ASQ-2 is a developmental screening tool for children between 4 and 60 months, comprising 19 age-specific questionnaires designed to screen infants and young children for developmental delays during the first 5 years of life (Ages & Stages Questionnaires, Second Edition. A Parent-Completed, Child Monitoring System, by Diane Bricker and Jane Squires. Copyright © 1999 by Paul H. Brookes Publishing Co.). questionnaires’ language was adapted to the local con- text of Bolivia. Each item had a score of 10, 5 or 0 depending on whether the child can perform the task always, some- times, or never, respectively. Raw scores were con- structed for each domain as the sum of scores across items. Because the population on which the ASQ was standardized (US children) was not considered an appro- priate reference population for our sample, we con- structed within sample or internally standardized scores adjusted by age. Following standard procedures, internal z-scores were constructed within the eleven age groups to have a mean of 0 and a SD of 1 (by subtracting the age-group specific mean of the raw score and dividing by the age-group specific SD). Measures of socioeconomic status There are several ways to measure socioeconomic status from household survey information, including direct monetary measures (income or expenditure), and proxy measures such as composite indices derived from own- ership of household assets and living conditions. Follow- ing existing literature on health inequalities [23], we used the rich information collected by ESNUT on household expenditures and asset ownership to con- struct alternative measures of SES and check the sensi- tivity of our results to the choice of the SES indicator. The question of whether the choice of SES measure matters in the analysis of socioeconomic inequalities has been explored in the literature, yet without a conclusive answer. While some studies find that the choice between consumption and the asset index makes little difference to the measured degree of inequality [24], others argue that results are actually sensitive to the choice of SES measure [25], or even to the choice of assets and charac- teristics that are included in the wealth index [26]. Con- sistent with previous findings, the relationship between our consumption and wealth index is relatively low, with a correlation coefficient of 0.45; therefore, as suggested in O ’Donnell et al. [23], we use both measures of SES to test the robustness of our results. To report our main findings, however, household consumption was used as our preferred direct measure of SES. A detailed descrip- tion of the construction of SES indicators is presented in the Technical Appendix. Measuring SES gradients SES gradients in child development were evaluated by comparing development outcomes across five quintiles of the population ranked by the level of consumption or wealth index. Quintiles were constructed based on the distribution of the household population rather than on the distribution of households; therefore, Quintile 1 (Q1) the corresponds to children in the poorest 20% of Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 5 of 25 Table 1 Estimated means and standard errors by SES Indicator Height for age z-score (HAZ) Weight for age z-score (WAZ) Weight for height z-score (WHZ) Stunting (%) (HAZ<-2SD) Underweight (%) (WAZ<-2SD) Overweight (%) (WHZ>+2SD) Anemia (%) RBP level (mmol/l) sTfR level (mg/l) Vitamin A deficiency (%) (<0.7 mmol/l) Iron deficiency (%) (>8.3mg/l) Gross motor z-score Communication z-score All Mean -0.979 -0.185 0.523 0.181 0.016 0.074 0.540 0.943 7.965 0.391 0.240 -0.001 -0.005 Poorest 20% (Q1) Richest 20% (Q5) Q5-Q1 SE 0.024 0.021 0.022 0.007 0.002 0.004 0.010 0.105 0.239 0.023 0.017 0.032 0.031 N 10,870 Mean -1.528 10,469 -0.582 10,453 10,870 10,453 10,453 9,414 1,609 1,609 1,609 1,609 5,753 5,753 0.411 0.333 0.025 0.057 0.686 1.040 8.247 0.390 0.238 -0.061 0.007 SE 0.036 0.034 0.034 0.013 0.004 0.006 0.016 0.238 0.518 0.031 0.030 0.056 0.049 N 3,895 3,665 3,660 3,895 3,660 3,660 3,260 516 516 516 516 2,031 2,031 Mean -0.577 0.114 0.624 0.099 0.010 0.101 0.409 1.032 7.407 0.299 0.164 0.151 0.115 SE 0.049 0.044 0.043 0.010 0.003 0.010 0.017 0.130 0.438 0.035 0.027 0.047 0.057 N 1,334 1,321 1,317 1,334 1,317 1,317 1,157 216 216 216 216 680 680 diff 0.952 0.696 0.213 -0.234 -0.015 0.045 -0.278 -0.009 -0.840 -0.091 -0.074 0.212 0.108 p-value 0.000 0.000 0.000 0.000 0.007 0.000 0.000 0.957 0.226 0.044 0.062 0.003 0.160 Notes: Data are from the ESNUT 2012. Means and standard errors estimated considering survey sampling design, including sample weights and clustering effects. population, whereas Quintile 5 (Q5) to children in the richest 20%15. The same classification of quintiles was used for the analysis of all child development outcomes. Given its straightforward interpretation, comparisons of outcomes by quintile have been widely used to characterize gradients in child’s health and development and are considered a preferable approach when other more complex measures of inequality do not provide additional insight of the problem [28]. In addition to the descriptive approach, we used a re- gression framework to assess the adjusted associations between child development outcomes and socioeco- nomic status, after controlling for other demographic factors that might affect child development. The SES gap in child development was estimated by OLS using the following equation: Y i ¼ α þ X 5 k¼2 β kQki þ γagei þ δfemalei þ εi ð1Þ where Yi is the development outcome for child i; agei are semi-parametric controls for child’s age in months using 3-month bins; femalei is a dummy variable equal one if child is a girl and zero otherwise. Qki is a binary indicator for the kth quintile of the SES distribution where the omitted category is Q1. Hence, the estimated 15Following Rutstein and Johnson [27], the quintile cutoffs are based on total household population, instead of the households themselves. Quintiles were constructed using a weighted frequency distribution of households where weights were the product of the number of de jure household members and the sampling weight of the household. Consequently, a tabulation of the unweighted sample will not produce five quintile groups of equal size, but the weighted sample will. ^β k represents the difference in the average coefficient outcome obtained by children in quintile k with respect to the average outcome in Q1. Joint hypothesis testing was used to test the association between the outcome and socioeconomic status across all quintiles16. In all analytical approaches, survey sampling design, including sample weights and clustering effects, were considered when computing point estimates and standard errors. Results Descriptive overview This section presents descriptive statistics and un- adjusted mean differences between high and low con- sumption quintiles for our outcomes of interest. For each measure, Table 1 shows mean, standard error and sample size for the whole sample of children and disag- gregated for children at the bottom quintile (Q1) and the top quintile (Q5) of household consumption. The last two columns show the unadjusted difference in means between high and low quintile children and the p-value for the test of equality of means. There were large and statistically significant differ- ences in all anthropometrics z-scores between chil- dren in Q5 and children in Q1: height for age (0.95 SD), weight for age (0.70 SD), and weight for height (0.21 SD). Differences were also demonstrated for prevalence of stunting: while one in every ten chil- dren were stunted in the richest quintile, this propor- tion more than tripled to one in every three children 16We test whether the coefficients for the quintile indicators are jointly zero. Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 6 of 25 in the poorest quintile. Although the prevalence of underweight is low in Bolivia, the percentage of poor children that were underweight (2.5%) more than doubled that of the richest group. On the other hand, the prevalence of overweight for children in the top quintile (10.1%) almost doubled that of children at the bottom quintile (5.7%). Results for biomarkers showed that the prevalence of anemia in Bolivia was high, with more than half of chil- dren (54%) suffering from anemia in 2012. When dividing the sample in children at the bottom and top consump- tion quintiles the results showed that 68.6% of children in the poorest quintile were anemic, compared to 40.9% in Interestingly, although anemia is the richest quintile. Fig. 1 Child nutrition and development indicators by consumption quintile across age cohorts Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 7 of 25 significantly lower among the richest group, it is still higher than average values in most other countries in the region. Similarly, the prevalence of vitamin A deficiency was 29.9% for children in the richest quintile and almost 10 percentage points higher for those in the poorest quin- tile (39.0%). By contrast, the SES gap in iron deficiency was somewhat smaller with prevalence varying from 16.4% in the richest group to 23.8% in the poorest group. The socioeconomic gaps for these indicators were all sta- tistically significant at the 5% confidence level, except for iron deficiency that was significant at the 10% level. Finally, Table 1 shows mean standardized gross motor and communication z-scores which are centered at zero since they were constructed within sample. The results showed that children in the richest quintile have a gross motor score 0.21 SD higher than children in the poorest quintile. For the communication z-scores, the difference between children in the poorest and richest quintiles was 0.11 SD, but not statistically significant. Overall, the descriptive analysis indicates that there were large and significant gaps in child development and its associated nutritional risk factors by socioeconomic status. Non-Parametric relation of SES gap and age To explore the onset and evolution of the socioeco- nomic gap in child development, we analyzed the differ- ences in outcomes between the bottom and top quintiles across different age groups. In Fig. 1 we plotted kernel weighted local polynomial regressions of standardized scores for all child nutrition and development indicators and reported 95% confidence intervals for each group. The first panel shows that even at very young ages there was a significant SES gap in height for age, and that this gap widened for children in older age groups. In the first five months of life, the gap was approximately 0.5 SD. Among older children, the z-score of height for age de- creased for children in both the poorest and richest quin- tiles; however, the gap between them started to broaden markedly at 6 months until it reached its maximum at 24 months, stabilizing at around 1 SD. The gap path over age is similar when we analyze prevalence of stunting (Panel 4). At the first five months children in the poorest house- holds were approximately 7 percentage points more likely to be stunted than those in the richest households. This gap reached a noticeable peak of more than 25 percentage points for 24-month-old children. While the gap tends to decrease slightly for older cohorts, it remained at 20 per- centage points among five-year-old children. The next panels show weight for age and weight for height z-scores as well as prevalence of children under- weight and overweight. In the case of weight for age (Panel 2), the socioeconomic gap widened from approxi- mately 0.5 SD during the first 6 months of age to 0.65 SD at 18 months, stabilizing around this number at older ages. For prevalence of underweight (Panel 5) the evidence suggests that any initial (non-significant) gap between children in the top and bottom quintiles tends to disappear with age. The gap between rich and poor children in the weight for height z-score showed that while it tended to increase significantly from birth until 15 months of age, it reduced again in the following two years (Panel 3). Finally, while differences in prevalence of overweight between the richest and poorest children were less clear during the first two years of life, the gap increased and became more relevant among three-year- olds, remaining relatively stable at approximately 6 per- centage points (Panel 6). In Panels 7 to 9 we analyzed the SES gap of nutrition in- dicators obtained from biomarkers. Panel 7 shows how the gap in anemia between children in the richest and poorest quintile changed by age group. Noticeably, the lar- gest SES gap in anemia occurred in early infancy (around 30 percentage points), declined for older cohorts and tended to increase again among children between three and five years of age. Panels 8 and 9 present vitamin A de- ficiency and iron deficiency by age, respectively. Since DBS were taken only for a random sample of children aged 6 to 24 months, confidence intervals for indicators based on DBS were substantially larger. The results for VAD showed a difference of approximately 20 percentage points between the poorest and richest quintile for chil- dren 6 to 9 months old; however, the gap closed rapidly at older ages. The prevalence of iron deficiency showed no significant differences between poorest and richest chil- dren when we compared within age groups. The last two panels of Fig. 1 show socioeconomic dif- ferences in standardized scores of gross motor and com- munication skills across age groups. The gap in gross motor skills became positive around the age of 15 months, stabilizing at about 0.30 SD in favor of the rich- est group. For communication skills, results showed no significant differences between children in Q1 and Q5 within age groups, although for some multivariate re- gression models discussed below the SES gaps in com- munication become apparent. Overall, the nonparametric analysis of gaps within age subgroups showed that, for key nutritional indicators, there were significant SES gaps that started early in life (e.g. anemia, stunting). In some cases, the gaps increased markedly for older children (e.g. stunting) and in other cases they remained relatively stable over time. The gap, while it exists, was less noticeable for measures of vita- min A and iron deficiency. Parametric estimation of SES gaps in child development Results of parametric regressions defined in equation (1) for the whole sample and disaggregated by ecological re- gions are presented in Table 2. After adjusting for sex Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 8 of 25 ) 4 9 0 0 ( . 4 4 0 0 - . ) 8 7 0 0 ( . * * 7 7 1 0 . ) 5 6 0 0 ( . 5 8 0 0 . ) 5 4 0 0 ( . 3 7 0 0 . ) 2 7 0 0 ( . * * 6 4 1 0 . ) 1 7 0 0 ( . * * * 8 6 2 0 . ) 9 7 0 0 ( . * * 3 9 1 0 . ) 5 4 0 0 ( . * * * 9 2 2 0 . ) 8 8 0 0 ( . * * * 1 9 2 0 . ) 7 7 0 0 ( . * * * 0 2 2 0 . ) 4 8 0 0 ( . * * * 2 5 2 0 . ) 2 5 0 0 ( . * * * 7 9 2 0 . ) 9 9 0 0 ( . 7 2 0 0 . ) 8 7 0 0 ( . * * 7 9 1 0 . ) 9 7 0 0 ( . 5 9 0 0 . ) 0 5 0 0 ( . * * 0 1 1 0 . ) 6 7 0 0 ( . * * * 2 8 3 0 . ) 4 8 0 0 ( . * * * 1 0 5 0 . ) 1 8 0 0 ( . * * * 1 3 3 0 . ) 9 4 0 0 ( . * * * 1 5 4 0 . ) 6 8 0 0 ( . * * * 3 1 6 0 . ) 1 7 0 0 ( . * * * 6 5 6 0 . ) 0 8 0 0 ( . * * * 1 5 4 0 . ) 1 5 0 0 ( . * * * 7 4 6 0 . ) 3 0 1 0 ( . 8 6 1 0 . ) 1 7 0 0 ( . 6 5 0 0 . ) 4 0 1 0 ( . * * 5 6 2 0 . ) 9 4 0 0 ( . * * * 2 6 1 0 . ) 7 8 0 0 ( . * * * 6 5 5 0 . ) 0 8 0 0 ( . * * * 9 0 4 0 . ) 8 8 0 0 ( . * * * 0 5 4 0 . ) 0 5 0 0 ( . * * * 1 4 5 0 . ) 6 8 0 0 ( . * * * 5 7 7 0 . ) 4 8 0 0 ( . * * * 4 8 6 0 . ) 7 0 1 0 ( . * * * 6 2 5 0 . ) 8 5 0 0 ( . * * * 0 8 7 0 . ) 1 0 1 0 ( . * * 3 5 2 0 . ) 6 8 0 0 ( . 3 4 1 0 . ) 1 4 1 0 ( . 2 8 1 0 . ) 5 5 0 0 ( . * * * 4 1 2 0 . ) 6 9 0 0 ( . * * * 6 4 7 0 . ) 7 8 0 0 ( . * * * 3 6 5 0 . ) 6 2 1 0 ( . * * * 5 5 5 0 . ) 8 5 0 0 ( . * * * 7 9 6 0 . ) 7 0 1 0 ( . * * * 4 1 0 1 . ) 9 7 0 0 ( . * * * 9 1 8 0 . ) 6 9 0 0 ( . * * * 2 2 7 0 . ) 1 6 0 0 ( . * * * 2 6 9 0 . ) 6 5 0 0 ( . 7 3 0 0 - . ) 6 4 0 0 ( . 9 0 0 0 . ) 3 5 0 0 ( . * * * 9 6 1 0 . ) 0 3 0 0 ( . 3 3 0 0 . ) 6 5 0 0 ( . 8 1 0 0 - . ) 7 4 0 0 ( . 4 6 0 0 . ) 3 4 0 0 ( . * * * 0 2 2 0 . ) 9 2 0 0 ( . * * 5 7 0 0 . ) 8 5 0 0 ( . 4 1 0 0 . ) 6 5 0 0 ( . * * 2 2 1 0 . ) 2 5 0 0 ( . * * * 9 7 1 0 . ) 4 3 0 0 ( . * * * 9 9 0 0 . ) 2 9 2 0 ( . 0 3 0 0 . ) 2 7 1 0 ( . 6 5 2 0 . ) 8 4 1 0 ( . 0 1 1 0 . ) 1 2 1 0 ( . 2 3 1 0 . ) 9 6 1 0 ( . * * 3 3 4 0 - . * * * 1 9 8 0 - . ) 1 5 1 0 ( . * * * 4 1 1 1 - . ) 8 3 1 0 ( . * * * 3 4 8 0 - . ) 6 9 0 0 ( . * * * 6 0 7 0 - . ) 3 3 1 0 ( . * * * 3 3 4 1 - . ) 3 5 1 0 ( . * * * 4 3 5 1 - . ) 9 8 1 0 ( . * * * 1 8 2 1 - . ) 8 9 0 0 ( . s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A e r o c s - z t h g e h i r o f t h g e W i e r o c s - z e g a r o f t h g e W i e r o c s - z e g a r o f t h g e H i i n o g e r y b s r o t a c d n i i t n e m p o e v e d l d n a n o i t i r t u n d l i h c n i i s t n e d a r g S E S d e t s u d A j 2 e l b a T ) 7 1 0 0 ( . 0 1 0 0 - . ) 6 1 0 0 ( . 7 0 0 0 . ) 4 1 0 0 ( . 9 0 0 0 . ) 9 0 0 0 ( . 5 0 0 0 . ) 3 1 0 0 ( . 3 0 0 0 - . ) 9 0 0 0 ( . * 7 1 0 0 - . ) 9 0 0 0 ( . 2 1 0 0 - . ) 6 0 0 0 ( . 9 0 0 0 - . ) 2 3 0 0 ( . * * 0 7 0 0 - . ) 1 2 0 0 ( . 9 0 0 0 - . ) 7 1 0 0 ( . 2 2 0 0 . ) 8 1 0 0 ( . 6 1 0 0 . ) 0 1 0 0 ( . 4 1 0 0 . ) 9 0 0 0 ( . 5 0 0 0 - . ) 0 1 0 0 ( . 2 0 0 0 - . ) 0 1 0 0 ( . 4 0 0 0 - . ) 5 0 0 0 ( . 3 0 0 0 - . ) 2 2 0 0 ( . 7 1 0 0 . ) 6 1 0 0 ( . 3 0 0 0 - . ) 4 2 0 0 ( . * * 9 4 0 0 . ) 1 1 0 0 ( . * * 3 2 0 0 . * * * 2 2 0 0 - . ) 8 0 0 0 ( . ) 0 1 0 0 ( . 6 0 0 0 - . * * * 0 2 0 0 - . ) 7 0 0 0 ( . * * * 6 1 0 0 - . ) 5 0 0 0 ( . ) 5 2 0 0 ( . * * 1 6 0 0 . ) 6 1 0 0 ( . 1 1 0 0 . ) 2 3 0 0 ( . 9 4 0 0 . ) 2 1 0 0 ( . * * * 6 4 0 0 . ) 9 0 0 0 ( . 2 1 0 0 - . ) 9 0 0 0 ( . * * 7 1 0 0 - . ) 2 1 0 0 ( . 3 1 0 0 - . ) 5 0 0 0 ( . * * 3 1 0 0 - . * * * 0 2 1 0 - . ) 6 2 0 0 ( . * * * 1 6 1 0 - . ) 9 2 0 0 ( . * * * 3 6 1 0 - . ) 0 3 0 0 ( . s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l 5 6 6 3 , 4 4 0 0 . 6 4 1 0 0 . 1 9 3 3 , 7 1 0 0 . 8 2 7 0 0 . 7 9 3 3 , 8 3 0 0 . 9 4 6 1 0 . 3 5 4 0 1 , 0 2 0 0 . 0 2 0 0 0 . 0 7 6 3 , 6 7 0 0 . 0 0 0 0 0 . 5 9 3 3 , 1 5 0 0 . 0 0 0 0 0 . 4 0 4 3 , 1 7 0 0 . 0 0 0 0 0 . 9 6 4 0 1 , 7 6 0 0 . 0 0 0 0 0 . 3 2 7 3 , 7 5 1 0 . 0 0 0 0 0 . ) % ( i t h g e w r e v O ) % ( i t h g e w r e d n U 1 0 7 3 , 5 1 1 0 . 0 0 0 0 0 . * * * 7 7 0 0 - . ) 9 2 0 0 ( . * * * 6 8 1 0 - . ) 5 2 0 0 ( . * * * 6 0 2 0 - . ) 5 2 0 0 ( . * * * 9 2 2 0 - . ) 2 2 0 0 ( . s y e l l a V ) 6 8 0 0 ( . 4 7 0 0 - . ) 4 7 0 0 ( . 2 3 0 0 . ) 7 7 0 0 ( . 5 9 0 0 . ) 8 4 0 0 ( . 9 2 0 0 . ) 3 1 1 0 ( . 4 3 1 0 . ) 0 9 0 0 ( . * * 6 1 2 0 . ) 5 8 0 0 ( . 2 2 1 0 - . ) 5 5 0 0 ( . 7 6 0 0 . ) 6 3 0 0 ( . * * 0 8 0 0 - . ) 2 3 0 0 ( . 6 4 0 0 - . ) 8 7 0 0 ( . 1 8 0 0 - . ) 3 6 0 0 ( . 7 9 0 0 . ) 4 9 0 0 ( . 3 0 1 0 . ) 3 4 0 0 ( . 0 2 0 0 . ) 0 0 1 0 ( . * 7 7 1 0 . ) 7 7 0 0 ( . 8 6 0 0 . ) 3 6 0 0 ( . * * 1 4 1 0 - . ) 9 4 0 0 ( . 3 7 0 0 . * * * 1 6 1 0 - . ) 6 3 0 0 ( . * * 2 9 0 0 - . s d n a w o L l s y e l l a V l s d n a h g H i ) % ( y c n e i c i f e d n o r I l l A s d n a w o L l s y e l l a V l s d n a h g H i ) % ( D A V l l A s d n a w o L l s y e l l a V ) 0 9 0 0 ( . 5 0 1 0 - . ) 0 6 0 0 ( . 0 9 0 0 . ) 9 0 1 0 ( . * * * 2 2 3 0 . ) 7 4 0 0 ( . 9 1 0 0 . ) 6 2 1 0 ( . 0 5 0 0 . ) 1 6 0 0 ( . * 6 0 1 0 - . ) 1 9 0 0 ( . * * 4 8 1 0 - . ) 4 6 0 0 ( . 3 2 0 0 - . * * * 2 5 1 0 - . ) 5 4 0 0 ( . ) 3 3 0 0 ( . ) 6 3 0 0 ( . * * * 1 7 1 0 - . ) 7 4 0 0 ( . * * 7 9 0 0 - . 5 6 6 3 , 7 2 0 0 . 7 9 0 0 0 . 1 9 3 3 , 3 1 0 0 . 5 5 6 5 0 . 7 9 3 3 , 5 2 0 0 . 2 7 0 3 0 . 3 5 4 0 1 , 1 1 0 0 . 1 5 0 0 0 . 5 6 6 3 , 1 2 0 0 . 8 0 0 0 0 . 1 9 3 3 , 2 2 0 0 . 1 7 0 2 0 . 7 9 3 3 , 2 2 0 0 . 3 8 0 0 0 . 3 5 4 0 1 , 4 1 0 0 . 4 6 1 0 0 . 3 2 7 3 , 5 4 0 0 . 0 0 0 0 0 . 1 0 7 3 , 3 7 0 0 . 0 0 0 0 0 . 6 4 4 3 , 4 7 0 0 . 0 0 0 0 0 . * * 6 2 0 0 - . ) 3 1 0 0 ( . ) 2 1 0 0 ( . 0 2 0 0 - . ) 1 1 0 0 ( . 1 1 0 0 . ) 8 0 0 0 ( . * 3 1 0 0 - . ) 6 0 0 0 ( . 0 1 0 0 - . ) 6 0 0 0 ( . * 2 1 0 0 - . * * * 8 1 0 0 - . ) 6 0 0 0 ( . * * * 2 1 0 0 - . ) 3 0 0 0 ( . ) 6 1 0 0 ( . 8 0 0 0 - . ) 8 1 0 0 ( . * * 1 4 0 0 - . ) 4 2 0 0 ( . * * 0 6 0 0 - . ) 2 7 0 0 ( . * 6 3 1 0 . ) 3 3 0 0 ( . * * * 6 3 1 0 . ) 3 4 0 0 ( . * * 3 9 0 0 . ) 9 2 0 0 ( . * * * 9 1 1 0 . ) 5 2 0 0 ( . * * * 6 7 0 0 . ) 1 3 0 0 ( . * * * 9 8 0 0 . ) 1 2 0 0 ( . * * 2 5 0 0 . ) 6 1 0 0 ( . * * * 5 7 0 0 . ) 6 3 0 0 ( . * * * 9 6 1 0 . ) 4 4 0 0 ( . * * * 5 4 3 0 . ) 9 4 0 0 ( . * * * 0 1 3 0 . ) 6 2 0 0 ( . * * * 5 9 2 0 . l s d n a h g H i * * * 3 4 1 0 - . ) 0 3 0 0 ( . * * * 8 9 1 0 - . ) 1 3 0 0 ( . * * * 1 0 2 0 - . ) 7 3 0 0 ( . * * * 0 6 2 0 - . ) 4 3 0 0 ( . l s d n a h g H i * * * 4 4 1 0 - . ) 1 3 0 0 ( . * * * 7 6 1 0 - . ) 4 4 0 0 ( . * * * 2 1 1 0 - . ) 8 1 0 0 ( . * * * 4 8 1 0 - . ) 6 1 0 0 ( . * * * 6 1 2 0 - . ) 7 1 0 0 ( . * * * 6 3 2 0 - . ) 6 1 0 0 ( . * * * 4 3 0 0 - . ) 1 1 0 0 ( . l l A ) % ( 0 7 8 0 1 , 7 6 0 0 . 0 0 0 0 0 . a i m e n A l l A * * * 3 0 1 0 - . ) 0 2 0 0 ( . * * * 4 6 1 0 - . ) 3 2 0 0 ( . * * * 6 8 1 0 - . ) 6 2 0 0 ( . 6 4 4 3 , 4 9 0 0 . 0 0 0 0 0 . 0 7 8 0 1 , 2 3 1 0 . 0 0 0 0 0 . ) % ( g n i t n u t S e l i t n u q i 1 = 2 e l i t n u Q i 1 = 3 e l i t n u Q i 1 = 4 e l i t n u Q i 1 = 5 e l i t n u Q i S E L B A R A V I t n a t s n o C l e a m e F r o f t s e t - F ( l e u a v - p ) s t c e f f e s n o i t a v r e s b O d e r a u q s - R 1 = 2 e l i t n u Q i S E L B A R A V I 1 = 3 e l i t n u Q i 1 = 4 e l i t n u Q i e l i t n u q i r o f t s e t - F ( l e u a v - p ) s t c e f f e S E L B A R A V I 1 = 5 e l i t n u Q i s n o i t a v r e s b O d e r a u q s - R t n a t s n o C l e a m e F 1 = 2 e l i t n u Q i 1 = 3 e l i t n u Q i 1 = 4 e l i t n u Q i Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 9 of 25 ) 0 9 0 0 ( . 9 9 0 0 - . ) 6 5 0 0 ( . 2 5 0 0 - . ) 7 7 0 0 ( . 7 1 0 0 - . ) 8 3 0 0 ( . * * 4 8 0 0 - . ) 8 0 1 0 ( . 7 7 0 0 - . ) 1 6 0 0 ( . 0 7 0 0 - . ) 1 8 0 0 ( . * 4 4 1 0 - . ) 4 4 0 0 ( . * 5 8 0 0 - . * * * 4 4 2 0 - . ) 1 3 0 0 ( . s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l * * * 6 0 2 0 - . ) 3 3 0 0 ( . s y e l l a V l s d n a h g H i * * * 4 3 3 0 - . ) 4 4 0 0 ( . * * * 6 8 2 0 - . ) 2 2 0 0 ( . l l A 1 = 5 e l i t n u Q i S E L B A R A V I ) 7 4 0 0 ( . 9 6 0 0 - . * * * 2 2 1 0 - . ) 4 4 0 0 ( . * * * 0 5 1 0 - . ) 0 5 0 0 ( . * * * 9 0 1 0 - . ) 9 2 0 0 ( . ) 6 5 0 0 ( . 9 0 0 0 - . ) 9 3 0 0 ( . 7 1 0 0 - . ) 0 5 0 0 ( . 7 1 0 0 - . ) 9 2 0 0 ( . 2 1 0 0 - . ) 2 2 0 0 ( . 4 1 0 0 - . ) 3 2 0 0 ( . * 5 4 0 0 - . ) 2 3 0 0 ( . * 8 5 0 0 - . ) 5 1 0 0 . * * * 9 3 0 0 - . l e a m e f e r o c s - z t h g e h i r o f t h g e W i e r o c s - z e g a r o f t h g e W i e r o c s - z e g a r o f t h g e H i ) d e u n i t n o C ( i n o g e r y b s r o t a c d n i i t n e m p o e v e d l d n a n o i t i r t u n d l i h c n i i s t n e d a r g S E S d e t s u d A j 2 e l b a T ) 5 9 0 0 ( . * * 6 1 2 0 . ) 7 5 0 0 ( . * * 8 4 1 0 . ) 5 7 0 0 ( . * * * 8 5 3 0 . ) 0 4 0 0 ( . * * * 9 2 2 0 . ) 2 0 1 0 ( . * * * 7 4 5 0 . ) 4 6 0 0 ( . * * * 3 5 3 0 . ) 8 8 0 0 ( . * * * 1 0 6 0 . ) 5 4 0 0 ( . * * * 6 6 4 0 . ) 8 5 0 0 ( . * * * 6 2 5 0 . ) 7 5 0 0 ( . * * * 3 4 5 0 . ) 0 7 0 0 ( . * * * 2 0 8 0 . ) 7 3 0 0 ( . * * * 1 2 6 0 . 6 4 0 0 . 6 2 5 0 8 0 8 0 . 0 9 0 0 . 5 3 5 9 9 4 0 0 . 1 9 0 0 . 8 4 5 1 1 0 0 0 . 9 0 6 1 , 6 4 0 0 . 9 3 2 0 0 . 2 7 0 0 . 6 2 5 9 4 1 0 0 . 0 6 0 0 . 5 3 5 2 6 0 0 0 . 6 8 0 0 . 8 4 5 2 6 2 1 0 . 9 0 6 1 , 4 3 0 0 . 9 7 0 0 0 . 5 1 3 3 , 1 0 1 0 . 0 0 0 0 0 . 5 0 2 3 , 0 1 1 0 . 0 0 0 0 0 . e r o c s - z n o i t a c i n u m m o C 4 9 8 2 , 6 9 0 0 . 0 0 0 0 0 . 4 1 4 9 , 7 9 0 0 . 0 0 0 0 0 . e r o c s - z r o t o m s s o r G e l i t n u q i ) 6 7 1 0 ( . 2 9 0 0 - . ) 2 2 1 0 ( . 6 5 0 0 . ) 3 1 1 0 ( . 2 1 1 0 - . ) 6 5 1 0 ( . 9 5 1 0 . ) 3 6 0 0 ( . * 0 1 1 0 . ) 0 4 1 0 ( . 9 5 1 0 . s d n a w o L l 2 3 9 1 , 8 4 0 0 . 3 3 0 1 0 . ) 6 0 1 0 ( . 1 0 1 0 - . ) 0 0 1 0 ( . 9 3 0 0 . ) 0 1 1 0 ( . 1 0 1 0 - . ) 8 0 1 0 ( . 6 2 1 0 . ) 2 6 0 0 ( . 5 7 0 0 . ) 9 3 1 0 ( . 9 9 0 0 . s y e l l a V 2 1 9 1 , 1 3 0 0 . 0 1 1 3 0 . l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A ) 2 8 0 0 ( . * * 1 7 1 0 - . ) 1 7 0 0 ( . * 3 3 1 0 - . ) 7 2 2 0 ( . 6 6 0 0 . ) 6 8 0 0 ( . 9 1 0 0 - . ) 5 8 0 0 ( . 5 3 0 0 . ) 2 7 0 0 ( . 8 2 0 0 . ) 6 0 1 0 ( . 2 4 0 0 . ) 9 5 0 0 ( . 7 3 0 0 . ) 0 5 1 0 ( . 6 0 1 0 . ) 9 7 0 0 ( . 9 9 0 0 - . ) 2 2 1 0 ( . 0 0 0 0 - . ) 1 6 0 0 ( . 6 1 0 0 . ) 9 4 1 0 ( . 8 3 0 0 - . ) 7 6 0 0 ( . 4 8 0 0 - . ) 4 4 1 0 ( . 6 2 1 0 . ) 2 1 1 0 ( . 1 8 0 0 - . ) 1 6 1 0 ( . 8 7 0 0 - . ) 9 6 0 0 ( . 1 4 0 0 . ) 6 3 1 0 ( . 5 1 0 0 - . ) 3 7 0 0 ( . 3 1 1 0 . ) 7 9 1 0 ( . * 7 6 3 0 . ) 0 9 0 0 ( . 1 5 0 0 . ) 9 4 1 0 ( . 5 1 0 0 - . ) 0 7 0 0 ( . * * * 0 1 2 0 . ) 0 6 0 0 ( . 8 9 0 0 . ) 5 3 0 0 ( . * * * 1 9 0 0 . ) 7 7 0 0 ( . 8 8 0 0 - . ) 7 5 0 0 ( . 0 8 0 0 - . ) 6 6 0 0 ( . 9 5 0 0 - . ) 0 4 0 0 ( . * 6 7 0 0 - . ) 9 2 1 0 ( . 9 1 0 0 - . ) 8 7 0 0 ( . 8 9 0 0 . ) 6 0 2 0 ( . 9 5 1 0 - . ) 4 0 1 0 ( . 2 0 1 0 - . ) 5 3 1 0 ( . 3 5 0 0 - . ) 6 7 0 0 ( . 4 1 1 0 - . 9 0 9 1 , 7 1 0 0 . 1 8 0 2 0 . 3 5 7 5 , 6 1 0 0 . 2 8 0 0 0 . 2 3 9 1 , 5 2 0 0 . 9 3 0 0 0 . 2 1 9 1 , 3 3 0 0 . 9 3 2 5 0 . 9 0 9 1 , 7 2 0 0 . 3 1 6 9 0 . 3 5 7 5 , 5 1 0 0 . 6 7 0 0 0 . e l i t n u q i r o f t s e t - F ( l e u a v - p ) s t c e f f e s n o i t a v r e s b O d e r a u q s - R t n a t s n o C 1 = 2 e l i t n u Q i 1 = 3 e l i t n u Q i 1 = 4 e l i t n u Q i 1 = 5 e l i t n u Q i S E L B A R A V I s n o i t a v r e s b O d e r a u q s - R t n a t s n o C l e a m e F r o f t s e t - F ( l e u a v - p ) s t c e f f e d l i h c f o t e s a d n a x e s ’ s d l i h c e d u l c n i s l o r t n o C . n o i t a m i t s E S L O . s e s e h t n e r a p n i ) t n e m g e s r o r o t c e s s u s n e c ( l e v e l U S P e h t t a d e r e t s u l c E S . l e v e l % 0 1 e h t t a t n a c i f i n g i s * ; l e v e l % 5 e h t t a t n a c i f i n g i s * * ; l e v e l % 1 e h t t a t n a c i f i n g i s * * * : s e t o N a t i p a c r e p l y h t n o m g n i s u s e l i t n u q i l n o i t a u p o p e r a s e l i t n u Q i . ) 9 5 - 7 5 , 6 5 - 4 5 , 3 5 - 1 5 , 0 5 - 8 4 , 7 4 - 5 4 , 4 4 - 2 4 , 1 4 - 9 3 , 8 3 - 6 3 , 5 3 - 3 3 , 2 3 - 0 3 , 9 2 - 7 2 , 6 2 - 4 2 , 3 2 - 1 2 , 0 2 - 8 1 , 7 1 - 5 1 , 4 1 - 2 1 , 1 1 - 9 , 8 - 6 . 0 = 5 Q = 4 Q = 3 Q = 2 Q f o t s e t - F e h t f o l e u a v - p e h t s i ) s t c e f f e e l i t n u q i r o f t s e t - F ( l e u a v - p , 5 - 3 , 2 - 0 ( s h t n o m n i s e i r o g e t a c e g a r o f i s e m m u d . l e b a i r a v i g n k n a r e h t s a n o i t p m u s n o c l d o h e s u o h Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 10 of 25 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 S E L B A R A V I e r o c s - z t h g e h i r o f t h g e W i e r o c s - z e g a r o f t h g e W i e r o c s - z e g a r o f t h g e H i e g a ’ s d l i h c y b s r o t a c d n i i t n e m p o e v e d l d n a n o i t i r t u n d l i h c n i i s t n e d a r g S E S d e t s u d A j 3 e l b a T ) 7 6 0 0 ( . 5 2 0 0 . ) 5 8 0 0 ( . 7 9 0 0 . ) 9 8 0 0 ( . * * 6 8 1 0 . ) 0 0 1 0 ( . 5 6 1 0 . ) 2 6 0 0 ( . * * * 4 0 2 0 . ) 0 8 0 0 ( . * * * 7 6 2 0 . ) 1 9 0 0 ( . * * * 6 9 2 0 . ) 6 0 1 0 ( . * 4 0 2 0 . ) 9 6 0 0 ( . * * * 3 1 3 0 . ) 7 8 0 0 ( . * * * 3 8 3 0 . ) 5 9 0 0 ( . * * * 8 8 2 0 . ) 7 2 1 0 ( . 3 3 1 0 . ) 5 6 0 0 ( . 9 7 0 0 . ) 3 9 0 0 ( . 5 1 0 0 . ) 8 8 0 0 ( . * * * 3 9 2 0 . ) 2 1 1 0 ( . * 1 1 2 0 . ) 6 6 0 0 ( . * * * 8 0 5 0 . ) 6 8 0 0 ( . * * * 8 8 3 0 . ) 7 8 0 0 ( . * * * 8 3 5 0 . ) 8 0 1 0 ( . * * * 1 8 3 0 . ) 9 6 0 0 ( . * * * 9 4 7 0 . ) 0 8 0 0 ( . * * * 2 9 6 0 . ) 1 9 0 0 ( . * * * 7 1 6 0 . ) 4 3 1 0 ( . * * * 9 2 4 0 . ) 3 6 0 0 ( . * * 7 4 1 0 . ) 0 9 0 0 ( . * * 8 9 1 0 . ) 3 0 1 0 ( . * * 6 2 2 0 . ) 9 2 1 0 ( . 6 1 1 0 . ) 1 6 0 0 ( . * * * 8 3 5 0 . ) 7 8 0 0 ( . * * * 6 7 6 0 . ) 0 0 1 0 ( . * * * 1 6 5 0 . ) 8 2 1 0 ( . * * * 9 4 4 0 . ) 4 7 0 0 ( . * * * 8 6 7 0 . ) 8 8 0 0 ( . * * * 2 6 9 0 . ) 4 0 1 0 ( . * * * 1 2 8 0 . ) 6 3 1 0 ( . * * * 6 0 6 0 . ) 6 7 0 0 ( . * * 7 6 1 0 . ) 6 8 0 0 ( . * 9 5 1 0 . ) 8 0 1 0 ( . * * * 0 9 3 0 . ) 3 2 1 0 ( . * 1 4 2 0 . ) 2 8 0 0 ( . * * * 5 5 7 0 . ) 6 8 0 0 ( . * * * 9 3 7 0 . ) 7 0 1 0 ( . * * * 1 8 7 0 . ) 1 3 1 0 ( . * * * 1 2 5 0 . ) 4 8 0 0 ( . * * * 7 6 0 1 . ) 9 9 0 0 ( . * * * 4 2 1 1 . ) 7 2 1 0 ( . * * * 8 8 9 0 . ) 6 4 1 0 ( . * * * 8 0 6 0 . 1 = 2 1 = 3 1 = 4 1 = 5 e l i t n u Q i e l i t n u Q i e l i t n u Q i e l i t n u Q i ) 2 9 0 0 ( . * * * 3 5 5 0 . ) 4 7 0 0 ( . * * * 6 7 3 0 . ) 2 9 0 0 ( . 7 0 1 0 . ) 4 0 1 0 ( . * * * 0 8 2 0 . ) 2 9 0 0 ( . * * * 9 3 5 0 - . ) 5 7 0 0 ( . * * * 1 2 7 0 - . ) 9 8 0 0 ( . * * * 8 2 7 0 - . ) 9 0 1 0 ( . * * * 2 5 3 0 - . ) 2 9 0 0 ( . * * * 6 4 6 1 - . ) 7 8 0 0 ( . * * * 4 5 8 1 - . ) 9 9 0 0 ( . * * * 5 6 6 1 - . ) 8 1 1 0 ( . * * * 8 9 7 0 - . t n a t s n o C ) 8 4 0 0 ( . 1 5 0 0 - . ) 4 5 0 0 ( . 9 3 0 0 . ) 1 6 0 0 ( . * 5 0 1 0 . ) 3 7 0 0 ( . 8 9 0 0 . ) 2 5 0 0 ( . 4 8 0 0 - . ) 3 5 0 0 ( . * * 1 1 1 0 . ) 5 6 0 0 ( . * * * 3 9 1 0 . ) 4 7 0 0 ( . * * 5 6 1 0 . ) 3 5 0 0 ( . 1 6 0 0 - . ) 2 6 0 0 ( . * * 7 5 1 0 . ) 2 7 0 0 ( . * * * 4 4 2 0 . ) 4 8 0 0 ( . * * 3 7 1 0 . 1 = l e a m e F s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 S E L B A R A V I ) % ( i t h g e w r e v O ) % ( i t h g e w r e d n U ) % ( g n i t n u t S 2 0 7 3 , 8 0 0 0 . 4 6 8 0 0 . 3 2 3 2 , 9 0 0 0 . 7 0 4 1 0 . 5 8 2 2 , 3 2 0 0 . 7 2 0 0 0 . 9 8 6 1 , 8 0 0 0 . 1 9 9 1 0 . 7 0 7 3 , 7 7 0 0 . 0 0 0 0 0 . 5 2 3 2 , 0 8 0 0 . 0 0 0 0 0 . 5 8 2 2 , 1 7 0 0 . 0 0 0 0 0 . 3 9 6 1 , 4 3 0 0 . 1 0 0 0 0 . 2 8 8 3 , 9 2 1 0 . 0 0 0 0 0 . 1 1 4 2 , 5 2 1 0 . 0 0 0 0 0 . 1 9 3 2 , 7 9 0 0 . 0 0 0 0 0 . 8 1 7 1 , 5 4 0 0 . 0 0 0 0 0 . ) s t c e f f e e l i t n u q i r o f t s e t - F ( l e u a v - p s n o i t a v r e s b O d e r a u q s - R ) 4 1 0 0 ( . 4 0 0 0 . ) 0 2 0 0 ( . 4 1 0 0 . ) 7 1 0 0 ( . 2 2 0 0 . ) 2 2 0 0 ( . 2 0 0 0 . ) 1 1 0 0 ( . 8 0 0 0 . ) 5 0 0 0 ( . * 9 0 0 0 - . ) 2 1 0 0 ( . * * * 3 3 0 0 - . ) 4 1 0 0 ( . 9 0 0 0 - . ) 7 2 0 0 ( . * * * 6 1 1 0 - . ) 0 4 0 0 ( . * * * 0 2 1 0 - . ) 9 3 0 0 ( . * * * 9 4 1 0 - . ) 1 3 0 0 ( . * 1 6 0 0 - . ) 7 1 0 0 ( . 7 1 0 0 . ) 7 1 0 0 ( . 6 0 0 0 . ) 8 1 0 0 ( . 3 2 0 0 . ) 7 2 0 0 ( . 9 1 0 0 . ) 7 0 0 0 ( . 6 0 0 0 . ) 9 0 0 0 ( . 3 0 0 0 . ) 3 1 0 0 ( . * * * 5 3 0 0 - . ) 7 1 0 0 ( . 0 0 0 0 - . ) 6 2 0 0 ( . * * * 9 8 1 0 - . ) 6 3 0 0 ( . * * * 5 2 2 0 - . ) 4 3 0 0 ( . * * * 7 0 2 0 - . ) 0 3 0 0 ( . * * * 3 0 1 0 - . ) 8 1 0 0 ( . 5 2 0 0 . ) 2 2 0 0 ( . 5 3 0 0 . ) 4 1 0 0 ( . 4 0 0 0 . ) 5 2 0 0 ( . 5 1 0 0 . ) 4 0 0 0 ( . 7 0 0 0 - . ) 5 0 0 0 ( . * 9 0 0 0 - . ) 4 1 0 0 ( . * * 2 3 0 0 - . ) 2 1 0 0 ( . * * 7 2 0 0 - . ) 6 2 0 0 ( . * * * 4 2 2 0 - . ) 3 3 0 0 ( . * * * 0 6 2 0 - . ) 4 3 0 0 ( . * * * 1 6 2 0 - . ) 9 2 0 0 ( . * * * 8 2 1 0 - . ) 0 2 0 0 ( . * * * 9 6 0 0 . ) 3 2 0 0 ( . * 2 4 0 0 . ) 3 2 0 0 ( . * * * 1 6 0 0 . ) 0 3 0 0 ( . 5 1 0 0 . ) 6 0 0 0 ( . 0 0 0 0 - . ) 7 0 0 0 ( . 6 0 0 0 - . ) 4 1 0 0 ( . * * 0 3 0 0 - . ) 1 1 0 0 ( . * * 9 2 0 0 - . ) 3 2 0 0 ( . * * * 5 5 2 0 - . ) 4 3 0 0 ( . * * * 5 9 2 0 - . ) 2 3 0 0 ( . * * * 6 8 2 0 - . ) 4 3 0 0 ( . * * * 4 9 0 0 - . 1 = 2 1 = 3 1 = 4 1 = 5 e l i t n u Q i e l i t n u Q i e l i t n u Q i e l i t n u Q i ) 3 1 0 0 ( . * 5 2 0 0 - . ) 8 1 0 0 ( . 1 2 0 0 - . ) 4 1 0 0 ( . 2 0 0 0 - . ) 9 1 0 0 ( . 6 0 0 0 . ) 6 0 0 0 ( . 8 0 0 0 - . ) 4 0 0 0 ( . 4 0 0 0 - . ) 7 0 0 0 ( . 7 0 0 0 - . ) 0 1 0 0 ( . * * * 5 3 0 0 - . ) 8 1 0 0 ( . 6 2 0 0 . ) 3 2 0 0 ( . * * * 0 6 0 0 - . ) 4 2 0 0 ( . * * * 0 9 0 0 - . ) 0 2 0 0 ( . * * 5 4 0 0 - . 1 = l e a m e F ) 4 2 0 0 ( . * * * 8 7 0 0 . ) 9 1 0 0 ( . * * * 0 5 0 0 . ) 7 1 0 0 ( . * * 6 3 0 0 . ) 9 1 0 0 ( . * * * 7 9 0 0 . ) 5 0 0 0 ( . 7 0 0 0 . ) 7 0 0 0 ( . * * 8 1 0 0 . ) 4 1 0 0 ( . * * * 0 4 0 0 . ) 4 1 0 0 ( . * * * 1 6 0 0 . ) 0 3 0 0 ( . * * * 7 2 3 0 . ) 5 3 0 0 ( . * * * 5 6 4 0 . ) 2 3 0 0 ( . * * * 7 0 4 0 . ) 8 2 0 0 ( . * * * 6 0 2 0 . t n a t s n o C s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 6 ) 7 5 0 0 ( . 5 1 0 0 . ) 0 8 0 0 ( . 8 4 0 0 . ) 4 5 0 0 ( . 5 1 0 0 . ) 8 6 0 0 ( . 6 2 0 0 . ) 7 6 0 0 ( . 3 0 0 0 . ) 0 7 0 0 ( . 6 4 0 0 . ) 2 5 0 0 ( . 0 8 0 0 - . ) 5 5 0 0 ( . * 3 9 0 0 - . y c n e i c i f e d n o r I ) 5 3 0 0 ( . * * * 0 1 1 0 - . ) 7 4 0 0 ( . * * 9 0 1 0 - . ) 7 5 0 0 ( . * * * 7 4 3 0 . ) 5 4 0 0 ( . * * * 0 2 2 0 . 2 0 7 3 , 3 1 0 0 . 9 4 1 0 0 . 3 2 3 2 , 8 0 0 0 . 1 2 8 2 0 . 5 8 2 2 , 0 1 0 0 . 9 3 5 0 0 . 9 8 6 1 , 7 0 0 0 . 6 2 4 9 0 . 2 0 7 3 , 1 1 0 0 . 1 5 2 1 0 . 3 2 3 2 , 5 0 0 0 . 8 8 1 2 0 . s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 6 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 S E L B A R A V I 5 8 2 2 , 3 1 0 0 . 3 1 9 0 0 . 9 8 6 1 , 3 2 0 0 . 6 6 4 0 0 . y c n e i c i f e d A n m a t i V i 2 8 8 3 , 1 6 0 0 . 0 0 0 0 0 . 1 1 4 2 , 0 7 0 0 . 0 0 0 0 0 . 1 9 3 2 , 3 7 0 0 . 0 0 0 0 0 . 8 1 7 1 , 9 2 0 0 . 2 0 0 0 0 . a i m e n A ) s t c e f f e e l i t n u q i r o f t s e t - F ( l e u a v - p s n o i t a v r e s b O d e r a u q s - R ) 0 6 0 0 ( . 8 6 0 0 . ) 9 9 0 0 ( . 7 6 0 0 . ) 4 3 0 0 ( . * * * 7 4 1 0 - . ) 5 4 0 0 ( . * * 8 1 1 0 - . ) 6 3 0 0 ( . 2 3 0 0 - . ) 4 4 0 0 ( . 2 6 0 0 - . ) 9 5 0 0 ( . 5 8 0 0 . ) 8 7 0 0 ( . 7 5 0 0 . ) 7 3 0 0 ( . * * * 1 0 2 0 - . ) 5 4 0 0 ( . * * * 4 4 2 0 - . ) 6 3 0 0 ( . 8 5 0 0 - . ) 9 4 0 0 ( . * * 7 1 1 0 - . ) 2 7 0 0 ( . 4 0 0 0 - . ) 9 8 0 0 ( . 4 5 0 0 - . ) 6 3 0 0 ( . * * * 0 5 2 0 - . ) 7 4 0 0 ( . * * * 6 1 2 0 - . ) 3 4 0 0 ( . 2 6 0 0 - . ) 7 5 0 0 ( . * * * 5 6 1 0 - . ) 4 5 0 0 ( . 4 6 0 0 - . ) 8 7 0 0 ( . 8 1 1 0 - . ) 5 3 0 0 ( . * * * 4 2 3 0 - . ) 1 5 0 0 ( . * * * 8 5 2 0 - . ) 4 4 0 0 ( . * * * 6 3 2 0 - . ) 2 5 0 0 ( . * * * 8 9 2 0 - . 1 = 2 1 = 3 1 = 4 1 = 5 e l i t n u Q i e l i t n u Q i e l i t n u Q i e l i t n u Q i ) 4 5 0 0 ( . * * * 3 8 2 0 . ) 9 5 0 0 ( . * * * 7 8 4 0 . ) 2 4 0 0 ( . * * * 6 2 7 0 . ) 2 4 0 0 ( . * * * 1 9 7 0 . ) 8 3 0 0 ( . * * * 9 2 8 0 . ) 1 4 0 0 ( . * * * 1 1 6 0 . t n a t s n o C ) 9 3 0 0 ( . 6 0 0 0 - . ) 9 4 0 0 ( . 1 2 0 0 - . ) 6 2 0 0 ( . * * 7 5 0 0 - . ) 0 3 0 0 ( . 2 0 0 0 . ) 6 2 0 0 ( . 2 3 0 0 - . ) 6 3 0 0 ( . * 2 6 0 0 - . 1 = l e a m e F Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 11 of 25 d l i h c f o t e s a d n a x e s ’ s d l i h c 9 0 0 1 , 1 4 0 0 . 3 1 1 3 0 . 7 5 0 0 . 0 0 6 3 3 9 1 0 . s h t n o M 6 3 - 4 2 ) 7 9 0 0 ( . 7 5 0 0 . ) 0 9 0 0 ( . * * 3 3 2 0 . ) 1 1 1 0 ( . 5 2 0 0 - . ) 5 2 1 0 ( . * * 2 5 2 0 . ) 0 6 0 0 ( . 1 9 0 0 . ) 9 8 0 0 ( . * * 9 2 2 0 - . 2 9 0 2 , 9 1 0 0 . 3 8 2 0 0 . ) d e u n i t n o C ( e g a ’ s d l i h c y b s r o t a c d n i i t n e m p o e v e d l d n a n o i t i r t u n d l i h c n i i s t n e d a r g S E S d e t s u d A j 3 e l b a T e r o c s - z n o i t a c i n u m m o C e r o c s - z r o t o m s s o r G s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 S E L B A R A V I 9 0 0 1 , 9 1 0 0 . 5 6 5 0 0 . 1 2 0 0 . 0 0 6 9 2 8 2 0 . 0 8 5 3 , 7 6 0 0 . 0 0 0 0 0 . 9 7 1 2 , 0 5 0 0 . 0 0 0 0 0 . 7 6 1 2 , 7 3 0 0 . 0 0 0 0 0 . 8 8 4 1 , 4 9 0 0 . 0 0 0 0 0 . ) s t c e f f e e l i t n u q i r o f t s e t - F ( l e u a v - p s n o i t a v r e s b O d e r a u q s - R ) 0 9 0 0 ( . * * * 3 9 2 0 - . ) 5 1 1 0 ( . 1 6 1 0 - . ) 6 1 1 0 ( . 9 7 1 0 . ) 3 0 1 0 ( . 7 6 0 0 - . ) 8 0 1 0 ( . 8 5 0 0 - . ) 5 8 0 0 ( . * * * 9 5 2 0 - . ) 8 9 0 0 ( . * * 9 3 2 0 . ) 1 2 1 0 ( . 5 3 1 0 . ) 7 8 0 0 ( . 1 5 0 0 - . ) 7 0 1 0 ( . 6 6 0 0 - . ) 4 0 1 0 ( . * 9 9 1 0 - . ) 8 0 1 0 ( . 6 0 0 0 . ) 7 1 1 0 ( . 6 4 1 0 . ) 9 1 1 0 ( . 2 0 0 0 . ) 4 0 1 0 ( . 0 6 0 0 - . ) 8 8 0 0 ( . 5 9 0 0 - . ) 9 1 1 0 ( . * 1 2 2 0 . ) 4 1 1 0 ( . * * * 9 3 4 0 . ) 8 9 0 0 ( . 3 4 1 0 . ) 1 1 1 0 ( . 2 3 0 0 - . 1 = 2 1 = 3 1 = 4 1 = 5 e l i t n u Q i e l i t n u Q i e l i t n u Q i e l i t n u Q i ) 5 6 0 0 ( . * * 8 4 1 0 . ) 6 6 0 0 ( . 3 0 0 0 - . ) 9 6 0 0 ( . 9 7 0 0 - . ) 6 6 0 0 ( . 5 3 0 0 - . ) 2 6 0 0 ( . * * 5 3 1 0 - . 1 = l e a m e F ) 1 9 0 0 ( . 2 3 0 0 . ) 1 9 0 0 ( . 7 6 0 0 . ) 3 1 1 0 ( . * 0 1 2 0 - . ) 8 8 0 0 ( . 6 6 0 0 - . ) 9 8 0 0 ( . 4 2 0 0 . t n a t s n o C 5 8 1 2 , 8 1 0 0 . 9 1 0 0 0 . 6 7 4 1 , 7 2 0 0 . 7 0 0 0 0 . 2 9 0 2 , 4 3 0 0 . 6 1 0 0 0 . 5 8 1 2 , 8 0 0 0 . 5 3 5 2 0 . 6 7 4 1 , 0 1 0 0 . 0 8 6 9 0 . ) s t c e f f e e l i t n u q i r o f t s e t - F ( l e u a v - p s n o i t a v r e s b O d e r a u q s - R e d u l c n i s l o r t n o C . n o i t a m i t s E S L O . s e s e h t n e r a p n i ) t n e m g e s r o r o t c e s s u s n e c ( l e v e l U S P e h t t a d e r e t s u l c E S . l e v e l % 0 1 e h t t a t n a c i f i n g i s * ; l e v e l % 5 e h t t a t n a c i f i n g i s * * ; l e v e l % 1 e h t t a t n a c i f i n g i s * * * : s e t o N a t i p a c r e p l y h t n o m g n i s u s e l i t n u q i l n o i t a u p o p e r a s e l i t n u Q i . ) 9 5 - 7 5 , 6 5 - 4 5 , 3 5 - 1 5 , 0 5 - 8 4 , 7 4 - 5 4 , 4 4 - 2 4 , 1 4 - 9 3 , 8 3 - 6 3 , 5 3 - 3 3 , 2 3 - 0 3 , 9 2 - 7 2 , 6 2 - 4 2 , 3 2 - 1 2 , 0 2 - 8 1 , 7 1 - 5 1 , 4 1 - 2 1 , 1 1 - 9 , 8 - 6 . 0 = 5 Q = 4 Q = 3 Q = 2 Q f o t s e t - F e h t f o l e u a v - p e h t s i ) s t c e f f e e l i t n u q i r o f t s e t - F ( l e u a v - p , 5 - 3 , 2 - 0 ( s h t n o m n i s e i r o g e t a c e g a r o f i s e m m u d . l e b a i r a v i g n k n a r e h t s a n o i t p m u s n o c l d o h e s u o h Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 12 of 25 Table 4 Estimated Q1-Q5 gap using alternative SES measures Regression Coefficient for Q5 Consumption Height for age z-score (HAZ) Weight for age z-score (WAZ) Weight for height z-score (WHZ) Stunting (HAZ<-2SD) Underweight (WAZ<-2SD) Overweight (WHZ>+2SD) Anemia (%) Vitamin A deficiency (%) Iron deficiency (%) Gross motor (z-score) Communication (z-score) 0.962* 0.697* 0.214* -0.236* -0.013 0.046* -0.286*** -0.085* -0.084 0.210* 0.113 Wealth Index 0.934* 0.665* 0.210* -0.220* -0.010* 0.034* -0.280* -0.167* 0.030 0.176 0.174 Notes: significant at the 1% level; * significant at the 5% level; significant at the 10% level. OLS Estimation. Controls include child’s sex and a set of child dummies for age categories in months (0-2, 3-5, 6-8, 9-11, 12-14, 15-17, 18-20, 21-23, 24-26, 27-29, 30-32, 33-35, 36-38, 39-41, 42-44, 45-47, 48-50, 51-53, 54-56, 57-59). and age, there was a strong and significant association between SES and almost every anthropometric indicator. Compared to children in the poorest quintile Q1, children were 0.30 SD and 0.65 SD taller for their age in Q2 and Q3, respectively. This gap increases to 0.96 SD for children in the richest quintile. The re- sults followed a similar SES gradient for weight for age and weight for height z-scores, where higher quin- tiles were associated with higher z-scores. The results for prevalence of stunting showed that 30% of children in Q1 were stunted, after adjusting for age and sex. This number reduced by one third (11.2 percentage points) for children in the next consumption quintile and decreased even more markedly across higher quintiles. We found a similar pattern of inequality for the prevalence of underweight. Children in Q5 had a 21% lower probability of being underweight than children in Q1. In the case of preva- lence of overweight, children in the richest quintile were 4.6 percentage points more likely to be overweight than children in the poorest quintile, a relative difference of 37%. Our results also showed a clear SES gradient in preva- lence of anemia. When compared to children in the poorest quintile, anemia was 10.3 and 16.4 percentage points lower in Q2 and Q3, respectively. Among chil- dren in the richest quintile, the prevalence dropped by 28.6 percentage points (compared to Q1). For vitamin A deficiency, iron deficiency and gross motor development the association with socioeconomic status was less strong. Although the percentage of children with deficits in vitamin A or iron did not differ for children in Q2- Q4, relative to children in Q1, it reduced significantly for children in Q5 (8.5 and 8.4 percentage points for vitamin A and iron, respectively). Similarly, gross motor z-score among children in the richest quintile was 0.21 SD higher compared to children in the poorest quintile. in all regions Table 2 also presents adjusted SES gaps for eco- logical regions. While for some indicators SES gradi- ents were strongly significant (e.g. height for age, weight for age, stunting, underweight and anemia), other indicators presented mixed results. Notably, there were significant SES gaps in the preva- lence of overweight children in the highlands (Q4 vs Q1) and lowlands (Q5 vs Q1), but not in the valleys. Socioeconomic related inequalities in vitamin A defi- ciency were more evident in the highlands than in the valleys or lowlands. for age, weight Table 3 presents estimated coefficients disaggregat- ing the sample by age subgroups. The results for for age, and prevalence of height stunting showed that the SES gradient reached its peak in children between 24 and 36 months . Preva- lence of stunting was 9.4 percentage points lower for children in Q5 with respect to Q1 in the first 3-11 months of age, whereas this difference became 29.5 percentage points for children aged 24 to 36 months. Because stunting is an indicator of chronic malnutri- tion and is affected by multidimensional factors, this pattern could be the result of children in poor fam- ilies being exposed to multiple and cumulative haz- ards somewhat they grow. This pattern was different for anemia where SES gaps were largest in children less than one and at four years of age. For vitamin A and iron deficiency the sample was too small to efficiently estimate SES gaps within age groups. as Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 13 of 25 Regarding gross motor skills, we found that differences between children in the poorest and richest quintiles be- came more evident among older children: while gross motor z-score did not vary by quintile at very young ages, the Q1-Q5 gap was large and statistically signifi- cant among children aged 24-36 months (0.44 SD). Socioeconomic gaps in communication development have been largely analyzed in the literature [4, 5, 20]. While our estimated coefficient on communication z- score for Q5 in Table 2 was positive for the full sample, it was not statistically significant at conventional levels. In Table 3 we observe a heterogenous relationship, with large and significant SES gaps in communication for Q3 and Q5 (relative to Q1) for children in the 3-11- and 24- 36-month categories, and a reverse association for inter- mediate quintiles in the 12-23-month group. However, the relationship between communication z-score and al- ternative measures of SES depicted in Table 4 showed large and statistically significant differences between Q1 and Q5. The lack of consistency in the communication dimension might be related to the choice of instrument used for this analysis (ASQ-2), which in previous studies presented low internal validity compared to the “gold standard” Bayley-III [21, 22]17. Finally, we analyzed the association between SES and development outcomes controlling for nutritional status to partial out nutritional status in the association of SES and child development. Tables 5 and 6 present the esti- mated SES gradients in gross motor and communicative development, respectively. Model (1) estimates SES gaps without nutritional controls, whereas models (2) to (7) include various nutritional outcomes as covariates. In general, results show an expected association between poor nutritional status and lower gross motor and com- munication development. This association is particularly significant with indicators of height-for-age and weight- for-age. However, after including nutritional status as controls, the association between socioeconomic status and child gross motor development remained significant and of similar magnitude to the unadjusted models. The estimated SES gaps for communication development also remained similar before and after accounting for nutritional status in the model. Robustness checks As a check of robustness, we re-estimated all our results using a wealth index as the measure of socioeconomic status. All the analyses produced similar results, which suggests that our findings were consistent across differ- ent measures of socioeconomic status. A summary of these results is presented in Table 4, and further detailed 17Note, however, that Rubio-Codina et al (2016) internal validity as- sessment is based on the ASQ-3. in the Web Appendix (Tables 9). We also ran all the analyses for the subsample of children with DBS data. The results were similar, although standard errors were larger due to reduced sample size. This suggests that the comparison of results across indicators was not biased due to sample composition. Additionally, we reweighted the sample for differences in missing data in our out- comes of interest since children for which we have in- formation may be systematically different from those whose parents agree to health measures (Web Appendix Tables 7 and 8). Finally, we ran all our analysis using In- verse Probability Weighting to account for sample selec- tion of missing data [29]. Results, shown in Web Appendix Table 10 and 11, remained similar and consist- ent to main results. is one of that poverty Discussion Developmental delays in early childhood have life- long consequences for an individual’s future health, school performance, productivity, earnings and well- being [8–10, 30, 31] and for a society at large [7]. Re- search shows the most risk factors associated with a child’s detrimental health and development, with poorer children failing to reach their developmental potential [2, 20]. Some of the mediating factors contributing to this negative relationship are illness, nutritional deficiencies, low paren- tal education and poor home environments [2, 21, 22, 32]. SES gradients in child health and development are present and extensively documented across countries. However, analyzing disparities within countries is particularly im- portant as it provides information to identify, prioritize, implement and evaluate more efficient policies and inter- ventions child reducing development. disparities aimed in at Our study explored SES disparities in child nutri- tion and development outcomes in the context of Bolivia using a broadly representative sample of the population and analyzing a more comprehensive set of outcomes than considered in the literature to date. Measures of child development included gross motor and communicative development, whereas nutritional status was assessed using anthropometric measures and indicators of prevalence of anemia, vitamin A and iron deficiency. In our main analysis we used household consumption to measure household socio- economic status, though results were robust to alter- native SES measures. We found large disparities in child nutrition indicators by socioeconomic status. A nonparametric analysis within age groups indicates that the gap between chil- dren from rich and poor families started very early in life (stunting, anemia). In measures of height-for-age and stunting, the SES gap tended to increase rapidly from Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 14 of 25 Table 5 Adjusted SES gradients in Gross Motor z-score controlling for nutritional status VARIABLES Model (1) Model (2) Model (3) Model (4) Model (5) Model (6) Model (7) Quintile 2 = 1 0.028 (0.072) -0.010 (0.071) 0.013 (0.073) 0.040 (0.074) -0.002 (0.074) 0.041 (0.075) 0.002 (0.078) Quintile 3 = 1 0.016 (0.061) -0.073 (0.059) -0.029 (0.061) 0.033 (0.062) -0.042 (0.062) 0.036 (0.063) -0.032 (0.064) Quintile 4 = 1 0.041 (0.069) -0.074 (0.069) -0.016 (0.069) 0.053 (0.070) -0.023 (0.070) 0.055 (0.071) -0.009 (0.072) Quintile 5 = 1 0.210** (0.070) 0.084 (0.070) 0.136* (0.071) 0.216*** (0.072) 0.142* (0.074) 0.221* (0.072) 0.188 (0.080) female -0.076* (0.040) -0.105* (0.039) -0.100 (0.040) -0.082 (0.040) -0.099 (0.041) -0.082** (0.041) -0.070* (0.042) 0.129* (0.017) 0.127* (0.020) 0.030 (0.021) Height for age z score (hfa) Weight for age z-score (wfa) Weight for height z-score (wfh) Stunted (hfa <2SD) Underweight (wfa < 2SD) Anemia (any level) -0.261* (0.058) -0.137 (0.201) -0.113 (0.050) Constant -0.114 (0.076) 0.003 (0.081) -0.082 (0.080) -0.140* (0.078) -0.038 (0.081) -0.124 (0.080) -0.043 (0.091) Observations R-squared 5,753 0.015 5,641 0.036 5,464 0.030 5,460 0.016 5,641 0.023 5,460 0.015 5,098 0.018 Notes: *significant at the 1% level; significant at the 5% level; significant at the 10% level. SE clustered at the PSU level (census sector or segment) in parentheses. OLS Estimation. Additional controls include child’s sex and a set of child dummies for age categories in months (0-2, 3-5, 6-8, 9-11, 12-14, 15-17, 18- 20, 21-23, 24-26, 27-29, 30-32, 33-35, 36-38, 39-41, 42-44, 45-47, 48-50, 51-53, 54-56, 57-59). Quintiles are population quintiles using monthly per capita household consumption as the ranking variable. Table 6 Adjusted SES gradients in Communication z-score controlling for nutritional status VARIABLES Quintile 2 = 1 Quintile 3 = 1 Quintile 4 = 1 Quintile 5 = 1 female Height for age z score (hfa) Weight for age z-score (wfa) Weight for height z-score (wfh) Stunted (hfa <2SD) Underweight (wfa < 2SD) Anemia (any level) Constant Observations R-squared Model (1) Model (2) Model (3) Model (4) Model (5) Model (6) Model (7) -0.133 (0.071) -0.155** (0.072) -0.134* (0.073) -0.123* (0.073) -0.155** (0.074) -0.121* (0.073) -0.167 (0.078) 0.037 (0.059) -0.013 (0.061) 0.039 (0.062) 0.061 (0.061) -0.002 (0.061) 0.061 (0.061) 0.010 (0.062) -0.084 (0.067) -0.156 (0.068) -0.093 (0.069) -0.065 (0.068) -0.135** (0.068) -0.065 (0.069) -0.114* (0.068) 0.113 (0.073) 0.033 (0.073) 0.097 (0.075) 0.131* (0.075) 0.057 (0.076) 0.130* (0.075) 0.093 (0.080) 0.091* (0.035) 0.078 (0.036) 0.087 (0.036) 0.094* (0.036) 0.079 (0.036) 0.095* (0.037) 0.095* (0.036) 0.073* (0.017) 0.044 (0.018) -0.017 (0.018) -0.183* (0.051) 0.174 (0.174) -0.049 (0.045) 0.098 (0.078) 0.164** (0.079) 0.093 (0.080) 0.084 (0.080) 0.151* (0.079) 0.071 (0.081) 0.151* (0.084) 5,753 0.016 5,641 0.024 5,464 0.020 5,460 0.018 5,641 0.021 5,460 0.019 5,098 0.020 Notes: *significant at the 1% level; significant at the 5% level; significant at the 10% level. SE clustered at the PSU level (census sector or segment) in parentheses. OLS Estimation. Additional controls include child’s sex and a set of child dummies for age categories in months (0-2, 3-5, 6-8, 9-11, 12-14, 15-17, 18- 20, 21-23, 24-26, 27-29, 30-32, 33-35, 36-38, 39-41, 42-44, 45-47, 48-50, 51-53, 54-56, 57-59). Quintiles are population quintiles using monthly per capita household consumption as the ranking variable. Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 15 of 25 the sixth month of age, reaching its peak at 24 months and stabilizing at older ages. Socioeconomic gaps were less noticeable for other nutrition outcomes including vitamin A and iron deficiencies. This results was to some extend unexpected as most anemia in countries with high prevalence is caused largely or in whole by iron deficiency [33], though our sample size and age range is more limited for these measures. The results for our early child development outcomes showed that the gap between children from families in the upper and lower consumption quintiles became apparent between 24 and 36 months when it was 0.44 standard deviations in gross motor and 0.25 standard deviations in commu- nication z-scores. The parametric analysis suggested that SES inequality persisted after adjusting for demographic factors that affect child development. The analysis disaggregated by subnational regions shed light on the heterogenous rela- tionship between socioeconomic status and child devel- opment. While some indicators (height for age, weight for age, stunting, anemia) demonstrate common large and significant SES gaps in all regions, others show sig- nificant disparities in some regions and not in others (overweight, vitamin A deficiency). This study presents some limitations. First, in the absence of exogenous variation in SES within the study population, the relationships estimated between SES and children’s nutrition and development indica- tors are considered associations, and not causal ef- fects. Second, while the results presented in this study are largely consistent with existing studies from other countries, and the study sample is nationally repre- sentative for Bolivia, these findings cannot be directly extrapolated to other country contexts. Third, al- though this study includes a rich set of nutritional outcomes, a subset of those outcomes (namely Vita- min A and iron deficiency) are measured on a sub- sample of children in the 6-23 month age range, re- ducing the external validity and limiting statistical precision for these analysis relative to outcomes mea- sured for the entire sample. such as undernutrition, anemia However, the richness of our data allows us to look at child development outcomes, including gross motor and communication development, as well as nutrition risk factors, and micronutrient deficiencies, within the same represen- tative sample of children. In addition, the survey col- lected novel data to measure vitamin A and iron deficiencies. This is un- usual for health surveys in developing countries and this study helps to motivate other countries to follow similar strategies. from dried blood samples Our study complements the existing body of litera- subpopulations of focuses on particular ture that disadvantaged children [4] or analyzes SES gaps in child development across countries using a single SES indicator and measure of child development [5]. We show that SES gaps are consistent across different measurements of SES status of children using data that that allows estimating SES gap at the national level as well as in subpopulations of interest, such as urban and rural households. Furthermore, unlike other related studies we compare, within sample, the SES gradient across child development, anthropomet- ric measures, biomarkers, and micronutrient deficien- cies. With these data we show that the relationship between SES and child development remains strong and significant once we control for risk factors related to nutritional status, which shows that there is and important association between wealth or income and child cognition development independent of nutrition. Further research should assess other channels by which income affects child development, such as en- vironmental factors or parental behavior. The evidence provided in this study shows that al- though a large proportion of young children in Bolivia are affected by specific developmental risks (anemia affects around 50% of children and 39% have it’s the poorest children that vitamin A deficiency), face the greatest threats that compromise their devel- opment. These disparities are evident at birth and need to be addressed urgently to reduce developmen- tal inequities. From a policy perspective, the large so- cioeconomic gaps in nutrition outcomes documented here reinforce the need to strengthen efforts that tackle the multiple causes of malnutrition for the poorest. On this topic, the country has been imple- menting nation scale programs to incentivize the use of preventive health services for children and women during pregnancy (Program Bono Juana Azurduy), and programs that contribute to the prevention and care of malnutrition through multisectoral actions (Food and Nutrition Multisectoral Program in the Life Cycle). Although these programs have some level of prioritization, their national scope and limited target- ing reduce their effectiveness to close socioeconomic gaps. Interventions in other areas related to child de- velopment are scarce in Bolivia, as there is still a need for national and subnational governments to prioritize early childhood development in their pro- grams of work. An important step forward was taken in 2014 with the implementation of a pilot ECD pro- gram that sought to improve early child development by strengthening child-stimulation practices at home (Program Grow Well to Live Well). The program’s impact evaluation reported positive effects on cogni- tive, communication and fin motor development of children in poor families in rural areas [34]. Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 16 of 25 Appendix 1 Web Appendix Table 7 Adjusted SES gradients in child development by subnational region: DBS Subsample Height for age z-score Weight for age z-score Weight for height z-score VARIABLES All Highlands Valleys Lowlands All Highlands Valleys Lowlands All Highlands Valleys Lowlands Quintile 2 = 1 0.162 (0.125) Quintile 3 = 1 0.481 (0.121) Quintile 4 = 1 0.742* (0.132) Quintile 5 = 1 0.847*** (0.131) -0.087 (0.168) 0.155 (0.179) 0.736 (0.469) 0.526* (0.267) Female Constant 0.275* (0.073) 0.314* (0.112) -0.899* (0.119) -1.327* (0.187) -0.792* (0.157) Observations 1,597 544 R-squared 0.154 0.118 530 0.150 Stunting (%) 0.090 (0.170) 0.407 (0.195) 0.360 (0.162) 0.794* (0.184) 0.177* (0.097) 0.562 (0.255) 0.612 (0.240) 0.817* (0.231) 0.911* (0.259) 0.448* (0.135) -0.577 (0.259) 523 0.237 0.218* (0.111) 0.409* (0.107) 0.478* (0.112) 0.686* (0.116) 0.160 (0.073) 0.080 (0.177) 0.354 (0.178) 0.491 (0.335) 0.567 (0.217) 0.207 (0.140) -0.397* (0.108) -0.624* (0.196) 1,553 0.073 534 0.074 Underweight (%) 0.226 (0.168) 0.462 (0.181) 0.260 (0.178) 0.644* (0.213) 0.101 (0.112) -0.191 (0.171) 502 0.077 0.336* (0.189) 0.250 (0.177) 0.410 (0.182) 0.603* (0.174) 0.246* (0.128) -0.333* (0.196) 517 0.086 0.169 (0.107) 0.203* (0.104) 0.126 (0.113) 0.329** (0.136) 0.023 (0.082) 0.212* (0.120) 1,553 0.012 0.146 (0.174) 0.340* (0.176) 0.154 (0.273) 0.367 (0.293) 0.044 (0.147) 0.263 (0.216) 534 0.034 Overweight (%) 0.234 (0.185) 0.333 (0.167) 0.081 (0.194) 0.302 (0.241) 0.026 (0.147) 0.401 (0.199) 502 0.024 0.074 (0.177) -0.082 (0.167) 0.011 (0.169) 0.216 (0.174) 0.017 (0.134) 0.030 (0.194) 517 0.035 VARIABLES All Highlands Valleys Lowlands All Highlands Valleys Lowlands All Highlands Valleys Lowlands Quintile 2 = 1 -0.089 (0.044) -0.089 (0.076) Quintile 3 = 1 -0.164* (0.036) -0.221* (0.063) Quintile 4 = 1 -0.202* (0.039) -0.187 (0.115) Quintile 5 = 1 -0.224* (0.035) -0.226* (0.075) Female Constant -0.089* (0.023) -0.079 (0.050) 0.261* (0.033) 0.322*** (0.073) Observations 1,597 544 R-squared 0.104 0.099 Anemia (%) -0.063 (0.072) -0.113* (0.066) -0.146* (0.055) -0.193* (0.048) -0.050 (0.038) 0.188* (0.041) 530 0.098 -0.140 (0.086) -0.156 (0.067) -0.223* (0.076) -0.227* (0.076) -0.139* (0.038) 0.288* (0.076) 523 0.136 -0.022 (0.030) -0.050* (0.026) -0.051 (0.025) -0.049 (0.024) -0.024 (0.018) 0.084** (0.042) 534 0.047 -0.018 (0.018) -0.030* (0.017) -0.026 (0.018) -0.024 (0.018) -0.008 (0.009) 0.053*** (0.019) 1,553 0.011 VAD (%) -0.036 (0.025) -0.021 (0.032) -0.006 (0.035) -0.022 (0.031) -0.003 (0.019) 0.023 (0.026) 502 0.020 0.008 (0.024) -0.012 (0.016) -0.014 (0.016) 0.005 (0.027) -0.002 (0.009) 0.053* (0.029) 517 0.039 -0.012 (0.016) 0.005 (0.017) 0.001 (0.018) 0.034 (0.024) 0.001 (0.014) 0.062** (0.024) 1,553 0.012 0.002 (0.026) -0.003 (0.013) 0.058 (0.053) 0.070 (0.057) 0.012 (0.022) 0.080* (0.046) 534 0.064 Iron deficiency (%) -0.036 (0.027) 0.010 (0.036) -0.020 (0.032) 0.025 (0.041) 0.006 (0.029) 0.081* (0.046) 502 0.012 -0.013 (0.026) -0.010 (0.029) -0.021 (0.026) -0.001 (0.030) -0.015 (0.020) 0.042 (0.028) 517 0.026 VARIABLES All Highlands Valleys Lowlands All Highlands Valleys Lowlands All Highlands Valleys Lowlands Quintile 2 = 1 Quintile 3 = 1 -0.041 (0.045) -0.025 (0.042) Quintile 4 = 1 -0.096** (0.045) -0.122* (0.067) -0.001 (0.056) 0.041 (0.064) Quintile 5 = 1 -0.182*** (0.049) -0.114 (0.088) Female Constant -0.048* (0.028) -0.063 (0.051) 0.741* (0.045) 0.816* (0.077) Observations 1,593 545 R-squared 0.036 0.062 0.088 (0.069) -0.048 (0.077) -0.091 (0.067) -0.182*** (0.069) -0.085* (0.051) 0.767*** (0.071) 528 0.089 0.056 (0.098) 0.109 (0.095) 0.010 (0.105) -0.054 (0.109) -0.015 (0.042) 0.067 (0.055) 0.073 (0.049) -0.023 (0.064) -0.085* (0.044) -0.012 (0.029) 0.539* (0.106) 0.466* (0.045) 520 0.021 1,609 0.034 -0.122 (0.085) -0.141 (0.063) -0.184 (0.091) -0.144* (0.081) -0.017 (0.050) 0.601* (0.088) 548 0.086 0.216 (0.090) 0.068 (0.077) -0.106* (0.061) -0.070 (0.061) -0.017 (0.039) 0.134 (0.113) 0.177* (0.100) 0.050 (0.126) -0.077 (0.108) -0.009 (0.056) 0.353* (0.064) 0.547* (0.102) 535 0.060 526 0.072 0.029 (0.048) 0.020 (0.043) 0.019 (0.047) -0.084 (0.038) -0.109* (0.029) 0.229* (0.040) 1,609 0.046 0.095 (0.077) 0.103 (0.094) 0.322* (0.109) -0.017 (0.077) -0.150* (0.050) 0.358* (0.075) 548 0.091 0.032 (0.074) 0.097 (0.063) 0.090 (0.060) -0.052 (0.056) -0.122* (0.044) 0.148 (0.057) 535 0.090 -0.074 (0.086) -0.081 (0.078) -0.105 (0.090) -0.099 (0.090) -0.069 (0.047) 0.216** (0.095) 526 0.046 Gross motor z-score Communication z-score VARIABLES All Highlands Valleys Lowlands All Highlands Valleys Lowlands Quintile 2 = 1 -0.168 (0.110) -0.204 (0.168) Quintile 3 = -0.084 -0.133 -0.057 (0.168) -0.162 -0.417* (0.240) -0.302 (0.123) -0.499* (0.181) -0.078 (0.186) -0.409 (0.250) -0.189 (0.197) -0.086 -0.350** 0.017 -0.149 (0.197) Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 17 of 25 Table 7 Adjusted SES gradients in child development by subnational region: DBS Subsample (Continued) 1 (0.106) (0.221) (0.165) (0.187) (0.111) (0.175) Quintile 4 = 1 Quintile 5 = 1 Female Constant -0.072 (0.148) 0.173* (0.100) -0.107 (0.071) -0.020 (0.101) -0.362 (0.293) 0.154 (0.208) 0.007 (0.137) -0.133 (0.165) Observations 1,473 515 R-squared 0.021 0.031 -0.246 (0.250) 0.193 (0.162) -0.119 (0.109) -0.036 (0.144) 489 0.044 -0.088 (0.271) -0.097 (0.215) -0.183 (0.111) 0.280 (0.251) 469 0.051 -0.152 (0.127) -0.016 (0.124) 0.105 (0.085) 0.147 (0.109) 1,473 0.022 -0.083 (0.252) -0.098 (0.195) 0.174 (0.130) -0.019 (0.190) 515 0.056 -0.117 (0.190) 0.203 (0.223) 0.079 (0.141) 0.000 (0.160) 489 0.020 -0.328 (0.222) -0.290 (0.204) 0.097 (0.141) 0.558 (0.214) 469 0.062 Note: significant at the 1% level; * significant at the 5% level; significant at the 10% level. SE clustered at the PSU level (census sector or segment) in parentheses. OLS Estimation. Controls include child’s sex and a set of child dummies for age categories in months (0-2, 3-5, 6-8, 9-11, 12-14, 15-17, 18-20, 21-23, 24-26, 27-29, 30-32, 33-35, 36-38, 39-41, 42-44, 45-47, 48-50, 51-53, 54-56, 57-59). Quintiles are population quintiles using monthly per capita household consump- tion as the ranking variable. p-value (F-test for quintile effects) is the p-value of the F-test of Q2=Q3=Q4=Q5=0. Table 8 Adjusted SES gradients in child development by child’s age: DBS Subsample Stunting (%) Height for age z-score Weight for age z- Weight for height z- score Underweight (%) Overweight (%) VARIABLES 6-11 Months Quintile 2 = 1 0.126 (0.217) Quintile 3 = 1 Quintile 4 = 1 Quintile 5 = 1 Female = 1 0.581 (0.214) 0.632* (0.210) 0.615* (0.204) 0.173 (0.130) 12-23 Months 0.199 (0.153) 0.458* (0.131) 0.817* (0.163) 0.991* (0.153) 0.331* (0.101) Constant -0.777* (0.144) -1.614* (0.138) Observations 598 R-squared 0.055 999 0.118 score 6-11 Months 0.068 (0.157) 0.448* (0.146) 0.355 (0.179) 0.364 (0.159) 0.001 (0.115) -0.185 (0.119) 589 0.031 12-23 Months 0.343 (0.152) 0.434* (0.136) 0.578* (0.151) 0.896* (0.151) 0.253 (0.101) 6-11 Months 0.003 (0.158) 0.181 (0.142) 0.027 (0.187) 0.033 (0.200) -0.138 (0.123) -0.815* (0.145) 0.428* (0.131) 964 0.078 589 0.008 12-23 Months 0.305 (0.147) 0.257* (0.138) 0.214 (0.152) 0.526* (0.174) 0.118 (0.104) -0.027 (0.152) 964 0.026 6-11 Months -0.045 (0.051) -0.127* (0.031) -0.109* (0.038) -0.125* (0.035) -0.036 (0.026) 0.176*** (0.030) 598 0.048 12-23 Months -0.125* (0.065) -0.197* (0.051) -0.263* (0.054) -0.290* (0.051) -0.121* (0.034) 0.412*** (0.048) 999 0.081 6-11 Months 12-23 Months 0.012 (0.026) -0.016 (0.010) -0.018* (0.009) 0.001 (0.022) -0.006 (0.014) 0.036*** (0.013) 589 0.024 -0.038 (0.026) -0.041 (0.027) -0.033 (0.028) -0.041 (0.027) -0.011 (0.012) 0.056* (0.031) 964 0.011 6-11 Months -0.056 (0.027) -0.052 (0.026) 0.012 (0.044) -0.008 (0.043) -0.003 (0.025) 0.090*** (0.032) 589 0.021 12-23 Months 0.018 (0.019) 0.039* (0.020) 0.000 (0.012) 0.063** (0.028) 0.005 (0.017) 0.029 (0.019) 964 0.021 Anemia VAD (%) Iron deficiency (%) Gross motor z-score VARIABLES Quintile 2 = 1 Quintile 3 = 1 6-11 Months -0.053 (0.070) -0.064 (0.070) Quintile 4 = 1 -0.157* (0.082) Quintile 5 = 1 -0.251* (0.071) Female = 1 Constant -0.036 (0.062) 0.769* (0.052) Observations 598 R-squared 0.043 12-23 Months 6-11 Months 12-23 Months 6-11 Months 12-23 Months -0.035 (0.059) -0.001 (0.058) -0.059 (0.060) -0.141 (0.068) -0.055 (0.042) 0.831* (0.052) 995 0.035 0.067 (0.099) 0.057 (0.078) -0.054 (0.089) -0.118 (0.078) -0.021 (0.049) 0.068 (0.060) 0.085 (0.059) -0.004 (0.072) -0.064 (0.054) -0.006 (0.039) 0.487* (0.059) 0.283* (0.054) 600 0.021 1,009 0.019 0.048 (0.080) 0.026 (0.068) 0.046 (0.070) -0.093* (0.055) -0.109 (0.047) 0.220* (0.045) 600 0.057 0.015 (0.057) 0.015 (0.054) 0.003 (0.067) -0.080 (0.052) -0.110* (0.035) 0.347* (0.057) 1,009 0.041 6-11 Months -0.474* (0.165) -0.261 (0.177) -0.223 (0.177) -0.087 (0.150) -0.157 (0.108) 0.189 (0.121) 556 0.038 12-23 Months 0.044 (0.144) 0.042 (0.126) 0.041 (0.183) 0.344* (0.130) -0.062 (0.092) -0.210 (0.131) 917 0.022 Communication z- score 6-11 Months -0.405 (0.205) 0.073 (0.173) -0.074 (0.204) 0.200 (0.200) 0.009 (0.130) 0.137 (0.128) 556 0.044 12-23 Months -0.220 (0.139) -0.164 (0.131) -0.191 (0.158) -0.133 (0.146) 0.164 (0.101) 0.010 (0.143) 917 0.016 Note: significant at the 1% level; * significant at the 5% level; significant at the 10% level. SE clustered at the PSU level (census sector or segment) in parentheses. OLS Estimation. Controls include child’s sex and a set of child dummies for age categories in months (0-2, 3-5, 6-8, 9-11, 12-14, 15-17, 18-20, 21-23, 24-26, 27-29, 30-32, 33-35, 36-38, 39-41, 42-44, 45-47, 48-50, 51-53, 54-56, 57-59). Quintiles are population quintiles using monthly per capita household consump- tion as the ranking variable. p-value (F-test for quintile effects) is the p-value of the F-test of Q2=Q3=Q4=Q5=0. Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 18 of 25 ) 3 7 0 0 ( . 8 0 0 0 - . ) 8 7 0 0 ( . 6 1 0 0 - . ) 8 4 0 0 ( . * 8 9 0 0 . ) 3 5 0 0 ( . * 2 0 1 0 - . ) 2 2 0 0 ( . * * * 5 7 0 0 - . ) 9 0 0 0 ( . 7 0 0 0 . ) 5 0 0 0 ( . 6 0 0 0 - . ) 8 1 0 0 ( . * * * 5 0 1 0 - . ) 5 4 0 0 ( . * * 4 0 1 0 . ) 9 4 0 0 ( . * * * 7 1 2 0 . ) 6 5 0 0 ( . * * * 4 4 2 0 . 1 = 2 e l i t n u Q i ) 2 7 0 0 ( . 1 8 0 0 . ) 6 6 0 0 ( . * * 0 5 1 0 . ) 1 5 0 0 ( . 9 3 0 0 . ) 5 6 0 0 ( . * 7 1 1 0 - . ) 4 2 0 0 ( . * * * 7 5 1 0 - . ) 1 1 0 0 ( . * * 5 2 0 0 . ) 4 0 0 0 ( . * * * 3 1 0 0 - . ) 9 1 0 0 ( . * * * 2 7 1 0 - . ) 7 5 0 0 ( . * * 1 3 1 0 . ) 1 5 0 0 ( . * * * 2 9 3 0 . ) 0 6 0 0 ( . * * * 9 1 5 0 . 1 = 3 e l i t n u Q i ) 4 7 0 0 ( . 7 0 0 0 . ) 4 7 0 0 ( . 2 0 0 0 . ) 4 4 0 0 ( . 4 2 0 0 . ) 3 5 0 0 ( . 6 5 0 0 - . ) 5 2 0 0 ( . * * * 6 1 2 0 - . ) 4 1 0 0 ( . * * * 9 4 0 0 . ) 6 0 0 0 ( . 4 0 0 0 - . ) 8 1 0 0 ( . * * * 7 8 1 0 - . ) 4 5 0 0 ( . * * * 9 8 1 0 . ) 4 5 0 0 ( . * * * 0 3 5 0 . ) 8 6 0 0 ( . * * * 3 1 7 0 . 1 = 4 e l i t n u Q i ) 0 8 0 0 ( . * * 4 7 1 0 . ) 9 7 0 0 ( . * * 6 7 1 0 . ) 3 4 0 0 ( . 0 3 0 0 . ) 3 5 0 0 ( . * * * 7 6 1 0 - . ) 6 2 0 0 ( . * * * 0 8 2 0 - . ) 2 1 0 0 ( . * * * 4 3 0 0 . ) 5 0 0 0 ( . * 0 1 0 0 - . ) 7 1 0 0 ( . * * * 0 2 2 0 - . ) 7 5 0 0 ( . * * * 0 1 2 0 . ) 6 5 0 0 ( . * * * 5 6 6 0 . ) 7 6 0 0 ( . * * * 4 3 9 0 . 1 = 5 e l i t n u Q i ) 5 3 0 0 ( . * * 7 8 0 0 . ) 9 3 0 0 ( . * * 3 8 0 0 - . ) 9 2 0 0 ( . * * * 5 0 1 0 - . ) 8 2 0 0 ( . 1 1 0 0 - . ) 5 1 0 0 ( . * * * 2 4 0 0 - . ) 8 0 0 0 ( . 2 1 0 0 - . ) 3 0 0 0 ( . * * * 2 1 0 0 - . ) 2 1 0 0 ( . * * * 5 3 0 0 - . ) 9 2 0 0 ( . 5 3 0 0 . ) 9 2 0 0 ( . * * * 3 8 0 0 . ) 4 3 0 0 ( . * * * 0 1 1 0 . l e a m e F ) 4 9 0 0 ( . 8 4 0 0 . ) 3 7 0 0 ( . 5 0 1 0 - . ) 1 4 0 0 ( . * * * 4 8 1 0 . ) 5 4 0 0 ( . * * * 3 5 5 0 . ) 3 3 0 0 ( . * * * 2 1 6 0 . ) 0 3 0 0 ( . * * * 4 1 1 0 . ) 6 1 0 0 ( . * * * 3 7 0 0 . ) 6 2 0 0 ( . * * * 6 7 2 0 . ) 3 2 1 0 ( . 7 1 1 0 . ) 8 9 0 0 ( . * * * 5 0 8 0 - . ) 1 0 1 0 ( . * * * 2 0 2 1 - . t n a t s n o C 1 6 7 5 , 3 1 0 0 . 1 6 7 5 , 6 1 0 0 . 0 1 6 1 , 0 4 0 0 . 0 1 6 1 , 1 3 0 0 . 7 2 4 9 , 0 0 1 0 . 6 6 4 0 1 , 1 1 0 0 . 6 6 4 0 1 , 3 1 0 0 . 4 8 8 0 1 , 9 5 0 0 . 6 6 4 0 1 , 0 2 0 0 . 1 8 4 0 1 , 3 6 0 0 . 4 8 8 0 1 , s n o i t a v r e s b O 6 2 1 0 . d e r a u q s - R . n u m m o C e r o c s - z r o t o m s s o r G i y c n e c i f e d n o r I ) % ( D A V i a m e n A ) % ( i t h g e w r e v O ) % ( i t h g e w r e d n U ) % ( g n i t n u t S t h g e h i r o f t h g e W i e g a r o f t h g e W i e r o c s - z ) % ( e r o c s - z e r o c s - z e r o c s - z e g a r o f t h g e H i S E L B A R A V I S E S f o e r u s a e m s a x e d n i h t l a e w g n i s u t n e m p o e v e d l d l i h c n i i s t n e d a r g S E S d e t s u d A j 9 e l b a T Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 19 of 25 ) 6 9 0 0 ( . 3 2 0 0 - . ) 0 8 0 0 ( . * * 4 8 1 0 . ) 6 6 0 0 ( . 3 8 0 0 . ) 5 4 0 0 ( . * 0 8 0 0 . ) 5 7 0 0 ( . * * 7 6 1 0 . ) 9 6 0 0 ( . * * * 7 6 2 0 . ) 1 8 0 0 ( . * * 3 9 1 0 . ) 6 4 0 0 ( . * * * 7 3 2 0 . ) 5 9 0 0 ( . * * * 2 0 3 0 . ) 5 7 0 0 ( . * * * 9 0 2 0 . ) 6 8 0 0 ( . * * * 0 4 2 0 . ) 4 5 0 0 ( . * * * 8 9 2 0 . 1 = 2 e l i t n u Q i ) 4 0 1 0 ( . 0 4 0 0 . ) 6 7 0 0 ( . * * * 8 9 1 0 . ) 2 8 0 0 ( . 0 8 0 0 . ) 0 5 0 0 ( . * * 3 1 1 0 . ) 7 7 0 0 ( . * * * 0 9 3 0 . ) 9 7 0 0 ( . * * * 6 8 4 0 . ) 4 8 0 0 ( . * * * 3 1 3 0 . ) 9 4 0 0 ( . * * * 9 4 4 0 . ) 2 9 0 0 ( . * * * 1 1 6 0 . ) 1 7 0 0 ( . * * * 8 0 6 0 . ) 1 8 0 0 ( . * * * 7 1 4 0 . ) 2 5 0 0 ( . * * * 6 2 6 0 . 1 = 3 e l i t n u Q i ) 8 0 1 0 ( . 5 7 1 0 . ) 3 7 0 0 ( . 2 6 0 0 . ) 5 0 1 0 ( . * * 8 4 2 0 . ) 9 4 0 0 ( . * * * 0 6 1 0 . ) 6 8 0 0 ( . * * * 2 6 5 0 . ) 2 8 0 0 ( . * * * 6 0 4 0 . ) 1 9 0 0 ( . * * * 6 3 4 0 . ) 2 5 0 0 ( . * * * 9 3 5 0 . ) 8 8 0 0 ( . * * * 7 6 7 0 . ) 5 8 0 0 ( . * * * 9 5 6 0 . ) 7 1 1 0 ( . * * * 6 1 5 0 . ) 8 5 0 0 ( . * * * 6 6 7 0 . 1 = 4 e l i t n u Q i ) 8 0 1 0 ( . * * 2 6 2 0 . ) 9 8 0 0 ( . 5 4 1 0 . ) 8 4 1 0 ( . 1 9 1 0 . ) 7 5 0 0 ( . * * * 2 1 2 0 . ) 9 9 0 0 ( . * * * 9 3 7 0 . ) 8 8 0 0 ( . * * * 4 4 5 0 . ) 8 3 1 0 ( . * * * 4 6 5 0 . ) 9 5 0 0 ( . * * * 1 9 6 0 . ) 1 1 1 0 ( . * * * 6 8 9 0 . ) 4 8 0 0 ( . * * * 2 7 7 0 . ) 0 0 1 0 ( . * * * 5 1 7 0 . ) 3 6 0 0 ( . * * * 1 4 9 0 . 1 = 5 e l i t n u Q i ) 4 5 0 0 ( . 0 4 0 0 - . ) 6 4 0 0 ( . 2 1 0 0 . ) 3 5 0 0 ( . * * * 7 7 1 0 . ) 0 3 0 0 ( . 8 3 0 0 . ) 6 5 0 0 ( . 5 0 0 0 - . ) 9 4 0 0 ( . 3 6 0 0 . ) 7 4 0 0 ( . * * * 3 2 2 0 . ) 0 3 0 0 ( . * * * 5 8 0 0 . ) 0 6 0 0 ( . 0 4 0 0 . ) 0 6 0 0 ( . * * 0 3 1 0 . ) 5 5 0 0 ( . * * * 3 8 1 0 . ) 5 3 0 0 ( . * * * 9 1 1 0 . l e a m e F ) 6 2 1 0 ( . 1 0 2 0 . ) 5 7 0 0 ( . * * * 7 4 2 0 . ) 7 7 0 0 ( . * * 9 5 1 0 . ) 1 5 0 0 ( . * * * 0 0 2 0 . ) 4 9 0 0 ( . * * * 4 0 3 0 - . ) 4 7 0 0 ( . * * * 5 6 5 0 - . ) 1 8 0 0 ( . * * * 1 8 7 0 - . ) 8 4 0 0 ( . * * * 3 7 5 0 - . ) 7 9 0 0 ( . * * * 0 0 8 0 - . ) 5 7 0 0 ( . * * * 7 2 3 1 - . ) 2 9 0 0 ( . * * * 9 9 5 1 - . ) 3 5 0 0 ( . * * * 8 7 2 1 - . t n a t s n o C s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A S E L B A R A V I e r o c s - z t h g e h i r o f t h g e W i e r o c s - z e g a r o f t h g e W i e r o c s - z e g a r o f t h g e H i a t a d g n i s s i m r o f s t n e m j t s u d a W P I h t i w n o g e r i y b t n e m p o e v e d l d l i h c n i i s t n e d a r g S E S d e t s u d A j 0 1 e l b a T 5 6 6 3 , 9 2 0 0 . 1 9 3 3 , 3 1 0 0 . s d n a w o L l s y e l l a V 7 9 3 3 , 0 2 0 0 . 3 5 4 0 1 , 5 1 0 0 . ) % ( i t h g e w r e v O 0 7 6 3 , 7 5 0 0 . 5 9 3 3 , 7 3 0 0 . 4 0 4 3 , 4 5 0 0 . 9 6 4 0 1 , 6 5 0 0 . ) % ( i t h g e w r e d n U 3 2 7 3 , 3 1 1 0 . 1 0 7 3 , 0 8 0 0 . 6 4 4 3 , 9 6 0 0 . 0 7 8 0 1 , s n o i t a v r e s b O 2 0 1 0 . d e r a u q s - R ) % ( g n i t n u t S l l A S E L B A R A V I l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i ) 9 1 0 0 ( . 8 0 0 0 - . ) 6 1 0 0 ( . 9 0 0 0 . ) 4 1 0 0 ( . 5 0 0 0 . ) 9 0 0 0 ( . 4 0 0 0 . ) 3 1 0 0 ( . 5 0 0 0 - . ) 9 0 0 0 ( . * 7 1 0 0 - . ) 0 1 0 0 ( . 0 1 0 0 - . ) 6 0 0 0 ( . 0 1 0 0 - . ) 3 3 0 0 ( . * * 0 7 0 0 - . ) 8 2 0 0 ( . * * * 4 7 0 0 - . ) 0 3 0 0 ( . * * * 1 4 1 0 - . ) 8 1 0 0 ( . * * * 2 1 1 0 - . 1 = 2 e l i t n u Q i ) 4 2 0 0 ( . 9 0 0 0 - . ) 7 1 0 0 ( . 6 2 0 0 . ) 7 1 0 0 ( . 3 1 0 0 . ) 0 1 0 0 ( . 3 1 0 0 . ) 0 1 0 0 ( . 7 0 0 0 - . ) 0 1 0 0 ( . 1 0 0 0 - . ) 0 1 0 0 ( . 6 0 0 0 - . ) 6 0 0 0 ( . 4 0 0 0 - . ) 6 2 0 0 ( . * * * 8 1 1 0 - . ) 4 2 0 0 ( . * * * 5 7 1 0 - . ) 1 3 0 0 ( . * * * 0 9 1 0 - . ) 6 1 0 0 ( . * * * 1 8 1 0 - . 1 = 3 e l i t n u Q i ) 5 2 0 0 ( . 2 1 0 0 . ) 6 1 0 0 ( . 1 0 0 0 - . ) 5 2 0 0 ( . * 8 4 0 0 . ) 1 1 0 0 ( . * 0 2 0 0 . ) 8 0 0 0 ( . * * * 6 2 0 0 - . ) 0 1 0 0 ( . 8 0 0 0 - . ) 7 0 0 0 ( . * * * 2 2 0 0 - . ) 5 0 0 0 ( . * * * 8 1 0 0 - . ) 0 3 0 0 ( . * * * 8 5 1 0 - . ) 5 2 0 0 ( . * * * 0 0 2 0 - . ) 7 3 0 0 ( . * * * 6 9 1 0 - . ) 7 1 0 0 ( . * * * 3 1 2 0 - . 1 = 4 e l i t n u Q i ) 8 2 0 0 ( . * 5 5 0 0 . ) 7 1 0 0 ( . 1 1 0 0 . ) 3 3 0 0 ( . 0 5 0 0 . ) 3 1 0 0 ( . * * * 2 4 0 0 . ) 0 1 0 0 ( . 5 1 0 0 - . ) 9 0 0 0 ( . * * 8 1 0 0 - . ) 1 1 0 0 ( . 3 1 0 0 - . ) 6 0 0 0 ( . * * * 5 1 0 0 - . ) 1 3 0 0 ( . * * * 6 5 1 0 - . ) 2 2 0 0 ( . * * * 9 1 2 0 - . ) 4 3 0 0 ( . * * * 9 5 2 0 - . ) 6 1 0 0 ( . * * * 2 3 2 0 - . 1 = 5 e l i t n u Q i ) 3 1 0 0 ( . * 3 2 0 0 - . ) 2 1 0 0 ( . 8 1 0 0 - . ) 1 1 0 0 ( . 2 1 0 0 . ) 7 0 0 0 ( . 1 1 0 0 - . ) 6 0 0 0 ( . 9 0 0 0 - . ) 6 0 0 0 ( . * 2 1 0 0 - . ) 5 0 0 0 ( . * * * 7 1 0 0 - . ) 3 0 0 0 ( . * * * 2 1 0 0 - . ) 6 1 0 0 ( . 1 1 0 0 - . ) 8 1 0 0 ( . * * 3 4 0 0 - . ) 5 2 0 0 ( . * * * 7 6 0 0 - . ) 1 1 0 0 ( . * * * 9 3 0 0 - . l e a m e F ) 1 3 0 0 ( . * * 6 6 0 0 . ) 5 1 0 0 ( . * * * 7 8 0 0 . ) 7 1 0 0 ( . * * * 5 5 0 0 . ) 1 1 0 0 ( . * * * 7 6 0 0 . ) 1 1 0 0 ( . * * * 6 3 0 0 . ) 1 1 0 0 ( . * * * 9 4 0 0 . ) 0 1 0 0 ( . * * * 1 4 0 0 . ) 6 0 0 0 ( . * * * 2 4 0 0 . ) 1 3 0 0 ( . * * * 0 0 2 0 . ) 2 2 0 0 ( . * * * 0 4 3 0 . ) 1 3 0 0 ( . * * * 2 8 3 0 . ) 6 1 0 0 ( . * * * 3 2 3 0 . t n a t s n o C 5 6 6 3 , 2 1 0 0 . 1 9 3 3 , 4 0 0 0 . s d n a w o L l s y e l l a V 7 9 3 3 , 9 0 0 0 . 3 5 4 0 1 , 4 0 0 0 . ) % ( y c n e i c i f e d n o r I 5 6 6 3 , 7 0 0 0 . 1 9 3 3 , 2 1 0 0 . 7 9 3 3 , 0 1 0 0 . 3 5 4 0 1 , 8 0 0 0 . ) % ( D A V 3 2 7 3 , 1 3 0 0 . 1 0 7 3 , 2 5 0 0 . 6 4 4 3 , 4 5 0 0 . 0 7 8 0 1 , s n o i t a v r e s b O 2 5 0 0 . d e r a u q s - R ) % ( i a m e n A l l A S E L B A R A V I l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i ) 2 9 0 0 ( . 4 9 0 0 - . ) 0 7 0 0 ( . 8 1 0 0 . ) 8 7 0 0 ( . 2 0 1 0 . ) 7 4 0 0 ( . 4 2 0 0 . ) 0 1 1 0 ( . 1 1 1 0 . ) 0 8 0 0 ( . * * 4 0 2 0 . ) 4 8 0 0 ( . 7 1 1 0 - . ) 2 5 0 0 ( . 2 8 0 0 . ) 3 3 0 0 ( . * 0 6 0 0 - . ) 2 3 0 0 ( . 8 3 0 0 - . ) 1 3 0 0 ( . * * * 6 1 1 0 - . ) 0 2 0 0 ( . * * * 2 7 0 0 - . 1 = 2 e l i t n u Q i ) 8 7 0 0 ( . 5 6 0 0 - . ) 5 6 0 0 ( . * * 5 3 1 0 . ) 5 8 0 0 ( . 4 0 1 0 . ) 2 4 0 0 ( . 7 4 0 0 . ) 4 0 1 0 ( . 8 5 1 0 . ) 7 7 0 0 ( . 5 2 0 0 . ) 8 6 0 0 ( . * * 8 5 1 0 - . ) 2 5 0 0 ( . 0 6 0 0 . ) 6 3 0 0 ( . * * * 7 9 1 0 - . ) 6 3 0 0 ( . * * 3 7 0 0 - . ) 6 4 0 0 ( . * * * 9 3 1 0 - . ) 4 2 0 0 ( . * * * 9 4 1 0 - . 1 = 3 e l i t n u Q i ) 5 9 0 0 ( . 8 9 0 0 - . ) 7 6 0 0 ( . 8 8 0 0 . ) 7 0 1 0 ( . * * * 8 1 4 0 . ) 0 5 0 0 ( . 7 4 0 0 . ) 6 2 1 0 ( . 6 3 0 0 . ) 0 6 0 0 ( . * 5 0 1 0 - . ) 2 9 0 0 ( . * * 3 8 1 0 - . ) 3 6 0 0 ( . 2 2 0 0 - . ) 2 4 0 0 ( . * * * 0 9 1 0 - . ) 5 3 0 0 ( . * * * 7 3 1 0 - . ) 2 4 0 0 ( . * * 4 8 0 0 - . ) 5 2 0 0 ( . * * * 0 7 1 0 - . 1 = 4 e l i t n u Q i ) 2 9 0 0 ( . 8 9 0 0 - . ) 3 5 0 0 ( . 2 2 0 0 - . ) 9 6 0 0 ( . 2 2 0 0 - . ) 7 3 0 0 ( . 9 5 0 0 - . ) 5 0 1 0 ( . 6 9 0 0 - . ) 1 6 0 0 ( . 8 5 0 0 - . ) 3 8 0 0 ( . * * 2 8 1 0 - . ) 5 4 0 0 ( . * 0 8 0 0 - . ) 8 3 0 0 ( . * * * 2 7 2 0 - . ) 2 3 0 0 ( . * * * 2 3 2 0 - . ) 9 4 0 0 ( . * * * 7 6 2 0 - . ) 3 2 0 0 ( . * * * 8 7 2 0 - . 1 = 5 e l i t n u Q i ) 3 5 0 0 ( . 8 5 0 0 - . ) 3 4 0 0 ( . * * * 2 3 1 0 - . ) 5 5 0 0 ( . * * * 8 4 1 0 - . ) 1 3 0 0 ( . * * * 0 1 1 0 - . ) 9 5 0 0 ( . 6 0 0 0 - . ) 9 3 0 0 ( . 5 0 0 0 - . ) 0 5 0 0 ( . 1 1 0 0 . ) 0 3 0 0 ( . 1 0 0 0 - . ) 3 2 0 0 ( . 8 2 0 0 - . ) 1 2 0 0 ( . * * * 7 5 0 0 - . ) 9 2 0 0 ( . * * * 5 8 0 0 - . ) 4 1 0 0 ( . * * * 4 5 0 0 - . l e a m e F ) 0 0 1 0 ( . * * * 0 7 2 0 . ) 0 7 0 0 ( . * * * 3 5 3 0 . ) 8 8 0 0 ( . * * * 5 6 4 0 . ) 9 4 0 0 ( . * * * 9 4 3 0 . ) 3 3 1 0 ( . * * * 4 0 5 0 . ) 2 8 0 0 ( . * * * 5 8 3 0 . ) 4 0 1 0 ( . * * * 1 1 6 0 . ) 9 5 0 0 ( . * * * 4 5 4 0 . ) 9 3 0 0 ( . * * * 9 4 7 0 . ) 9 2 0 0 ( . * * * 3 8 7 0 . ) 2 3 0 0 ( . * * * 7 0 9 0 . ) 1 2 0 0 ( . * * * 3 1 8 0 . t n a t s n o C Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 20 of 25 1 1 0 0 . 8 1 5 1 5 0 0 . 0 3 5 1 1 1 0 . 2 4 5 0 9 5 1 , 8 2 0 0 . 4 3 0 0 . 8 1 5 6 4 0 0 . 0 3 5 ) d e u n i t n o C ( a t a d g n i s s i m r o f s t n e m j t s u d a W P I h t i w n o g e r i y b t n e m p o e v e d l d l i h c n i i s t n e d a r g S E S d e t s u d A j 0 1 e l b a T 9 3 0 0 . 2 4 5 0 9 5 1 , 7 1 0 0 . e r o c s - z n o i t a c i n u m m o C 5 1 3 3 , 4 3 0 0 . 5 0 2 3 , 0 4 0 0 . 4 9 8 2 , 5 4 0 0 . 4 1 4 9 , s n o i t a v r e s b O 2 4 0 0 . d e r a u q s - R e r o c s - z r o t o m s s o r G d l i h c f o t e s a d n a x e s ’ s d l i h c e d u l c n i s l o r t n o C . n o i t a m i t s E S L O . s e s e h t n e r a p n i ) t n e m g e s r o r o t c e s s u s n e c ( l e v e l U S P e h t t a d e r e t s u l c E S . l e v e l % 0 1 e h t t a t n a c i f i n g i s * ; l e v e l % 5 e h t t a t n a c i f i n g i s * * ; l e v e l % 1 e h t t a t n a c i f i n g i s * * * : e t o N ) 9 2 1 0 ( . * * 4 8 2 0 . ) 0 0 1 0 ( . 3 0 0 0 - . ) 2 1 1 0 ( . 4 3 0 0 - . ) 5 6 0 0 ( . 4 8 0 0 . ) 4 3 1 0 ( . 4 8 0 0 . ) 6 8 0 0 ( . 7 4 0 0 - . ) 3 1 1 0 ( . 6 7 0 0 - . ) 2 6 0 0 ( . 1 3 0 0 - . t n a t s n o C ) 1 7 0 0 ( . 6 9 0 0 . ) 9 5 0 0 ( . 4 4 0 0 . ) 4 7 0 0 ( . 5 5 0 0 . ) 0 4 0 0 ( . 5 6 0 0 . ) 6 7 0 0 ( . * * 4 5 1 0 - . ) 1 6 0 0 ( . 4 6 0 0 - . ) 6 7 0 0 ( . 3 7 0 0 - . ) 1 4 0 0 ( . * * 1 0 1 0 - . l e a m e F 2 3 9 1 , 0 4 0 0 . 2 1 9 1 , 4 1 0 0 . 9 0 9 1 , 8 0 0 0 . 3 5 7 5 , 2 1 0 0 . 2 3 9 1 , 3 1 0 0 . 2 1 9 1 , 5 1 0 0 . 9 0 9 1 , 2 0 0 0 . 3 5 7 5 , s n o i t a v r e s b O 6 0 0 0 . d e r a u q s - R ) 3 7 1 0 ( . 9 6 1 0 - . ) 8 0 1 0 ( . 6 5 1 0 - . ) 8 0 1 0 ( . 8 7 1 0 - . ) 4 7 0 0 ( . * * 6 6 1 0 - . ) 6 8 1 0 ( . 5 1 0 0 - . ) 2 9 0 0 ( . 9 6 0 0 - . ) 3 0 1 0 ( . 3 2 0 0 . ) 0 7 0 0 ( . 8 0 0 0 - . 1 = 2 ) 4 1 1 0 ( . 8 7 0 0 . ) 4 0 1 0 ( . 2 1 0 0 . ) 7 1 1 0 ( . 1 3 0 0 - . ) 3 6 0 0 ( . 1 4 0 0 . ) 1 1 1 0 ( . 3 1 0 0 . ) 9 8 0 0 ( . 5 4 1 0 - . ) 1 4 1 0 ( . 7 2 0 0 . ) 6 6 0 0 ( . 7 0 0 0 - . 1 = 3 ) 9 1 1 0 ( . 9 2 1 0 - . ) 9 0 1 0 ( . 5 5 1 0 - . ) 6 2 1 0 ( . 2 1 0 0 . ) 1 7 0 0 ( . 3 8 0 0 - . ) 9 0 1 0 ( . 9 3 0 0 . ) 1 2 1 0 ( . 9 7 1 0 - . ) 2 4 1 0 ( . 5 2 0 0 - . ) 1 7 0 0 ( . 6 0 0 0 . 1 = 4 ) 3 5 1 0 ( . 1 6 0 0 . ) 1 0 1 0 ( . 3 1 1 0 . ) 3 6 1 0 ( . 1 5 0 0 . ) 8 7 0 0 ( . 9 9 0 0 . ) 0 5 1 0 ( . 5 6 1 0 . ) 1 0 1 0 ( . 3 3 0 0 - . ) 3 7 1 0 ( . 4 1 0 0 - . ) 3 7 0 0 ( . 0 1 1 0 . 1 = 5 e l i t n u Q i e l i t n u Q i e l i t n u Q i e l i t n u Q i s d n a w o L l s y e l l a V l s d n a h g H i l l A s d n a w o L l s y e l l a V l s d n a h g H i l l A S E L B A R A V I - e s u o h a t i p a c r e p l y h t n o m g n i s u s e l i t n u q i l n o i t a u p o p e r a s e l i t n u Q i . ) 9 5 - 7 5 , 6 5 - 4 5 , 3 5 - 1 5 , 0 5 - 8 4 , 7 4 - 5 4 , 4 4 - 2 4 , 1 4 - 9 3 , 8 3 - 6 3 , 5 3 - 3 3 , 2 3 - 0 3 6 2 - 4 2 , 3 2 - 1 2 , 0 2 - 8 1 , 7 1 - 5 1 , 4 1 - 2 1 , 1 1 - 9 , 8 - 6 , 5 - 3 , 2 - 0 ( s h t n o m n i s e i r o g e t a c e g a r o f i s e m m u d t s e t - F , 9 2 - 7 2 , e h t f o . 0 = 5 Q = 4 Q = 3 Q = 2 Q f o l e u a v - p e h t s i ) s t c e f f e e l i t n u q i r o f t s e t - F ( l e u a v - p . l e b a i r a v i g n k n a r e h t s a n o i t p m u s n o c l d o h Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 21 of 25 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 S E L B A R A V I e r o c s - z t h g e h i r o f t h g e W i e r o c s - z e g a r o f t h g e W i e r o c s - z e g a r o f t h g e H i ) 8 6 0 0 ( . 7 2 0 0 . ) 3 8 0 0 ( . 2 0 1 0 . ) 0 9 0 0 ( . * * 6 9 1 0 . ) 1 0 1 0 ( . 8 5 1 0 . ) 1 6 0 0 ( . * * * 9 0 2 0 . ) 8 7 0 0 ( . * * * 9 6 2 0 . ) 1 9 0 0 ( . * * * 8 0 3 0 . ) 4 0 1 0 ( . * 9 9 1 0 . ) 8 6 0 0 ( . * * * 6 1 3 0 . ) 6 8 0 0 ( . * * * 8 7 3 0 . ) 1 9 0 0 ( . * * * 0 9 2 0 . ) 6 2 1 0 ( . 0 3 1 0 . 1 = 2 ) 5 6 0 0 ( . 9 7 0 0 . ) 1 9 0 0 ( . 0 2 0 0 . ) 6 8 0 0 ( . * * * 3 9 2 0 . ) 2 1 1 0 ( . * 8 0 2 0 . ) 6 6 0 0 ( . * * * 1 1 5 0 . ) 5 8 0 0 ( . * * * 3 9 3 0 . ) 7 8 0 0 ( . * * * 3 3 5 0 . ) 6 0 1 0 ( . * * * 7 8 3 0 . ) 9 6 0 0 ( . * * * 8 5 7 0 . ) 9 7 0 0 ( . * * * 5 9 6 0 . ) 1 9 0 0 ( . * * * 1 0 6 0 . ) 6 3 1 0 ( . * * * 5 3 4 0 . 1 = 3 ) 3 6 0 0 ( . * * 1 4 1 0 . ) 8 8 0 0 ( . * * 5 9 1 0 . ) 1 0 1 0 ( . * * 1 3 2 0 . ) 8 2 1 0 ( . 6 0 1 0 . ) 1 6 0 0 ( . * * * 8 3 5 0 . ) 5 8 0 0 ( . * * * 2 7 6 0 . ) 0 0 1 0 ( . * * * 6 6 5 0 . ) 7 2 1 0 ( . * * * 5 4 4 0 . ) 2 7 0 0 ( . * * * 4 7 7 0 . ) 8 8 0 0 ( . * * * 8 5 9 0 . ) 8 0 1 0 ( . * * * 7 2 8 0 . ) 4 3 1 0 ( . * * * 0 1 6 0 . 1 = 4 ) 6 7 0 0 ( . * * 6 6 1 0 . ) 5 8 0 0 ( . * 5 5 1 0 . ) 7 0 1 0 ( . * * * 1 9 3 0 . ) 5 2 1 0 ( . * 3 2 2 0 . ) 2 8 0 0 ( . * * * 4 5 7 0 . ) 5 8 0 0 ( . * * * 3 4 7 0 . ) 6 0 1 0 ( . * * * 8 7 7 0 . ) 1 3 1 0 ( . * * * 3 1 5 0 . ) 4 8 0 0 ( . * * * 6 6 0 1 . ) 9 9 0 0 ( . * * * 9 3 1 1 . ) 6 2 1 0 ( . * * * 4 7 9 0 . ) 5 4 1 0 ( . * * * 6 1 6 0 . 1 = 5 e l i t n u Q i e l i t n u Q i e l i t n u Q i e l i t n u Q i ) 7 4 0 0 ( . 4 5 0 0 - . ) 3 5 0 0 ( . 6 4 0 0 . ) 1 6 0 0 ( . * 3 0 1 0 . ) 3 7 0 0 ( . 0 9 0 0 . ) 1 5 0 0 ( . 3 8 0 0 - . ) 2 5 0 0 ( . * * 6 1 1 0 . ) 6 6 0 0 ( . * * * 0 9 1 0 . ) 3 7 0 0 ( . * * 8 5 1 0 . ) 2 5 0 0 ( . 7 5 0 0 - . ) 3 6 0 0 ( . * * 0 6 1 0 . ) 4 7 0 0 ( . * * * 9 3 2 0 . ) 2 8 0 0 ( . * * 4 7 1 0 . 1 = l e a m e F ) 8 2 2 0 ( . * * 4 8 5 0 . ) 1 3 2 0 ( . 2 1 0 0 . ) 4 8 1 0 ( . 6 9 1 0 - . ) 8 2 1 0 ( . * * 5 7 2 0 . ) 4 0 2 0 ( . 3 9 2 0 - . ) 5 2 2 0 ( . * * * 0 0 0 1 - . ) 4 7 1 0 ( . * * * 1 1 8 0 - . ) 4 2 1 0 ( . * * * 9 6 3 0 - . ) 3 7 1 0 ( . * * * 2 7 4 1 - . ) 2 0 3 0 ( . * * * 4 7 0 2 - . ) 1 8 1 0 ( . * * * 7 6 5 1 - . ) 3 5 1 0 ( . * * * 8 5 6 0 - . t n a t s n o C a t a d g n i s s i m r o f s t n e m j t s u d a W P I h t i w e g a ’ s d l i h c y b t n e m p o e v e d l d l i h c n i i s t n e d a r g S E S d e t s u d A j 1 1 e l b a T 2 0 7 3 , 5 0 0 0 . 3 2 3 2 , 9 0 0 0 . 5 8 2 2 , 9 1 0 0 . 9 8 6 1 , 7 0 0 0 . ) D S 2 > ( i t h g e w r e v O 7 0 7 3 , 6 7 0 0 . 5 2 3 2 , 1 8 0 0 . 5 8 2 2 , 6 6 0 0 . 3 9 6 1 , 3 3 0 0 . ) D S 2 - < ( i t h g e w r e d n U 2 8 8 3 , 9 2 1 0 . 1 1 4 2 , 6 2 1 0 . 1 9 3 2 , 2 9 0 0 . 8 1 7 1 , s n o i t a v r e s b O 7 4 0 0 . d e r a u q s - R ) D S 2 - < ( g n i t n u t S s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 S E L B A R A V I ) 4 1 0 0 ( . 6 0 0 0 . ) 9 1 0 0 ( . 3 1 0 0 . ) 7 1 0 0 ( . 2 2 0 0 . ) 2 2 0 0 ( . 2 0 0 0 . ) 1 1 0 0 ( . 8 0 0 0 . ) 5 0 0 0 ( . * 9 0 0 0 - . ) 2 1 0 0 ( . * * * 2 3 0 0 - . ) 4 1 0 0 ( . 0 1 0 0 - . ) 6 2 0 0 ( . * * * 8 1 1 0 - . ) 9 3 0 0 ( . * * * 8 1 1 0 - . ) 8 3 0 0 ( . * * * 0 5 1 0 - . ) 2 3 0 0 ( . * 1 6 0 0 - . 1 = 2 ) 7 1 0 0 ( . 9 1 0 0 . ) 7 1 0 0 ( . 7 0 0 0 . ) 7 1 0 0 ( . 2 2 0 0 . ) 7 2 0 0 ( . 9 1 0 0 . ) 7 0 0 0 ( . 7 0 0 0 . ) 9 0 0 0 ( . 4 0 0 0 . ) 2 1 0 0 ( . * * * 3 3 0 0 - . ) 8 1 0 0 ( . 2 0 0 0 - . ) 6 2 0 0 ( . * * * 3 9 1 0 - . ) 4 3 0 0 ( . * * * 8 2 2 0 - . ) 4 3 0 0 ( . * * * 5 0 2 0 - . ) 0 3 0 0 ( . * * * 8 0 1 0 - . 1 = 3 ) 8 1 0 0 ( . 5 2 0 0 . ) 2 2 0 0 ( . 6 3 0 0 . ) 5 1 0 0 ( . 6 0 0 0 . ) 5 2 0 0 ( . 3 1 0 0 . ) 4 0 0 0 ( . * 8 0 0 0 - . ) 5 0 0 0 ( . * 9 0 0 0 - . ) 3 1 0 0 ( . * * 1 3 0 0 - . ) 2 1 0 0 ( . * * 7 2 0 0 - . ) 5 2 0 0 ( . * * * 8 2 2 0 - . ) 2 3 0 0 ( . * * * 0 6 2 0 - . ) 3 3 0 0 ( . * * * 1 6 2 0 - . ) 9 2 0 0 ( . * * * 2 3 1 0 - . 1 = 4 ) 0 2 0 0 ( . * * * 0 7 0 0 . ) 3 2 0 0 ( . * 0 4 0 0 . ) 4 2 0 0 ( . * * * 2 6 0 0 . ) 0 3 0 0 ( . 2 1 0 0 . ) 6 0 0 0 ( . 0 0 0 0 - . ) 7 0 0 0 ( . 6 0 0 0 - . ) 4 1 0 0 ( . * * 9 2 0 0 - . ) 2 1 0 0 ( . * * 9 2 0 0 - . ) 3 2 0 0 ( . * * * 5 5 2 0 - . ) 3 3 0 0 ( . * * * 8 9 2 0 - . ) 1 3 0 0 ( . * * * 5 8 2 0 - . ) 4 3 0 0 ( . * * * 8 9 0 0 - . 1 = 5 e l i t n u Q i e l i t n u Q i e l i t n u Q i e l i t n u Q i ) 3 1 0 0 ( . * 5 2 0 0 - . ) 7 1 0 0 ( . 0 2 0 0 - . ) 4 1 0 0 ( . 2 0 0 0 - . ) 9 1 0 0 ( . 5 0 0 0 . ) 6 0 0 0 ( . 8 0 0 0 - . ) 4 0 0 0 ( . 4 0 0 0 - . ) 7 0 0 0 ( . 6 0 0 0 - . ) 0 1 0 0 ( . * * * 5 3 0 0 - . ) 7 1 0 0 ( . 5 2 0 0 . ) 3 2 0 0 ( . * * * 1 6 0 0 - . ) 3 2 0 0 ( . * * * 9 8 0 0 - . ) 1 2 0 0 ( . * * 7 4 0 0 - . 1 = l e a m e F ) 2 5 0 0 ( . * 9 8 0 0 . ) 4 6 0 0 ( . 2 0 0 0 . ) 4 3 0 0 ( . 0 2 0 0 . ) 7 2 0 0 ( . * * * 9 0 1 0 . ) 1 2 0 0 ( . 0 1 0 0 - . ) 1 2 0 0 ( . * * 7 4 0 0 . ) 3 2 0 0 ( . * * 9 4 0 0 . ) 6 1 0 0 ( . * * * 9 6 0 0 . ) 7 5 0 0 ( . * * * 0 7 3 0 . ) 0 9 0 0 ( . * * * 6 6 5 0 . ) 0 6 0 0 ( . * * * 8 5 3 0 . ) 0 3 0 0 ( . * * * 6 1 2 0 . t n a t s n o C 2 0 7 3 , 1 1 0 0 . 3 2 3 2 , 6 0 0 0 . 5 8 2 2 , 0 1 0 0 . 9 8 6 1 , 3 0 0 0 . 2 0 7 3 , 5 0 0 0 . 3 2 3 2 , 5 0 0 0 . 5 8 2 2 , 0 1 0 0 . 9 8 6 1 , 2 2 0 0 . 2 8 8 3 , 9 5 0 0 . 1 1 4 2 , 2 7 0 0 . 1 9 3 2 , 3 7 0 0 . 8 1 7 1 , s n o i t a v r e s b O 1 3 0 0 . d e r a u q s - R y c n e i c i f e d n o r I y c n e i c i f e d A n m a t i V i i a m e n A s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 6 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 6 s h t n o M 9 5 - 7 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 S E L B A R A V I ) 5 5 0 0 ( . 2 2 0 0 . ) 8 7 0 0 ( . 2 5 0 0 . ) 9 5 0 0 ( . 9 7 0 0 . ) 5 9 0 0 ( . 1 7 0 0 . ) 3 3 0 0 ( . * * * 6 9 0 0 - . ) 4 3 0 0 ( . * * * 7 9 0 0 - . ) 6 3 0 0 ( . 5 2 0 0 - . ) 0 5 0 0 ( . 9 3 0 0 - . 1 = 2 ) 0 5 0 0 ( . 5 3 0 0 . ) 0 7 0 0 ( . 6 4 0 0 . ) 0 6 0 0 ( . 7 7 0 0 . ) 0 8 0 0 ( . 8 2 0 0 . ) 0 4 0 0 ( . * * * 5 6 1 0 - . ) 1 5 0 0 ( . * * * 5 4 2 0 - . ) 8 3 0 0 ( . * 3 7 0 0 - . ) 4 5 0 0 ( . * 3 0 1 0 - . 1 = 3 ) 6 6 0 0 ( . 4 4 0 0 . ) 1 7 0 0 ( . 2 5 0 0 . ) 9 6 0 0 ( . 1 1 0 0 - . ) 9 8 0 0 ( . 6 5 0 0 - . ) 9 3 0 0 ( . * * * 4 3 2 0 - . ) 3 4 0 0 ( . * * * 8 2 2 0 - . ) 1 4 0 0 ( . 7 5 0 0 - . ) 6 5 0 0 ( . * * 7 3 1 0 - . 1 = 4 ) 8 4 0 0 ( . 4 5 0 0 - . ) 5 5 0 0 ( . * 4 9 0 0 - . ) 4 5 0 0 ( . 1 6 0 0 - . ) 8 7 0 0 ( . 9 1 1 0 - . ) 8 3 0 0 ( . * * * 5 8 2 0 - . ) 2 4 0 0 ( . * * * 4 8 2 0 - . ) 4 4 0 0 ( . * * * 4 5 2 0 - . ) 2 5 0 0 ( . * * * 0 1 3 0 - . 1 = 5 e l i t n u Q i e l i t n u Q i e l i t n u Q i e l i t n u Q i ) 6 3 0 0 ( . * * * 9 9 0 0 - . ) 7 4 0 0 ( . * * 0 2 1 0 - . ) 8 3 0 0 ( . 3 0 0 0 - . ) 8 4 0 0 ( . 2 1 0 0 - . ) 4 2 0 0 ( . * * * 2 8 0 0 - . ) 0 3 0 0 ( . 0 0 0 0 - . ) 6 2 0 0 ( . 8 3 0 0 - . ) 3 3 0 0 ( . * * 4 7 0 0 - . 1 = l e a m e F ) 6 8 0 0 ( . * * * 3 4 5 0 . ) 9 0 1 0 ( . 0 8 0 0 . ) 6 0 1 0 ( . * * 6 1 2 0 . ) 1 5 1 0 ( . * * 1 8 3 0 . ) 6 8 0 0 ( . * * * 2 0 1 1 . ) 6 2 1 0 ( . * * * 5 0 2 1 . ) 1 7 0 0 ( . * * * 1 2 9 0 . ) 3 6 0 0 ( . * * * 0 4 4 0 . t n a t s n o C Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 22 of 25 d l i h c f o t e s a d n a x e s ’ s d l i h c e d u l c n i s l o r t n o C . n o i t a m i t s E S L O . s e s e h t n e r a p n i ) t n e m g e s r o r o t c e s s u s n e c ( l e v e l U S P e h t t a d e r e t s u l c E S . l e v e l % 0 1 e h t t a t n a c i f i n g i s * ; l e v e l % 5 e h t t a t n a c i f i n g i s * * ; l e v e l % 1 e h t t a t n a c i f i n g i s * * * : e t o N ) 9 5 0 0 ( . 6 7 0 0 . ) 3 6 0 0 ( . * * 1 4 1 0 . ) 4 6 0 0 ( . 4 0 0 0 . ) 0 7 0 0 ( . 4 9 0 0 - . ) 6 6 0 0 ( . 8 4 0 0 - . ) 0 6 0 0 ( . * * 5 3 1 0 - . 1 = l e a m e F ) 0 9 2 0 ( . 1 7 4 0 - . ) 8 5 1 0 ( . 3 5 1 0 . ) 0 2 1 0 ( . 8 7 0 0 . ) 8 4 2 0 ( . 7 4 1 0 - . ) 4 5 1 0 ( . 4 6 1 0 - . ) 2 1 1 0 ( . 9 4 1 0 - . t n a t s n o C 2 9 0 2 , 4 1 0 0 . 5 8 1 2 , 7 1 0 0 . 6 7 4 1 , 1 2 0 0 . 2 9 0 2 , 9 1 0 0 . 5 8 1 2 , 7 0 0 0 . 6 7 4 1 , s n o i t a v r e s b O 1 1 0 0 . d e r a u q s - R s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 s h t n o M 6 3 - 4 2 s h t n o M 3 2 - 2 1 s h t n o M 1 1 - 3 S E L B A R A V I e r o c s - z n o i t a c i n u m m o C e r o c s - z r o t o m s s o r G ) 9 8 0 0 ( . * * 5 2 2 0 . ) 5 8 0 0 ( . * * * 9 5 2 0 - . ) 7 9 0 0 ( . * * 0 0 2 0 . ) 0 2 1 0 ( . 7 2 1 0 . ) 5 8 0 0 ( . 6 3 0 0 - . ) 7 0 1 0 ( . 9 2 0 0 - . 1 = 3 ) 1 1 1 0 ( . 2 3 0 0 - . ) 3 0 1 0 ( . * * 5 1 2 0 - . ) 7 0 1 0 ( . 4 1 0 0 - . ) 3 1 1 0 ( . 0 3 1 0 . ) 8 1 1 0 ( . 2 1 0 0 . ) 2 0 1 0 ( . 0 4 0 0 - . 1 = 4 ) 5 2 1 0 ( . * 8 2 2 0 . ) 5 8 0 0 ( . 4 1 1 0 - . ) 3 2 1 0 ( . 0 0 2 0 . ) 6 1 1 0 ( . * * * 4 0 4 0 . ) 5 9 0 0 ( . 4 4 1 0 . ) 1 1 1 0 ( . 3 1 0 0 - . 1 = 5 ) 2 9 0 0 ( . 8 6 0 0 . ) 9 8 0 0 ( . * * * 0 9 2 0 - . ) 8 0 1 0 ( . 7 5 1 0 - . ) 1 1 1 0 ( . 3 8 1 0 . ) 9 9 0 0 ( . 8 6 0 0 - . ) 1 0 1 0 ( . 5 3 0 0 - . 1 = 2 e l i t n u Q i e l i t n u Q i e l i t n u Q i e l i t n u Q i 7 3 0 0 . 4 9 9 8 4 0 0 . 6 9 5 6 1 0 0 . 4 9 9 ) d e u n i t n o C ( a t a d g n i s s i m r o f s t n e m j t s u d a W P I 0 2 0 0 . 6 9 5 h t i w e g a ’ s d l i h c y b t n e m p o e v e d l d l i h c n i i s t n e d a r g S E S d e t s u d A j 1 1 e l b a T 0 8 5 3 , 7 5 0 0 . 9 7 1 2 , 7 5 0 0 . 7 6 1 2 , 0 4 0 0 . 8 8 4 1 , s n o i t a v r e s b O 9 8 0 0 . d e r a u q s - R - e s u o h a t i p a c r e p l y h t n o m g n i s u s e l i t n u q i l n o i t a u p o p e r a s e l i t n u Q i . ) 9 5 - 7 5 , 6 5 - 4 5 , 3 5 - 1 5 , 0 5 - 8 4 , 7 4 - 5 4 , 4 4 - 2 4 , 1 4 - 9 3 , 8 3 - 6 3 , 5 3 - 3 3 , 2 3 - 0 3 6 2 - 4 2 , 3 2 - 1 2 , 0 2 - 8 1 , 7 1 - 5 1 , 4 1 - 2 1 , 1 1 - 9 , 8 - 6 , 5 - 3 , 2 - 0 ( s h t n o m n i s e i r o g e t a c e g a r o f i s e m m u d t s e t - F , 9 2 - 7 2 , e h t f o . 0 = 5 Q = 4 Q = 3 Q = 2 Q f o l e u a v - p e h t s i ) s t c e f f e e l i t n u q i r o f t s e t - F ( l e u a v - p . l e b a i r a v i g n k n a r e h t s a n o i t p m u s n o c l d o h Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 23 of 25 Fig. 2 Distribution of households by wealth scores, Bolivia 2012. Notes. Cases weighted by household weight Appendix 2 Technical Appendix: Construction of SES indicators Direct measure of SES: household consumption Household consumption was used as our preferred direct measure of SES. In our context, we believe consumption is a better measure of SES than other monetary measures such as income, considering the problems with measuring income in settings with a high proportion of self-employed and informal workers. In addition, irregular and intermittent earnings from informal employment make income measurements more volatile than consumption and, therefore, more directly related to current living standards [22, 35]. Our consumption index is based on the aggregation of pay- ments for goods and services collected using monthly, quar- terly, and yearly reference periods. For food consumption, we used information about monthly purchases of a list of 41 food items as well as spending on food consumed outside the house. Because home-produced foods are an important part of food consumption in rural areas, we included self-reported valuations of consumption from household production. Non- food consumption includes payments in housing, household services, education, personal goods, health, transport, recre- ation, and financial services. For non-renters, housing con- sumption was imputed using a hedonic price model. We excluded spending on durable goods and other lumpy spend- ing such as hospitalizations18. To construct total consumption, individual items based on quarterly and yearly recall periods were converted into monthly figures and then all food and nonfood items were added together. We adjusted for 18The ESNUT did not collect information about the use value of durables to construct a complete measure of household consumption. household size by dividing total consumption by the number of household members and obtaining a per capita measure. Proxy measure of SES: wealth index Additionally, we developed an alternative proxy measure of SES by estimating a composite indicator “wealth index” that com- bines information about ownership of household assets, physical characteristics of the dwelling and access to basic services using principal components analysis [26, 36]19. Asset-based indices have several advantages over direct measures of living standards: data on assets and dwelling characteristics are less expensive and easier to collect; they are more robust to measurement error and reporting biases than income and expenditures; and an asset-based index reflects the notion of permanent income more closely, and is based on a long term conceptualization of wealth, which is more relevant for inequality analysis ([36, 37]; O’Donnell et al., 2008 [26, 38];). Wealth indices based on princi- pal components analysis have also been used in analyses of so- cioeconomic gaps in health and child development outcomes in low- and middle-income countries [3, 4, 6, 26, 39]. For this study, a wealth index was estimated based on assets ownership, including refrigerator, radio, television, fixed phone line, car, motorcycle, bicycle; and information about water sources, toilet facilities, electricity, and the construction material of wall, roof, and floor in the household20. The resulting score is a standard- ized wealth index with a mean of zero and a standard deviation 19Principal components analysis uses statistical methods to determine the weights of items in the index. Individual items are weighted to maximize the variability of the new composite variable. 20Principal components were computed using the Stata® command factor with the pcf option specified. Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 24 of 25 of one. All members in the household receive the same wealth index score. Figure 2 shows the distribution of households by the value of the wealth index. Acknowledgements All opinions in this paper are those of the authors and do not necessarily represent the views of the National Commission for Social Protection in Health, of the Government of Mexico or the Inter-American Development Bank, their Executive Directors or the governments they represent. The au- thors acknowledge that a working paper version of this paper is available in the link: https://publications.iadb.org/en/socioeconomic-gaps-child-develop- ment-evidence-national-health-and-nutrition-survey-bolivia. Authors’ contributions Pablo Celhay (PC), Sebastian Martinez (SM), Cecilia Vidal (CV). Conceived and designed the study: PC, SM. Wrote the paper: PC, CV, SM. Discussed, critically revised, and approved the study protocol: PC, SM, CV. Responsible for the organization and conduct of the study: SM, PC, CV. Supervised the study: SM, PC. Responsible for the statistical analysis: CV. Contributed to data analysis: PC. Drafted the first version of the report: PC, SM, CV. Elaborated, discussed and approved the final version of the paper for publication: PC, SM, CV. Funding Celhay thanks institutional support from the Millennium Nucleus Initiative and acknowledges financial support from CONICYT, FONDECYT Iniciación 11180416. Availability of data and materials The data used in this study are available at http://ghdx.healthdata.org/ record/bolivia-health-and-nutrition-assessment-survey-2012 Ethics approval and consent to participate The analysis of the biological data was approved by the Comisión de Ética de la Investigación del Comité Naicional de Bioética (CEI-CNB). No number is provided. Consent for publication The data used in this study is secondary and de-identified. Competing interests Dr. Celhay has nothing to disclose. Dr. Martinez has nothing to disclose. Ms. Vidal has nothing to disclose. Author details 1School of Government, Pontificia Universidad Católica de Chile and Millennium Nuclei for the Study of the Life Course and Vulnerability, Avda. Vicuña Mackenna 4860 – Macul, Santiago, Chile. 2Inter-American Development Bank, 1300 New York Avenue, NW, Washington, DC 20577, USA. Received: 19 December 2019 Accepted: 19 May 2020 References 1. Lu C, Black MM, Richter LM. Risk of poor development in young children in low-income and middle-income countries: an estimation and analysis at the global, regional, and country level. Lancet Glob Health. 2016;4(12):e916–22. Grantham-McGregor S, Cheung YB, Cueto S, Glewwe P, Richter L, Strupp B. Developmental potential in the first 5 years for children in developing countries. Lancet. 2007;369(9555):60–70. Hamadani JD, Tofail F, Huda SN, Alam DS, Ridout DA, Attanasio O, Grantham-McGregor SM. Cognitive Deficit and Poverty in the First 5 Years of Childhood in Bangladesh. Pediatrics. 2014;134(4). Rubio-Codina M, Attanasio O, Meghir C, Varela N, Grantham-McGregor S. The socio-economic gradient of child development children 6-42 months in Bogota the socio-economic gradient of. J Human Res. 2015;50(2):464–83. Schady N, Behrman JR, Araujo MC, Azuero R, Bernal R, Bravo D, et al. Wealth Gradients in Early Childhood Cognitive Development in Five Latin American Countries. J Human Res. 2015;50(2):446–63. 2. 3. 4. 5. 6. 7. Van De Poel E, Hosseinpoor AR, Speybroeck N, Van Ourti T, Vega J. Socioeconomic inequality in malnutrition in developing countries. Bull World Health Organ. 2008;86(4):282–91. Naudeau S, Kataoka N, Valerio A, Neuman MJ, Elder LK. Investing in young children: an early childhood development guide for policy dialogue and project preparation. Washington, DC: The World Bank; 2011. 8. Walker SP, Chang SM, Vera-Hernandez M, Grantham-McGregor S. Early Childhood Stimulation Benefits Adult Competence and Reduces Violent Behavior. Pediatrics. 2011a;127(5):849–57. 9. Walker SP, Wachs TD, Grantham-Mcgregor S, Black MM, Nelson CA, Huffman SL, et al. Inequality in early childhood: Risk and protective factors for early child development. Lancet. 2011b;378(9799):1325–38. 10. Almond D, Currie J. Chapter 15 – Human capital development before age five. In: Handbook of Labor Economics, vol. 4; 2011. p. 1315–486. 11. Walker SP, Wachs TD, Meeks Gardner J, Lozoff B, Wasserman GA, Pollitt E, Carter JA. Child development: risk factors for adverse outcomes in developing countries. Lancet. 2007;369(9556):145–57 https://doi.org/10.1016/S0140-6736(07)60076-2. 12. Black RE, Victora CG, Walker SP, Bhutta ZA, Christian P, De Onis M, et al. Maternal and child undernutrition and overweight in low-income and middle-income countries. Lancet. 2013;382(9890):427–51. Lozoff B, Beard J, Connor J, Felt B, Georgieff M, Schallert T. Long-Lasting Neural and Behavioral Effects of Iron Deficiency in Infancy. Nutr Rev. 2006; 64(5 Pt 2):S34–43 discussion S72-91. 13. 14. West KP Jr, Darnton-Hill I. Vitamin A Deficiency. In: Semba RD, Bloem MW, 15. editors. Nutrition and Health in Developing Countries; 2001. p. 267–306. INE-Ministerio de Salud. Encuesta de Demografía y Salud EDSA 2016. La Paz. 2017. 16. Grantham-McGregor S, Ani C. A Review of Studies on the Effect of Iron 18. Deficiency on Cognitive Development in Children. J Nutr. 2001;131:649–68. 17. World Health Organization. WHO child growth standards: lenght/height-for- age, weight-for-age, weight-for-length, weight-for-height and body mass index-for-age: methods and development. Geneva: World Health Organization; 2006. Erhardt JG, Estes JE, Pfeiffer CM, Biesalski HK, Craft NE. Combined measurement of ferritin, soluble transferrin receptor, retinol binding protein, and C-reactive protein by an inexpensive, sensitive, and simple sandwich enzyme-linked immunosorbent assay technique. J Nutr. 2004;134(11):3127–32. Sommer A, Davidson FR. Assessment and Control of Vitamin A Deficiency: The Annecy Accords. J Nutr. 2002;75(0022):2857–2866S. Fernald LCH, Kariger P, Hidrobo M, Gertler PJ. Socioeconomic gradients in child development in very young children: Evidence from India, Indonesia, Peru, and Senegal. Proc Natl Acad Sci. 2012;109(Supplement 2):17273–80. 21. Rubio-Codina M, Araujo MC, Attanasio O, Muñoz P, Grantham-McGregor S. 19. 20. Concurrent validity and feasibility of short tests currently used to measure early childhood development in large scale studies. PLoS ONE. 2016a;11(8):1–17. 22. Rubio-Codina M, Attanasio O, Grantham-McGregor S. Mediating pathways in the socio-economic gradient of child development. Int J Behav Dev. 2016; 40(6):483–91. 23. O’Donnell O, Van Doorslaer E, Wagstaff A, Lindelow M. Analyzing health equity using household survey data: a guide to techniques and their implementation. Washington, DC: The World Bank; 2008. 24. Wagstaff A, Watanabe N. What difference does the choice of SES make in health inequality measurement? Health Econ. 2003;12(10):885–90. Lindelow M. Sometimes more equal than others: How health inequalities depend on the choice of welfare indicators. Health Econ. 2006;15(3):263–79. 25. 26. Houweling TAJ, Kunst AE, Mackenbach JP. Measuring health inequality among children in developing countries: Does the choice of the indicator of economic status matter? Int J Equity Health. 2003;2:1–12. 27. Rutstein SO, Johnson K. The DHS Wealth Index. DHS Comparative Reports No. 6; 2004. p. 1–71. 28. World Health Organization. Handbook on health inequality monitoring: with a special focus on low- and middle-income countries. Geneva: World Health Organization; 2013. 29. Wooldridge JM. Inverse probability weighted estimation for general missing data problems. J Econ. 2007;141:1281–301. 30. Campbell F, Conti G, Heckman JJ, Moon SH, Pinto R, Pungello E, Pan Y. Early Childhood Investments Substantially Boost Adult Health. Science. 2014; 343(6178):1478–85. 31. Gertler P, Heckman J, Pinto R, Zanolini A, Vermeersch C, Walker S, et al. Labor market returns to an early childhood stimulation intervention in Jamaica. Science. 2014;344(6187):998–1001. Celhay et al. International Journal for Equity in Health (2020) 19:122 Page 25 of 25 32. Krawinkel MB, Assad-García J. Estimación de los niveles de hierro y vitamina A en niños de 6 a 23 meses de edad en Bolivia. Reporte de la encuesta nutricional realizada de mayo a diciembre 2012 (unpublished report); 2012. 34. 33. Yip R. Iron Deficiency and Anemia. In: Semba R, Bloem MW, editors. Nutrition and Health in Developing Countries; 2001. p. 327–42. Johannsen J, Martinez S, Vidal C, Yarygina A. Evaluación de impacto del programa de desarrollo infantil temprano “Crecer Bien para Vivir Bien”: modalidad visitas domiciliarias (Technical Notes No. IDB-TN-1790). Washington, D.C; 2019. 35. Deaton A, Grosh M. Consumption. In: Grosh M, Glewwe P, editors. Designing Household Survey Questionnaires for Developing Countries: Lessons from Ten Years of LSMS Experience. Washington, DC: World Bank; 2000. p. 91–133. Filmer D, Pritchet L. Estimating Wealth Effects Without Expenditure Data—Or Tears: An Application to Educational Enrollments in States of India. Demography. 2001;38(1):115–32. 36. 37. Gakidou E, Oza S, Vidal Fuertes C, Li AY, Lee DK, Sousa A, et al. Improving Child Survival Through Environmental and Nutritional Interventions. The importance of Targeting Interventions Toward the Poor. Jama. 2007;298(16). 38. Vyas S, Kumaranayake L. Constructing socio-economic status indices: How to use principal components analysis. Health Policy Plan. 2006;21(6):459–68. 39. Gwatkin DR, Rutstein S, Johnson K, Suliman E, Wagstaff A, Amouzou A. Socio-economic differences in health, nutrition, and population within developing countries. An overview. Country reports on HNP and poverty. Washington, DC: The World Bank; 2007. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s13054-023-04505-7	Bianchi et al. Critical Care (2023) 27:207 https://doi.org/10.1186/s13054-023-04505-7 RESEARCH Critical Care Open Access Identification of an optimal threshold to define oliguria in critically ill patients: an observational study Nathan Axel Bianchi1,2, Marco Altarelli1, Céline Monard1, Tatiana Kelevina1, Aziz Chaouch3 and Antoine Guillaume Schneider1,2* Abstract Background The relevance of current consensus threshold to define oliguria has been challenged by small observa- tional studies. We aimed to determine the optimal threshold to define oliguria in critically-ill patients. Methods Cohort study including adult patients admitted within a multi-disciplinary intensive care unit between January 1st 2010 and June 15th 2020. Patients on chronic dialysis or who declined consent were excluded. We extracted hourly urinary output (UO) measurements along with patient’s characteristics from electronic medical records and 90-day mortality from the Swiss national death registry. We randomly split our data into a training (80%) and a validation (20%) set. In the training set, we developed multivariable models to assess the relationship between 90-day mortality and the minimum average UO calculated over time windows of 3, 6, 12 and 24 h. Optimal thresholds were determined by visually identifying cut-off values for the minimum average UO below which predicted mortality increased substantially. We tested models’ discrimination and calibration on the entire validation set as well as on a subset of patients with oliguria according to proposed thresholds. Results Among the 15,500 patients included in this analysis (training set: 12,440, validation set: 3110), 73.0% (95% CI [72.3–73.8]) presented an episode of oliguria as defined by consensus criteria (UO < 0.5 ml/kg/h for 6 h). Our models had excellent (AUC > 85% for all time windows) discrimination and calibration. The relationship between minimum average UO and predicted 90-day mortality was nonlinear with an inflexion point at 0.2 ml/kg/h for 3 and 6 h win- dows and 0.3 ml/kg/h for 12 and 24 h windows. Considering a threshold of < 0.2 ml/kg/h over 6 h, the proportion of patients with an episode of oliguria decreased substantially to 24.7% (95% CI [24.0–25.4]). Contrary to consensus definition, this threshold identified a population with a higher predicted 90-day mortality. Conclusions The widely used cut-off for oliguria of 0.5 ml/kg/h over 6 h may be too conservative. A cut-off of 0.2 ml/ kg/h over 3 or 6 h is supported by the data and should be considered in further definitions of oliguria. Keywords Acute kidney injury, Urinary output, Oliguria, Thresholds, ICU mortality, Critical illness Correspondence: Antoine Guillaume Schneider antoine.schneider@chuv.ch Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Bianchi et al. Critical Care (2023) 27:207 Page 2 of 10 Background Urinary output (UO) is the most obvious and accessible window to renal function in the intensive care unit (ICU). It is a major component of acute kidney injury (AKI) diagnosis and staging [1]. Low UO (oliguria) is associated with 90-day mortality irrespective of changes in serum creatinine (sCr) [2–4]. However, the interpretation of oliguria remains a daily challenge for clinicians through- out the world who struggle to differentiate a physiological adaptation to stress from kidney damage. According to consensus definition, a urinary output of less than 0.5 ml/ kg/h for a minimal duration of 6 h defines AKI[1]. This threshold is based on old physiological data demonstrat- ing that urine was maximally concentrated when its vol- ume output fell below 0.35 ml/min (21 ml/h) [5]. The clinical relevance of this definition, however, has recently been challenged [6–9]. In a series of > 3500 patients, intra-operative episodes of oliguria defined by a UO between 0.3 and 0.5 ml/kg/h were not associ- ated with post-operative AKI but only those defined by a UO < 0.3 ml/kg/h [10]. In 239 critically ill patients, olig- uria, defined by a cut-off of 0.5 ml/kg/h for > 4 h, had a very low positive predictive value (11%) for AKI (defined as an increased serum creatinine) [11]. In 725 critically ill patients, a threshold of 0.3 ml/kg/h for 6 h outper- formed the current definition to predict mortality or the need for dialysis [6]. Those findings were confirmed in a larger cohort study conducted in Finland, where, com- pared to standard thresholds, stricter (< 0.3 ml/kg/h for 6 h or < 0.1 ml/kg/h for 3 h) cut-offs increased the predic- tive values of oliguria for 90-day mortality [12]. Hence, the current definition of oliguria appears too conservative. However, the best cut-off to define oligu- ria remains to be determined. We aimed to define the best time weighted intensity cut-offs for UO to predict 90-day mortality in critically ill patients. Accordingly, we designed a retrospective observational study to assess the association between the minimum average UO and 90-day mortality, to identify potential cut-offs in this relationship and to assess their clinical relevance. we considered only the first eligible ICU admission. Data were extracted from electronic medical records [Metavision®(IMD Soft, Tel Aviv, Israel) and Soarian® (Cerner, North Kansas City, USA)]. In particular, we collected baseline characteristics, comorbidities, pre- admission weight, illness severity scores and hourly UO measurements. The primary outcome was 90-day mortal- ity. This outcome was assessed by cross-referencing our dataset with the Swiss national death registry. Continuous data are reported as mean (standard devia- tion, SD) or median (interquartile range, IQR) according to underlying data distribution. Categorical variables are expressed as number (percentage). All statistical analyses and modeling were carried out in R version 4.1.2 [13]. The level of statistical significance was set at 5%. Average standardized urinary output calculation vis ) by pre-admission We standardized hourly UO (ml/h) ( wi ) when available (missing values were body weight ( imputed using multiple imputations see below). We then used a sliding window to calculate the average standard- ized urinary output ( hours pre- ceding time t, such that vit ) of a patient i over d vit (d) = 1 d t s=t−d+1 vis wi For convenience, we restricted the analysis to slid- d = {3, 6, 12, 24} hours. Note that ing windows of width vit (d) cannot be calculated when . We then calcu- lated the minimum value among all moving averages of width d that can be computed over the whole ICU stay of each patient, that is ui(d) = min t≥d (vit (d)) t < d Hence, ui(d) corresponds to the minimal average UO that patient i experienced over a period of d hours during his ICU stay. Of note, we considered pre-admission body weight when available. Modeling 90‑day mortality Methods Design, setting, participants, data sources and statistics For this observational study, we have used a high-reso- lution cohort described in details elsewhere [2]. Briefly, we collected data from all adult (> 18 years old) patients admitted to our tertiary ICU between January 1st 2010 and June 15th 2020. We excluded patients with docu- mented or expressed wishes of non-participation to clinical research, those with end-stage renal disease (ESRD), whose stay lasted for less than 6 h or with miss- ing outcome or AKI defining data. For each patient, Logistic regression was used to predict 90-day mortal- ity as a function of the minimum average urine output of patients, separately for medical, scheduled surgical and unscheduled surgical admissions. Within each admis- sion type, control variables included patient’s age at ICU admission, SAPS II score (corrected to not account for daily UO) and Charlson comorbidities index. All predic- tors were continuous and flexibly modeled using penal- ized thin plate regression splines within the framework of generalized additive models [14, 15]. Alternative candidate models also included smooth terms for the Bianchi et al. Critical Care (2023) 27:207 Page 3 of 10 interaction between minimal average UO and other con- tinuous predictors. A model which did not include mini- mum average UO as predictor (“base model”) was also fitted for comparative purposes. The model featuring the highest overall calibration performance (i.e., lowest mean squared error of prediction) across all time windows was selected using tenfold cross-validation. For each value of d, predicted 90-day mortality was plotted as a function of the minimal average UO for fixed covariate patterns (e.g., corresponding to median values of control vari- ables). This allowed visual identification of suitable oligu- ria thresholds, i.e., UO thresholds below which mortality increases substantially, and compare these with thresh- olds used in current practice. Data were randomly split into a training (80%) and a validation set (20%). Training data were used to develop and fit prognostic models (including selection of the best model using tenfold cross-validation). Validation data were exclusively used to evaluate the final models’ dis- crimination and calibration properties. Discrimination was assessed using the area under the receiver operator characteristic curve (AU-ROC). Calibration was assessed using Hosmer–Lemeshow test [16] and calibration belt [17]. Discrimination and calibration performances were assessed on all validation data as well as on subsets of patients whose minimum average urine output fell below proposed thresholds for oliguria. Finally, we performed sensitivity analyses on the train- ing set, to assess the influence of body weight, gender, ill- ness severity and time periods on our results. Handling missing covariate data Missing covariate values were imputed using multiple imputations [18] with a total of 50 complete datasets being reconstructed. Multiple imputations were car- ried out separately within training and validation sets. Additional variables such as gender, body weight, base- line creatinine, need for noradrenaline were used in the wi was imputation process. Note that when body weight missing, imputed weights were used to calculate imputed ui(d) as well. Rubin rules [18] values for were used to pool predicted mortalities, AU-ROC esti- mates and/or calibration results obtained within each complete dataset. vit (d) and thus Urinary output is collected manually on an hourly basis by ICU nurses. The management of missing UO values is described in details in the Additional file 1. Ethics This study was approved by the Ethics Committee Vaud (CER-VD 2017-00008, Lausanne, Switzerland). In accord- ance with the Swiss Federal Act on Research involving Human Beings (article 34) [19], retrospective utilization of non-genetic health-related personal data was permit- ted, provided that the patient (or its legal representative) had not expressed wishes of non-participating to clini- cal research. This study followed “The Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) guidelines for reporting observational studies. Results Population Among the 18,314 patients admitted to our ICU dur- ing the study period. 2584 were excluded: 1116 declined consent for data utilization for research, 84 were under 18 years old, 747 had an ICU stay of less than 6 h, 217 had ESRD, 45 had no administrative data available, 305 had incomplete UO or sCr data and 70 had unknown 90-day mortality status. Therefore, our main analysis included 15,550 patients (Table 1). Patients were predominantly male (10,283, 66.1%) with a median (IQR) age of 65 (53.0–75.0) years, a median body weight of 75.0 (65.0– 87.0) kg, a median Charlson score of 4 (2.0–6.0) and a median corrected SAPS II score of 37 (28–48). Among those, 73.0% (95% CI [72.3–73.8]) presented a minimal average UO below 0.5 ml/kg/h for 6 h and therefore ful- filled criteria for oliguria according to consensus defini- tion. Clinical outcomes and missing values are reported in the Additional file 1 (Additional file 1: Tables S1 and S2). Pre-admission body weight was not available and had to be imputed in 3188 (20.5%) patients (Cf. multiple imputations above). Relationship between UO and 90‑day mortality The training and validation sets included data from 12,440 and 3110 patients, respectively (Table 1). Figure 1 illustrates the dependence between the mini- mum average UO (computed over time windows of 3, 6, 12 or 24 h) and 90-day predicted mortality in the differ- ent categories of admissions. Predicted mortality is esti- mated for median values of age, corrected SAPS II score and Charlson index. This figure suggests that adjusted 90-day mortality increases substantially in all categories of admission when the minimum average urine output is below 0.2 ml/kg/h over 3 or 6 h, or below 0.3 ml/kg/h over 12 or 24 h. Unadjusted data are presented in Addi- tional file 1: Figure S1. Similar associations and thresholds between minimum average UO and mortality were observed when using the same model to predict 30-day mortality (Additional file 1: Figure S2). Similarly, in sensitivity analyses, extreme patients’ weights (< 43 kg or > 130 kg; Additional file 1: Figure S3), gender (Additional file 1: Figures S4 and S5), SAPS II scores (Additional file 1: Figures S6 and S7) and time periods (Additional file 1: Figures S8, S9 and S10) did not appear to influence these results. Bianchi et al. Critical Care (2023) 27:207 Page 4 of 10 Table 1 Patients and ICU stay characteristics Demographics Male, n (%) Female, n (%) All patients (n = 15,550) Training set (n = 12,440) Validation set (n = 3110) P value 10,283 (66.1) 8221 (66.1) 2062 (66.3) Age at ICU admission, median (IQR), years Pre-admission body weight, median (IQR), kg Retrieved from medical records Imputed (multiple imputations) 65.0 (53.0, 75.0) 75.0 (65.0, 87.0) 65.0 (53.0, 75.0) 75.0 (65.0, 87.0) 12,362 (79.5) 3188 (20.5) 9897 (79.6) 2543 (20.4) 65.0 (53.0, 75.0) 75.0 (65.0, 87.0) 2465 (79.3) 645 (20.7) Baseline creatinine, median (IQR), μmol/L 69.0 (54.0, 90.0) 69.0 (53.0, 90.0) 70.0 (54.0, 91.0) Comorbidities Charlson score, median (IQR) Chronic kidney disease, n (%) Hypertension, n (%) Diabetes, n (%) Heart failure, n (%) Chronic obstructive pulmonary disease, n (%) Myocardial infarction, n (%) Chronic liver disease, n (%) Cancer, n (%) ICU admission characteristics Type of admission Medical Surgical scheduled Surgical unscheduled Main ICU diagnosis Cardiovascular surgery Other heart disease Septic shock Neurologic Trauma Acute respiratory insufficiency Cardiopulmonary arrest Other 4.0 (2.0, 6.0) 4.0 (2.0, 6.0) 4.0 (2.0, 6.0) 1610 (10.4) 7286 (47.0) 2926 (18.9) 3454 (22.3) 1806 (11.7) 5430 (35.0) 1188 (7.7) 2900 (18.7) 6534 (42.8) 4374 (28.7) 4345 (28.5) 3505 (23.1) 1963 (12.9) 1762 (11.6) 1614 (10.6) 1236 (8.1) 1313 (8.6) 897 (5.9) 2908 (19.1) 1255 (10.1) 5804 (46.8) 2307 (18.6) 2747 (22.2) 1455 (11.7) 4357 (35.1) 957 (7.7) 2312 (18.6) 5248 (43.0) 3487 (28.6) 3461 (28.4) 2812 (23.1) 1602 (13.2) 1387 (11.4) 1293 (10.6) 958 (7.9) 1061 (8.7) 731 (6.0) 2306 (19.0) 355 (11.5) 1482 (47.8) 619 (20.0) 707 (22.8) 351 (11.3) 1073 (34.6) 231 (7.5) 588 (19.0) 1286 (42.1) 887 (29.0) 884 (28.9) 693 (22.7) 361 (11.8) 375 (12.3) 321 (10.5) 278 (9.1) 252 (8.3) 166 (5.4) 602 (19.8) sCr at ICU admission, median (IQR), μmol/L 83.0 (65.0, 112.0) 83.0 (65.0, 112.0) 83.0 (65.0, 114.0) Mechanical ventilation within 24 h of ICU admission Noradrenaline within 24 h of ICU admission Modified SAPS II scorea, median (IQR) 9152 (58.9) 10,702 (68.8) 7346 (59.1) 8547 (68.7) 1806 (58.1) 2155 (69.3) 37.0 (28.0, 48.0) 37.0 (28.0, 48.0) 37.0 (28.0, 48.0) 0.84 0.50 0.37 0.84 0.07 0.19 0.03 0.33 0.09 0.45 0.54 0.60 0.65 0.70 0.63 0.08 0.32 0.33 0.54 0.97 Presented percentage exclude missing values. P value provided is for comparison between patients in the training set and the validation set. Number of missing values are reported in Additional file 1 ICU, intensive care unit; sCr, serum creatinine; SAPS, simplified acute physiology score a Modified SAPS II does not include points (0 to + 11) for urine output Proposal for a new definition of oliguria Based on these findings, we propose to consider a mini- mum average UO of < 0.2 ml/kg/h for 6 h as the new threshold to define oliguria. In our cohort, 24.7% [95% CI 24.0–25.4] of the patients fulfilled these revised criteria for oliguria (vs 73.0% [95% CI 72.3–73.8]) for consensus criteria). Unlike consensus criteria, this stricter definition is significantly associated with a higher 90-day adjusted mortality (OR 1.98 [95% CI 1.57–2.49] vs 1.27 [95% CI 0.95–1.70]). As shown in Fig. 2, patients with a minimum average UO of < 0.2 ml/kg/h for 6 h had a lower 3-month survival compared to other groups of patients through- out categories of illness severity and age, and this per- sisted at 12 months, (Additional file 1: Figure S11). Bianchi et al. Critical Care (2023) 27:207 Page 5 of 10 Fig. 1 Adjusted 90-day mortality as a function of the minimum average urine output for time windows of 3 h (a), 6 h (b), 12 h (c) and 24 h (d). Data is stratified by type of admission (medical and scheduled/unscheduled surgical admissions). Colored areas refer to 95% confidence intervals around the regression lines. Vertical dashed lines refer to thresholds below which the adjusted mortality increases substantially. Predictions are carried out for a fictive patient with continuous predictors fixed at their median value (i.e., 65 years old at ICU admission, corrected SAPS II score of 37 and Charlson index of 4) Confirmatory analyses Our prognostic models achieved good discrimination on validation data, with AU-ROC exceeding 85% for all time windows (Table 2). However, prior univariate analy- ses identified the corrected SAPS II score as the strong- est predictor of 90-day mortality, with AU-ROC already reaching 82%. As a consequence, only minor improve- ments in discrimination were observed when accounting for the minimal average urine output in prognostic mod- els over the base model (Table 2). When considering all validation data, calibration per- formance was very good (Additional file 1: Figure S12) both for the base model and the model including mini- mal average UO as a predictor of mortality. On the other hand, calibration belts presented in Fig. 3 suggest that, when restricting the attention to patients complying with the proposed oliguria definitions, the calibration of predicted mortalities improved substantially when accounting for the minimal average urine output. This was further supported by statistically significant (p < 0.001) results returned by Hosmer–Lemeshow tests on validation data for all models that omitted urine out- put from the list of predictors of mortality, suggesting that predicted probabilities of death derived from such models were miscalibrated in patients with oliguria. On the other hand, models which included UO as a predic- tor of mortality showed no evidence of miscalibration in these patients, with p values > 0.05 for 3, 6, 12 and 24 h time windows. Discussion Summary of key findings We conducted a large observational study including more than 15,000 patients admitted to a tertiary ICU during Bianchi et al. Critical Care (2023) 27:207 Page 6 of 10 Fig. 2 Kaplan–Meier 3-month survival curves according to minimum average urinary output (6 h windows). Data is stratified by tertiles of age and corrected SAPS score. Analyses are restricted to patients with available body weight (no imputation) n = 12,658. Urine output corresponds to the minimum average urinary output over a period of 6 h Table 2 Area under the receiver operating characteristic curve (AU-ROC) for all patients in the validation set as well as for validation patients complying with suggest oliguria thresholds Time window Suggested oliguria threshold (ml/kg/h) All validation patients Validation patients with oliguria Base model Full model Base model Full model 3 h 6 h 12 h 24 h < 0.2 < 0.2 < 0.3 < 0.3 85.3 85.3 85.3 85.3 86.1 85.9 86.1 86.1 82.1 80.0 82.5 79.8 83.0 81.5 83.7 82.0 Compared with base model that does not include minimum average urine output as predictor, the full model demonstrates only minimal increase in prediction as illustrated by minor improvement in AU-ROC across all time windows Bianchi et al. Critical Care (2023) 27:207 Page 7 of 10 Fig. 3 Calibration belts for patients in the validation set complying with the proposed oliguria definitions (UO < 0.2 ml/kg/h) for time windows of 3 h (a), 6 h (b), 12 h (c) and 24 h (d). Blue: base model (unadjusted for minimum average UO); red final model (adjusted for minimum average UO). Colored areas refer to 95% pointwise confidence intervals for observed mortality a 10-year period. We assessed the association between minimum average UO (calculated over time windows of 3, 6, 12 and 24 h) and 90-day mortality after controlling for age on admission, comorbidities and illness severity. When assessed on validation data, prognostic models including minimum average UO as a predictor of 90-day mortality resulted in good calibration and discrimina- tion. We observed a nonlinear relationship between minimum average UO and 90-day mortality. Our results suggest that the widely used cut-off for oliguria of 0.5 ml/ kg/h over 6 or 12 h may be too conservative. We found that values of 0.2 ml/kg/h (3 and 6 h time-windows) and 0.3 ml/kg/h (12 and 24 h time-windows) had higher prognostic implications. These values were similar for patients admitted for medical or surgical reasons. Comparisons with previous studies Only a handful of studies have evaluated the relation- ship between oliguria and clinical outcomes. Two have considered the occurrence of AKI (defined as a sCr rise) as an outcome [10, 11]. In those studies, oliguria, defined as a UO < 0.5 ml/kg/h for 6 h, only had a weak associa- tion with AKI and very limited positive predictive value. Intrinsically, however, this design implies that only a rise in sCr could define true AKI while UO only represents either an associated finding or a biomarker. This is in contradiction with recent data which suggest that oligu- ria might be associated with mortality irrespective of sCr rise [2, 3]. Therefore, patients’ centered outcomes such as mortality or need for renal replacement therapy (RRT) appear more suitable to assess the relevance of oliguria. Our data are in agreement with other smaller observa- tional studies conducted in other health systems such as New Zealand or Finland [6, 12]. Ralib et al. have deter- mined optimal UO thresholds for prediction of mortality or the need for RRT for time windows of 1 to 12 h. These cut-offs were derived from AUC-ROC curves at a univar- iate level. For the 3- 6- and 12-h time windows, identified Bianchi et al. Critical Care (2023) 27:207 Page 8 of 10 cut-offs were, respectively, 0.2, 0.3 and 0.5 ml/kg/h. Our analyses, based on a much larger dataset and modeling UO as a continuous predictor with smooth functional forms while accounting for age, comorbidities and sever- ity score, lead to slightly different values particularly at 12 h. However, we confirm the authors’ impression that current definition of oliguria is too liberal. In an analy- sis of the FINNAKI cohort, episodes of severe (< 0.1 ml/ kg/h) oliguria lasting more than 3 consecutive hours were independently associated with the development of sCr- AKI or RRT. In this analysis, consistent with our data, oliguria defined as a UO between 0.3 and 0.5 ml/kg/h for 6 h, was not significantly associated with adjusted 90-day mortality (OR 1.65 [95% CI 1.0–2.72]). The shortest peri- ods of consecutive oliguria associated with an increased risk for 90-day mortality on multivariate analyses were 6 h (for 0.1 to 0.3 ml/kg/h) and 3 h (for < 0.1 ml/kg/h). In contrast, an analysis from the MIMIC-2 cohort suggested that mortality increased rapidly as UO decreased < 0.5 ml/kg/h [20]. These analyses used mul- tivariate logistic regression analysis and contour plots. The difference in appreciation with our study is most likely explained by different modeling strategies for UO. Indeed, the authors have dichotomized UO whereas it was treated as a continuous predictor in the present study. Due to the nonlinear effect of UO on 90-day mor- tality (observed in our data), a dichotomous treatment of UO would likely lead to an overestimation of the optimal oliguria threshold. Indeed, a dichotomous threshold of 0.5 ml/kg/h to define oliguria would already lead to a dif- ference in estimated mortality between oliguric and non- oliguric patients while this mortality signal is only driven by the fraction of patients with UO < 0.2 ml/kg/h. Hence, we argue that treating the minimal average UO as a con- tinuous predictor is a preferable strategy to define mean- ingful oliguria thresholds. Finally, in contrast to previous studies of comparable size [12, 20], we have flexibly modeled UO as a continuous predictor of 90-day mortality. Our study, however, has limitations. As a single center study, the external validity of our findings may be chal- lenged. However, although relying on different types of statistical modeling, our findings are in line with previ- ous smaller or un-adjusted studies and can be viewed as confirmatory. We have elected to restrict our analyses to convenience time windows of 3, 6, 12 and 24 h. Perhaps different time windows would have provided different results. However, due to the similarity of data obtained across all reported time windows, this appears unlikely. In addition, proposed time windows are similar to those currently utilized (except for the 3 h window) and appear clinically relevant and applicable. We have not consid- ered other outcomes of potential importance such as increased sCr or need for RRT. Indeed, previous analyses have demonstrated that severe (stage 2 or 3) oliguria was an independent risk factor for mortality irrespective of sCr [2]. We believe that the demonstration of an impact on 90-day mortality is the most meaningful way to define oliguria. In addition, KDIGO criteria and the sCr crite- ria in particular were validated through their relevance on mortality. As this is a real-life study, we have not used an automated urimeter which might have improved the accuracy of the data collection. However, the low rate of missing values for UO suggests a high level of validity in our data. Lastly, UO was normalized to patients’ body weight, a parameter known to be highly inaccurate in critical illness. This was minimized by the primary con- sideration of pre-admission weight which was available in 79.5% of the patients and imputed for the remaining 20.5% of patients. In addition, sensitivity analyses sug- gested that extreme weights did not impact our results [21]. Strengths and limitations Implications for clinicians and policy makers This study is the largest (> 15,000 patients) attempt- ing to assess the relationship between minimal average UO and mortality. The sample includes the vast major- ity of patients admitted to a large multidisciplinary ICU. Hence, we were able to assess the influence of the type of admission (medical vs surgical) on the described rela- tionship. In addition, the large sample size enabled to separate our cohort into a derivation and validation sub- cohorts and test the discrimination and calibration of our models. We used all UO data points across the entire patient stay. Hence, our data is representative of the entire patients’ admission. The inclusion of illness sever- ity and comorbidity scores as well as age in our models enabled to control for the effect of important confound- ers on the relationship between UO and 90-day mortality. Our data provide an explanation for the perceived lack of relevance of oliguria as defined by a minimum aver- age UO of less than 0.5 ml/kg/h for 6 h. Indeed, this definition appears to be too liberal and to include many patients for which oliguria has no prognostic implication. By considering UO as a continuous predictor of 90-day mortality, we propose to revise the current definition of oliguria and use 0.2 ml/kg/h for 6 h as a cut-off. Our data further suggests that such cut-off might be applied indif- ferently to patients with medical and surgical (emergent and elective) admissions despite the fact that these pop- ulations have very different baseline adjusted mortality rates. Furthermore, we show that an increased mortality can already be identified when averaging UO over 3 h while Bianchi et al. Critical Care (2023) 27:207 Page 9 of 10 using the same cut-off for oliguria (< 0.2 ml/kg/h). Using a time window of three hours to monitor urine output may thus allow early identification of patients at risk of excess mortality. Conclusions In this large observational study, we confirmed that the current consensus threshold of 0.5 ml/kg/h over 6 h to define oliguria may be too conservative. A cut-off of 0.2 ml/kg/h over 3 or 6 h is supported by the data and should be considered in further definitions of oliguria. Abbreviations AKI AU-ROC ESRD ICU IQR SAPS sCr SD STROBE RRT UO Acute kidney injury Area under the receiver operator characteristic curve End-stage renal disease Intensive care unit Interquartile range Simplified acute physiology score Serum creatinine Standard deviation Strengthening the reporting of observational studies in epidemiology Renal replacement therapy Urinary output UO Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s13054- 023- 04505-7. Additional file 1. Supplementary Table 1 and 2 and Figures S1–S12. Acknowledgements The authors want to thank Isabelle Cristiani and the Laus’ I Care Research Team for her precious support in all regulatory aspects of this work. Author contributions NAB: participated in study design, acquired study data, analyzed the study results and wrote the first draft of the manuscript. MA: helped acquiring and managing study data and critically reviewed the manuscript. TK: helped acquiring and managing study data and critically reviewed the manuscript. AC: performed statistical analyses and critically reviewed the manuscript. AGS: participated in study design, supervised data collection, participated in data interpretation and analyses and critically reviewed the manuscript. All authors read and approved the final manuscript and agree to be personally account- able for their contribution. Funding Open access funding provided by University of Lausanne The study was funded by the intensive care unit research fund. Availability of data and materials The datasets used and/or analyzed as well as R code used in the current study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate The study protocol was approved by the Ethics Committee of Vaud, Switzer- land (CER-VD 2017-00008). Consent for publication Not applicable. Competing interests CM received a grant from the Societé Francaise d’Anesthésie-Réanimation (SFAR) and speaker’s fees from Fresenius Medical Care outside the submit- ted work. AGS received grants from the Leenaards Foundation and B Braun Avitum during the conduct of the study and speaking honorarium from Frese- nius Medical Care, CytoSorbents Corporation and Jafron. No other disclosures were reported. Author details 1 Adult Intensive Care Unit, Centre Hospitalier Universitaire Vaudois (CHUV), 46, Avenue du Bugnon, 1011 Lausanne, Switzerland. 2 Faculty of Biology and Medicine, University of Lausanne, Lausanne, Switzerland. 3 Department of Epidemiology and Health Systems, Center for Primary Care and Public Health (Unisanté), University of Lausanne, Lausanne, Switzerland. Received: 31 March 2023 Accepted: 23 May 2023 References 1. Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. KDIGO clinical practice guideline for acute kidney injury. Kidney Int Suppl. 2012;2:1–138. 2. Bianchi NA, Stavart LL, Altarelli M, Kelevina T, Faouzi M, Schneider AG. Association of oliguria with acute kidney injury diagnosis, severity assess- ment, and mortality among patients with critical illness. JAMA Netw Open. 2021;4:e2133094. 3. Qin J-P, Yu X-Y, Qian C-Y, Li S-S, Qin T-H, Chen E-Z, et al. Value of kidney disease improving global outcomes urine output criteria in critically ill patients: a secondary analysis of a multicenter prospective cohort study. Chin Med J (Engl). 2016;129:2050–7. 4. Vincent J-L, Ferguson A, Pickkers P, Jakob SM, Jaschinski U, Almekhlafi GA, et al. The clinical relevance of oliguria in the critically ill patient: analysis of a large observational database. Crit Care. 2020;24:171. 5. Chesley LC. Renal excretion at low urine volumes and the mechanism of oliguria. J Clin Investig. 1938;17:591–7. 6. Md Ralib A, Pickering JW, Shaw GM, Endre ZH. The urine output definition of acute kidney injury is too liberal. Crit Care. 2013;17:R112. 7. Klein SJ, Lehner GF, Forni LG, Joannidis M. Oliguria in critically ill patients: a narrative review. J Nephrol. 2018;31:855–62. 8. Koeze J, Keus F, Dieperink W, van der Horst ICC, Zijlstra JG, van Meurs M. Incidence, timing and outcome of AKI in critically ill patients varies with the definition used and the addition of urine output criteria. BMC Nephrol. 2017;18:70. 9. Allen JC, Gardner DS, Skinner H, Harvey D, Sharman A, Devonald MAJ. Definition of hourly urine output influences reported incidence and stag- ing of acute kidney injury. BMC Nephrol. 2020;21:19. 10. Mizota T, Yamamoto Y, Hamada M, Matsukawa S, Shimizu S, Kai S. Intra- operative oliguria predicts acute kidney injury after major abdominal surgery. Br J Anaesth. 2017;119:1127–34. 11. Prowle JR, Liu Y-L, Licari E, Bagshaw SM, Egi M, Haase M, et al. Oliguria as predictive biomarker of acute kidney injury in critically ill patients. Crit Care. 2011;15:R172. 12. Vaara ST, Parviainen I, Pettilä V, Nisula S, Inkinen O, Uusaro A, et al. Asso- ciation of oliguria with the development of acute kidney injury in the critically ill. Kidney Int. 2016;89:200–8. 13. R Core Team. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2020. 14. Hastie T, Tibshirani R. Generalized additive models. Stat Sci. 1986. https:// doi. org/ 10. 1214/ ss/ 11770 13604. full. 15. Wood SN. Generalized additive models: an introduction with R. 2nd ed. Boca Raton: Chapman and Hall; 2017. 16. Hosmer DW, Lemeshow S, Sturdivant RX. Applied logistic regression. 3rd ed. Hoboken: Wiley; 2013. 17. Finazzi S, Poole D, Luciani D, Cogo PE, Bertolini G. Calibration belt for quality-of-care assessment based on dichotomous outcomes. PLoS ONE. 2011;6:e16110. Bianchi et al. Critical Care (2023) 27:207 Page 10 of 10 18. Rubin DB. Multiple imputation for nonresponse in surveys. New York: Wiley; 1987. 19. Swiss Federal Act on Research involving Human Beings, HRA. 2011 Jan 30th. (810.30.4.34). 20. Mandelbaum T, Lee J, Scott DJ, Mark RG, Malhotra A, Howell MD, et al. Empirical relationships among oliguria, creatinine, mortality, and renal replacement therapy in the critically ill. Intensive Care Med. 2013;39:414–9. 21. Katayama S, Koyama K, Goto Y, Koinuma T, Tonai K, Shima J, et al. Body weight definitions for evaluating a urinary diagnosis of acute kidney injury in patients with sepsis. BMC Nephrol. 2018;19:101. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s13059-021-02387-y	Sherrill-Mix et al. Genome Biology (2021) 22:169 https://doi.org/10.1186/s13059-021-02387-y R E S E A R C H Open Access Detection of SARS-CoV-2 RNA using RT- LAMP and molecular beacons Scott Sherrill-Mix1,2, Young Hwang1, Aoife M. Roche1, Abigail Glascock1, Susan R. Weiss1, Yize Li1, Leila Haddad3, Peter Deraska3, Caitlin Monahan3, Andrew Kromer3, Jevon Graham-Wooten2, Louis J. Taylor1, Benjamin S. Abella4, Arupa Ganguly3, Ronald G. Collman2, Gregory D. Van Duyne5* and Frederic D. Bushman1* * Correspondence: vanduyne@ pennmedicine.upenn.edu; bushman@pennmedicine.upenn. edu 5Department of Biochemistry and Biophysics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA 1Department of Microbiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA Full list of author information is available at the end of the article Abstract Background: Rapid spread of SARS-CoV-2 has led to a global pandemic, resulting in the need for rapid assays to allow diagnosis and prevention of transmission. Reverse transcription-polymerase chain reaction (RT-PCR) provides a gold standard assay for SARS-CoV-2 RNA, but instrument costs are high and supply chains are potentially fragile, motivating interest in additional assay methods. Reverse transcription and loop-mediated isothermal amplification (RT-LAMP) provides an alternative that uses orthogonal and often less expensive reagents without the need for thermocyclers. The presence of SARS-CoV-2 RNA is typically detected using dyes to report bulk amplification of DNA; however, a common artifact is nonspecific DNA amplification, which complicates detection. Results: Here we describe the design and testing of molecular beacons, which allow sequence-specific detection of SARS-CoV-2 genomes with improved discrimination in simple reaction mixtures. To optimize beacons for RT-LAMP, multiple locked nucleic acid monomers were incorporated to elevate melting temperatures. We also show how beacons with different fluorescent labels can allow convenient multiplex detection of several amplicons in “single pot” reactions, including incorporation of a human RNA LAMP-BEAC assay to confirm sample integrity. Comparison of LAMP- BEAC and RT-qPCR on clinical saliva samples showed good concordance between assays. To facilitate implementation, we developed custom polymerases for LAMP- BEAC and inexpensive purification procedures, which also facilitates increasing sensitivity by increasing reaction volumes. Conclusions: LAMP-BEAC thus provides an affordable and simple SARS-CoV-2 RNA assay suitable for population screening; implementation of the assay has allowed robust screening of thousands of saliva samples per week. Keywords: COVID-19, SARS-CoV-2, Coronavirus, Loop-mediated isothermal amplification, Molecular beacon, LAMP-BEAC, RT-LAMP © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 2 of 17 Background Since its first detection in December 2019, the beta-coronavirus SARS-CoV-2 has spread around the world, at this writing infecting over 150 million people and causing over 3 million deaths. Frequent asymptomatic spread of this virus means that frequent, rapid, and affordable screening and surveillance testing are essential to controlling this pandemic [1–3]. Numerous methods have been developed to detect SARS-CoV-2 infection. The most common method is RT-qPCR to detect SARS-CoV-2 RNA [4]. RT-qPCR has the ad- vantage of providing accurate and sensitive detection, but supply chain issues have at times limited testing, motivating the development of additional methods using orthog- onal materials. RT-LAMP has been widely studied as an alternative [5–9]. LAMP assays use a “rolling hairpin” mechanism to allow amplification at a single temperature using polymerase enzymes different from those used for PCR, helping avoid supply chain bot- tle necks. In addition, RT-LAMP can be implemented on neat saliva, or on RNA puri- fied using simple reagents available in bulk [9], again helping bypass supply chain issues and adding robustness to assays. RT-LAMP assays are typically not as sensitive as RT-qPCR assays [1], but the import- ance of this varies with the application. Clinical diagnostic tests typically require high sensitivity; however, studies suggest that infected individuals are far more infectious during periods of peak viral loads, so methods for screening asymptomatic populations can be adequate even with lesser sensitivities [1, 2]. A recent study emphasized that fre- quency of testing and speed of reporting results are much more important than assay sensitivity for reducing transmission, emphasizing the value of assays like RT-LAMP that may be implemented efficiently and inexpensively [1]. However, a complication is that RT-LAMP reactions often result in non-specific amplification in the absence of target, particularly at longer reaction times, limiting sen- sitivity. This off-target amplification is especially problematic because LAMP reactions are commonly quantified using colorimetric or fluorescent dyes reporting only bulk DNA synthesis. To address these problems, improvements based on sequence-specific detection have been proposed such as incorporating DNA sequencing (LAMP-seq) [10] or CAS enzymes (DETECTR) [11]. These methods are promising, but as presently de- signed they typically require opening of RT-LAMP tubes and secondary manipulation of reaction products, which has the potential to result in contamination of subsequent reactions with amplification products from previous assays. In another detection method, a quencher-fluorophore duplex can be created by adding a short oligonucleo- tide complementary to a standard LAMP primer which is then displaced upon amplifi- cation [12–15]. These methods are more specific than bulk reporters but are still potentially vulnerable to false positives from spurious amplification. Previous research has shown the potential for molecular beacons [16] to allow sequence-specific detection of LAMP products in “single-pot” assays [17, 18]. Here, we adapt molecular beacons to detect SARS-CoV-2 sequences, a method we have named LAMP-BEAC (Fig. 1a). Molecular beacons are target-specific oligonucleotides labeled with a fluorophore on one end and a quencher on the other. The beacons are designed to incorporate complementary sequences on their 5′ and 3′ ends such that at low tem- peratures the ends anneal to form a hairpin, bringing the quencher and fluorophore into close proximity and quenching fluorescence. When the target of interest is present, Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 3 of 17 Fig. 1 LAMP-BEAC: RT-LAMP assayed using molecular beacons. a Product from a LAMP reaction is depicted emphasizing its loop region which forms single-stranded loops during amplification along with an example molecular beacon in its annealed hairpin form, which is quenched. Binding of the beacon to the target complementary sequence LAMP amplicon separates the fluorescent group and the quencher, allowing detection of fluorescence. The red loops on the beacon indicate locked nucleic acids used to increase binding affinity. b A genome map of SARS-CoV-2 showing the locations of primer binding sites for the LAMP primer sets used in this study. c Example of visual detection of LAMP-BEAC fluorescence using an orange filter with blue illumination. Multiplexed SARS-CoV-2-targeted Penn and human control STATH primer sets were used to amplify samples consisting of water or inactivated saliva with or without synthetic SARS-CoV-2 RNA (10,000 copies per reaction). Molecular beacons Penn_LF_S1 conjugated to a FAM fluorophore fluorescing green and Stath_LB_S2 conjugated to Cy3 fluorescing yellow were included in the reaction. The image was captured using the “Night Sight” mode of a Google Pixel 2 cell phone the complementary target-specific beacon sequence anneals to its target, separating the fluorophore from the quencher and greatly increasing the fluorescent signal. The bind- ing sites for beacons can be targeted to amplicon sequences not present in oligonucleo- tides used for priming, thereby enhancing specificity. The increase in fluorescence resulting from annealing of the beacon probe can be detected without manipulation of the product or opening the reaction tube. Here we describe (1) development of molecu- lar beacons for detection of SARS-CoV-2 RNA in LAMP-BEAC reactions, (2) develop- ment of a LAMP-BEAC method to detect human RNA to validate sample integrity, (3) combinations of LAMP-BEAC assays for single-pot multiplex detection, (4) develop- ment of custom polymerases allowing inexpensive expression and purification of re- quired enzymes, (5) use of LAMP-BEAC to screen infected subjects for viral RNA in saliva, and (6) increased sensitivity accessible using the high specificity of molecular beacons. Results Designing molecular beacons for SARS-CoV-2 RT-LAMP Several beacons were tested for detection of SARS-CoV-2 RNA in RT-LAMP reactions (Additional file 2: Table S1). Optimization required identifying sequence designs that performed properly under the conditions of the RT-LAMP reaction, which is typically run at temperatures around 65°C. Function of the beacon requires that the hairpin re- main mostly folded in the hairpin structure at this temperature, while still opening suf- ficiently often to allow annealing to the target RT-LAMP cDNA product. The annealed beacon-target cDNA duplex must then be sufficiently stable at 65°C to result in unquenching and an increase in fluorescence. To increase beacon affinity for use at Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 4 of 17 higher temperatures, we substituted multiple dNTP positions within the target se- quence of each beacon with locked nucleic acids [17]. Locked nucleic acids reduce the conformational flexibility of dNTPs and make the free energy of nucleic acid annealing more favorable [19]. We tested the performance of 28 molecular beacons using five previously reported SARS-CoV-2 (Fig. 1b) and three human control RT-LAMP ampli- cons (Additional file 1: Fig. S1, Additional file 2: Table S1). Testing LAMP-BEAC An example of a successful beacon design is Penn_LFMB_S1 (Additional file 2: Table S1). The RT-LAMP amplicon targets the orf1ab coding region and was first reported by El-Tholoth and coworkers at the University of Pennsylvania (named “Penn”) [7]. The favored beacon was designed to target sequences within the forward DNA loop generated during LAMP; thus, the beacon is designated Penn loop forward beacon, contracted to Penn_LFMB_S1. Detection can be accomplished with laboratory plate readers or PCR machines (below), and even visually with a simple blue light and orange filter (Fig. 1c). Figure 2 shows use of the Penn_LFMB_S1 system to detect synthetic SARS-CoV-2 RNA. Tests were carried out with commercial LAMP polymerase and reverse tran- scriptase preparations. In addition, to avoid possible supply chain problems and allow potential production of reagents in resource limited settings, we produced and purified novel DNA polymerase and reverse transcriptase enzymes, which were assayed in paral- lel with commercial preparations for some tests (described below). To compare standard LAMP amplification with LAMP-BEAC, reactions were pre- pared containing both fluorescent dye (Fig. 2a), which detects bulk DNA by inter- calation, and the molecular beacon Penn_LFMB_S1 (Fig. 2b). Reaction products were detected at two wavelengths, allowing separate quantification of the bulk dye and the molecular beacon in single reactions. The non-specific intercalating dye re- ported bulk DNA production in positive samples earlier than the water controls, but the negative controls did amplify shortly after. This spurious late amplification is commonly seen with RT-LAMP, though the mechanism is unclear. The primers may interact with each other to form products and launch amplification, or per- haps the reaction results from amplification of adventitious environmental DNA. In separate tests, synthesis of DNA products was shown to depend on addition of LAMP primers (data not shown). Molecular beacon Penn_LFMB_S1 in the same reactions showed more clear-cut dis- crimination (Fig. 2b). The positive samples showed positive signal, but no signal was detected for the negative water controls. Lack of amplification in negative controls has been reproducible over multiple independent reactions (examples below). The nature of the products could be assessed using thermal denaturation (Fig. 2c and d). Reactions were first cooled to allow full annealing of complementary DNA strands, then slowly heated while recording fluorescence intensity. The fluorescent started high but dropped with increasing signal of temperature in all samples (Fig. 2c), consistent with denaturation of the duplex and release of the intercalating dye into solution. In contrast, the beacon’s fluores- in the water controls started at low fluorescence (Fig. 2d), consistent cent signal intercalating dye the Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 5 of 17 Fig. 2 Reaction progression curves comparing RT-LAMP using the Penn primer set assayed using an intercalating dye and the Penn_LFMB_S1 molecular beacon in the same reactions. a Conventional RT-LAMP assay using non-specific dye to detect amplification of synthetic SARS-CoV-2 RNA diluted in water. Time after reaction initiation (x-axis) is compared to the relative fluorescence intensity (y-axis). The copy numbers of SARS-CoV-2 RNA in the reaction mixtures are shown in the key at the bottom. b Detection of the amplification of SARS-CoV-2 RNA using a LAMP-BEAC molecular beacon in the same reactions shown in A. Lines are colored as in A. c, d Thermal melting curves to characterize amplification products. The results shown are for the same reactions as in a and b. Reaction products were cooled to room temperature, then slowly heated for the melt curve analysis. c Characterization of the fluorescence intensity produced by non- specific intercalating dye (y-axis) with RT-LAMP end products over varying temperatures (x-axis). Lines are colored as in a. d Characterization of the fluorescence intensity produced by a LAMP-BEAC molecular beacon with RT-LAMP end products over varying temperatures. Markings as in c with annealing of the beacon DNA termini to form the hairpin structure (Fig. 1a). At temperatures above 70°C, the fluorescence modestly increased, consistent with opening of the hairpin and reptation of the beacon as a random coil in solution. For reactions containing the RT-LAMP product and Penn_LFMB_S1 beacon, fluor- escence values were high at lower temperatures, consistent with formation of the annealed duplex, then at temperature sufficient for denaturation, the fluorescence values fell to match those of the random coil (Fig. 2d). Thus, the LAMP-BEAC assay generates strong fluorescence signals during LAMP amplification in the pres- ence of target RNA but not in negative controls, and the thermal melting proper- ties are consistent with formation of the expected products. Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 6 of 17 Multiplex LAMP-BEAC assays We next sought to develop additional LAMP-BEAC assays to allow multiplex detection of SARS-CoV-2 RNA, and to allow parallel analysis of human RNA controls as a check on sample integrity, and so developed several additional beacons (Additional file 2: Table S1). E1_LBMB_S1 recognizes an amplicon targeting the viral E gene reported in [20], As1e_LBMB_S2 recognizes the As1e amplicon reported in [9] targeting the orf1ab coding region, N2_LBMB_S3 recognizes N2 amplicon reported in [20] targeting the N coding region, and N-A_LFMB_S2 recognizes N-A amplicon reported in [21] targeting the N coding region (Fig. 1b). We also developed positive control beacons, STATH_ LFMB_S1 and a later brighter iteration STATH_LBMB_S2, to detect a LAMP amplicon targeting the human statherin mRNA [22]. Additional control beacons included ACTB to detect beta-actin mRNA [20] and RNaseP to detect ribonuclease P subunit p20 POP7 mRNA or DNA [23] (Additional file 2: Table S1, Additional file 1: Figure S1). We chose to focus on STATH for further testing because it is abundantly expressed in the human saliva and spans an exon junction to allow selective detection of RNA and not DNA. To allow independent detection of each amplicon as a quadruplex assay, each beacon was labeled using fluorophores with different wavelengths of maximum emission. For example, E1_LBMB_S1 was labeled with FAM and detected at 520 nm, STATH_ LFMB_S1 was labeled with hexachlorofluorescein (Hex) and detected at 587 nm, As1e_ LBMB_S2 was labeled with Tex615 and detected at 623 nm, and Penn_LFMB_S1 was labeled with cyanine-5 (Cy5) and detected at 682 nm. This quadruplex LAMP-BEAC assay was tested with contrived samples, in which the saliva was doped with synthetic SARS-CoV-2 RNA (Fig. 3). Prior to dilution, the saliva was treated with TCEP and EDTA, followed by heating at 95°C, which inactivates both SARS-CoV-2 and cellular RNases [9], and so is part of our sample processing pipeline. The STATH_LFMB_S1 amplicon detected the human RNA control in all saliva sam- ples (Fig. 3a). The E1_LBMB_S1 and As1e_LBMB_S2 amplicons both consistently de- tected SARS-CoV-2 RNA down to ~250 copies per reaction (Fig. 3bc). Samples were called positive if either E1 or As1e showed amplification. Using this scoring method, the combination consistently detected SARS-CoV-2 down to 125 copies, and even de- tected 2/3 positives at 16 copies per reaction. The Penn_LFMB_S1 amplicon was least sensitive, detecting SARS-CoV-2 RNA con- sistently only at ~1000 copies per reaction (Fig. 3d). In the multiplex setting, the Penn amplicon sensitivity was lower than that observed when run in isolation (Fig. 2), likely indicating competition between amplicons during multiplexed reactions. Thus, the use of the Penn_LFMB_S1 assay in the multiplex format selectively reports particularly high RNA copy numbers. A useful feature of the STATH control amplicon used here is that it amplifies more slowly than the SARS-CoV-2 amplicons. Slower amplification of human controls is de- sirable to avoid exhaustion of reaction components due to competition, which could prevent viral detection. Melt curve analysis was also carried out to verify reaction products (Fig. 3e–h). Melt curve profiles were distinctive for each beacon, but the overall pattern included high fluorescence in the positive samples and low values in negative samples at lower tem- peratures, then convergence of positive and negative samples at high temperatures Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 7 of 17 Fig. 3 A multiplex LAMP-BEAC method assaying four amplicons. Assays were carried out using a LAMP primers and molecular beacon to detect human STATH RNA (a) and three primer sets and corresponding beacons to detect SARS-CoV-2: E1 (b), As1e (c), and Penn (d). For these assays, synthetic SARS-CoV-2 RNA was diluted into the saliva (inactivated as described [9]); copies per reaction are shown by the color code below the figures. a–d Amplification curves of the reactions showing fluorescent intensity (x-axis) over time (y-axis). e–h Endpoint melt curves of the reactions shown in a–d showing changes in fluorescent intensity (y-axis) as the temperatures (x-axis) of the final reaction products were raised from 25 to 95°C associated with full melting of the beacon and reptation in solution. The melt curve data for each beacon supported correct function and the expected structures of the amplification products. Assessing LAMP-BEAC performance on clinical saliva samples We next tested the LAMP-BEAC assay on a set of 82 saliva samples collected during surveillance for potential SARS-CoV-2 infection. Samples were from a clinical site, where subjects were tested by nasopharyngeal (NP) swabbing and clinical RT-qPCR, and also donated saliva for comparison. Saliva samples were treated with TCEP and EDTA and heated at 95°C for 5 min to inactivate RNase and SARS-CoV-2 [9]. We per- formed a triplex LAMP-BEAC assay using Penn_LFMB_S1, N2_LBMB_S3, and STAT H_LBMB_S2 beacons in a set of five 20 μl reactions. As an additional check, RNA was Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 8 of 17 purified from the same set of saliva samples and RT-qPCR carried out using the CDC- recommend N1 primer set. The absence of STATH amplification can indicate potential RNA degradation or in- hibitors in the sample, but another reason for lost signal can be competition between amplicons. STATH tends to amplify more slowly than the SARS-CoV-2-targeted ampli- cons and so can be suppressed by robust amplification of viral amplicons (Fig. 3). In practice, we suggest that a sample with SARS-CoV-2 amplification should be called as positive regardless of STATH results, a sample with no SARS-CoV-2 amplification should be called as negative if STATH amplification is observed, and a sample with no STATH amplification and no viral amplification should be called indeterminant. In these clinical samples, some degradation was apparent and STATH amplification was detected in only 56 out of 82 samples (Additional file 1: Figure S2, Additional file 3: Table S2). These saliva samples had been stored for months and frozen and thawed multiple times, so some attrition is not surprising. SARS-CoV-2 amplification was ob- served in 6 of these STATH failures suggesting potential competition between ampli- cons or a greater robustness of viral RNA. Where STATH or SARS-CoV-2 amplification was detectable, the LAMP-BEAC assay correlated perfectly with the amp- lification of SARS-CoV-2 above the limit of detection by laboratory RT-qPCR on the same saliva samples, i.e., a sensitivity and specificity of 1 (Fig. 4). Performance was simi- lar in quadruplex and duplex LAMP-BEAC assays using Penn_LFMB_S1, E1_LBMB_ S1, As1e_LBMB_S2, and STATH_LFMB_S1 performed on subsets of the same samples (Additional file 3: Table S2). Comparison to the results of clinical RT-qPCR testing on NP swabs from the same patients was complicated by disagreements with the laboratory RT-qPCR testing on matched saliva samples. Of the 24 samples scored as positive by clinical testing on NP Fig. 4 Validation of multiplexed LAMP-BEAC on 82 clinical saliva samples. Inactivated saliva samples were assayed by LAMP-BEAC using molecular beacons STATH_LBMB_S2, N2_LBMB_S3, and Penn_LFMB_S1. Samples were called “Positive” if they had detectable amplification in any SARS-CoV-2 amplicon; if human control STATH amplification was detected but not SARS-CoV-2, they were called “Negative”; if no STATH or SARS-CoV-2 amplification was detected they were called “Inconclusive.” SARS-CoV-2 targeted N2 and Penn fluorescence were quantified as the fold difference from a threshold set at two times the highest fluorescence observed in negative controls (dashed line). The maximum Penn or N2 fluorescence observed among the 5 reactions for each sample (y-axis) was compared to viral copy numbers estimates by laboratory qPCR (x-axis). Samples with qPCR copy numbers below the inferred limit of detection of 100 copies per μl are arbitrarily spaced apart for visualization. Sample integrity was assessed as STATH end point fluorescence greater than 120% the greatest fluorescence observed in water controls. Blue points indicate samples with detected STATH amplification and red indicates samples with no detectable STATH amplification, i.e., potentially indicating degraded samples or competition between amplicons Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 9 of 17 swabs, only 17 had detectable amplification by laboratory qPCR on the saliva and an additional 9 samples with detectable amplification by laboratory qPCR had been marked negative by clinical testing. All but one disagreement (see below) occurred in samples with concentrations inferred as less than 100 copies per microliter by labora- tory qPCR (clinical quantifications were not available). We thus inferred the laboratory qPCR had a practical limit of detection of 100 copies per microliter. The LAMP-BEAC assay did not detect amplification in any of these discrepant samples. For samples with greater than 100 inferred copies per microliter, the clinical test results and LAMP- BEAC agreed perfectly with the exception of a single saliva sample called positive by LAMP-BEAC but negative by clinical NP testing. This sample was also estimated at 200,000 viral RNA copies per microliter by laboratory RT-qPCR and as positive in 23 LAMP-BEAC amplifications in 14 separate reactions across 4 different primer sets (Additional file 3: Table S2). A recent study has documented differences between the loads of SARS-CoV-2 RNA at different body sites [24], including oral and nasal sites, potentially accounting at least in part for the observed differences. We note that the detection shown in Fig. 4, using end point fluorescence values and not reaction progression curves, offers a simplified read out for reaction results. That is, advanced qPCR machines are not needed for amplification or quantification of prod- uct formation using LAMP-BEAC, but rather reactions can be performed using a sim- ple heat block or incubator and reaction end points can be read out using a simpler fluorescent plate reader or even visual/cell phone detection (Fig. 1c). This may help by- pass possible supply chain bottlenecks and expenses associated with purchasing qPCR machines for SARS-CoV-2 assays. Laboratory-based production of polymerases required for RT-LAMP Polymerase enzymes are expensive and potentially subject to supply chain disruptions, so we engineered novel reverse transcriptase and DNA polymerase enzymes and de- vised simple purification protocols, allowing inexpensive local production of the re- quired enzymes. HIV-2 reverse transcriptase and the polA large fragment from Geobacillus stearothermophilus were each engineered to contain several amino acid substitutions expected to stabilize enzyme folding at higher temperatures (RT) or im- prove strand displacement activity (Bst). Enzymes were purified and tested as described in the methods. Side-by-side assays using lab-purified polymerases and commercial en- zyme preparations indicated that our novel polymerase enzymes are at least as efficient as commercial preparations (Additional file 1: Figure S3). Increased sensitivity through increased reaction volume The combination of affordable enzyme and low probability of false positives suggests that it could be possible to increase testing sensitivity by increasing reaction volume and sample input. To test this, we quantified sensitivity versus reaction volume using the N2 primer set, comparing detection with nonspecific dye and the N2_LBMB_S3 beacon. To test the relationship of reaction volume and sensitivity, we first compared the per- formance of 10 μl reactions with 4 μl of saliva input versus 20 μl reactions with 8 μl of the same saliva input. Samples were contrived using varying concentrations of synthetic Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 10 of 17 SARS-CoV-2 RNA in the inactivated saliva. We observed that the larger 20-μl reaction volume and correspondingly larger saliva input increased the detection rate for the mo- lecular beacon and non-specific dye (Fig. 5a–f). However, interpretation of results by non-specific dye was complicated by the variable distribution of cycle thresholds ob- served in negative controls, making a clean distinction of positive amplification difficult. In contrast, the LAMP-BEAC molecular beacon detected no sequence-specific amplifi- cation in any negative sample. The clear binary threshold provided by molecular bea- cons simplifies interpretation, enables endpoint detection, and suggests that sensitivity could be heightened by further increasing reaction volume. Fig. 5 Comparison of nonspecific dye and LAMP-BEAC performance in detecting SARS-CoV-2 RNA with increased reaction volume. Times to threshold was estimated in reactions amplifying synthetic SARS-CoV-2 RNA diluted in inactivated saliva with the N2 primer set with modified LB primer in total reaction volumes of 10 μl (a–c), 20 μl (d–f), and 200 μl (g–i). Time to threshold was calculated for both nonspecific dye (a, d, g) and molecular beacon N2_LBMB_S3 labeled with a cyanine-3 fluorophore (b, e, h) within each reaction. Dashed horizontal line indicates threshold used to call a well positive or negative; less than 20 min for nonspecific dye in 10 or 20 μl reactions, less than 17 min for nonspecific dye in 200 μl reactions and 80 min (the time the reactions were terminated) for the molecular beacon. For ease of comparison, dots are colored blue if called positive by nonspecific dye and molecular beacon, yellow if called positive by molecular beacon alone, and gray if called negative by both methods (no sample was called positive by nonspecific dye and negative by molecular beacon). Times to threshold for molecular beacon wells with no observed increase in fluorescence were arbitrarily set to 85 min. Points are offset slightly on the x-axis for visualization. g–i A comparison of the proportion of wells called positive in the reaction shown in a–b, d–e, and g–h by nonspecific dye and molecular beacon (blue) or molecular beacon alone (yellow) in 10 μl (g), 20 μl (h), or 200 μl (i) reactions (n=24 for each dilution) Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 11 of 17 We thus ran a series of 200 μl reactions with 80 μl of saliva input (Fig. 5g–i). Amplifi- cation detection by non-specific dye performed poorly in these large reaction volumes, with rapid amplification observed in all negative wells. With this strong background amplification, setting a threshold to call as positive was problematic. The fastest time to threshold in a negative well was 17 min while the fastest time to threshold in a positive well was 15 min, leaving little room for discrimination (Fig. 5g). Even using the unreal- istically tight threshold of calling any amplification detected in less than a 17-min posi- tive, nonspecific dye did not achieve high sensitivities in reactions with low copy numbers (Fig. 5i). In contrast, the N2_LBMB_S3 molecular beacon showed almost perfect discrimin- ation in the 200-μl reactions (Fig. 5h). No amplification was detected by molecular bea- con in any of the 24 negative reactions (4800 μl of total reaction volume) over the 80- min reaction time. In wells containing synthetic SARS-CoV-2 RNA, the molecular bea- con detected 100% of reactions containing 0.5 or 0.25 copies of RNA per μl of saliva in- put and 23/24 reactions containing 0.1 copy of RNA per μl of saliva (Fig. 5i). Note that even with these low target concentrations, the absence of signal in negative wells means that a real-time quantification is not necessary and simple endpoint read out is just as discriminative. Discussion Standard RT-LAMP is an attractive method for assay of SARS-CoV-2 RNA in patient samples due to the simplicity of the method and the use of a supply chain orthogonal to the clinical assay supply chain. However, conventional LAMP typically detects only the presence of amplified bulk DNA, and thus, assays can be complicated by nonspe- cific amplification. Improved specificity can be achieved by sequence-specific detection, and multiple methods have been proposed [10–15]. Here, we introduce a particularly convenient and effective method for sequence-specific detection of SARS-CoV-2 RNA in unpurified saliva using molecular beacons—LAMP-BEAC—that does not require manipulation of reaction products, can be carried out in a multiplex format in a “single tube,” greatly reduces the potential for false positives, and allows increased sensitivity. In simple small volume assays, the LAMP-BEAC method on straight saliva may not be as sensitive as RT-qPCR on purified RNA, but it can be implemented inexpensively, potentially allowing frequent population screening. The reaction set up and incubation can be done in less than an hour with a simple heat block, allowing rapid and high throughput turnaround. The assay can even be read out visually with quite inexpensive equipment (Fig. 1c), e.g., a ~$25 p51 viewer (from miniPCR). Thus, the LAMP-BEAC assay meets the needs articulated by modeling studies for effective surveys of asymp- tomatic populations [1]. When greater sensitivity is needed, a simple increase in reaction volume allows LAMP-BEAC to detect SARS-CoV-2 RNA down to a 0.1 copy per microliter of the sal- iva (Fig. 5). This high sensitivity did not require purification or concentration of RNA. Comparison of LAMP-BEAC to RT-qPCR showed better concordance between the RT-qPCR assay carried out on the same saliva samples than for RT-qPCR carried out on eluates from NP swabs. For the assays on the saliva, all samples with detectable STATH or SARS-CoV-2 amplification agreed between LAMP-BEAC and RT-qPCR, suggesting that both are similarly effective at identifying samples with higher viral RNA Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 12 of 17 copy numbers. The reason for divergence with some results for RT-qPCR on NP swabs is unknown; however, differences in viral RNA loads within patients at different body sites is well documented, possibly accounting for some differences [24]. Recently, Vogels et al. reported SalivaDirect, an RT-qPCR assay run on inactivated but unpurified saliva [25]. SalivaDirect uses a duplex single-tube analytical method, with one amplicon targeting SARS-CoV-2 RNA and another targeting a human RNA. This parallels our triplex LAMP-BEAC targeting viral RNA and human STATH RNA (Fig. 4). The LAMP-BEAC provides an effective complement to SalivaDirect since it does not rely on commercial enzyme mixes, potentially providing more resilience to possible supply chain disruptions, and can be carried out as an end-point assay using a heat block and fluorescent viewer, thus bypassing the need for quantitative real-time PCR machines. Additionally, LAMP-BEAC allows identification of variants of concern using targeted molecular beacons [26]. A protocol based on LAMP-BEAC has been implemented in a Clinical Laboratory Improvement Amendments (CLIA) certified laboratory and has been used to screen thousands of samples per week. To date more than 40 asymptomatic subjects have been identified as positive and referred for follow-up care. LAMP-BEAC thus enables rapid, affordable, and scalable screening programs. Conclusions Affordable, fast and robust testing is a necessity for the control of SARS-CoV-2 and fu- ture pandemics. LAMP-BEAC meets all these criteria while allowing sensitive detection when needed, all while using simple isothermal amplification and supply chains inde- pendent of commercial qPCR assays. Methods Design of molecular beacons Beacons were designed to detect amplification product generated using previously pub- lished LAMP primer sets. To design beacons targeting the loop region of the LAMP product, we mapped the FIP and BIP primers to the SARS-CoV-2 genome to find the entire forward and backward loop regions of the amplicon (potentially including re- gions outside the original LF and LB primers). We then selected the most GC-rich sub- sequences within these loops and selected bases for LNA modification based on the predicted change in melting temperature using a stepwise greedy heuristic of consecu- tively adding the LNA with the highest predicted Tm. Additional nucleotides were then added to the 5′ and 3′ ends, avoiding strings of 4 guanine or a guanine next to the fluorophore, to form a hairpin with predicted melting temperature between 57–65°C. Melting temperatures were predicted using OligoAnalzyer v3.1 (IDT), where possible terminal bases of the target sequence were used as part of the hairpin. To allow easy and relatively affordable synthesis, beacons were kept shorter than 25 nt with 6 locked nucleic acids. With these constraints, commercial synthesis (IDT) provided an average yield ~50 nmoles at a cost of ~$400 giving a final cost per 15-μl reaction of US $0.03. Local synthesis was also implemented on a BioAutomation MerMade 4 oligonucleotide synthesizer in case of supply chain failure. Successful design often required several Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 13 of 17 iterations. These earlier iterations are listed in Table S1, and their fluorescence charac- teristics are compared in Figure S1. As an additional optimization, we created a new N2 loop primer to reduce overlap with molecular beacons. The loop primers of the N2 primer set span almost the entir- ety of the loop regions and thus present the possibility for false amplification involving the primer to generate targets for beacons landing in the same region. As a work around, we shortened the backward loop primer while maintaining binding affinity using locked nucleic acids (N2_LB_2, Additional file 2: Table S1) and used this modi- fied primer set for further testing. Design and purification of polymerases We initially chose the D720A mutant of Geobacillus stearothermophilus PolA for LAMP due to the high-strand displacement activity observed for a closely related polymerase [27]. Using strain DSM 13240, the coding sequence corresponding to the large polymerase fragment (Bst-LF; residues 290–878) was amplified from gen- omic DNA, ligated into CDFDuet (Novagen) in-frame with an N-terminal hexahis- tidine tag and the D720A substitution was incorporated. To explore alternative Bst-LF variants for LAMP, we generated the R433A and R433P variants, each of which results in disruption of the salt-bridge formed with Asp720 in the wild-type enzyme (protein data bank: 1XWL). Bst-LF was expressed in strain BL21(DE3) at 37°C with 2xYT medium and IPTG in- duction for 3 h. Pelleted cells were stored at −70°C prior to purification. Cells were lysed using an Avestin cell disrupter in a 50-mM sodium phosphate, pH 8, 300-mM NaCl, 2-mM MgCl2, 5-mM 2-mercaptoethanol, and protease inhibitors. After centrifu- gation, the cleared lysate was purified at 4°C using a 5-ml Talon column (Clontech) fol- lowing the manufacturer’s protocol. Eluted fractions from Talon were diluted 1:1 with buffer HepA (20 mM Tris-HCl, pH 7.4, 5 mM MgCl2, 10 mM 2-mercaptoethanol) and purified on an 8 mL heparin sepharose column (GE), and at 20°C using a 0–500-mM NaCl gradient. Heparin fractions were pooled and dialyzed overnight vs 20 mM Tris- HCl, pH 8, 50 mM NaCl, and 10 mM 2-mercaptoethanol followed by anion exchange chromatography at 20°C using an 8-ml MonoQ (GE) column. Bst-LF eluted as sharp peaks from a 15–30% sodium chloride gradient in buffer containing 20 mM TrisHCl, pH 8, and 10 mM 2-mercaptoethanol. Purified Bst-LF mutants were concentrated using Millipore centrifugal concentrators, glycerol added to 10%, and aliquots were flash fro- zen and stored at −80°C. Primer extension assays using M13 DNA template and 3H- dTTP-labeled dNTPs were used to establish specific activity as described for commer- cially prepared Bst (NEB). To demonstrate that RT-LAMP can be performed using a reverse transcriptase gen- erated in-house, we first constructed a synthetic gene for the HIV1 RT p66 (strain NL4-3) subunit containing substitutions expected to confer thermal stability (RTx; NEB). The p66 sequence was inserted into pET29b, and the p51 subunit coding se- quence was amplified by PCR and inserted in frame with an N-terminal hexahistidine tag in CDFDuet. An alternative RT (RT2m) was produced using a similar approach with HIV2 RT as the template (Genbank AAB25033), where thirteen naturally occur- ring substitutions were incorporated. The full-length subunit was inserted into Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 14 of 17 pCDFDuet, and the smaller subunit was fused to a C-terminal hexahistidine tag after Thr436 in pETDuet. For both RTs, the subunits were co-expressed in BL21(DE3) cells and 2xYT media, with IPTG induction at 20°C for 5h. Pelleted cells were stored at −70°C prior to purifi- cation. Cells were lysed using an Avestin cell disrupter in a 50-mM sodium phosphate, pH 8, 300-mM NaCl, 2-mM MgCl2, 5-mM 2-mercaptoethanol, and protease inhibitors. After centrifugation, the cleared lysate was purified at 4°C using a 5-ml Talon column following the manufacturer’s protocol. Eluted fractions from Talon were dialyzed vs buffer A (20 mM Tris-HCl, pH 7.4, 50 mM NaCl, 5 mM MgCl2, 10 mM 2- mercaptoethanol) and purified on an 8-mL heparin sepharose column at 20°C using a 150–400-mM NaCl gradient in buffer A. Heparin fractions were pooled and concen- trated on 5-ml Ni-NTA resin (Qiagen), followed by elution with 20 mM TrisHCl, 100 mM NaCl, 10 mM 2-mercaptoethanol, 250 mM imidazole, and pH 8. Purified RT was dialyzed vs 20 mM Tris-HCl pH 8, 50 mM NaCl, 10 mM 2-mercaptoethanol, and con- centrated using Millipore centrifugal concentrators. Glycerol was added to 10%, and ali- quots were flash frozen and stored at −80°C. Primer extension assays using poly-A template and 3H-labeled dTTP were used to determine specific activity at 50°C as de- scribed for commercial RTx (NEB). RT-LAMP reaction mixtures RT-LAMP reactions were prepared by mixing 7.5 μl commercial 2x LAMP master mix (NEB E1700L) or our own LAMP mix (40 mM TrisHCl, pH 8.5, 20 mM (NH4)2SO4, 100 mM KCl, 16 mM MgSO4, 0.2% Tween-20, 2.8 mM each dNTP, 16 μg/ml polA LF, and 2.6–7.7 μg/ml RT) with 1.5 μl of 10x primer/beacon master mix (final concentra- tion: 1.6 μM FIP/BIP, 0.2 μM F3/B3, 0.4 μM LF/LB, 0.25 μM beacon) and 6 μl of sam- ple and/or water. Larger 20 and 200 μl and smaller 10 μl reactions were scaled proportionally. For multiplexed LAMP reactions, total concentration of primers was preserved while maintaining the individual beacon concentrations, e.g., when two primer sets were added to the same reaction, the individual primer concen- trations were halved while beacon concentrations remained at 0.25 μM. For non- specific amplification detection, LAMP fluorescent dye (NEB B1700S) was used at a 100–300x final dilution. the final Assays using LAMP-BEAC LAMP-BEAC reactions were performed at 60–65°C with fluorescent quantification every 30 s on a ThermoFisher QuantStudio 5 or ThermoFisher QuantStudio 6. Reac- tions typically completed within 45 min, but for research purposes, data was collected for additional time spans. The synthetic SARS-CoV-2 RNA used as a standard during assay development was obtained from Twist (MT007544.1). After reaction completion, for melt curve analysis, the reaction was heated to 95°C for 5 min to inactivate any remaining enzyme, cooled to 25°C (at a rate of 0.2°C/s) and then slowly heated to 95°C with fluorescence measured every degree. Time to threshold was calculated as the time required for fluorescence from nonspe- cific dye detected at 520 nm to reach 200,000 relative fluorescence units (RFU) for 10 or 20 μl reactions higher than baseline or 400,000 RFU for 200 μl reactions. For Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 15 of 17 fluorescence from molecular beacon detected at 587 nm, time to threshold was calcu- lated as the time required to reach 20,000 RFU higher than baseline for 10 or 20 μl re- actions or 40,000 RFU for 200 μl reactions. Baseline was calculated as the average RFU over the first 2.5 min of the reaction for 10 or 20 μl reactions or the average over the second 2.5 min for 200 μl reactions to allow the reactions to completely equilibrate. RT-qPCR to characterize saliva samples RNA was extracted from ~140 μl saliva using the Qiagen QIAamp Viral RNA Mini Kit. The RT-qPCR assay used the CDC 2019-nCoV_N1 primer-probe set (2019-nCoV_N1-F: GACC 2019-nCoV_N1-R: TCTGGTTACTGCCAGTTGAATCTG, CCAAAATCAGCGAAAT, 2019_nCoV_N1-P: FAM-ACCCCGCATTACGTTTGGTGGACC-IBFQ). The RT-qPCR master mix contained: 8.5 μl dH2O, 0.5 μl N1-F (20 μM), 0.5 μl N1-R (20 μM), 0.5 μl N1-P (5 μM), and 5.0 μl TaqMan™ Fast Virus 1-Step Master Mix per reaction. Five microliters of the extracted RNA was added to 15 μl of the prepared master mix for a final volume of 20 μl per reaction. The final concentrations of both 2019-nCoV_N1-F and 2019-nCoV_N1-R primers were 500nM, and the final concentration of the 2019-nCoV_N1-P probe was 125nM. The assay was performed using a ThermoFisher QuantStudio 5. The thermocycler conditions were 5 min at 50°C, 20 s at 95°C, and 40 cycles of 3 s at 95°C and 30 s at 60°C. Abbreviations RT-qPCR: Reverse transcription quantitative polymerase chain reaction; COVID-19: Coronavirus disease from 2019; SARS- CoV-2: Severe acute respiratory syndrome coronavirus number two; RT-LAMP: Reverse transcription loop-mediated iso- thermal amplification; RFU: Relative fluorescence units Supplementary Information The online version contains supplementary material available at https://doi.org/10.1186/s13059-021-02387-y. Additional file 1: Figures S1-S3. Additional file 2: Table S1. Oligonucleotides used in the LAMP-BEAC assay. Additional file 3: Table S2. Clinical samples and results of assays for SARS-CoV-2 and human RNA. Additional file 4. Review history. Acknowledgements We are grateful to the members of the Bushman, Van Duyne, and Collman laboratories for their help and suggestions. We acknowledge the assistance of the Penn Medicine BioBank, including JoEllen Weaver and Daniel Rader. This work was supported by the Penn Center for Research on Coronaviruses and Other Emerging Pathogens. Review history The review history is available as Additional file 4. Peer review information Andrew Cosgrove was the primary editor of this article and managed its editorial process and peer review in collaboration with the rest of the editorial team. Authors’ contributions SS-M, BDA, RGC, GVD, and FDB designed the study; JG-W, LJT, BSA, and RGC collected clinical specimens; SS-M, YH, AMR, AGl, AGa, LH, PD, CN, AK, and GVD carried out assays; SRW and YL grew SARS-CoV-2 in tissue culture; SS-M, GVD, and FDB wrote the paper. The authors read and approved the final manuscript. Funding This work was supported by the Center for Research on Coronaviruses and Other Emerging Pathogens. Availability of data and materials All data newly generated in this study is disclosed in the published manuscript. The plasmids used for local production of LAMP enzymes have been deposited on Addgene (#170277, 170278, 170279). Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 16 of 17 Declarations Ethics approval and consent to participate All sample collection was carried out under IRB-approved protocols (IRB protocol #842613 and #813913) and complied with the Helsinki Declaration. Salivary samples were collected from possible SARS-CoV-2-positive patients at one of three locations: (1) Penn Presbyterian Medical Center Emergency Department, (2) Hospital of the University of Pennsyl- vania Emergency Department, and (3) Penn Medicine COVID-19 ambulatory testing center. Inclusion criteria including any adult (age>17 years) who underwent SARS-CoV-2 testing via standard nasopharyngeal swab at the same visit. Pa- tients with known COVID-19 disease who tested positive previously were excluded. After verbal consent was obtained by a trained research coordinator, patients were instructed to self-collect the saliva into a sterile specimen container which was then placed on ice until further processing for analysis. Consent for publication All authors have reviewed the manuscript and consented to allow publication. Competing interests The authors declare that they have no competing interests. Author details 1Department of Microbiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA. 2Department of Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA. 3Department of Genetics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA. 4Department of Emergency Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA. 5Department of Biochemistry and Biophysics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA. Received: 6 October 2020 Accepted: 24 May 2021 References 1. Larremore DB, Wilder B, Lester E, Shehata S, Burke JM, Hay JA, et al. Test sensitivity is secondary to frequency and turnaround time for COVID-19 screening. Sci Adv. 2021;7(1). https://doi.org/10.1126/sciadv.abd5393. 2. Mina MJ, Andersen KG. COVID-19 testing: one size does not fit all. Science. 2021;371(6525):126–7. https://doi.org/10.112 3. 4. 5. 6. 7. 8. 9. 10. 6/science.abe9187. Lyng GD, Sheils NE, Kennedy CJ, Griffin DO, Berke EM. Identifying optimal COVID-19 testing strategies for schools and businesses: balancing testing frequency, individual test technology, and cost. PLoS One. 2021;16(3):e0248783. https://doi. org/10.1371/journal.pone.0248783. Lieberman JA, Pepper G, Naccache SN, Huang ML, Jerome KR, Greninger AL. Comparison of commercially available and laboratory-developed assays for in vitro detection of SARS-CoV-2 in clinical laboratories. J Clin Microbiol. 2020;58(8). https://doi.org/10.1128/JCM.00821-20. Lamb LE, Bartolone SN, Ward E, Chancellor MB. Rapid detection of novel coronavirus/severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) by reverse transcription-loop-mediated isothermal amplification. PLoS One. 2020;15(6): e0234682. https://doi.org/10.1371/journal.pone.0234682. Lee SH, Baek YH, Kim YH, Choi YK, Song MS, Ahn JY. One-pot reverse transcriptional loop-mediated isothermal amplification (RT-LAMP) for detecting MERS-CoV. Front Microbiol. 2016;7:2166. El-Tholoth M, Bau HH, Song J. A single and two stage, closed-tube, molecular test for the 2019 novel coronavirus (COVID-19) at home, clinic, and points of entry. chemrxiv. 2020. Butler D, Mozsary C, Meydan C, Foox J, Rosiene J, Shaiber A, et al. Shotgun transcriptome, spatial omics, and isothermal profiling of SARS-CoV-2 infection reveals unique host responses, viral diversification, and drug interactions. Nat Commun. 2021;12(1):1660. https://doi.org/10.1038/s41467-021-21361-7. Rabe BA, Cepko C. SARS-CoV-2 detection using isothermal amplification and a rapid, inexpensive protocol for sample inactivation and purification. Proc Natl Acad Sci U S A. 2020;117(39):24450–8. https://doi.org/10.1073/pnas.2011221117. Schmid-Burgk JL, Schmithausen RM, Li D, Hollstein R, Ben-Shmuel A, Israeli O, et al. Lamp-Seq: population scale COVID- 19 diagnostics using combinatorial barcoding. bioRxiv. 2020. 11. Broughton JP, Deng X, Yu G, Fasching CL, Servellita V, Singh J, et al. CRISPR-Cas12-based detection of SARS-CoV-2. Nat Biotechnol. 2020;38(7):870–4. https://doi.org/10.1038/s41587-020-0513-4. 12. Ball CS, Light YK, Koh CY, Wheeler SS, Coffey LL, Meagher RJ. Quenching of unincorporated amplification signal reporters in reverse-transcription loop-mediated isothermal amplification enabling bright, single-step, closed-tube, and multiplexed detection of RNA viruses. Anal Chem. 2016;88(7):3562–8. https://doi.org/10.1021/acs.analchem.5b04054. Tanner NA, Zhang Y, Evans TC Jr. Simultaneous multiple target detection in real-time loop-mediated isothermal amplification. Biotechniques. 2012;53(2):81–9. https://doi.org/10.2144/0000113902. 13. 14. Zhang Y, Tanner NA. Development of multiplexed reverse-transcription loop-mediated isothermal amplification for detection of SARS-CoV-2 and influenza viral RNA. Biotechniques. 2021;70(3):167–74. https://doi.org/10.2144/btn-2020-0157. 15. Bektas A, Covington MF, Aidelberg G, Arce A, Matute T, Nunez I, et al. Accessible LAMP-enabled rapid test (ALERT) for detecting SARS-CoV-2. Viruses. 2021;13(5):742. 16. Antony T, Subramaniam V. Molecular beacons: nucleic acid hybridization and emerging applications. J Biomol Struct Dyn. 2001;19(3):497–504. https://doi.org/10.1080/07391102.2001.10506757. 17. Bakthavathsalam P, Longatte G, Jensen SO, Manefield M, Gooding JJ. Locked nucleic acid molecular beacons for multiplex detection of loop mediated isothermal amplification. Sensors Actuators B Chem. 2018;268:255–63. https://doi. org/10.1016/j.snb.2018.04.081. Sherrill-Mix et al. Genome Biology (2021) 22:169 Page 17 of 17 18. Liu W, Huang S, Liu N, Dong D, Yang Z, Tang Y, et al. Establishment of an accurate and fast detection method using molecular beacons in loop-mediated isothermal amplification assay. Sci Rep. 2017;7(1):40125. https://doi.org/10.1038/ srep40125. 19. Petersen M, Wengel J. LNA: a versatile tool for therapeutics and genomics. Trends Biotechnol. 2003;21(2):74–81. https:// doi.org/10.1016/S0167-7799(02)00038-0. 20. Zhang Y, Ren G, Buss J, Barry AJ, Patton GC, Tanner NA. Enhancing colorimetric loop-mediated isothermal amplification speed and sensitivity with guanidine chloride. Biotechniques. 2020;69(3):178–85. https://doi.org/10.2144/btn-2020-0078. 21. Zhang Y, Odiwuor N, Xiong J, Sun L, Nyaruaba RO, Wei H, et al. Rapid molecular detection of SARS-CoV-2 (COVID-19) 22. virus RNA using colorimetric LAMP. medRxiv. 2020. Satoh T, Kouroki S, Ogawa K, Tanaka Y, Matsumura K, Iwase S. Development of mRNA-based body fluid identification using reverse transcription loop-mediated isothermal amplification. Anal Bioanal Chem. 2018;410(18):4371–8. https://doi. org/10.1007/s00216-018-1088-5. 23. Curtis KA, Morrison D, Rudolph DL, Shankar A, Bloomfield LSP, Switzer WM, et al. A multiplexed RT-LAMP assay for detection of group M HIV-1 in plasma or whole blood. J Virol Methods. 2018;255:91–7. https://doi.org/10.1016/j. jviromet.2018.02.012. 24. Wyllie AL, Fournier J, Casanovas-Massana A, Campbell M, Tokuyama M, Vijayakumar P, et al. Saliva or nasopharyngeal swab specimens for detection of SARS-CoV-2. N Engl J Med. 2020;383(13):1283–6. https://doi.org/10.1056/NEJMc2016359. 25. Vogels CBF, Watkins AE, Harden CA, Brackney DE, Shafer J, Wang J, et al. SalivaDirect: a simplified and flexible platform 26. to enhance SARS-CoV-2 testing capacity. Med (N Y). 2021;2:263–80 e266. Sherrill-Mix S, Van Duyne G, Bushman FD. Molecular beacons allow specific RT-LAMP detection of B.1.1.7 variant SARS- CoV-2. medRxiv. 2021. 27. Piotrowski Y, Gurung MK, Larsen AN. Characterization and engineering of a DNA polymerase reveals a single amino-acid substitution in the fingers subdomain to increase strand-displacement activity of A-family prokaryotic DNA polymerases. BMC Mol Cell Biol. 2019;20(1):31. https://doi.org/10.1186/s12860-019-0216-1. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s12917-019-1890-0	Lambert et al. BMC Veterinary Research (2019) 15:135 https://doi.org/10.1186/s12917-019-1890-0 R E S E A R C H A R T I C L E Open Access Impact of alignment algorithm on the estimation of pairwise genetic similarity of porcine reproductive and respiratory syndrome virus (PRRSV) Marie-Ève Lambert1,2* Sylvie D’Allaire1,2 , Julie Arsenault1,2, Benjamin Delisle1,2, Pascal Audet1,2, Zvonimir Poljak3 and Abstract Background: Porcine reproductive and respiratory syndrome (PRRS) is a major threat to the swine industry. It is caused by the PRRS virus (PRRSV). Determination and comparison of the nucleotide sequences of PRRSV strains provides useful information in support of control initiatives or epidemiological studies on transmission patterns. The alignment of sequences is the first step in analyzing sequence data, with multiple algorithms being available, but little is known on the impact of this methodological choice. Here, a study was conducted to evaluate the impact of different alignment algorithms on the resulting aligned sequence dataset and on practical issues when applied to a large field database of PRRSV open reading frame (ORF) 5 sequences collected in Quebec, Canada, from 2010 to 2014. Five multiple sequence alignment programs were compared: Clustal W, Clustal Omega, Muscle, T-Coffee and MAFFT. Results: The resulting alignments showed very similar results in terms of average pairwise genetic similarity, proportion of pairwise comparisons having ≥97.5% genetic similarity and sum of pairs (SP) score, except for T-Coffee where increased length of aligned datasets as well as limitation to handle large datasets were observed. Conclusions: Based on efficiency at minimizing the number of gaps in different dataset sizes with default open gap values as well as the capability to handle a large number of sequences in a timely manner, the use of Clustal Omega might be recommended for the management of PRRSV extensive database for both research and surveillance purposes. Keywords: Porcine reproductive and respiratory syndrome virus, PRRS, Alignment algorithm, Sequence, Genetic similarity Background Porcine reproductive and respiratory syndrome virus (PRRSV) infection has a major economic impact on the swine production with annual cost estimated at $664 M for the US industry [1]. The virus causes reproductive fail- ure as well as respiratory problems, impaired growth per- formance and increased mortality in growing pigs [2]. The important heterogeneity observed among North American PRRSV strains, combined with the absence of complete cross-protection following infection with heterologous * Correspondence: marie-eve.lambert@umontreal.ca 1Laboratoire d’épidémiologie et de médecine porcine (LEMP), Faculty of Veterinary Medicine, Université de Montréal, St. Hyacinthe, Quebec, Canada 2Swine and Poultry Infectious Diseases Research Center (CRIPA), Faculty of Veterinary Medicine, Université de Montréal, St. Hyacinthe, Quebec, Canada Full list of author information is available at the end of the article PRRSV strains complicates disease management [3, 4]. the disease mostly relies on limiting Prevention of between-herd transmission that could occur through sev- eral direct and indirect pathways. In that regards, the gen- etic diversity in PRRSV can be used to support epidemiological investigations of a likely common source of infection or transmission events between herds in a re- search, surveillance or control context. A pairwise nucleo- tide sequence similarity ≥97.5% is the threshold often used to indicate if two sequences are considered similar and likely to originate from a same source [5, 6]. This thresh- old is also used into molecular-based interactive tools for field investigations on sources of contamination [7]. These tools are used to generate hypotheses about how a specific herd got infected which can orient the implementation of © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Lambert et al. BMC Veterinary Research (2019) 15:135 Page 2 of 10 specific preventive measures to avoid further introduction and spread of the virus. they databases; progressively The alignment of sequences is a prerequisite for the estimation of genetic distances between pairs of se- quences. Several algorithms to align sequences are avail- able but they differ in terms of computing approaches. Dynamic programming is an exact method evaluating each possible alignment to determine the best solution. Unfortunately, it is too cumbersome to be run for more than a few sequences. Thus, heuristic methods, progres- sive or iterative, are preferably used to manage large se- incorporate quence pairwise alignments into multiple alignment which con- siderably decrease computing time [8]. Also, global or local methods can be chosen according to how similar- ities are maximized throughout the alignment process; global methods consider the entire sequence length whereas local ones focus on highly homologous areas of the gene [9]. Finally, algorithms can be distinguished by the number of sequences considered in the alignment process and the purpose of analysis. A particular se- quence can be aligned to each sequence found in a data- base to find the most similar one using a pairwise sequence alignment, or a large group of sequences can be aligned simultaneously with multiple sequence align- ment to better take into account genetic evolution [10]. For most algorithms, the alignment process generally arranges gene sequences one over the other to maximize identical matches of nucleotides between sequences [8]. Generally, algorithms try to optimize an objective func- tion minimizing mismatches and gaps. In fact, gaps can be inserted during the alignment process if deletion or insertion sites are detected by the algorithm, so that sites with identical nucleotides align together. Using large penalties (cost) for opening a gap and a much smaller one for extending it result in programs adding fewer and/or shorter gaps [11]. Due to differences in objective functions or approaches used to optimize them, as well as several other parameter settings such as gap penalty, variations can be observed in final alignments obtained from different algorithms, sometimes leading to differ- ences in inferred phylogenetic trees [12]. Even if preliminary evaluation of several alignment al- gorithms or gap penalty settings on biological datasets is often suggested, this is rarely done, and studies on PRRSV are not an exception [13–15]. In fact, the choice of the best algorithm is not a straightforward task when using field data, since no reference alignment is available contrarily to simulated dataset or reference alignment based on the three-dimensional superposition of the proteins [16, 17]. Some studies have compared algo- rithms for highly divergent sequences belonging to dif- ferent families of genes, or based on a collection of known protein genes (e.g. ribosomal 16S or 23S subunit) across many species and showing important variation of sequence length [16–18]. However, studies on PRRSV North American genotype generally focus on relation- ships among sequences from a single viral gene (ORF5) that is expected to have considerably less diversity (≤25%) and to be relatively well conserved in length [19–21], which are two characteristics reported to influ- ence the alignment process [11]. Although one could ex- pect that default parameters set to align distantly related sequences should also work on less complex dataset, it has been suggested that these parameters should be evaluated on biological sequence datasets used in a spe- cific context [11, 22]. Also, it has been recommended to test simultaneously different algorithms, particularly on large-scale phylogenetic studies [23]. The rationale being that congruent results among several techniques should give better support to the accuracy of the final alignment [24]. This will also provide useful information on the comparability of results from PRRSV diversity studies based on different alignment programs. This study was conducted to evaluate the impact of the alignment algorithm on the resulting aligned se- quence dataset as well as on practical issues such as the capability to handle large dataset in a timely manner when applied to a large database of PRRSV ORF5 se- quences used for molecular-based epidemiological stud- ies and surveillance program. Methods Data collection and study population The PRRSV ORF5 sequence database of the Laboratoire d’épidémiologie et de médecine porcine (LEMP) of the Université de Montréal was used for the study. This database comprises sequences from field samples sub- mitted by veterinarians to different laboratories on a vol- untary basis as part of their herd surveillance or control programs. Since January 2010, a sharing agreement with 97% of all Quebec swine veterinarians have ensured that all PRRSV ORF5 sequences obtained from their field submissions to the veterinary diagnostic laboratory of the Université de Montréal or to two other private la- boratories were automatically transferred to the LEMP. All sequences gathered between January 1st 2010 and December 31st 2014 inclusively (n = 2383), with the ex- ception of one 606 base pair (bp) sequence, were used for the evaluation of alignment algorithms. This latter sequence was removed to ensure a balance of sequence length (600, 603 bp) in further replicates. Laboratory analyses RT-PCR and sequencing of the gene ORF5 coding for the major envelope protein GP5 was performed on all samples to identify a PRRSV sequence. Approximately 55% of sequences gathered in the LEMP sequence Lambert et al. BMC Veterinary Research (2019) 15:135 Page 3 of 10 database between 2010 and 2014 were submitted to the Veterinary Diagnostic Laboratory of the Université de Montréal. RNA was first extracted from serum or tissues (e.g. lungs, tonsils) with different extraction kits accord- ing to manufacturer’s instructions. RT-PCR was per- formed using Qiagen OneStep RT-PCR Kit using various primers. Prior to ORF5 sequencing, purification of PCR products was done using EZNA Cycle Pure Kit (Omega Bio-tek inc, Norcross, Georgia, US). Afterwards, both strands of PCR amplicons were sequenced using the same RT-PCR primers with BigDye terminator on ABI Genetic analyzer (Applied Biosystems Canada, Streets- ville, Ontario, Canada). The remaining sequences were obtained from private diagnostic laboratories which have used their routine protocols. Detection of recombinant sequences Detection of recombinant sequences was carried out by doing an exploratory scan for mosaic signals using de- fault detection methods implemented into Recombin- ation Detection Program (RDP) version 4.76 [25]. Primary scan was performed using only RDP, Geneconv and MaxChi. These latter methods were then used in addition to Bootscan, SisScan, Chimaera and 3-Seq for secondary scan. For each detection method, default par- ameter settings were used. Sequences identified by at least one primary method considering a 0.05 p-value using a Bonferroni correction for multiple testing were considered as significant recombinants. Alignment algorithms Selection Considering that a high level of similarity was expected over the entire ORF5 gene and that the overall objective was to manage a large PRRSV sequence database, only global mul- tiple sequence alignment methods available in freeware were considered for selection. Five algorithms were selected con- sidering their accuracy and popularity: Clustal W v.2.1 [26], Clustal Omega v.1.2.0 [27], Muscle v.3.8.31 [28], T-Coffee v. 11,00,8cbe486 [29] and MAFFT v.7.215 [18]. Parameter settings – other than open gap penalty value When possible, options were set to obtain the maximal accuracy reachable by the algorithms according to the user manual provided by their authors. For Clustal W, sequences were aligned in pairs using dynamic program- ming to generate a DNA weight matrix using Inter- national Union of Biochemistry (IUB) scoring matrix (used in BESTFIT). Scores were then converted into dis- tances and used to build a neighbor-joining guide tree (option Clustering = NJ). Iteration refinements were per- formed throughout the progressive approach (option It- eration = Tree). Default settings were used for Clustal Omega except for the use of full distance matrix in guide-tree calculation and iteration. Default settings were used in Muscle and T-Coffee. For MAFFT, G-INS-I was chosen based on highest accuracy and suit- ability for the study of sequences having similar lengths (MAFFT manual 2007-06-09, https://mafft.cbrc.jp/align- ment/software/manual/manual.html). Needleman-Wunsch algorithm computed pairwise align- ments (globalpair option) in combination with a max- imum of or of convergence of scoring alignment (option maxiterate = 1000). Default values were attributed to other parame- ters except for the open gap penalty value. 1000 cycles refinement iterative Parameter settings for open gap penalty value For each alignment algorithm, a sensitivity analysis was used to determine the open gap penalty parameter to be used for subsequent comparison of algorithms. The only exception was for Clustal Omega, for which the gap param- eter is directly handled by the algorithm. Sequences includ- ing recombinants (n = 2383) were randomly selected without replacement to form replicates of different sizes: ten replicates of 238 sequences, five replicates of 476, two replicates of 1191 and one including all 2383 sequences. For each algorithm and replicate, alignment was attempted for 11 different open gap penalty values i.e. from baseline to upper limit by equal increment. The open gap penalty values were the following: Clustal W (0 to 100 by 10), MAFFT (0 to 10 by 1), Muscle (0 to − 1000 by − 100) and T-Coffee (0 to − 1000 by − 100). The impact of gap penalty value was evaluated according to three criteria: average pairwise genetic similarity, proportion of pairwise compari- sons having ≥97.5% genetic similarity (Fig. 1), as well as the maximal number of gaps introduced per sequence. The open gap value from which a plateau was reached for the three criteria (i.e. minimum average pairwise similarity, minimum proportion of pairwise comparison with ≥97.5% genetic similarity, minimum number of gaps) was selected for further analyses based on visual assessment. Dataset sizes unable to run on all algorithms in less than 2 weeks were not considered for the choice of the open gap value. Implementation All alignments were run in a Linux environment (Ubuntu 14.04 LTS) on a Dell Precision T7610 workstation with 10 Intel Xeon Processor E5–2670 @ 2.5 GHz, 128 GB of RAM (DDR3) and a 2 TB HD. All computer resources were solely attributed to the alignment process. Selection and evaluation of comparison criteria Analytical criteria To compare performances of the five algorithms, two replicates of 1191 sequences were created from the data- set. A stratified random selection was used for sequence similar allocation to each replicate to ensure a Lambert et al. BMC Veterinary Research (2019) 15:135 Page 4 of 10 Fig. 1 Operational workflow used for computations of analytical criteria. Analytical criteria results are pictured in blue whereas pairwise matrices represent intermediary steps involved in computations. Illustrated with a fictive dataset of 4 sequences aligned with 2 algorithms (A and B) proportion of sequences with 600 and 603 bp as well as recombinant sequences in each replicate. The average pairwise genetic similarity, the proportion of pairwise comparisons having ≥97.5% genetic similarity and the length of final aligned dataset were computed for the two replicates for each algorithm using either SAS version 9.3 software (SAS Institute Inc., Cary, North Carolina, USA) or scripts written in Python. Characteris- tics of gap insertions, if they were introduced in singleton or triplets, were noted for each aligned dataset. i.e. Two additional criteria were evaluated: the sum of pairs (SP) score and the percentage of congruent cells having ≥97.5% similarity. SP-score is a measure of accur- acy defined as the proportion of shared homologies by estimated and reference alignments over the total num- ber of homologies in the reference alignment [22]. Since the reference alignment was unknown in the current study, alignment from each algorithm was by turns con- sidered as the reference, and the SP-score was used as a measure of agreement between algorithms. SP-score was Lambert et al. BMC Veterinary Research (2019) 15:135 Page 5 of 10 computed using FastSP, an open-source executable writ- ten in Java available online [23]. The percentage of con- gruent cells having ≥97.5% similarity among algorithms was computed as follows. For each alignment, a pairwise similarity matrix was calculated and transformed into a binary matrix: 0 for < 97.5% similarity, 1 for ≥97.5%. For each combination of two alignment algorithms, the bin- ary matrices were compared and the proportion of cells in agreement (having the same binary value) over total number of cells was computed. The operational flow- chart used in computations of all analytical criteria is de- scribed in Fig. 1. Technical criteria Four technical criteria were also used to compare per- formance of algorithms, i.e. handling capability of large dataset, rapidity, multi-platform availability and manage- ment of IUB ambiguity symbol characters (symbols other than A, T, C and G). The first two criteria were evaluated for 10 replicates of 238 sequences, 5 of 476, 2 of 1191 and 1 of 2383. Results were averaged over replicates. Data were analyzed in SAS. Sensitivity analysis on recombinant inclusion All analyses described for the analytical criteria assessment were reconducted to evaluate the impact of recombinant sequences using the same two replicates of 1191 se- quences, but without detected recombinants (n = 1183). Results Following the sensitivity analysis for the determination of open gap penalty values, the open gap penalty parameter was set at 30 for Clustal W, 7 for MAFFT, − 200 for T-Coffee, − 1000 for Muscle and default value for Clustal Omega. In general, the open gap penalty value had only a minimal impact on average pairwise similarity, proportion of pairwise comparisons having ≥97.5% similarity and maximal number of gaps per sequence for MAFFT and T-Coffee, but was more influential for Muscle and to a lesser extent for Clustal W (Fig. 2). The impact of the gap penalty value on the pairwise similarity and number of gap introduced tended to increase with dataset size, but convergence was obtained at approximately the same open gap penalty value whatever the size of the dataset. For each algorithm, a plateau was ob- served generally first (i.e. at lower gap penalty value) for the proportion of pairwise comparisons having ≥97.5%, followed by the average pairwise similarity and number of gaps. Once the plateau was reached for the three parameters, all algo- rithms converged to a similar number of gaps introduced for (i.e. 3) T-Coffee which introduced more gaps (up to 9 for one repli- cate of 1191 sequences). for datasets with ≤1191 sequences, except A total of 17 recombinant sequences were identified within 12 distinct recombination events. In order to allow even number of recombinants (n = 8) in each replicate (n = 1191) formed to investigate analytical criteria, one 603 bp recombinant sequence was excluded. The evaluation of the analytical criteria revealed a high and very similar aver- age pairwise genetic similarity across all algorithms and replicates, ranging from 88.28 to 88.84% (Table 1). The pro- portion of pairwise comparisons of sequences having ≥97.5% genetic similarity was also very similar, ≤0.25% vari- ation between algorithms within replicates and a slightly larger variation (0.5%) between replicates for the same algo- rithm. The sequence length of aligned dataset differed ac- cording to algorithm. Whereas Clustal W, Muscle and Clustal Omega introduced the minimal number of gaps (n = 3) on 600 bp sequences to integrate them with the 603 bp sequences in the final alignment, MAFFT introduced 3 to 6 gaps and T-Coffee, 7 to 9 gaps depending on replicates. For Clustal W, Clustal Omega, MAFFT and Muscle algo- rithms, all gaps were introduced as triplets, representing the code frame shift. Most gaps (> 98%) introduced by T-Coffee were singletons. Based on the SP-score, more than 99.7% of all pair homologies were shared between each combination of two algorithms, with T-Coffee showing a slightly higher disagreement with all other algorithms. A similar finding was observed for the proportion of congru- ent cells (≥99.86%). All algorithms were able to handle datasets of up to 1191 sequences, whereas only MAFFT, Muscle and Clustal Omega could process a 2383 sequence dataset in less than 2 weeks (Table 2). The rapidity mirrored the same tendency, as MAFFT, Muscle and Clustal Omega were the fastest, in- dependently of dataset size, aligning sequences in less than 20 s on the smallest dataset (238 sequences) and less than 29 min on the largest dataset (2383 sequences). T-Coffee and Clustal W were generally very slow, varying from 13 min for the smallest dataset (238 se- quences) to over 9–17 h for the largest dataset proc- essed (1191 sequences). All algorithms were available on the Web, Windows and Linux platforms. MAFFT and T-Coffee managed a greater number of IUB am- biguity symbols, followed by Muscle and Clustal series. Two replicates of 1183 sequences were formed by re- moving the 16 recombinants from the initial two repli- cates of 1191 sequences. The exclusion of recombinant sequences had a very minor impact on the results. The average pairwise similarity and the proportion of pairwise comparisons having ≥97.5% similarity slightly increase when excluding recombinant sequences (Table 3). For these two criteria, the greater difference was observed for T-Coffee. The length of aligned dataset, SP-score and pro- portion of congruent cells were very similar regardless of the presence of recombinants for all algorithms except T-Coffee for which small differences were observed, and this was mainly associated with the second replicate. Lambert et al. BMC Veterinary Research (2019) 15:135 Page 6 of 10 Fig. 2 Impact of open gap penalty value on average pairwise similarity, proportion of pairwise comparison having ≥97.5% genetic similarity and maximal number of gaps introduced per sequence for Clustal W, MAFFT, T-Coffee and Muscle. The different statistics were computed on each aligned dataset. Results obtained for dataset sizes of 238, 476 and 1191 were averaged over 10, 5 and 2 replicates, respectively. Results for the 2383 dataset are shown only for algorithms that generated results in less than two weeks. Recombinants were included in the datasets. Arrows indicate open gap penalty value selected for further analyses. *Default value of open gap penalty as defined by the algorithm user manual Discussion We investigated the impact of the choice of alignment algorithm when applied on a PRRSV North American genotype 2 sequence dataset. The dataset of 2383 se- quences employed was rather homogenous in regards to both similarity (≥79.1% minimum pairwise similarity ob- tained with different algorithms and open gap settings) and sequence length (603, 600 bp) reflecting viral popu- lation field studies as opposed to benchmark datasets such as BAliBASE [17, 18]. A priori, the multiple se- quence alignment did not appear to face specific hur- dles, and good accuracy from most algorithms was to be expected. In this study, although it was not possible to determine which alignment algorithm was more accur- ate due to the absence of a reference alignment [23], we compared algorithms by quantifying the variation in genetic similarity, which is important for molecular epi- demiology studies on PRRSV. Moreover, algorithms were compared from a practical perspective, namely for sur- veillance of PRRSV which requires timely analyses. The sensitivity analysis on open gap values revealed the algorithms in gap management differences for evaluated. In the study, the gap value parameter was op- timized to minimize the number of introduced gaps. This decision seemed biologically sound since the aligned sequences were from one gene with no non-coding DNA, and that fewer gap insertions usually gives better alignment accuracy [30]. Globally, Muscle and Clustal W were the most affected by variation of the open gap parameter value and inserted a large number of gaps especially when the penalty was low. The results therefore supported that empirical investigations should be conducted before using default open gap value on a large PRRSV dataset. For all algorithms, default values were inadequate to minimize the number of gaps intro- duced into the resulting alignment, particularly on the datasets with more than 1000 sequences. However, using default open gap values on smaller dataset (n = 238 or 476) had a negligible effect for Clustal W and Muscle, and practically no effect for MAFFT and T-Coffee. For PRRSV field investigations and molecular epidemi- ology studies, a pairwise genetic similarity threshold (e.g. ≥97.5%) is often used to determine whether two herds have similar strains [5, 6]. Results showed that most Lambert et al. BMC Veterinary Research (2019) 15:135 Page 7 of 10 Table 1 Results on analytical criteria investigated in a comparative study on PRRSV sequence alignment algorithmsa Criterion Algorithm Clustal W MAFFT T-Coffee Muscle Clustal 0mega 1. Similarity: average pairwise genetic similarity (%) of aligned sequences within the dataset (mean ± standard deviation) Replicate 1 (1191 sequences) Replicate 2 (1191 sequences) 88.77 ± 4.19 88.68 ± 4.11 88.84 ± 4.17 88.69 ± 4.11 88.71 ± 4.23 88.28 ± 4.31 88.78 ± 4.19 88.69 ± 4.11 88.78 ± 4.19 88.69 ± 4.11 2. Proportion of pairwise comparisons of sequences having ≥ 97.5% genetic similarity (%) Replicate 1 (1191 sequences) Replicate 2 (1191 sequences) 5.17 4.91 5.17 4.91 3. Length of aligned dataset: number of sites per sequence in the aligned dataset Replicate 1 (1191 sequences) Replicate 2 (1191 sequences) 603 603 606 603 4. Average sum of pairs (SP) score: proportion of shared homologies with reference alignment (%)b Clustal W as reference MAFFT as reference T-Coffee as reference Muscle as reference Clustal Omega as reference Average – 99.93 99.92 99.91 99.94 99.92 99.93 – 99.96 99.97 99.97 99.95 5.19 4.66 607 609 99.74 99.78 – 99.76 99.78 99.76 5. Congruent cells ≥ 97.5% similarity: proportion of cells between two pairwise similarity matrices having the same binary value (0: < 97.5%; 1: ≥97.5%) for genetic similarityb Clustal W as reference MAFFT as reference T-Coffee as reference Muscle as reference Clustal Omega as reference Average – 100.00 99.86 99.99 99.99 99.96 100.00 – 99.86 99.99 99.99 99.96 99.86 99.86 – 99.86 99.86 99.86 5.17 4.91 603 603 99.91 99.97 99.94 – 99.95 99.94 99.99 99.99 99.86 – 99.99 99.95 5.17 4.91 603 603 99.94 99.97 99.97 99.95 – 99.95 99.99 99.99 99.86 99.99 – 99.95 aThe open gap penalties used was 30 for Clustal W, 7 for MAFFT, −200 for T-Coffee, −1000 for Muscle and default for Clustal Omega. The dataset included 2383 sequences collected in 2010–2014 divided in two replicates bAverage of 2 replicates of 1191 sequences Table 2 Results on technical criteria investigated in a comparative study on PRRSV sequence alignment algorithmsa Criterion Algorithm Clustal W MAFFT T-Coffee Muscle Clustal 0mega 1. Handling capability of large dataset: capacity to generate results in less than 2 weeks (yes/no) 10 replicates of 238 sequences 5 replicates of 476 sequences 2 replicates of 1191 sequences Full dataset (2383 sequences) yes yes yes no yes yes yes yes 2. Rapidity: average time (minutes) necessary to align (Linux platform, 10 physical cores) 10 replicates of 238 sequences 5 replicates of 476 sequences 2 replicates of 1191 sequences Full dataset (2383 sequences) 3. Multiplatform availability (yes/no) Web, Windows and Linux 12.8 57.1 1040.5 n/a yes 0.2 1.0 7.0 28.5 yes yes yes yes no 13.1 56.1 540.0 n/a yes 4. Management of IUB ambiguity symbol characters: ability to manage symbols other than A, T, C and G List of managed symbols N N, R, Y, W, S, K, M, D, V, H, B N, R, Y, W, S, K, M, D, V, H, B yes yes yes yes 0.2 0.7 3.9 17.0 yes N, R, Y yes yes yes yes 0.2 0.4 1.2 2.9 yes N aThe open gap penalties used was 30 for Clustal W, 7 for MAFFT, −200 for T-Coffee, −1000 for Muscle and default for Clustal Omega. The dataset included 2383 sequences collected in 2010–2014 Lambert et al. BMC Veterinary Research (2019) 15:135 Page 8 of 10 Table 3 Differences in results for analytical criteria when excluding or not recombinants for the different algorithmsa Criterion Algorithm Clustal W MAFFT T-Coffee Muscle Clustal 0mega 1. Difference in similarity: average pairwise genetic similarity (%) of aligned sequences within the dataset (mean) Replicate 1 Replicate 2 0.01 0.02 −0.05 0.01 0.01 0.15 2. Difference in proportion of pairwise comparisons of sequences having ≥ 97.5% genetic similarity (%) Replicate 1 Replicate 2 0.07 0.05 0.07 0.05 0.07 0.27 3. Difference in length of aligned dataset: number of sites per sequence in the aligned dataset Replicate 1 Replicate 2 0 0 −3 0 0 −1 0.01 0.01 0.07 0.05 0 0 4. Difference in average sum of pairs (SP) score: proportion of shared homologies with reference alignment (%)b Clustal W as reference MAFFT as reference T-Coffee as reference Muscle as reference Clustal Omega as reference Average – 0.01 0.01 0.01 0.01 0.01 0.01 – 0.01 0.01 0.00 0.01 0.04 0.04 – 0.04 0.04 0.04 0.01 0.01 0.01 – 0.00 0.01 0.01 0.01 0.07 0.04 0 0 0.01 0.00 0.00 0.00 – 0.01 5. Difference in congruent cells ≥ 97.5% similarity: proportion of cells between two pairwise similarity matrices having the same binary value (0: < 97.5%; 1: ≥97.5%) for genetic similarityb – Clustal W as reference 0.11 0.00 0.00 MAFFT as reference T-Coffee as reference Muscle as reference Clustal Omega as reference Average −0.01 0.11 0.00 0.00 0.02 −0.01 – 0.11 0.01 0.00 0.02 0.11 – 0.11 0.11 0.11 0.01 0.11 – 0.00 0.03 0.00 0.11 0.00 – 0.03 aThe open gap penalties used was 30 for Clustal W, 7 for MAFFT, −200 for T-Coffee, −1000 for Muscle and default for Clustal Omega. The five criteria presented in Table 1 for the two replicates including recombinants (Replicates 1 and 2, n = 1191) were re-evaluated for each replicate without recombinants (Replicates 1 and 2, n = 1183). Then, differences in results were computed (i.e. the result obtained with recombinant was subtracted from the result obtained without recombinant bAverage of 2 replicates of 1183 sequences algorithms provided highly similar results in terms of aver- age pairwise similarity of sequences, both when using the ≥97.5% similarity threshold or on continuous scale, and thus the use of different algorithms should not signifi- cantly affect epidemiological conclusions. This is also sup- ported by the SP-score, which revealed that almost all pair homologies were conserved from one aligned dataset to the others, even when gaps were introduced. T-Coffee seemed to behave differently compared to other algo- rithms and showed more variation between replicates. PRRSV usually evolves through punctual mutations, but recombination events are also a part of the virus evolution [31–34]. The detection of recombinant sequences should be an important concern for molecular epidemiology study using classic phylogenetic analyses since the evolutionary histories of recombinants are not correctly taken into ac- count by these methods [33, 35]. Since the recombinants are not necessarily identified before the alignment process, either because of the need of a timely analysis of sequences for surveillance purposes or considering that alignment of sequences is the first step in investigating the presence of recombinants, it was therefore advisable to determine their influence on alignment according to the algorithm used. As expected, the inclusion of recombinants led to a decrease in pairwise genetic similarity; however, the results were highly similar between algorithms. Moreover, the number of gaps introduced stayed stable no matter if recombinants were included or not with the exception of T-Coffee. From an end-user perspective, even if excluding recombinants from PRRSV sequence dataset is favorable before performing phylogenetic analyses, their presence at the initial alignment step does not seem to influence the behavior of most algorithms, at least when they represent a small proportion of the sequences as we observed in our field database. The reasons underlying the greater influ- ence of recombinants on T-Coffee were not investigated in our study; however, as these differences were mostly seen in one replicate, they could have resulted from Lambert et al. BMC Veterinary Research (2019) 15:135 Page 9 of 10 specific characteristics of sub-datasets. the recombinants or the Ethics approval and consent to participate Not applicable. Finally, technical aspects showed major differences in speed and capability of handling large datasets. Since runtimes vary according to genetic diversity observed in datasets, number and length of sequences, as well as processor and memory allowed for computing align- ments, results obtained from different studies are not directly comparable. Considering the current datasets, algorithm settings and computational resources, the abil- ity to timely align large sequence datasets (2383 se- quences) by Clustal Omega, MAFFT and Muscle is a significant advantage. Conclusion The different algorithms compared for the analysis of a PRRSV ORF5 sequence dataset provided very similar alignments, but differed in their ability to handle large datasets. Results from most algorithms were not affected by the presence of recombinants detected in our field database. Our study also revealed that prior investigations to set open gap parameter are advisable, especially when used on more than 1000 sequences. Muscle and Clustal W inserted many gaps when the open gap parameter was left at default or near zero values. Based on the efficiency at minimizing the number of gaps on different dataset sizes with default open gap value, the congruency of sev- eral analytical criteria with other algorithms as well as the capability to handle a large number of sequences in a timely manner, Clustal Omega might be warranted to manage large PRRSV database for both research and on- going disease surveillance purposes. Abbreviations LEMP: Laboratoire d’épidémiologie et de médecine porcine; ORF: Open reading frame; PRRSV: Porcine reproductive and respiratory syndrome virus Acknowledgments The authors would like to acknowledge the significant support of Darren Patrick Martin in regards to RDP software and of Jean-Charles Côté for careful manuscript revision. Funding This research was funded by Swine Cluster 2 (project #1343), the Éleveurs de Porcs du Québec and Zoetis Canada inc. These organizations were not involved in the design of the study and collection, analysis and interpretation of data and in writing the manuscript. Availability of data and materials The datasets analyzed during the current study are not publicly available due to sharing agreements. These were necessary for the transfer of sequences and data from the diagnostic laboratories to the LEMP, to analyze data and report results. Authors’ contributions MEL, JA, BD, PA, ZP and SD designed the experiments. PA coded all programs for comparing aligned datasets, BD performed the recombinant detection analyses and MEL analyzed the data. MEL wrote the manuscript. All authors revised and approved the final manuscript. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Author details 1Laboratoire d’épidémiologie et de médecine porcine (LEMP), Faculty of Veterinary Medicine, Université de Montréal, St. Hyacinthe, Quebec, Canada. 2Swine and Poultry Infectious Diseases Research Center (CRIPA), Faculty of Veterinary Medicine, Université de Montréal, St. Hyacinthe, Quebec, Canada. 3Department of Population Medicine, Ontario Veterinary College, University of Guelph, Guelph, Ontario, Canada. Received: 9 November 2018 Accepted: 29 April 2019 References 1. Holtkamp DJ, Kliebenstein JB, Neumann EJ, Zimmerman JJ, Rotto HF, Yoder TK, et al. Assessment of the economic impact of porcine reproductive and respiratory syndrome virus on United States pork producers. J Swine Health Prod. 2013;21:72–84. Christianson WT, Joo HS. Porcine reproductive and respiratory syndrome virus: a review. J Swine Health Prod. 1994;2:10–28. Lager KM, Mengeling WL, Brockmeier SL. Evaluation of protective immunity in gilts inoculated with the NADC-8 isolate of porcine reproductive and respiratory syndrome virus (PRRSV) and challenge-exposed with an antigenically distinct PRRSV isolate. Am J Vet Res. 1999;60:1022–7. 2. 3. 4. Meng XJ. Heterogeneity of porcine reproductive and respiratory syndrome 5. 6. 7. 8. 9. 10. virus: implications for current vaccine efficacy and future vaccine development. Vet Microbiol. 2000;74:309–29. Lambert ME, Arsenault J, Poljak Z, D'Allaire S. Correlation among genetic, Euclidean, temporal, and herd ownership distances of porcine reproductive and respiratory syndrome virus strains in Quebec, Canada. BMC Vet Res. 2012;8:76. Larochelle R, D'Allaire S, Magar R. Molecular epidemiology of porcine reproductive and respiratory syndrome virus (PRRSV) in Quebec. Virus Res. 2003;96:3–14. Lambert ME, Audet P, Delisle B, Arsenault J, D'Allaire S. Porcine reproductive and respiratory syndrome virus: web-based interactive tools to support surveillance and control initiatives. Porcine Health Manag. 2019;5:10 Salemi M, Vandamme A-M. The phylogenetic handbook, a practical approach to DNA and protein phylogeny. New York: Cambridge University Press; 2003. Kumar S, Filipski A. Multiple sequence alignment: in pursuit of homologous DNA positions. Genome Res. 2007;17:127–35. Thompson SM. Multiple sequence alignment and analysis: part I - an introduction to the theory and application of multiple sequence analysis. In: Grant RP, editor. Computational genomics: theory and application. Norfolk: horizon scientific press; 2004. p. 1–53. 11. Morrison DA. Multiple sequence alignment for phylogenetic purposes. Aust Syst Bot. 2006;19:479–539. 12. Ogden TH, Rosenberg MS. Multiple sequence alignment accuracy and phylogenetic inference. Syst Biol. 2006;55:314–28. 14. 13. Mateu E, Diaz I, Darwich L, Casal J, Martin M, Pujols J. Evolution of ORF5 of Spanish porcine reproductive and respiratory syndrome virus strains from 1991 to 2005. Virus Res. 2006;115:198–206. Fang Y, Schneider P, Zhang WP, Faaberg KS, Nelson EA, Rowland RRR. Diversity and evolution of a newly emerged north American type 1 porcine arterivirus: analysis of isolates collected between 1999 and 2004. Arch Virol. 2007;152:1009–17. Li B, Fang L, Guo X, Gao J, Song T, Bi J, et al. Epidemiology and evolutionary characteristics of the porcine reproductive and respiratory syndrome virus in China between 2006 and 2010. J Clin Microbiol. 2011;49:3175–83. 15. Lambert et al. BMC Veterinary Research (2019) 15:135 Page 10 of 10 16. 17. 18. 19. Liu K, Linder CR, Warnow T. Multiple sequence alignment: a major challenge to large-scale phylogenetics. PLoS Curr. 2010;2:RRN1198. Thompson JD, Plewniak F, Poch O. A comprehensive comparison of multiple sequence alignment programs. Nucleic Acids Res. 1999;27:2682–90. Katoh K, Misawa K, Kuma K, Miyata T. MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform. Nucleic Acids Res. 2002;30:3059–66. Shi M, Lam TTY, Hon CC, Murtaugh MP, Davies PR, Hui RKH, et al. Phylogeny-based evolutionary, demographical, and geographical dissection of north American type 2 porcine reproductive and respiratory syndrome viruses. J Virol. 2010;84:8700–11. 20. Delisle B, Gagnon CA, Lambert ME, D'Allaire S. Porcine reproductive and respiratory syndrome virus diversity of eastern Canada swine herds in a large sequence dataset reveals two hypervariable regions under positive selection. Infect Genet Evol. 2012;12:1111–9. 21. Murtaugh MP. Use and interpretation of sequencing in PRRSV control 22. programs. In: Allen D. Leman swine conference. Vol. 39. Minnesota: veterinary continuing education; 2012. p. 49–55. Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013; 30:772–80. 23. Mirarab S, Warnow T. FASTSP: linear time calculation of alignment accuracy. Bioinformatics. 2011;27:3250–8. 24. Pennington T, Scotland RW. Systematics association homology and systematics : coding characters for phylogenetic analysis, 1st edn. London. New York: Taylor & Francis; 2000. 27. 26. 25. Martin DP, Murrell B, Golden M, Khoosal A, Muhire B. RDP4: detection and analysis of recombination patterns in virus genomes. Virus Evol. 2015;1:1–5. Larkin MA, Blackshields G, Brown NP, Chenna R, McGettigan PA, McWilliam H, et al. Clustal W and Clustal X version 2.0. Bioinformatics. 2007;23:2947–8. Sievers F, Wilm A, Dineen D, Gibson TJ, Karplus K, Li W, et al. Fast, scalable generation of high-quality protein multiple sequence alignments using Clustal omega. Mol Syst Biol. 2011;7:539. Edgar RC. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 2004;32:1792–7. 28. 29. Notredame C, Higgins DG, T-Coffee HJ. A novel method for fast and accurate multiple sequence alignment. J Mol Biol. 2000;302:205–17. 30. Rosenberg MS. Evolutionary distance estimation and fidelity of pair wise sequence alignment. BMC Bioinformatics. 2005;6:102. 31. Yuan SS, Nelsen CJ, Murtaugh MP, Schmitt BJ, Faaberg KS. Recombination between north American strains of porcine reproductive and respiratory syndrome virus. Virus Res. 1999;61:87–98. 32. Murtaugh MP, Stadejek T, Abrahante JE, Lam TTY, Leung FCC. The ever- expanding diversity of porcine reproductive and respiratory syndrome virus. Virus Res. 2010;154:18–30. 33. Martin-Valls GE, Kvisgaard LK, Tello M, Darwich L, Cortey M, Burgara-Estrella AJ, et al. Analysis of ORF5 and full-length genome sequences of porcine reproductive and respiratory syndrome virus isolates of genotypes 1 and 2 retrieved worldwide provides evidence that recombination is a common phenomenon and may produce mosaic isolates. J Virol. 2014;88:3170–81. Shi M, Holmes EC, Brar MS, Leung FC. Recombination is associated with an outbreak of novel highly pathogenic porcine reproductive and respiratory syndrome viruses in China. J Virol. 2013;87:10904–7. 34. 35. Martin DP, Lemey P, Posada D. Analysing recombination in nucleotide sequences. Mol Ecol Resour. 2011;11:943–55.
10.1186_s12920-023-01484-0	Afzaljavan et al. BMC Medical Genomics (2023) 16:72 https://doi.org/10.1186/s12920-023-01484-0 BMC Medical Genomics Genetic contribution of caspase-8 variants and haplotypes to breast cancer risk and prognosis: a case-control study in Iran Fahimeh Afzaljavan1, Elham Vahednia1, Matineh Barati Bagherabad1, Fatemeh Vakili2, Atefeh Moezzi1, Azar Hosseini3, Fatemeh Homaei Shandiz4, Mohammad Mahdi Kooshyar5, Mohammadreza Nassiri6 and Alireza Pasdar1,7 Abstract Purpose Multiple genome-wide and candidate-gene association studies have been conducted to search for common risk variants of breast cancer. Recent large meta-analyses and consolidating evidence have highlighted the role of the caspase-8 gene in breast cancer pathogenesis. Therefore, this study aimed to identify common variations and haplotypes associated with risk and overall survival of breast cancer with respect to underlying susceptibility variants in the CASP8 gene region in a group of the Iranian population. Methods In a case-control study with a total of 1008 samples (455 cases and 553 controls), genotyping of 12 candidate polymorphisms, consisting of rs3834129, rs2037815, rs7608692, rs12990906, rs3769821, rs6435074, rs3754934, rs3817578, rs10931936, rs1045485, rs1045487, and rs13113, were performed using PCR-based methods, including ARMS- PCR, AS-PCR, RFLP-PCR, HRM-PCR, and TaqMan-PCR. Results rs3834129, rs3754934, rs12990906, and rs10931936 were associated with the risk and overall survival of breast cancer. Several haplotypes were also identified an associated with a higher risk of breast cancer, including a three-SNP haplotype rs3817578-rs10931936-rs1045485 [p < 0.001, OR = 1.78(1.32–2.41)]. rs3754934-C allele showed an association with a lower risk of death in all patients [p = 0.022; HR = 0.46(0.23–0.89)] and in the hormone-receptor-positive group [p = 0.038; HR = 0.37(0.14–0.95)], as well as CC genotype in the hormone-receptor-positive group [p = 0.002; HR = 0.09(0.02–0.43)]. Conclusion The present study suggests a diagnostic and prognostic role of CASP8 gene variations in breast cancer. The risky haplotypes are likely to have one or more underlying breast cancer susceptibility alleles. Understanding the mode of action of these alleles will aid individual-level risk prediction. It also may help identify at-risk patients to provide them with better surveillance. Keywords Breast neoplasm, Biomarker, Caspase 8, Diplotype, Overall survival, Prognosis *Correspondence: Fahimeh Afzaljavan afzaljavanf2@mums.ac.ir; Fahimeh.afzaljavan@gmail.com Alireza Pasdar pasdara@mums.ac.ir Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessIntroduction Breast cancer is the most common cancer among women, accounting for 11.7% of all new cancers and 24.5% of all female cancers. Moreover, it is the fifth cause of cancer death and first-ranked in women [1]. Epidemiological studies indicated that progression in detection methods like mammographic screening led to increasing inci- dence rates of breast cancer during the 1980-90s decades in many countries. Conversely, widespread screening and reduced menopausal hormone therapy caused a decreased incidence during the early 2000s. However, breast cancer incidence is rising due to changes in life- style, sociocultural, and environmental issues. High Body Mass Index (BMI) resulting from a sedentary lifestyle and junk and high-calorie diet, night shift, and reproductive and gynecologic factors, including hormonal changes, reduced pregnancy, and lactation, have been identified as the risk factors of the disease. Therefore, identifying diag- nostic and prognostic markers of the disease is a promi- nent point of attention in oncology research [2]. It is estimated that 5–10% of breast cancers are heredi- tary; however, a high portion of the disease is sporadic type affected by genetic and environmental risk factors, although most of the underlying genetic mechanisms have not been fully defined [3]. Among genetic indica- tors, polymorphisms are common genomic variations in the general population identified as potential genetic markers for risk assessment. However, comparing high penetrance mutations, these are typically associated with moderate risk [4]. Although candidate gene studies have introduced various loci [5], in recent years, high- throughput genome-wide association studies have identi- fied many genetic loci associated with the risk of breast cancer, introducing breast cancer as a polygenic complex disease [6]. Caspase 8 protein (CASP8), a 55 kDa cysteine protease, is a member of the caspase family and a key apoptosis signaling molecule. It contributes to inducing cell death, particularly through the death receptor pathway. CASP8, one of the first low penetrance loci, has been identified to be associated with the risk of breast cancer in candi- date gene studies [7–10]. Furthermore, efforts to identify new variations in fine-mapping [11, 12] and genome- wide association [13] studies have provided evidence of the association of several variants of CASP8 with breast cancer risk. Given the importance of allelic variations associated with cancers, including breast cancer [7–13], this study aimed to investigate the association of CASP8 polymorphisms, haplotypes and diplotypes with breast cancer risk, prognosis, and clinicopathological features in a northeastern population of Iran. Page 2 of 11 Materials and methods Study population This study was approved by the Ethics Committee of Mashhad University of Medical Sciences under the ethi- cal approval number: IR.MUMS.REC.1394.188. All par- ticipants signed a written informed consent at the time of study entry. Due to the fact that CASP8 had not been assessed in previous research in Iran, we did not have access to the allele frequency of its variation in our population to cal- culate the exact sample size required for a decent power of the study (80%). Consequently, a pilot sample size was performed based on similar studies in this field, which mainly have suggested 200–400 samples in each group. However, the final study population included 1008 par- ticipants. The breast cancer group included 455 patients (152 new cases diagnosed between 2016 and 2018 and 303 patients diagnosed between 1987 and 2016 and fol- lowed in this period) referred to academic teaching hos- pitals of Mashhad University of Medical Sciences. The control group consisted of 553 healthy people referred to clinicians between 2016 and 2018 for screening, and their health was confirmed using the clinical breast exam (CBE) and mammography. Demographic information was collected using a questionnaire providing sociode- mographic data, including age, age of menarche, meno- pause and first gestation, BMI, history of lactation and abortion, and physical activity. Pedigree was drawn for all participants to check the family history of cancer and find participants’ relatives. Manchester Score (MS) was used to identify the prob- ability of harboring BRCA1/2 mutations [14]. As a result, the highly suspected hereditary cancer was excluded. After excluding five patients with probable hereditary breast cancer (with an MS of more than 10), 450 sporadic cancer subjects entered the study as the patient group. The histopathological data, including breast tumor subtype, stage, grade, and receptor status (ER, PR, and HER2), was extracted from patients’ medical records. Categorization was performed according to the standard protocols of the world health organization (WHO) [15], the American Joint Committee on Cancer (AJCC) [16], and the American Society of Clinical Oncology (ASCO) [17]. All cases were followed, and new events, includ- ing recurrence, secondary tumors, and metastasis, were documented. Blood collection and DNA extraction Five ml of peripheral blood was collected using a Vacuette K2-EDTA blood collection tube (Greiner Bio-One, USA). The salting-out method was utilized to isolate DNA [18]. The qualification and quantification of extracted DNA were evaluated by gel electrophoresis and Epoch™ Microplate Spectrophotometer (BioTek Instruments Inc., Afzaljavan et al. BMC Medical Genomics (2023) 16:72 Page 3 of 11 Winooski, VT, USA). Samples were aliquoted in a con- centration of 150 ng per microliter and stored at -20 until polymerase chain reaction (PCR) analysis. SNP selection Twelve validated polymorphisms of the CASP8 gene were selected in different gene regions, including 5’ UTR (promoter), exon, intron, and 3’ UTR regions. Selection of polymorphisms was performed based on several cri- teria, including validation of the association in numer- ous GWAS studies, which denotes a strong association with breast cancer risk in different populations. We also considered selecting SNPs that are located in the same region to be able to perform haplotype analysis to exam- ine the overall effect of these polymorphisms. We also considered selecting markers with an acceptable MAF and heterozygosity (minor allele frequency > 5% and het- erozygosity > 10%) to achieve the highest possible study power. Characteristics of the selected polymorphisms have been shown in (Additional file 1: Supplementary Table 1). Genotyping Genotyping was done using different PCR-based meth- ods. rs3834129, rs12990906, rs3754934, rs3817578 ,and rs10931936 were genotyped using Tetra-ARMS-PCR, rs2037815 and rs7608692 using allele-specific PCR, rs3769821 and rs1045485 using RFLP-PCR. Genotyping method for rs1045487 and rs6435074 was HRM (Light- Cycler® 96 Instrument (Roche Molecular Systems, Inc.)), and for rs13113 was TaqMan (SNP genotyping Assays (TaqMan®), Catalog number: 4,351,379; Rotor-Gene 6000™ real-time analyzer (Applied Biosystems)). Primers were designed using Primer1, Gene runner and WASP (Web-based Allele-Specific PCR assay), and evaluated using Oligoanalyzer and Mfold. The designed primer sequences have been shown in (Additional file 1: Supple- mentary Table 2). Amplification reactions and protocols are shown in (Additional file 1: Supplementary Tables 3 & 4). 5% of samples were randomly re-genotyped to verify genotyp- ing results for quality control purposes. In addition, three samples were randomly sanger sequenced to validate the genotyping method for each marker. Sequencing was done using outer primers for polymorphisms genotyped by Tetra-ARMS-PCR, and new primers, outer both sides of the genotyped region, were designed for the other variations. Haplotype and diplotype analysis Assessing the haplotypes and diplotypes distribution was carried out using the PHASE software version 2.1.1 for windows [19]. The linkage disequilibrium (LD) was cal- culated by 2LD program version 1.00 and evaluated by the D′ statistic as the deviation between the expected haplotype and observed frequency [20]. Statistical analysis The Hardy-Weinberg Equilibrium (HWE) assumption was assessed in the case and control samples using the χ2 with one degree of freedom. Data are shown in (Addi- tional file 1: Supplementary Table 5). Depending on the assessment of normality using the Kolmogorov-Smirnov (K-S) test, the normally distributed continuous variables were examined using the independent sample t-test and the Mann-Whitney U test was used to compare non- normally distributed variables between the two groups. ANOVA or Kruskal Wallis was also used to compare more than two groups. The categorical variables were compared appropriately with the chi-square or Fisher’s exact tests. Correlations between variables were tested using the Pearson correlation test for normally distrib- uted variables and the Spearman correlation test for non- normally distributed variables. The associations of alleles, genotypes, haplotypes, and diplotypes with breast cancer risk, breast cancer risk fac- tors, and histopathological status were judged by logistic regression. Odds ratios (ORs) and 95% confidence inter- vals (CIs) were calculated for the measured risk factors. Multivariate logistic regression was applied to identify the variables with independent association with the risk of breast cancer. The backward logistic regression (LR) model was implemented to select variables for multi- variable investigation. The results were also adjusted for potential confounders such as BMI, age at first gestation, and Menopause status in the logistic regression analysis. Overall survival (OS) time was considered the time between diagnosis according to the first biopsy confirm- ing the disease and the time of death due to cancer or last contact. Kaplan–Meier plots/Log-rank and Cox pro- portional hazards regression approaches were used to explain the associations between different covariates and overall survival. The hazard rate ratio (HR) and 95% CIs were calculated by the Cox models. Statistical analysis was performed using SPSS 16.0 (IBM, USA), and a P-value less than 0.05 was considered significant. Results Characteristics of the population After excluding 5 patients with hereditary breast cancer, 450 breast cancer patients (mean age = 47.20 ± 10.41) and 553 healthy individuals (mean age = 45.88 ± 11.51) were studied. The characteristics of breast cancer cases and cancer-free controls have been shown in Table 1. Fur- thermore, tumor features of breast cancer patients have been reported in Table 2. Afzaljavan et al. BMC Medical Genomics (2023) 16:72 Table 1 The characteristics of breast cancer cases and cancer-free controls Characteristic a Age Age of menarche Age of menopausec Age of first gestation BMI (Kg/m2)d BMI (Kg/m2) Menopause status History of lactation History of abortion Physical activity BMI < 25 BMI ≥ 25 Pri & pre Post Negative Positive Negative Positive Negative Positive Breast cancer 47.20 ± 10.41 13.05 ± 1.65 47.79 ± 5.61 21.39 ± 5.09 27.66 ± 5.04 117 (28.4%) 295 (71.6%) 238 (57.9%) 173 (42.1%) 18 (4.7%) 362 (95.3%) 236 (64.1%) 132 (35.9%) 125 (42.4%) 170 (57.6%) Page 4 of 11 OR (95%CI) 1.01 (0.99–1.02) 1.07 (0.98–1.16) 1.01 (0.97–1.06) 1.05 (1.02–1.08) 1.11 (1.08–1.14) 2.56 (1.94–3.37) 2.15 (1.63–2.84) 1.03 (0.54–1.98) 1.32 (0.98–1.79) 4.66 (3.20–6.80) Control 45.88 ± 11.51 13.23 ± 1.56 48.19 ± 5.21 22.55 ± 4.53 25.36 ± 4.36 260 (50.4%) 256 (49.6%) 397 (74.8%) 134 (25.2%) 20 (4.9%) 390 (95.1%) 281 (70.3%) 119 (29.8%) 51 (13.3%) 332 (86.7%) P-value b 0.065 0.116 0.545 0.001 < 0.001 Reference < 0.001 Reference < 0.001 Reference 0.926 Reference 0.071 Reference < 0.001 a Data are presented as mean ± SD for continuous variable or number (percentage, %) for categorical variables; b Significant data has been shown in bold c The age of menopause in individuals with natural menopause d BMI: Body Mass Index Menstrual status was significantly different between the two groups (p < 0.001). According to the findings of this study, there was no significant difference in lactation and abortion history between the groups (p > 0.05). BMI showed a significant difference (p < 0.001) with a mean of 27.65 ± 5.05 Kg/m2 in patients and 25.36 ± 4.36 Kg/m2 in healthy subjects. Also, the classification of this index into two groups of less and more than 25 showed that the percentage of people with a BMI above 25 in the patient group was higher than in the control group (p < 0.001). Evaluation of clinicopathologic features indicated the most common type of tumor in the study population was the invasive ductal type by 75.1% of the total speci- mens examined. In situ, lobular and metastatic tumors were less prevalent. Tumor grade and stage examination showed that more patients (56%) had low-grade tumors, and 50.7% of patients were identified in the early stages of the disease (1 and 2). In terms of tumor size, small tumors (with 64.9% of all specimens) ranked first. Find- ings related to lymph node status showed that 47.4% of patients were lymph node-positive, with the high- est number of involved nodes being between 1 and 3. Assessment of hormone receptor status showed that in more than 60% of patients, estrogen or progesterone hor- mone receptors were positive, and HER2 overexpression was observed in 22.9% of patients. Evaluation of overall survival in patients showed that 5-year overall survival was 90%, and 10-year overall sur- vival was 85%. Association of CASP8 genotypes, haplotypes and diplotypes with breast cancer risk Hardy–Weinberg equilibrium in the healthy controls is shown in (Additional file 1: Supplementary Table 5). For those polymorphisms which were not in Hardy– Weinberg equilibrium the genotyping results were veri- fied by regenotyping 5% of samples randomly and the results were consistent with the previously genotyped samples. The results of statistical analysis showed that rs3834129 was associated with breast cancer risk in dom- inant (II + ID vs. DD) (pAdj=0.034) and recessive (ID + DD vs. II) (pAdj=0.014) models. In the dominant model, rs2037815-G allele carriers (GA + GG) (pAdj=0.031), rs7608692-A-allele carriers (GA + AA) (pAdj=0.006), and rs10931936-T allele carriers (TT + CT) (pAdj<0.001) had a higher risk of breast cancer. On the other hand, carriers of the rs3754934-A allele (CA + AA) had a reduced risk of breast cancer in the dominant model (pAdj=0.004). We did not find a significant association between breast can- cer risk and rs3769821, rs6435074, rs3817578, rs1045485, rs1045487, and rs13113 in our study population. Alleles and genotypes frequencies have been reported in Table 3, for further information about the analyses based on dif- ferent genetic models see (Additional file 1: Supplemen- tary Table 6), and significant findings have been shown in Tables 4 and 5. The CTG haplotype of rs3817578-rs10931936- rs1045485, with a prevalence of 18.8%, among the hap- lotypes was associated with an increased risk of breast cancer (pAdj<0.001). Two 4-SNPs haplotypes, two 5-SNPs haplotypes and a 6-SNPs haplotype were also associ- ated with the risk of breast cancer in the study popu- lation. Since the frequency of identified haplotypes Afzaljavan et al. BMC Medical Genomics (2023) 16:72 Table 2 Distribution of tumour characteristics of Breast cancer cases Characteristics Tumor subtype Grade Tumor size Lymph node Metastasis Stage ER statusa PR statusb HER2c Receptor status Invasive Ductal Carcinoma Precursor lesions Invasive Lobular Carcinoma Invasive Medulary Carcinoma Metastatic Carcinoma Others Unreported Low grade (I & II) High grade (III) Unreported Small (I & II) Large (III & IV) Unreported Negative Positive (I, II & III) Unreported Negative Positive Unreported Early stage (I & II) Late stage (III & IV) Unreported Negative Positive Unreported Negative Positive Unreported Negative Positive Equivocal Unreported ER/PR + HER2 +/- ER/PR - HER2 + Triple negative (TNBC) Unreported Page 5 of 11 Percent 75.1 4.2 2.4 1.6 2.4 2.9 11.3 56 21.3 22.7 64.9 16.2 18.9 32 47.4 20.4 75.1 4.9 20 50.7 28.9 20.4 22.4 65.6 12 25.3 62.4 12.2 57.1 22.9 5.6 14.4 67.7 9.1 10 13.1 Number 338 19 11 7 11 13 51 252 96 102 292 73 85 144 213 93 338 22 90 128 130 92 101 295 54 114 281 55 257 103 25 65 305 41 45 59 a ER; Oestrogen receptor; b PR; Progesterone receptor c HER2; Human Epidermal growth factor Receptor 2 with more SNPs was lower than 10%, they were not investigated in this study. Diplotypes were also identi- fied using the haplotype data. Based on the identified diplotypes with a frequency of more than 10%, four diplotypes [rs3817578-rs10931936- rs1045485 (CCG- CTG), rs3817578-rs10931936- rs1045485 (CCC, CTG), rs3754934-rs3817578-rs10931936-rs1045485 (CCCG- CCTG) and rs3754934-rs381757-rs10931936-rs1045485- rs1045487 (CCCGG-CCTGG)] were associated with breast cancer risk. Significant results have been reported in Tables 4 and 5. Association of CASP8 polymorphisms, haplotypes and diplotypes with clinicopathological features and overall survival Genotypes, haplotypes, and diplotypes were extensively analyzed for a potential correlation/association with breast cancer clinicopathological characteristics and overall survival. Significant results have been presented in Tables 4 and 5. Evaluation of the genotypes with respect to clinicopath- ological features specified the association of rs3834129 (p = 0.034) and rs2037815 with menstrual age (p = 0.026), rs1045487 with the diagnosis age (p = 0.022), rs13113 Afzaljavan et al. BMC Medical Genomics (2023) 16:72 Table 3 The frequency of alleles and genotypes of CASP8 polymorphisms in breast cancer and healthy groups SNP ID rs3834129 rs2037815 rs7608692 rs12990906 rs3769821 rs6435074 rs3754934 rs3817578 rs10931936 rs1045485 Genotype DD ID II D I AA GA GG A G GG GA AA G A CC TC TT C T TT TC CC T C CC CA AA C A CC CA AA C A TT CT CC T C CC CT TT C T CC GC GG C G Breast cancer 58 (12.9%) 185 (41.1%) 207 (46.0%) 301 (33.4%) 599 (66.6%) 80 (17.8%) 261 (58.0%) 109 (24.2%) 421 (46.8%) 479 (53.2%) 161 (35.8%) 211 (46.9%) 78 (17.3%) 533 (59.2%) 367 (40.8%) 69 (15.3%) 196 (43.6%) 185 (41.1%) 334 (37.1%) 566 (62.9%) 273 (60.7%) 147 (32.7%) 30 (6.7%) 693 (77.0%) 207 (23.0%) 227 (50.4%) 181 (40.2%) 42 (9.3%) 635 (70.6%) 265 (29.4%) 385 (85.6%) 51 (11.3%) 14 (3.1%) 821 (91.2%) 79 (8.8%) 8 (1.8%) 119 (26.4%) 323 (71.8%) 135 (15.0%) 765 (85.0%) 245 (54.4%) 168 (37.3%) 37 (8.2%) 658 (73.1%) 242 (26.9%) 30 (6.7%) 127 (28.2%) 293 (65.1%) 187 (20.8%) 713 (79.2%) Control 95 (17.2%) 261 (47.2%) 197 (35.6%) 451 (40.8%) 655 (59.2%) 130 (23.5%) 301 (54.4%) 122 (22.1%) 561 (50.7%) 545 (49.3%) 249 (45.0%) 215 (38.9%) 89 (16.1%) 713 (64.5%) 393 (35.5%) 111 (20.1%) 248 (44.8%) 194 (35.1%) 470 (42.5%) 636 (57.5%) 361 (65.3%) 143 (25.9%) 49 (8.9%) 865 (78.2%) 241 (21.8%) 294 (53.2%) 218 (39.4%) 41 (7.4%) 806 (72.9%) 300 (27.1%) 464 (84.1%) 75 (13.6%) 13 (2.4%) 1005 (91.9%) 101 (8.1%) 16 (2.9%) 142 (25.7%) 395 (71.4%) 174 (15.7%) 932 (84.3%) 396 (71.6%) 122 (21.1%) 35 (6.3%) 914 (82.6%) 192 (17.4%) 56 (10.1%) 142 (25.7%) 355 (64.2%) 254 (23.0%) 852 (77.0%) P-value Adj. a Reference 0.158 0.008 Reference 0.011 Reference 0.514 0.502 Reference 0.684 Reference 0.006 0.136 Reference 0.179 Reference 0.172 0.712 Reference 0.026 Reference 0.389 0.92 Reference 0.533 Reference 0.841 0.924 Reference 0.856 Reference 0.002 0.856 Reference 0.051 Reference 0.555 0.88 Reference 0.993 Reference < 0.001 0.372 Reference < 0.001 Reference 0.744 0.28 Reference 0.296 Page 6 of 11 OR (95%CI) Adj. 1.49 (0.86–2.59) 2.14 (1.22–3.75) 1.43 (1.15–1.78) 1.15 (0.75–1.75) 0.84 (0.51–1.39) 1.04 (0.85–1.29) 1.52 (1.12–2.04) 1.35 (0.91-2.00) 1.16 (0.93–1.44) 1.41 (0.86–2.30) 1.09 (0.69–1.73) 1.27 (1.03–1.58) 1.18 (0.81–1.72) 1.03 (0.54–1.96) 1.08 (0.84-1-39) 1.06 (0.57–1.98) 1.03 (0.55–1.93) 1.02 (0.81–1.29) 0.41 (0.23–0.73) 0.89 (0.26–3.07) 1.50 (0.99–2.27) 0.89 (0.60–1.32) 0.91 (0.29–2.91) 1.00 (0.74–1.35) 2.31 (1.57–3.39) 1.34 (0.70–2.65) 1.73 (1.34–2.23) 1.07 (0.72–1.58) 0.69 (0.35–1.35) 1.15 (0.88–1.45) Afzaljavan et al. BMC Medical Genomics (2023) 16:72 Table 3 (continued) SNP ID rs1045487 rs13113 Genotype GG GA AA G A TT TA AA T A a significant data has been shown in bold Breast cancer 321 (71.3%) 111 (24.7%) 18 (4.0%) 753 (83.7%) 147 (16.3%) 170 (37.8%) 204 (45.3%) 76 (16.9%) 544 (60.4%) 356 (39.6%) Control 403 (72.9%) 135 (24.4%) 15 (2.7%) 941 (85.1%) 165 (14.9%) 228 (41.2%) 250 (45.2%) 75 (13.6%) 706 (63.8%) 400 (36.2%) P-value Adj. a Reference 0.912 0.113 Reference 0.15 Reference 0.804 0.285 Reference 0.133 Page 7 of 11 OR (95%CI) Adj. 1.02 (0.68–1.53) 2.30 (0.82–6.44) 1.24 (0.92–1.67) 1.04 (0.72–1.52) 1.33 (0.79–2.24) 1.18 (0.95–1.46) Table 4 Association of CASP8 polymorphism, haplotypes and diplotypes with breast cancer risk, the clinico-pathological features and overall survival Characteristics Polymorphism/ Haplotype/ Diplotype OR/HR (95%CI) D’ P-value Adj. a Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Breast cancer risk Age of diagnosis Age of menarche Age of menarche BMI ER (Pos. vs. Neg.) ER/PR + vs. TNBC ER (Pos. vs. Neg.) BMI rs3834129 (II + ID vs. DD) rs3834129 (ID + DD vs. II) rs2037815 (GA + GG vs. AA) rs7608692 (GA + AA vs. GG) rs3754934 (CA + AA vs. CC) rs10931936 (CT + TT vs. CC) Haplotype rs3817578-rs10931936- rs1045485 (CTG vs. Others) Haplotype rs3754934-rs3817578-rs10931936-rs1045485 (CCTG vs. Others) Haplotype rs3754934-rs3817578-rs10931936-rs1045485 (ATCG vs. Others) Haplotype rs3754934-rs381757-rs10931936-rs1045485-rs1045487 (CCCGG vs. Others) Haplotype rs3754934-rs381757-rs10931936-rs1045485-rs1045487 (CCTGG vs. Others) Haplotype rs12990906-rs3769821-rs6435074-rs3754934-rs3817578-rs10931936 (CTCCCC vs. Others) Diplotype rs3817578-rs10931936- rs1045485 (CCG-CTG) vs. Others) Diplotype rs3817578-rs10931936- rs1045485 (CCC, CTG) vs. Others) Diplotype rs3754934-rs3817578-rs10931936-rs1045485 (CCCG-CCTG vs. Others) Diplotype rs3754934-rs381757-rs10931936-rs1045485-rs1045487 (CCCGG-CCTGG) vs. Others) rs1045487(AA vs. GG) rs3834129 (Ins/Del vs. Ins/Ins) rs2037815 (AA vs. GG) rs13113 (AA vs. TT) rs3754934 (CA vs. CC) rs7608692 (AA vs. GG) Haplotype rs3754934-rs3817578-rs10931936-rs1045485 (ATCG vs. Others) Diplotype rs12990906- rs3769821-rs6435074-rs3754934-rs3817578-rs10931936 (TTCCCC- TTCCCC vs. Others) 0.034 0.014 0.031 0.006 0.004 < 0.001 < 0.001 < 0.001 0.007 0.03 0.004 0.011 0.004 < 0.001 0.007 0.019 0.022 0.034 0.026 0.029 0.008 0.039 < 0.001 0.004 Adj 1.76 (1.04–2.97) 1.58 (1.09–2.67) 1.44 (1.03–2.01) 1.47 (1.12–1.93) 0.49 (0.27–0.78) 2.06 (1.44–2.93) 1.78 (1.32–2.41) 1.75 (1.30–2.38) 0.43 (0.24–0.79) 0.77 (0.60–0.97) 1.58 (1.16–2.14) 0.71 (0.54–0.92) 2.01 (1.25–3.22) 5.04 (2.17–11.71) 1.93 (1.38–2.80) 1.78 (1.10–2.90) 0.37 (0.14–0.97) 0.83 (0.72–0.96) 0.79 (0.64–0.97) 0.92 (0.87–0.98) 0.40 (0.20–0.78) 1.56 (1.03–2.36) 0.25 (0.12–0.51) 1.04 (1.02–1.07) 0.52 0.61 0.61 0.62 0.62 0.59 0.61 Stage (Late vs. Early) Diplotype rs12990906-rs3769821-rs6435074-rs3754934-rs3817578-rs10931936 (TTCCCC- 0.017 3.21 (1.23–8.37) Her2 (Pos. vs. Neg.) CTCCCC vs. Others) Diplotype rs6435074-rs3754934-rs3817578-rs10931936-rs1045485-rs1045487 (CCCCGG, ACCTGG vs. Others) a The results were adjusted for BMI, age at first gestation, and Menopause status 0.043 1.96 (1.02–3.78) Table 5 Association of CASP8 polymorphism with overall survival Characteristics Overall survival Overall survival in Hormone receptor-positive patients Overall survival in Hormone receptor-positive patients a The results were adjusted for BMI, age at first gestation, and Menopause status Polymorphism rs3754934 (A vs. C) rs3754934 (A vs. C) rs3754934 (AA vs. CC) P-value Adj. a 0.022 0.038 0.002 OR/HR (95%CI) Adj 0.46 (0.23–0.89) 0.37 (0.14–0.95) 0.09 (0.0.2–0.43) Afzaljavan et al. BMC Medical Genomics (2023) 16:72 Page 8 of 11 with BMI (p = 0.029), rs7608692 with molecular category (p = 0.039) and rs3754934 with ER status (p = 0.008). Haplotype analysis identified a four-SNPs haplotype correlated with ER status (p < 0.001). Furthermore, three six-SNPs diplotypes were correlated with the stage of the disease (p = 0.017), HER2 status (p = 0.043), and BMI (p = 0.004). Evaluation of overall survival in patients showed that 10-year overall survival was 87% (Fig. 1A). Overall sur- vival comparison between different genetic models of rs3754934 polymorphism showed that the C allele was associated with a lower risk of death than the A allele [p = 0.022; HR = 0.46, 95% CI (0.23–0.89)] in all patients (Fig. 1B), as well as in hormone-positive group [p = 0.038; HR = 0.37, 95% CI (0.14–0.95)] (Fig. 1C). Furthermore, the CC genotype was associated with a lower risk of death than the AA genotype in the hormone-positive group [p = 0.002; HR = 0.09, 95% CI (0.02–0.43)] (Fig. 1D). However, we did not find any haplotypes and diplotypes associated with overall survival. Discussion Dysregulation of apoptosis has been well known in the pathogenesis of cancer. CASP8, as a key element of apop- tosis, has been represented with several genomic varia- tions in association with breast cancer [21]. Furthermore, its overexpression can lead to induced programmed cell death in breast tumors [22, 23]. Our results indicate variations in CASP8 are associated with the risk of breast cancer as well as clinicopathological features. Fig. 1 Overall Survival curves in total population (A and B) and in Hormone receptor-positive breast cancer patients (C and D) A: Kaplan-Meier overall survival curves of patients with breast cancer in total population; B: Kaplan-Meier overall survival curves for rs3754934 alleles (A vs. C) in all breast cancer patients; C: Kaplan-Meier overall survival curves for rs3754934 alleles (A vs. C) in Hormone receptor-positive breast cancer patients; D: Kaplan-Meier overall survival curves for rs3754934 genotypes (AA & AC vs. CC) in Hormone receptor-positive breast cancer patients Afzaljavan et al. BMC Medical Genomics (2023) 16:72 Page 9 of 11 behaviors. However, many of the molecular mechanisms of these effects are unknown and require functional stud- ies to identify common pathways and potential diagnos- tic and prognostic targets. The importance of polymorphisms is known as prog- nostic markers, as polymorphisms can play a leading role in altering the uptake and absorption of chemotherapy drugs and may influence the response to chemother- apy and, ultimately, the outcome of the disease [34, 35]. However, just CASP8 rs3754934 in the study population showed a relationship with prognosis. Previously, the association of rs3769821 [36] and rs1045485 [37] poly- morphisms with an increased risk of death in advanced lung adenocarcinoma and breast cancer, respectively, have been reported. Also, the rs3834129 deletion allele was associated with poor prognosis in the German pop- ulation, which contradicts the protective effect of this allele in breast cancer [37]. Conclusion The present study with a carefully selected range of genetic markers across the CASP8 gene region can add more evidence to the literature about the overall role of the gene in breast cancer and improve the information about the genetic basis of the disease. Based on the results of this study, which was conducted for the first time in the Northeastern female population of Iran, CASP8 gene polymorphisms, haplotypes, and diplotypes may be used as predictive markers for the risk and prognosis of breast cancer. In addition, identified haplotypes and diplotypes which carry certain risk-related alleles may have the abil- ity to be used in multigenic tests to calculate individual risk levels for personalized medicine purposes. These findings, however, suggest that there is a differ- ence in the allele frequency of considered variants in Ira- nian populations compared to Asian-related reports. This finding may indicate profound differences in the genetic background of populations and consequently different effects of alleles. Given that the eleven variants studied in this project were studied for the first time in Iran, highly- quality controlled frequencies obtained in this project can be used in calculating the appropriate sample size for future studies. However, identifying the mechanism of action of these haplotypes can also help to identify the tumorigenic process and may lead to opening new win- dows to the identification of therapeutic targets. Supplementary Information The online version contains supplementary material available at https://doi. org/10.1186/s12920-023-01484-0. Supplementary Material 1 Regarding the rs3834129, as the most prevalent vali- dated variant, I/D and D/D genotypes have been associ- ated with 1.32 times and 1.42 times lower risk of breast cancer, respectively, indicating a dose-dependent effect of deletion allele similar to the reports in a Chinese popu- lation [24]. While a large study on the Europeans found no significant outcome [25], a meta-analysis has con- firmed a reduced risk of breast cancer in association with the deletion allele, resulting in a reduction in the overall risk of cancer in the Asian and Caucasian populations but not in Africans [26]. Consistent with the association of rs7698692 A-allele carriers with a 47% increased risk of the disease in the dominant model, data from a meta- analysis study showed the association of A allele with a 35% increased risk of cancer in the Asian population [27]. In addition, rs10931936 may increase the risk of breast cancer by up to 73%, and carriers of the T allele in the dominant model also had a two-fold increased risk. In a GWAS in England, the association of rs10931936 with breast cancer was reported with a 13% increased risk (11). This result was again confirmed by a 7% increased risk in the European population [28]. However, a study on In Situ breast cancer patients reported no association between this polymorphism and breast cancer risk [29]. While A allele carriers of rs3754934 polymorphism in the dominant model had a 51% reduced risk of breast can- cer in our population, a study of this variant in the British population did not indicate a significant association [11]. Association studies have confirmed the higher statis- tical power of haplotype analyses compared with alleles or genotypes analysis itself [30, 31]. In this regard, hap- lotype analysis indicated combinations of multiple loci of CASP8, including a 3-SNPs, a 4-SNPs, and a 5-SNPs hap- lotypes, associated with 58–78% increased risk of breast cancer in the study population. In two previous studies considering different polymorphisms of CASP8, several haplotypes, including rs7608692, rs3834129, rs3817578, and rs1045485, have been reported to be associated with a 28–31% increased risk of breast cancer [11, 12]. In these studies, two polymorphisms rs3834129 and rs1045485 have been introduced as prominent risk-related variants in line with the present study. While a previous study has not provided such associa- tions [11], another research has reported some CASP8 variants related to pathological factors [32]. Considering age, associated markers may be favorable in setting up a direct-to-consumer test for early diagnosis in routine screening or assessment of prognosis. Previous findings have shown that patients diagnosed at lower ages had more aggressive features and worse prognoses than those at higher ages [33]. These results suggest that the genetic architecture of the disease may be different in older patients compared to younger, and possibly unknown genetic factors may be responsible for different tumor Afzaljavan et al. BMC Medical Genomics (2023) 16:72 Acknowledgements The authors thank all participants in this research. We would also like to thank Mashhad University of Medical Sciences, Omid, Ghaem and Imam Reza hospitals, Reza Radiotherapy & Oncology Center and Dr. Ejtehadi laboratory for supporting the project. Authors’ contributions Design the research: F. A., F. HS and A.P. Data collection: F. A., F. V., A. M., F. HS., MM. K., and MR. N. Laboratory work: F. A., E. V., M. BB., and A. H. Statistical analysis: F. A. and A. P. Manuscript draft: F. A. and A. P. All authors helped edit and approve the final version of this manuscript for submission. Funding This work was based on the PhD thesis of Dr. Fahimeh Afzaljavan and was financially supported by Mashhad University of Medical Sciences under grant 931185. Data availability The datasets generated and/or analyzed during the current study are not publicly available due Mashhad University of Medical Sciences research council rules, but are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The study was implemented in accordance with Declaration of Helsinki and relevant guidelines by the institutional ethics committee. The study was approved by the Ethical Committee of Mashhad University of Medical Sciences, Mashhad, Iran, with the Ethical approval number: IR.MUMS. REC.1394.188. Moreover, all participants signed a written informed consent approved in the Ethical Committee of Mashhad University of Medical Sciences. Consent for publication Not applicable. Competing interests The authors declare that they have no conflict of interest. Author details 1Department of Medical Genetics and Molecular Medicine, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran 2Midwifery department, Faculty of Nursing and Midwifery, Mashhad University of Medical Sciences, Mashhad, Iran 3Pharmacological Research Center of Medicinal Plants, Mashhad University of Medical Sciences, Mashhad, Iran 4Cancer research center, Mashhad University of Medical Sciences, Mashhad, Iran 5Department of Internal Medicine, Faculty of Medicine, Ghaem Medical Center, Mashhad University of Medical sciences, Mashhad, Iran 6Recombinant Protein Research Group, The Research Institute of Biotechnology, Ferdowsi University of Mashhad, Mashhad, Iran 7Bioinformatics Research Centre, Mashhad University of Medical Sciences, Mashhad, Iran Received: 27 May 2022 / Accepted: 9 March 2023 References 1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA: A Cancer Journal for Clinicians. 2021;71(3):209 – 49. 2. Nalejska E, Mączyńska E, Lewandowska MA. Prognostic and predictive bio- markers: tools in personalized oncology. Mol Diagn Ther. 2014;18(3):273–84. 3. Honrado E, Benitez J, Palacios J. The molecular pathology of hereditary breast cancer: genetic testing and therapeutic implications. Mod pathology: official J United States Can Acad Pathol Inc. 2005;18(10):1305–20. Page 10 of 11 4. 5. 6. Skol AD, Sasaki MM, Onel K. The genetics of breast cancer risk in the post- genome era: thoughts on study design to move past BRCA and towards clinical relevance. Breast Cancer Res. 2016;18(1):99. Fachal L, Dunning AM. From candidate gene studies to GWAS and post- GWAS analyses in breast cancer. Curr Opin Genet Dev. 2015;30:32–41. Lilyquist J, Ruddy KJ, Vachon CM, Couch FJ. Common Genetic Variation and Breast Cancer Risk-Past, Present, and Future. Cancer epidemiology, biomark- ers & prevention: a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. 2018;27(4):380–94. 8. 7. MacPherson G, Healey CS, Teare MD, Balasubramanian SP, Reed MW, Pharoah PD, et al. Association of a common variant of the CASP8 gene with reduced risk of breast cancer. J Natl Cancer Inst. 2004;96(24):1866–9. Cox A, Dunning AM, Garcia-Closas M, Balasubramanian S, Reed MW, Pooley KA, et al. A common coding variant in CASP8 is associated with breast cancer risk. Nat Genet. 2007;39(3):352–8. Barati Bagherabad M, Afzaljavan F, Vahednia E, Rivandi M, Vakili F, Hashemi Sadr S,Homaei Shandiz F, Pasdar A. Association of Caspase 8 promoter vari- ants and haplotypes with the risk of breast cancer and molecular profile in Iranian population: A case - control study. J Cell Biochem. 2019; https://doi. org/10.1002/jcb.28781 9. 10. Vahednia E, Homaei Shandiz F, Barati Bagherabad M, Moezzi A, Afzaljavan F, Tajbakhsh A, Kooshyar MM, Pasdar A. The Impact of CASP8 rs10931936 and rs1045485 Polymorphisms as well as the haplotypes on breast cancer Risk: A casecontrol study. Clinical Breast Cancer. 2019; pii: S1526-8209(18)30646-3. https://doi.org/10.1016/j.clbc.2019.02.011 11. Shephard ND, Abo R, Rigas SH, Frank B, Lin WY, Brock IW, et al. A breast cancer risk haplotype in the caspase-8 gene. Cancer Res. 2009;69(7):2724–8. 12. Camp NJ, Parry M, Knight S, Abo R, Elliott G, Rigas SH et al. Fine-mapping CASP8 risk variants in breast cancer. Cancer epidemiology, biomark- ers & prevention: a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. 2012;21(1):176–81. 13. Turnbull C, Ahmed S, Morrison J, Pernet D, Renwick A, Maranian M, et al. Genome-wide association study identifies five new breast cancer susceptibil- ity loci. Nature genetics. 2010;42(6):504–7. 14. Evans DGR, Eccles DM, Rahman N, Young K, Bulman M, Amir E, et al. A new scoring system for the chances of identifying a BRCA1/2 mutation outper- forms existing models including BRCAPRO. Journal of Medical Genetics. 2004;41(6):474–8. 15. Lakhani S, Ellis I, Schnitt S. al. e. WHO classification of Tumours of the breast. 4th ed. Lyon: IARC Press; 2012. 16. Kalli S, Semine A, Cohen S, Naber SP, Makim SS, Bahl M. American Joint Com- mittee on Cancer’s staging system for breast Cancer, Eighth Edition: what the Radiologist needs to know. Radiographics. 2018;38(7):1921–33. 17. Wolff AC, Hammond ME, Hicks DG, Dowsett M, McShane LM, Allison KH, et al. Recommendations for human epidermal growth factor receptor 2 testing in breast cancer: american society of clinical Oncology/College of american pathologists clinical practice guideline update. J Clin oncology: official J Am Soc Clin Oncol. 2013;31(31):3997–4013. 18. Suguna S, Nandal D, Kamble S, Bharatha A, Kunkulol R. Genomic DNA isolation from human whole blood samples by non enzymatic salting out method. Int J pharm pharm sci. 2014;6(6):198–9. 19. Stephens M, Donnelly P. A comparison of bayesian methods for haplo- type reconstruction from population genotype data. Am J Hum Genet. 2003;73(5):1162–9. 20. Zhao JH. 2LD, GENECOUNTING and HAP: computer programs for linkage disequilibrium analysis. Bioinf (Oxford England). 2004;20(8):1325–6. 21. Yin M, Yan J, Wei S, Wei Q. CASP8 polymorphisms contribute to cancer susceptibility: evidence from a meta-analysis of 23 publications with 55 individual studies. Carcinogenesis. 2010;31(5):850–7. 22. Aghababazadeh M, Dorraki N, Javan FA, Fattahi AS, Gharib M, Pasdar A. Downregulation of caspase 8 in a group of iranian breast cancer patients - a pilot study. J Egypt Natl Cancer Inst. 2017;29(4):191–5. 23. Wu Y, Alvarez M, Slamon DJ, Koeffler P, Vadgama JV. Caspase 8 and maspin are downregulated in breast cancer cells due to CpG site promoter methylation. BMC Cancer. 2010;10(1):32. 24. Sun T, Gao Y, Tan W, Ma S, Shi Y, Yao J, et al. A six-nucleotide insertion-deletion polymorphism in the CASP8 promoter is associated with susceptibility to multiple cancers. Nat Genet. 2007;39(5):605–13. Afzaljavan et al. BMC Medical Genomics (2023) 16:72 Page 11 of 11 Polymorphisms in Breast Carcinoma. Klinicka onkologie: casopis Ceske a Slovenske onkologicke spolecnosti. 2016;29(6):445–53. 33. Brandt J, Garne JP, Tengrup I, Manjer J. Age at diagnosis in relation to survival following breast cancer: a cohort study. World J Surg Oncol. 2015;13(1):33. 34. Li J, Bluth MH. Pharmacogenomics of drug metabolizing enzymes and transporters: implications for cancer therapy. Pharmgenomics Pers Med. 2011;4:11–33. 35. Sacco K, Grech G. Actionable pharmacogenetic markers for prediction and prognosis in breast cancer. EPMA J. 2015;6(1):15. 36. Liu D, Xu W, Ding X, Yang Y, Lu Y, Fei K, et al. Caspase 8 polymorphisms contribute to the prognosis of advanced lung adenocarcinoma patients after platinum-based chemotherapy. Cancer Biol Ther. 2017;18(12):948–57. 37. Kuhlmann J, Bankfalvi A, Schmid K, Callies R, Kimmig R, Wimberger P, et al. Prognostic relevance of caspase 8-652 6 N InsDel and Asp302His polymor- phisms for breast cancer. BMC Cancer. 2016;16(1):618. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 25. Frank B, Rigas SH, Bermejo JL, Wiestler M, Wagner K, Hemminki K, et al. The CASP8 -652 6 N del promoter polymorphism and breast cancer risk: a multi- center study. Breast Cancer Res Treat. 2008;111(1):139–44. 26. Chen D, Ma T, Liu X-W, Liu Z. CASP-8-652 6 N ins/del polymorphism and cancer risk: a literature-based systematic HuGE review and meta-analysis. Experimental and therapeutic medicine. 2012;4(4):762–70. 27. Zhang YJ, Zhong XP, Chen Y, Liu SR, Wu G, Liu YF. Association between CASP-8 gene polymorphisms and cancer risk in some asian population based on a HuGE review and meta-analysis. Genet Mol research: GMR. 2013;12(4):6466–76. 28. Barrdahl M, Canzian F, Joshi AD, Travis RC, Chang-Claude J, Auer PL, et al. Post- GWAS gene-environment interplay in breast cancer: results from the breast and prostate Cancer Cohort Consortium and a meta-analysis on 79,000 women. Hum Mol Genet. 2014;23(19):5260–70. 29. Campa D, Barrdahl M, Gaudet MM, Black A, Chanock SJ, Diver WR, et al. Genetic risk variants associated with in situ breast cancer. Breast Cancer Res. 2015;17(1):82. 30. Akey J, Jin L, Xiong M. Haplotypes vs single marker linkage disequilibrium tests: what do we gain? Eur J Hum Genet. 2001;9(4):291–300. 31. Zuo L, Wang K, Luo X. Use of diplotypes - matched haplotype pairs from homologous chromosomes - in gene-disease association studies. Shanghai Arch Psychiatry. 2014;26(3):165–70. 32. Brynychova V, Vaclavikova R, Kubackova K, Mrhalova M, Kodet R, Raus K et al. Clinical and Functional Importance of Selected CASP8 and CASP9 Afzaljavan et al. BMC Medical Genomics (2023) 16:72
10.1186_s12917-019-2170-8	Sirikaew et al. BMC Veterinary Research (2019) 15:419 https://doi.org/10.1186/s12917-019-2170-8 R E S E A R C H A R T I C L E Open Access Proinflammatory cytokines and lipopolysaccharides up regulate MMP-3 and MMP-13 production in Asian elephant (Elephas maximus) chondrocytes: attenuation by anti-arthritic agents Nutnicha Sirikaew1, Siriwadee Chomdej2, Siriwan Tangyuenyong3, Weerapongse Tangjitjaroen3, Chaleamchat Somgird3, Chatchote Thitaram3 and Siriwan Ongchai1* Abstract Background: Osteoarthritis (OA), the most common form of arthritic disease, results from destruction of joint cartilage and underlying bone. It affects animals, including Asian elephants (Elephas maximus) in captivity, leading to joint pain and lameness. However, publications regarding OA pathogenesis in this animal are still limited. Therefore, this study aimed to investigate the effect of proinflammatory cytokines, including interleukin-1 beta (IL- 1β), IL-17A, tumor necrosis factor-alpha (TNF-α), and oncostatin M (OSM), known mediators of OA pathogenesis, and lipopolysaccharides on the expression of cartilaginous degrading enzymes, matrix metalloproteinase (MMP)-3 and MMP-13, in elephant articular chondrocytes (ELACs) cultures. Anti-arthritic drugs and the active compounds of herbal plants were tested for their potential attenuation against overproduction of these enzymes. Results: Among the used cytokines, OSM showed the highest activation of MMP3 and MMP13 expression, especially when combined with IL-1β. The combination of IL-1β and OSM was found to activate phosphorylation of the mitogen-activated protein kinase (MAPK) pathway in ELACs. Lipopolysaccharides or cytokine-induced expressions were suppressed by pharmacologic agents used to treat OA, including dexamethasone, indomethacin, etoricoxib, and diacerein, and by three natural compounds, sesamin, andrographolide, and vanillylacetone. Conclusions: Our results revealed the cellular mechanisms underlying OA in elephant chondrocytes, which is triggered by proinflammatory cytokines or lipopolysaccharides and suppressed by common pharmacological or natural medications used to treat human OA. These results provide a more basic understanding of the pathogenesis of elephant OA, which could be useful for adequate medical treatment of OA in this animal. Keywords: Elephas maximus, Osteoarthritis, Proinflammatory cytokines, MMP-3, MMP-13 * Correspondence: siriwan.ongchai@cmu.ac.th 1Thailand Excellence Center for Tissue Engineering and Stem Cells, Department of Biochemistry, Faculty of Medicine, Chiang Mai University, 110 Intrawarorot Rd., Chiang Mai 50200, Thailand Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 2 of 13 Background Osteoarthritis (OA), the most prevalent arthritic disease, is characterized by cartilage degradation and consequent joint pain and disability [1, 2]. OA affects many species, including elephants, especially Asian elephants (Elephas maximus) kept in captivity. Excessive body weight along with the captive environment and trained behaviors are critical factors of OA pathogenesis in elephants [3, 4]. These factors disturb the equilibrium between the syn- thesis and degradation of the extracellular matrix (ECM) by chondrocytes, leading to further degradation of the ECM by matrix-degrading enzymes, especially matrix metalloproteinases (MMPs) [5]. The disturbance of this equilibrium is found particularly among captive ele- phants [6]. MMPs are a group of zinc-dependent endopeptidases that, when in excess, cause degeneration of the cartilage ECM. There has been a reported increase in MMP-3 and MMP-13 in humans and animals with OA, suggest- ing that these MMPs play a pivotal role in OA cartilage destruction [7–10]. It has previously been shown that the production of matrix-degrading enzymes is activated by proinflammatory cytokines, including interleukin-1 beta (IL-1β), IL-17A, tumor necrosis factor-alpha (TNF- α), and oncostatin M (OSM) [11–14]. In addition, the combination of OSM with other proinflammatory cyto- kines causes the greatest loss of cartilage matrix in OA [15–17]. Moreover, lipopolysaccharides (LPS), i.e., outer- membrane components of Gram-negative bacteria, con- tribute to septic arthritis and cartilage degeneration by upregulating the synthesis of catabolic factors, including proinflammatory cytokines and matrix-degrading en- zymes [18, 19]. In OA pathogenesis, cytokine-induced signal transduction involves the activation of several pathways, including those of the mitogen-activated pro- tein kinase (MAPK) family [20]. OA in elephants is caused by an imbalance of pressure on joints, which in turn is caused by a lack of exercise or an excessive body weight. This damages the cartilage, releasing inflammatory mediators and enzymes and, consequently, leading to joint inflammation. Affected el- ephants show signs of lameness and joint swelling and are reluctant to lay down because it will be difficult to stand up again. Swimming in a big pool to reduce weight bearing and administration of anti-inflammatory drugs are considered suitable treatments [21]. Current pharmacologic approaches for OA treatment aim at reducing inflammation and pain, improving joint function, and delaying disease progression. Commonly used medicines include steroids, non-steroidal anti- inflammatory drugs (NSAIDs), and disease-modifying OA drugs (DMOADs) [22], among which the most com- mon agents are dexamethasone, indomethacin, etori- coxib, and diacerein, which have been shown to inhibit the expression of MMPs such as MMP1, MMP2, MMP3, MMP9, and MMP13 [23–26]. However, these substances are associated with a high incidence of adverse effects, including gastrointestinal damage and heart failure [27]. Thus, natural product-derived compounds with anti- inflammatory activity and low toxicity have become alternative treatments for OA. Among such compounds, sesamin, andrographolide, and vanillylacetone or zinger- one have been reported to exhibit chondroprotective activity by inhibiting the expression of MMP1, MMP3, and MMP13 in chondrocytes [28–30]. It was reported that IL-1β stimulated the degradation of elephant cartilage in an explant culture model [31]. However, the existence of published studies on the cellular mechanisms of OA in elephants is limited. Therefore, the present study aimed to investigate the molecular mechanisms underlying the activation of ex- pression of MMP-3 and MMP-13 by proinflammatory cytokines and LPS in elephant articular chondrocytes (ELACs). Additionally, the ability of commonly used anti-OA medications and natural compounds to inhibit these mechanisms was investigated. The information gained from this study will be useful in improving the treatment of elephants with OA and in supporting fur- ther research on elephant degenerative arthritis, both of which are important for a better quality of life for the el- ephants and contribute to vital elephant conservation. Results Proinflammatory cytokines induced upregulation of MMP3 and MMP13 expression in ELACs culture Treatment with OSM alone resulted in a slight increase in MMP3 mRNA levels and a marked elevation of MMP13 levels. However, IL-1β, IL-17A, and TNF-α did not influence the expression of these genes in the mono- layer culture model (Fig. 1). The combination of cyto- kines OSM and TNF-α significantly induced MMP13 expression, whereas the combination of OSM and IL-1β or IL-17A tended to induce MMP3 expression. In the pellet culture model (Fig. 2), the results of individual cytokine treatments show that only TNF-α could signifi- cantly activate the expression of MMP13. Meanwhile, the results of treatments with combined cytokines dem- onstrate that OSM combined with IL-1β dramatically in- creased the expression of both MMP3 and MMP13, whereas OSM combined with TNF-α slightly induced the expression of MMP13 but not that of MMP3. Drugs and active compounds of medicinal plants inhibited cytokine-induced expression of MMP3 and MMP13 in ELACs culture The results show that medications used to treat OA in humans, such as diacerein, dexamethasone, indometh- acin, and etoricoxib, significantly attenuated MMP3 and Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 3 of 13 Fig. 1 Proinflammatory cytokines upregulate the mRNA expression of MMP3 (a) and MMP13 (b) in ELACs. The chondrocytes were treated with individual proinflammatory cytokines as follows: IL-1β (2.5 ng/mL); IL-17A (5 ng/mL); and TNF-α (5 ng/mL), or their combination with OSM (2 ng/ mL) or IL-17A (5 ng/mL), for 24 h. mRNA levels were assessed by real-time RT-PCR. Results are presented as mean ± SEM. * signifies statistical significance compared with control (p < 0.05), whereas # signifies statistical significance in relation to single-cytokine treatment (#p < 0.05) MMP13 mRNA levels in the ELACs culture (Fig. 3a and b). Likewise, natural active compounds, including sesa- min, andrographolide, and vanillylacetone, significantly suppressed the MMP3 and MMP13 mRNA levels in a dose-dependent manner (Fig. 4a and b). LPS induced the expression of MMP3 and MMP13 along with proinflammatory cytokine genes in ELACs culture The results show that LPS at a 0.125 μg/mL concentra- tion significantly increased MMP3 and MMP13 mRNA levels as well as the levels of IL1B and IL6 while increas- ing the expression of the TNF-α gene (TNFA) at a con- centration of only 0.25 μg/mL (Fig. 5). Co-treatment with LPS and anti-arthritic drugs such indomethacin, and etori- as diacerein, dexamethasone, coxib significantly suppressed MMP3 and MMP13 mRNA levels in a dose-dependent manner (Fig. 6a and b). Figure 6c illustrates the LPS-induced increase of MMP-13 protein levels in the culture media, which was significantly suppressed by dexamethasone and indo- methacin. However, the level of MMP-3 in the culture media could not be assessed using a human MMP-3 CLIA kit (data not shown). Activation of the MAPK pathway in ELACs by IL-1β combined with OSM The MAPK pathway, one of the molecular mechanisms involved in OA pathogenesis, was activated in ELACs treated with a combination of IL-1β and OSM. The re- sults show that the combined proinflammatory cytokines activated the maximum phosphorylation of p38, ERK, and JNK from 5 to 10 min, followed by its gradual de- crease after 15 min (Fig. 7). Discussion OA is the most prevalent musculoskeletal disorder in both humans and animals. Most studies on OA have fo- cused on humans, with few reports available on animals, Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 4 of 13 the mechanisms underlying OA pathogenesis. Although the pellet culture, a three-dimensional culture model, mimicked the chondrocytes’ microenvironment within cartilage tissue more accurately [32], two-dimensional monolayer cultures are a faster and simpler model for cell-based studies. They allowed for quick evaluation of the effects of several proinflammatory cytokines known to be involved in OA pathogenesis on the expressions of MMP3 and MMP13 in ELACs. The present results clearly demonstrate that ELACs are sensitive to activation by proinflammatory cytokines. Among the proinflammatory cytokines, the treatment with OSM alone strongly induced the expression of MMP13 in the monolayer cultures; TNF-α, which has been previously reported to induce the expression of MMP1, MMP3, and MMP13 in equine chondrocytes [11], caused a significant upregulation of MMP13 in the elephant chondrocyte pellet culture. IL-17A, alone or in combination with IL-1β or TNF-α, did not alter the ex- pression of MMP3 or MMP13. The treatment with a combination of IL-17A and OSM caused a slight upreg- ulation of MMP3 with no effect on MMP13. This result is inconsistent with previous studies on human cartilage cultures, which showed that the combination of IL-17A with TNF-α and OSM synergistically upregulates the ex- pression of enzymes MMP-1 and MMP-13 [33]. This cytokine is known to be increased in the serum of OA patients, in human OA pathogenesis [34]. suggesting its involvement Although IL-1β has been reported to play a key role in the OA pathogenesis of large animals by upregulating the expression of MMP-1, MMP-3, and MMP-13 en- zymes [13, 35, 36], our results clearly demonstrate that in the elephant chondrocyte pellet culture model, this cytokine could only induce the expression of MMP3 and MMP13 in combination with OSM. This result is con- sistent with a recent report suggesting that IL-1α and IL-1β are not crucial mediators of murine OA, which may explain the lack of success of IL-1-targeted therap- ies for OA [37]. Nevertheless, a previous report by our team demonstrated a great loss of hyaluronan from elephant cartilage explants treated with human recom- binant IL-1β, suggesting the catabolic potential of this cytokine via accelerating the processes of cleavage and release of ECM biomolecules from the affected cartilage tissue, leading to degenerative cartilage in OA [31]. Fig. 2 IL-1β in combination with OSM stimulates expression of MMP3 (a) and MMP13 (b) in ELAC pellets culture. ELAC pellets were treated with IL-1β or TNF-α, alone or in combination with OSM, for 3 days. The mRNA levels were assessed by real-time RT-PCR. Results are presented as mean ± SEM. signifies statistical significance compared with control (p < 0.05), whereas # signifies statistical significance in relation to single-cytokine treatment (#p < 0.05) especially elephants. Asian elephants kept in captivity frequently suffer from OA caused primarily by residing in damp buildings and being overworked by humans as well as by restricted movement, which leads to cartilage degeneration and lameness [3, 4]. Reports on the mecha- nisms underlying OA in elephants are rare. The present study used monolayer and pellet cultures of elephant chondrocytes as in vitro models to investigate OSM, which belongs to the IL-6 family, is one of the proinflammatory cytokines that contribute to inflamma- tion and cartilage destruction in degenerative arthritis [38]. OSM induces the expression of MMP1, MMP3, and MMP13 in bovine chondrocytes [12]. This cytokine has also been reported to synergize the action of other proinflammatory cytokines such as IL-1β, TNF-α, and IL- 17A, resulting in acceleration of cartilage degeneration Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 5 of 13 Fig. 3 Anti-arthritic drugs decrease the cytokines-induced expressions of MMP3 (a) and MMP13 (b) in ELACs. Chondrocytes were pre-treated with a combination of IL-1β (2.5 ng/mL) and OSM (2 ng/mL) for 2 h, after which they were treated with various concentrations of DIA (diacerein; 2.5– 10 μM), DEX (dexamethasone; 5–20 nM), INDO (indomethacin; 2.5–10 μM), and ETORI (etoricoxib; 2.5–10 μM), for 24 h. mRNA levels were assessed by real-time RT-PCR. Results are presented as mean ± SEM. signifies statistical significance compared with control (p < 0.05), whereas # signifies statistical significance in relation to the cytokines treatment group (#p < 0.05) [15–17]. In this study, in elephant chondrocytes, the com- bination of OSM with IL-1β exerted the strongest induc- tion of MMP3 and MMP13 expression in both the monolayer and pellet culture models, whereas the com- bined OSM with TNF-α only influenced the expression of MMP13. Our results suggest a cell-type specificity in response to the activation of cytokines. Additionally, all cytokines used in the present study were human recom- binant proteins, implying that their actions on elephant chondrocytes may not represent the actions of species- specific cytokines. Nevertheless, the significant enhance- ment of MMP3 and MMP13 expression achieved by the combination of OSM and IL-1β provides important infor- mation regarding the action of these cytokines in the cata- bolic processes of elephant OA, which are similar to OA pathogenesis in other animals [17, 39]. Enzymes MMP-3 and MMP-13 are members of a zinc-dependent group of endopeptidases and considered crucial for the destruction process of cartilage ECM that occurs in OA [7–10]. The present study reveals that the expression of elephant MMP13 is more sensitive to in- duction by cytokines than MMP3. Among MMPs, most studies have focused on MMP-13, a collagenase-3, which is suggested to play a critical role in both the early stages and progression of OA [9, 40]. It is overexpressed in pa- tients with OA but not in healthy patients. MMP-13 in- volves in cartilage degradation and also acts as a regulatory factor. It has been suggested that it plays a key role in controlling the onset of OA by leading chon- drocytes from a normal to a pathological state [41]. MMP-3, stromelysin-1, is a matrix-degrading enzyme found to be increased in the serum and plasma of Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 6 of 13 Fig. 4 Natural active compounds reduce the cytokines-induced mRNA levels MMP3 (a) and MMP13 (b) in ELACs. The chondrocytes were pre- treated with a combination of IL-1β (2.5 ng/mL) and OSM (2 ng/mL) for 2 h, after which they were treated with various concentrations of SE (sesamin; 0.25–1 μM), AD (andrographolide; 1.25–5 μM), and VA (vanillylacetone; 20–80 μM), for 24 h. The mRNA levels were assessed by real-time RT-PCR. Results are presented as mean ± SEM. signifies statistical significance compared with control (p < 0.05), whereas # signifies statistical significance in relation to the cytokines treatment group (#p < 0.05) humans with OA, although its levels are not directly as- sociated with OA severity [42]. Immunohistochemical assay of the synovium tissue of OA shows a high expres- sion of MMP-3, which is positively correlated to the se- verity of the disease [10]. Likewise, in this study, the high expression of these enzymes in elephant chondrocytes was demonstrated under activation by the proinflammatory cytokines re- sponsible for OA pathogenesis. Our results suggest that these enzymes, especially MMP-13, which exerts a strong response to cytokine activation, may be one of the key catabolic enzymes involved in elephant cartilage degeneration. Cytokine-induced upregulation of MMP13 mRNA levels was accompanied by an increase of MMP- 13 protein levels in the culture media. This protein was successfully measured by a test kit designed to deter- mine the level of human MMP-13, suggesting that the structures of elephant and human MMP-13 is closely re- lated. However, another test kit designed to analyze hu- man MMP-3 levels could not successfully be applied to measure the level of MMP-3 protein in elephant chon- drocytes. Therefore, we postulate that the MMP-3 pro- tein structure similarity between humans and elephants falls below the threshold of the recognizable capability of the human MMP-3 monoclonal antibody provided with the test kit. Currently, scientific evidence on OA pathogenesis in elephants is limited. Expanding information regarding the biomechanisms of the disease as well as the effective- ness of drugs will support the development of thera- peutic interventions to treat elephant OA. As such, the present study selected four drugs commonly prescribed to treat OA in humans and indomethacin, other animals, namely, dexamethasone, that may be helpful Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 7 of 13 Fig. 5 LPS induces expression of MMP3 and MMP13 (a), and proinflammatory cytokines (b) in ELACs culture. The chondrocytes were treated with LPS at various concentrations (0.125–1 μg/mL) for 24 h, then mRNA levels were assessed by real-time RT-PCR. Results are presented as mean ± SEM. signifies statistical significance compared with control (p < 0.05) etoricoxib, and diacerein. Dexamethasone is a synthetic corticosteroid previously shown to inhibit the expression of MMP3 and MMP13 in IL-1α-induced bovine chon- drocytes and suppress cytokine-induced inhibition of matrix biosynthesis in bovine cartilage [26]. NSAIDs are generally used to reduce pain and inflammation in arth- ritis through inhibition of cyclooxygenase (COX) [43]. Indomethacin is a non-selective inhibitor, whereas etori- coxib is in the COX2 selective class of NSAIDs. The former has been reported to reduce the expression of MMP1 and MMP3 in IL-1α-induced bovine chondro- cytes [23], whereas the latter has been found to decrease the levels of MMP-2 and MMP-9 [25]. Diacerein, a DMOADs, has been reported to decrease the production of IL-1-converting enzyme and IL-1β in human osteo- arthritic cartilage [44] as well as suppress the expression of MMP1, MMP3, MMP13, ADAMTS-4, and ADAMTS- 5 in IL-1β-induced bovine chondrocytes [24]. Our re- sults show that these drugs effectively suppress the ex- pression of MMP3 and MMP13 induced by the combination of IL-1β and OSM or LPS, suggesting that they exhibited an anti-arthritic potential in the elephant chondrocytes culture model. Moreover, this study demonstrates the protective ef- fect of natural compounds previously reported to have anti-arthritic properties such as sesamin, andrographo- lide, and vanillylacetone against cytokine-induced ex- pression of MMP3 and MMP13 in elephants, suggesting similarities in human and elephant OA pathogenesis, which is ameliorated by the action of these natural com- pounds. The concentration ranges of the natural com- pounds used in this study did not cause cell mortality but still effectively reduced the expression of MMP3 and MMP13 and were selected based on the results of the Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 8 of 13 Fig. 6 Anti-arthritic drugs suppressed mRNA levels of MMP3 (a) and MMP13 (b) and decreasing MMP13 protein levels (c). The chondrocytes were pre-treated with 0.5 μg/mL LPS for 2 h, after which they were treated with various concentrations of DIA (diacerein; 2.5–10 μM), DEX (dexamethasone; 5–20 nM), INDO (indomethacin; 2.5–10 μM), and ETORI (etoricoxib; 2.5–10 μM) for 24 h. mRNA levels were then assessed by real- time RT-PCR. Results are presented as mean ± SEM. signifies statistical significance compared with control (p < 0.05), whereas # signifies statistical significance in relation to the cytokines treatment group (#p < 0.05) MTT cytotoxic assay [see Additional file 1]. However, the therapeutic dose of these agents on human or animal arthritis remains unclear. Therefore, the application of these agents to human or animal arthritis must be further investigated to achieve the maximum therapeutic effect. It was reported that supplementation of sesame seed in patients with knee OA at a dose of 40 g daily for 2 Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 9 of 13 Fig. 7 Activation of the MAPK pathway in ELACs by IL-1β combined with OSM. ELACs were stimulated by the combination of IL-1β (2.5 ng/mL) and OSM (2.5 ng/mL) at the indicated time points. Cell lysates were immunoblotted to investigate the total and phosphorylated molecular forms, which indicated an active MAPK pathway. Immunoblots are represented in (a) and bar graphs (b) show the proportion between the band intensities of phosphorylated p38, ERK, and JNK over their total forms. Results are presented as mean ± SEM. signifies statistical significance compared with control (*p < 0.05) months, along with standard medical therapy, improved the disease activity by reducing serum IL-6 [45]. In papain-induced rat OA, intra-articular injection of 20 μL of 1 or 10 μM sesamin reduced cartilage distortion [28]. This compound is the most prominent lignan in sesame seed oil [46] and has been reported to exert anti- arthritic effects by reducing IL-1β-induced production of proinflammatory mediators and cartilage-degrading en- zymes MMP-1, MMP-3, and MMP-13, in human osteo- arthritic chondrocytes via suppressing phosphorylation of NF-κB p65 and IκB and activation of the Nrf2 signal- ing pathway [28, 47]. [48]. Vanillylacetone, also called zingerone, is the major component of ginger root and has known antioxidant and anti-inflammatory properties In cytokine- induced degradation of porcine cartilage explant, this compound decreased the release of MMP-13 and cartil- age matrix biomolecules into the culture media by sup- pressing the p38 and JNK MAPK signaling pathways [30]. Patients receiving one ginger extract capsule pre- pared from 2500 to 4000 mg of dried ginger rhizomes twice daily for 6 weeks showed a significant reduction of OA symptoms [49]. However, reports on the usage of vanillylacetone for anti-arthritic purposes in humans or animals are still limited. Andrographolide is a major bioactive compound of Andrographis paniculata (Burm.f.) that was found to in- hibit the expression of MMPs and inducible nitric oxide synthase in an IL-1β-induced OA model [29]. This agent reduced the productions of proinflammatory cytokines in vitro by suppressing the p38 MAPK and ERK1/2 pathways and alleviated arthritis severity in mice treated by oral administration of andrographolide 100 mg/kg/d [50]. It was reported that a combined administration of andrographolide 50 mg/kg/d and methotrexate 2 mg/kg/ week in rat arthritis induced by complete Freund’s adju- vant significantly attenuated inflammatory symptoms and reduced liver injury caused by methotrexate [51]. Andrographolide has been proposed as a new potential anti-arthritic agent [52]. Therefore, it is worth further investigating the optimal dose of this agent for arthritis treatments in animals or humans. LPS are known to in- duce infectious arthritis and contribute to low-grade inflammation in OA pathogenesis [19, 53, 54]. They en- hance the production of MMP-1, MMP-3, MMP-13, ni- tric oxide, and prostaglandin E2 in OA patients, leading to an increase in the area of cartilage destruction [55]. Likewise, the present study on elephant chondrocytes demonstrated a strong inducing effect of bacterial LPS on the expression of proinflammatory cytokine genes, including IL1B, TNFA, and IL6, together with matrix- degrading enzymes MMP3 and MMP13. These results shed light on the in vitro mechanisms of septic arthritis in an elephant chondrocyte culture model, which, when induced by LPS, showed an increased expression of pro- inflammatory cytokines and matrix-degrading enzymes. These effects were mitigated by dexamethasone, indo- methacin, etoricoxib, and diacerein. Our findings suggest Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 10 of 13 that these drugs attenuate LPS-induced inflammation and catabolic factors in both elephant and human chondrocytes. MAPK is one of the most important signaling path- ways regulating OA pathogenesis [56]. It is activated by including IL-1β and OSM proinflammatory cytokines, [12, 57], with consequent upregulation of cartilage- degrading enzyme production, including that of MMP-3 and MMP-13 [56, 58]. This study investigated the mech- anisms underlying elephant OA by treating elephant chondrocytes with a combination of IL-1β and OSM via a commercial test kit commonly used to detect cellular activation in human cells via the MAPK signaling path- way. The present study shows that this test kit was suc- cessful in revealing the effects of these cytokines on the activation of p38, ERK, and JNK phosphorylation within 5–10 min before the phosphorylated forms gradually weakened. Our results support the notion that signal transduction in elephants is similar to that in humans and that to elephant is chondrocytes. applicable test kit this Conclusions Overall, the findings of this study provide insight into in the molecular mechanisms of OA pathogenesis ELACs, which share similarities with those occurring in humans and other animals. In addition, anti-arthritic drugs commonly used to treat OA in humans and other animals were found to ameliorate the expression of fac- tors associated with arthritis, including proinflammatory cytokines and enzymes responsible for cartilage degener- ation. The present study provides data that contribute to the development of treatments for elephants with OA and support research into arthritis in this species. Methods Preparation of primary ELACs A stillborn elephant calf was caused by dystocia with no clinical appearance of joint disease in an elephant camp in Chiang Mai, Thailand. Cartilage samples from the femoral head of the stifle joint were aseptically collected within 6 h postmortem during the necropsy process, which was consented by the owner. Primary ELACs were isolated by overnight digestion with type II collagenase at 37 °C. The ELACs were washed with phosphate- buffered saline and grown in Dulbecco’s Modified Eagle Medium (DMEM) containing 10% v/v fetal calf serum (FCS), penicillin (100 U/mL), and streptomycin (100 μg/ mL) in a humidified incubator at 37 °C with 5% CO2 until confluence. Monolayer culture and cytokine treatment of ELACs ELACs at a 3 × 105 cells/well density were grown to con- fluence in DMEM containing 10% FCS. The ELACs were sustained in serum-free DMEM for 24 h, after which they were cytokines treated with proinflammatory (ProSpec, Rehovot, Israel), IL-1β (2.5 ng/mL), IL-17A (5 ng/mL), and TNF-α (5 ng/mL), either alone or in com- bination with OSM (2 ng/mL) for 24 h or with IL-17A (5 ng/mL) for 24 h. The ELACs were also treated with various concentrations of 0.125–1 μg/mL LPS (Sigma- Aldrich, U.S.A.). After 24 h, the cells were harvested, and the expression of MMP3 and MMP13 was investigated by real-time RT-PCR. Pellet culture and cytokine treatment of ELACs ELACs at 1 × 106 were centrifuged in 15 mL conical cul- ture tubes at 1500 rpm for 5 min. The pellets that formed at the bottom of the tube were cultured for seven days in 500 μl of chondrogenic medium (DMEM containing 10% FCS, 1X Insulin-Transferrin-Selenium − 7 M dexa- [59], 25 μg/mL ascorbic acid-2 phosphates, 10 methasone) in a humidified incubator at 37 °C and 5% CO2 to allow for spherical shape formation of each pel- let. The pellets were then further treated with IL-1β (5 ng/mL) and TNF-α (10 ng/mL), alone or in combination with OSM (4 ng/mL), for 3 days before being harvested for MMP3 and MMP13 mRNA expression analysis by real-time RT-PCR. Treatment with drugs and natural compounds ELACs in monolayer cultures were treated with a combin- ation of 2.5 ng/mL IL-1β and 2 ng/mL OSM or 0.5 μg/mL LPS for 2 h [60]. Following this, they were treated with drugs, including diacerein (2.5–10 μM; TRB Chemidica, Italy), dexamethasone (5–20 nM; Sigma-Aldrich, U.S.A.), indomethacin (2.5–10 μM; Sigma-Aldrich, U.S.A.), and etoricoxib (2.5–10 μM; Zuelling, Philippines) or with nat- ural bioactive compounds (Sigma-Aldrich, U.S.A.), includ- ing sesamin (0.25–1 μM), andrographolide (1.25–5 μM), and vanillylacetone (20–80 μM), for 24 h. The cells were then harvested to investigate the expression of MMP3 and MMP13 by real-time RT-PCR, and the culture media were analyzed for protein levels of MMP-3 and MMP-13. Real-time RT-PCR Total RNA was extracted from the ELACs obtained from the monolayer or pellet cultures using the Illustra RNAs- pin Mini RNA Isolation Kit (GE Healthcare Life Sciences, U.K.), according to the manufacturer’s protocol. The total (0.25 μg) the monolayer (0.5 μg) and pellet RNA of cultures was reverse transcribed into complementary DNA using the ReverTra Ace® qPCR RT Master Mix (TOYOBO, Japan). The elephant primer sequences were designed based on the NCBI Primer-BLAST tool in asso- ciation with GenBank accession numbers and synthesized by Bio Basic, Canada (Table 1). Real-time RT-PCR was performed using the SensiFAST™ SYBR No-ROX Kit Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 11 of 13 Table 1 Real-time RT-PCR primer sequences Gene MMP3 MMP13 IL1β IL6 TNFα GAPDH Primer sequence (5′-3′) Forward: AAAGGCAGGCATTTTTGGCG Reverse: AGGGTGAGGGTAGCTCTCG Forward: AGTTCCAAAGGCTACAACTT Reverse: CGCCAGAAGAATCTGTCTTT Forward: CTTGGTGCTTTCTGGTCCTTAT Reverse: AGACAAATCGCTTTTCCATCCT Forward: GGCACTGGCAGGAAACAATC Reverse: GCATTTGCAGTTGGGTCAGG Forward: ATCAGCCGTATCGCTGTCTC Reverse: CCAAAGTAGACCTGCCCAGA Forward: ATCACTGCCACCCAGAAGA Reverse: TTTCTCCAGGCGGCAGGTCAG (Bioline, U.K.). Gene expression quantification was −ΔΔCt method against the expression of based on the 2 the glyceraldehydes-3-phosphate dehydrogenase gene (GAPDH) as a housekeeping gene [61]. Measurement of MMP-3 and MMP-13 levels in the culture media The levels of MMP-3 and MMP-13 enzymes in the culture media were measured using human MMP-3 (catalog number: E-CL-H0931) and MMP-13 (catalog number: E-CL-H0127) sandwich ELISA kits (Elabscience, China), according to the manufacturer’s instructions. Briefly, 100 μl of MMP-3 or MMP-13 standard and sam- ple (culture media) was added to the monoclonal antibody against the proteins (MMP-3 or MMP-13) pre-coated mi- cro CLIA plate well, then incubated at 37 °C. After 90 min of incubation, the standard and sample were discarded, and 100 μl of a biotinylated detection antibody working solution was added to each well. The plate was incubated for 1 h at 37 °C, followed by three washings. A horseradish peroxidase conjugate (HRP) working solution was then added to each well (100 μl/well) and left to incubate at 37 °C for 30 min. After washing, 100 μl of substrate mix- ture solution was added to each well before being incu- bated in the dark for 5 min at 37 °C. The luminescence value was detected using a Synergy H4 hybrid multi-mode microplate reader (BioTek, U.S.A.), and the protein con- centrations were calculated by comparing the samples with standard curves. Western blot analysis of intracellular signaling molecules ELACs were treated with a combination of the cytokines IL-1β (2.5 ng/mL) and OSM (2.5 ng/mL) at various time points. To investigate the activation of the MAPK path- way, the cells were collected in a radioimmunoprecipita- tion assay buffer. The cell lysates were vortexed every few minutes before centrifugation at 14,000 g for 10 min at 4 °C, after which the supernatants of the cell lysate were transferred into new tubes. The cells were lysed with a sample buffer containing 5% mercaptoethanol. Equal amounts (25 μg protein) of the cell lysates were heated for 10 min at 95 °C then subjected to 13% SDS- PAGE and transferred to a nitrocellulose membrane. After blocking non-specific proteins with 5% skim milk in TBS containing 0.1% Tween 20 (TBS-T) for 1 h, the membranes were washed with TBS-T and probed with primary antibodies (Cell Signaling Technology, U.S.A.), including rabbit anti-phosphorylated-p38 MAPK anti- body, rabbit anti-phosphorylated-p44/42 MAPK anti- body, rabbit anti-phosphorylated-SAPK/JNK antibody, rabbit anti-p44/42 rabbit anti-p38 MAPK antibody, MAPK antibody, rabbit anti-SAPK/JNK antibody, and mouse anti-β-actin (Biolegend, CA), at 4 °C overnight. After being washed with TBS-T, the membranes were incubated for 1 h with the secondary antibody conju- gated with HRP anti-rabbit IgG or anti-mouse IgG at room temperature. The positive bands were visualized by enhanced chemiluminescence using the ChemiDoc system (Bio-Rad, U.S.A.). The intensity of the immuno- positive bands was calculated using the TotalLab TL120 software. Statistical analysis The results are presented as the mean ± standard error of the mean of three independent experiments. The stat- istical analysis was performed using one-way analysis of variance followed by LSD for post-hoc multiple compar- isons. A level of p < 0.05 was considered statistically significant. Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s12917-019-2170-8. Additional file 1. The effect of natural compounds on elephant articular chondrocytes viability by using MTT assay. Abbreviations ELACs: Elephant articular chondrocytes; FCS: Fetal calf serum; IL- 17A: Interleukin-17A; IL-1β: Interleukin-1beta; LPS: Lipopolysaccharides; MAPK: the mitogen-activated protein kinase; MMP: Matrix metalloproteinase; NSAIDs: Non-steroidal anti-inflammatory drugs; OA: Osteoarthritis; OSM: Oncostatin M; TNF-α: Tumor necrosis factor-alpha Acknowledgements The authors gratefully acknowledge all general support throughout the research process from Thailand Excellence Center for Tissue Engineering and Stem Cells, Department of Biochemistry, Faculty of Medicine, Chiang Mai University, Thailand. In addition, we wish to thank Miss Pianghathai Yavirach for her valuable suggestions in some parts of molecular analysis. Authors‘contributions SO and SC designed the experiments and applying for Grants; S.O. contributed as a project administrator; ST, WT, CS, and CT worked out for the ethical approval and collected the animal tissues; N.S. performed the experiments; SO and N.S. analyzed the data; N.S. and S.O. prepared the original draft of manuscript; SC, ST, WT, CS, and CT, NS and SO revised and edited the manuscript. All the authors mentioned in this manuscript have Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 12 of 13 agreed for authorship, read and approved the manuscript, and given consent for submission and subsequent publication of the manuscript. Funding This research work was supported by Thailand and the National Research Council of Thailand (Government Budget; 2015). The funder had no role in study. Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Ethics approval Animal use and all procedures in the present study were approved by the Animal Care and Use Committee, Faculty of Veterinary Medicine, Chiang Mai University, Thailand (FVM–ACUC; Ref. No. R22/2559). We obtained written informed consent to use the deceased elephant from the owner of the elephant camp. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1Thailand Excellence Center for Tissue Engineering and Stem Cells, Department of Biochemistry, Faculty of Medicine, Chiang Mai University, 110 Intrawarorot Rd., Chiang Mai 50200, Thailand. 2Department of Biology, Faculty of Science, Chiang Mai University, Chiang Mai 50200, Thailand. 3Department of Companion Animal and Wildlife Clinic, Faculty of Veterinary Medicine, Chiang Mai University, Chiang Mai 50100, Thailand. Received: 23 November 2018 Accepted: 8 November 2019 References 1. Kapoor M, Martel-Pelletier J, Lajeunesse D, Pelletier JP, Fahmi H. Role of proinflammatory cytokines in the pathophysiology of osteoarthritis. Nat Rev Rheumatol. 2011;7:1. Loeser RF, Goldring SR, Scanzello CR, Goldring MB. Osteoarthritis: a disease of the joint as an organ. Arthritis Rheum. 2012;64:6. Hittmair KM, Vielgrader HD. Radiographic diagnosis of lameness in African elephants (Loxodonta africana). Vet Radiol Ultrasound. 2000;41:6. Clubb R, Mason G. A review of the welfare of zoo elephants in Europe: RSPCA Horsham, UK; 2002. Im HJ, Li X, Muddasani P, Kim GH, Davis F, Rangan J, et al. Basic fibroblast growth factor accelerates matrix degradation via a neuro-endocrine pathway in human adult articular chondrocytes. J Cell Physiol. 2008;215:2. Egger GF, Witter K, Weissengruber G, Forstenpointner G. Articular cartilage in the knee joint of the African elephant, Loxodonta africana, Blumenbach 1797. J Morphol. 2008;269:1. Brama PA, TeKoppele JM, Beekman B, van El B, Barneveld A, van Weeren PR. Influence of development and joint pathology on stromelysin enzyme activity in equine synovial fluid. Ann Rheum Dis. 2000;59:2. Clements DN, Carter SD, Innes JF, Ollier WE, Day PJ. Analysis of normal and osteoarthritic canine cartilage mRNA expression by quantitative polymerase chain reaction. Arthritis Res Ther. 2006;8:6. 2. 3. 4. 5. 6. 7. 8. 9. Wang M, Sampson ER, Jin H, Li J, Ke QH, Im HJ, et al. MMP13 is a critical target gene during the progression of osteoarthritis. Arthritis Res Ther. 2013;15:1. 10. Chen JJ, Huang JF, Du WX, Tong PJ. Expression and significance of MMP3 in synovium of knee joint at different stage in osteoarthritis patients. Asian Pac J Trop Med. 2014;7:4. 11. Richardson DW, Dodge GR. Effects of interleukin-1β and tumor necrosis 12. factor-α on expression of matrix-related genes by cultured equine articular chondrocytes. Am J Vet Res. 2000;61:6. Li WQ, Dehnade F, Zafarullah M. Oncostatin M-induced matrix metalloproteinase and tissue inhibitor of metalloproteinase-3 genes expression in chondrocytes requires Janus kinase/STAT signaling pathway. J Immunol. 2001;166:5. 13. 14. 15. Tung J, Fenton J, Arnold C, Alexander L, Yuzbasiyan-Gurkan V, Venta P, et al. Recombinant equine interleukin-1ß induces putative mediators of articular cartilage degradation in equine chondrocytes. Can J Vet Res. 2002;66. Sylvester J, Liacini A, Li WQ, Zafarullah M. Interleukin-17 signal transduction pathways implicated in inducing matrix metalloproteinase-3, −13 and aggrecanase-1 genes in articular chondrocytes. Cell Signal. 2004;16:4. Koshy PJ, Henderson N, Logan C, Life PF, Cawston TE, Rowan AD. Interleukin 17 induces cartilage collagen breakdown: novel synergistic effects in combination with proinflammatory cytokines. Ann Rheum Dis. 2002;61:8. 16. Hui W, Rowan AD, Richards CD, Cawston TE. Oncostatin M in combination with tumor necrosis factor alpha induces cartilage damage and matrix metalloproteinase expression in vitro and in vivo. Arthritis Rheum. 2003;48:12. 17. Durigova M, Roughley PJ, Mort JS. Mechanism of proteoglycan aggregate degradation in cartilage stimulated with oncostatin M. Osteoarthr Cartil. 2008;16:1. 18. Campo GM, Avenoso A, Campo S, D'Ascola A, Traina P, Sama D, et al. Glycosaminoglycans modulate inflammation and apoptosis in LPS-treated chondrocytes. J Cell Biochem. 2009;106:1. 19. Huang ZY, Stabler T, Pei FX, Kraus VB. Both systemic and local lipopolysaccharide (LPS) burden are associated with knee OA severity and inflammation. Osteoarthr Cartil. 2016;24:10. Loeser RF, Erickson EA, Long DL. Mitogen-activated protein kinases as therapeutic targets in osteoarthritis. Curr Opin Rheumatol. 2008;20:5. 20. 21. West G. Musculoskeletal system. Biology, medicine, and surgery of elephants Ames. IA: Blackwell Publishing; 2006. 22. Qvist P, Bay-Jensen AC, Christiansen C, Dam EB, Pastoureau P, Karsdal MA. 23. 24. The disease modifying osteoarthritis drug (DMOAD): is it in the horizon? Pharmacol Res. 2008;58:1. Sadowski T, Steinmeyer J. Effects of non-steroidal antiinflammatory drugs and dexamethasone on the activity and expression of matrix metalloproteinase-1, matrix metalloproteinase-3 and tissue inhibitor of metalloproteinases-1 by bovine articular chondrocytes. Osteoarthr Cartil. 2001;9:5. Legendre F, Bogdanowicz P, Martin G, Domagala F, Leclercq S, Pujol JP, et al. Rhein, a diacerhein-derived metabolite, modulates the expression of matrix degrading enzymes and the cell proliferation of articular chondrocytes by inhibiting ERK and JNK-AP-1 dependent pathways. Clin Exp Rheumatol. 2007;25:4. 25. Yang SF, Hsieh YS, Lue KH, Chu SC, Chang IC, Lu KH. Effects of nonsteroidal anti-inflammatory drugs on the expression of urokinase plasminogen activator and inhibitor and gelatinases in the early osteoarthritic knee of humans. Clin Biochem. 2008;41:1–2. Li Y, Wang Y, Chubinskaya S, Schoeberl B, Florine E, Kopesky P, et al. Effects of insulin-like growth factor-1 and dexamethasone on cytokine-challenged cartilage: relevance to post-traumatic osteoarthritis. Osteoarthr Cartil. 2015;23:2. 26. 27. Varga Z, Sabzwari SRA, Vargova V. Cardiovascular risk of nonsteroidal anti- inflammatory drugs: an under-recognized public health issue. Cureus. 2017;9:4. 28. Phitak T, Pothacharoen P, Settakorn J, Poompimol W, Caterson B, Kongtawelert P. Chondroprotective and anti-inflammatory effects of sesamin. Phytochemistry. 2012;80. 29. Ding QH, Ji XW, Cheng Y, Yu YQ, Qi YY, Wang XH. Inhibition of matrix metalloproteinases and inducible nitric oxide synthase by andrographolide in human osteoarthritic chondrocytes. Mod Rheumatol. 2013;23:6. 31. 30. Ruangsuriya J, Budprom P, Viriyakhasem N, Kongdang P, Chokchaitaweesuk C, Sirikaew N, et al. Suppression of cartilage degradation by Zingerone involving the p38 and JNK MAPK signaling pathway. Planta Med. 2017;83:3–04. Tangyuenyong S, Viriyakhasem N, Aungsuchawan S, Peansukmanee S, Thitaram C, Kongtawelert P, et al. Catabolism of Asian elephant cartilage matrix biomolecules in explant culture. KKU Vet J. 2012;22:2. Edmondson R, Broglie JJ, Adcock AF, Yang L. Three-dimensional cell culture systems and their applications in drug discovery and cell-based biosensors. Assay Drug Dev Technol. 2014;12:4. 32. 33. Moran EM, Mullan R, McCormick J, Connolly M, Sullivan O, Fitzgerald O, et al. Human rheumatoid arthritis tissue production of IL-17A drives matrix and cartilage degradation: synergy with tumour necrosis factor-alpha, Oncostatin M and response to biologic therapies. Arthritis Res Ther. 2009;11:4. 34. Askari A, Naghizadeh MM, Homayounfar R, Shahi A, Afsarian MH, Paknahad A, et al. Increased serum levels of IL-17A and IL-23 are associated with decreased vitamin D3 and increased pain in osteoarthritis. PLoS One. 2016;11:11. Sirikaew et al. BMC Veterinary Research (2019) 15:419 Page 13 of 13 inhibition of MAPK signaling pathways in rat articular chondrocytes. Drug Dev Res. 2017;78:8. 58. Hellman NE, Spector J, Robinson J, Zuo X, Saunier S, Antignac C, et al. Matrix metalloproteinase 13 (MMP13) and tissue inhibitor of matrix metalloproteinase 1 (TIMP1), regulated by the MAPK pathway, are both necessary for Madin-Darby canine kidney tubulogenesis. J Biol Chem. 2008;283:7. Su SC, Tanimoto K, Tanne Y, Kunimatsu R, Hirose N, Mitsuyoshi T, et al. Celecoxib exerts protective effects on extracellular matrix metabolism of mandibular condylar chondrocytes under excessive mechanical stress. Osteoarthritis Cartilage. 2014;22:6. 59. 60. Zheng X, Xia C, Chen Z, Huang J, Gao F, Li G, et al. Requirement of the phosphatidylinositol 3-kinase/Akt signaling pathway for the effect of nicotine on interleukin-1beta-induced chondrocyte apoptosis in a rat model of osteoarthritis. Biochem Biophys Res Commun. 2012;423:3. 61. Al-Shanti N, Saini A, Stewart CE. Two-step versus one-step RNA-to-CT 2-step and one-step RNA-to-CT 1-step: validity, sensitivity, and efficiency. J Biomol Tech. 2009;20:3. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 35. Dvorak LD, Cook JL, Kreeger JM, Kuroki K, Tomlinson JL. Effects of carprofen and dexamethasone on canine chondrocytes in a three-dimensional culture model of osteoarthritis. Am J Vet Res. 2002;63:10. 36. Cortial D, Gouttenoire J, Rousseau C, Ronzière M-C, Piccardi N, Msika P, et al. Activation by IL-1 of bovine articular chondrocytes in culture within a 3D collagen-based scaffold. An in vitro model to address the effect of compounds with therapeutic potential in osteoarthritis. Osteoarthr Cartil. 2006;14:7. 37. Nasi S, Ea H-K, So A, Busso N. Revisiting the role of Interleukin-1 pathway in osteoarthritis: interleukin-1α and-1β, and NLRP3 Inflammasome are not involved in the pathological features of the murine Menisectomy model of osteoarthritis. Front Pharmacol. 2017;8. 38. Plater-Zyberk C, Buckton J, Thompson S, Spaull J, Zanders E, Papworth J, et al. Amelioration of arthritis in two murine models using antibodies to Oncostatin M. Arthritis Rheum. 2001;44:11. 40. 39. Rowan AD, Hui W, Cawston TE, Richards CD. Adenoviral gene transfer of interleukin-1 in combination with oncostatin M induces significant joint damage in a murine model. Am J Pathol. 2003;162:6. Kamekura S, Hoshi K, Shimoaka T, Chung U, Chikuda H, Yamada T, et al. Osteoarthritis development in novel experimental mouse models induced by knee joint instability. Osteoarthr Cartil. 2005;13:7. Li H, Wang D, Yuan Y, Min J. New insights on the MMP-13 regulatory network in the pathogenesis of early osteoarthritis. Arthritis Res Ther. 2017;19:1. 41. 42. Naito K, Takahashi M, Kushida K, Suzuki M, Ohishi T, Miura M, et al. Measurement of matrix metalloproteinases (MMPs) and tissue inhibitor of metalloproteinases-1 (TIMP-1) in patients with knee osteoarthritis: comparison with generalized osteoarthritis. Rheumatology (Oxford). 1999;38:6. 43. Yoon JB, Kim SJ, Hwang SG, Chang S, Kang SS, Chun JS. Non-steroidal anti- inflammatory drugs inhibit nitric oxide-induced apoptosis and dedifferentiation of articular chondrocytes independent of cyclooxygenase activity. J Biol Chem. 2003;278:17. 44. Moldovan F, Pelletier J, Jolicoeur F-C, Cloutier J-M, Martel-Pelletier J. 45. Diacerhein and rhein reduce the ICE-induced IL-1β and IL-18 activation in human osteoarthritic cartilage. Osteoarthr Cartil. 2000;8:3. Khadem Haghighian M, Alipoor B, Malek Mahdavi A, Eftekhar Sadat B, Asghari Jafarabadi M, Moghaddam A. Effects of sesame seed supplementation on inflammatory factors and oxidative stress biomarkers in patients with knee osteoarthritis. Acta Med Iran. 2015;53:4. 46. Murata J, Matsumoto E, Morimoto K, Koyama T, Satake H. Generation of 47. 48. triple-transgenic Forsythia cell cultures as a platform for the efficient, stable, and sustainable production of Lignans. PLoS One. 2015;10:12. Kong P, Chen G, Jiang A, Wang Y, Song C, Zhuang J, et al. Sesamin inhibits IL-1beta-stimulated inflammatory response in human osteoarthritis chondrocytes by activating Nrf2 signaling pathway. Oncotarget. 2016;7:50. Kim MK, Chung SW, Kim DH, Kim JM, Lee EK, Kim JY, et al. Modulation of age-related NF-κB activation by dietary zingerone via MAPK pathway. Exp Gerontol. 2010;45:6. 49. Altman RD, Marcussen KC. Effects of a ginger extract on knee pain in 50. 51. patients with osteoarthritis. Arthritis Rheum. 2001;44:11. Li ZZ, Tan JP, Wang LL, Li QH. Andrographolide benefits rheumatoid arthritis via inhibiting MAPK pathways. Inflammation. 2017;40:5. Li F, Li H, Luo S, Ran Y, Xie X, Wang Y, et al. Evaluation of the effect of andrographolide and methotrexate combined therapy in complete Freund's adjuvant induced arthritis with reduced hepatotoxicity. Biomed Pharmacother. 2018;106. 52. Hidalgo MA, Hancke JL, Bertoglio JC, Burgos RA. Andrographolide a new potential drug for the long term treatment of rheumatoid arthritis disease. Innovative Rheumatology: IntechOpen; 2013. 53. Blasioli DJ, Kaplan DL. The roles of catabolic factors in the development of osteoarthritis. Tissue Eng Part B Rev. 2014;20:4. 54. Huang Z, Kraus VB. Does lipopolysaccharide-mediated inflammation have a 55. 56. 57. role in OA? Nat Rev Rheumatol. 2016;12:2. Kim HA, Cho ML, Choi HY, Yoon CS, Jhun JY, Oh HJ, et al. The catabolic pathway mediated by toll-like receptors in human osteoarthritic chondrocytes. Arthritis Rheum. 2006;54:7. Shi J, Zhang C, Yi Z, Lan C. Explore the variation of MMP3, JNK, p38 MAPKs, and autophagy at the early stage of osteoarthritis. IUBMB Life. 2016;68:4. Li X, Guo Y, Huang S, He M, Liu Q, Chen W, et al. Coenzyme Q10 prevents the Interleukin-1 Beta induced inflammatory response via
10.1186_s12920-023-01452-8	Warren et al. BMC Medical Genomics (2022) 15:275 https://doi.org/10.1186/s12920-023-01452-8 BMC Medical Genomics RESEARCH Open Access Context matters in genomic data sharing: a qualitative investigation into responses from the Australian public Vanessa Warren1* , Christine Critchley1,2, Rebekah McWhirter1,3, Jarrod Walshe2 and Dianne Nicol1 From Personal Genomes Virtual. 28-30 April 2021. https://coursesandconferences.wellcomeconnectingscience.org/event/personal-genomes- virtual-conference-20210428/ Abstract Background Understanding public attitudes to genomic data sharing is widely seen as key in shaping effective governance. However, empirical research in this area often fails to capture the contextual nuances of diverse sharing practices and regulatory concerns encountered in real-world genomic data sharing. This study aimed to investigate factors affecting public attitudes to data sharing through responses to diverse genomic data sharing scenarios. Methods A set of seven empirically validated genomic data sharing scenarios reflecting a range of current practices in Australia was used in an open-ended survey of a diverse sample of the Australian public (n responses were obtained for each of the scenarios. Respondents were each allocated one scenario and asked five questions on: whether (and why/not) they would share data; what sharing would depend on; benefits and risks of sharing; risks they were willing to accept if sharing was certain to result in benefits; and what could increase their comfort about sharing and any potential risk. A thematic analysis was used to examine responses, coded and vali- dated by two blinded coders. 243). Qualitative = Results Participants indicated an overall high willingness to share genomic information, although this willingness varied considerably between different scenarios. A strong perception of benefits was reported as the foremost expla- nation for willingness to share across all scenarios. The high degree of convergence in the perception of benefits and the types of benefits identified by participants across all the scenarios suggests that the differentiation in intention to share may lie in perceptions of risk, which showed distinct patterns within and between the different scenarios. Some concerns were shared strongly across all scenarios, particularly benefit sharing, future use, and privacy. Conclusions Qualitative responses provide insight into popular assumptions regarding existing protections, concep- tions of privacy, and which trade-offs are generally acceptable. Our results indicate that public attitudes and concerns are heterogeneous and influenced by the context in which sharing takes place. The convergence of key themes such as benefits and future uses point to core concerns that must be centred in regulatory responses to genomic data sharing. *Correspondence: Vanessa Warren vanessa.warren@utas.edu.au Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Warren et al. BMC Medical Genomics (2022) 15:275 Page 2 of 16 Keywords Genomic data sharing, Benefit sharing, Future use, Commercialization, Public attitudes, Governance, Genetic data Background Significant advances have been made in human genom- ics research over the past few decades, albeit with a slower translation into tangible benefits in clinical prac- tice. Real improvements in clinical care will only arise if there is better understanding of the influences of genom- ics on health, whether at the level of the individual, the local community, or globally [1]. More work is therefore needed to better understand the consequences of genome variation on health both within and across populations. Progress relies on sharing of genomic data between indi- viduals, research laboratories and clinics globally. There are many reasons why individuals may be either enthu- siastic or concerned about having their genomic data shared [2]. Results from the global Your DNA, Your Say survey are already providing a useful account of how public views on the donation of genomic data can vary between coun- tries, and between different cohorts within countries [3]. Other findings published to date provide an indication of the types of factors that might influence public attitudes towards genomic data sharing (GDS), and how these can vary between jurisdictions. Across the board, individuals tend to have highest levels of trust in their own doctors and lowest in researchers working in private companies [4]. Between cohorts, it appears that people who consider genomic information to have special significance relative to other forms of health information (so-called genetic exceptionalists) may be more willing to donate genomic data than the rest of the population [5]. There is a large and growing body of other public opin- ion research on the attitudes of particular cohorts of individuals towards genomic data sharing. These cohorts include research participants [6], patients and their fami- lies [7, 8], particular groups of individuals within society (including Indigenous peoples [9, 10] and other margin- alised communities and cultures [11, 12]) and members of the broader public [13, 14]. A comprehensive review by Shabani et al. provides further detail on studies pub- lished before 2014 on research participant and public attitudes towards genomic data sharing [15]. There are a wide variety of locations where genomic data may be generated and used, ranging from the hos- pital clinic, to the university’s research laboratory, to the technology company’s laboratory and beyond. There is a body of public opinion research indicating that the con- text within which genomic data is generated and shared is likely to influence an individual’s attitude towards sharing [13]. In Australia for example, we already know that intention to participate in GDS changes if private companies are involved, compared with sharing within and between public research and clinical laboratories [16, 17]. Recognising the value of this extensive body of pub- lic opinion research, in formulating policy responses to GDS there is nevertheless an important gap in the lit- erature. There is a lack of research exploring how the factors influencing public attitudes towards GDS might vary depending on the circumstances within which genomic data is generated and shared. Does it mat- ter, for example, if genomic data is collected for clini- cal purposes, or for research purposes, or for a clinical trial, or for direct-to-consumer genetic testing? Does the way in which the data is used matter? Research exploring these questions has the capacity to play an important role in policy development and regulatory reform. Regulation of the collection, use and sharing of genomic data currently remains very much siloed along traditional lines. There are important regulatory distinctions in relation to: consent and non-consent; samples and data; deidentified and identified data; clin- ical and research data; public and private organisations [18]. These regulatory distinctions remain, even though in practice the hard borders between each of these cat- egories are dissolving. An understanding of differences in public attitudes in different data sharing scenarios will assist in guiding the much needed reform process. The research reported in this article focuses spe- cifically on a preliminary investigation into how broad public views on GDS can vary within a single country (Australia) depending on the context within which the genomic data is generated and shared. This research is part of a larger project exploring how Australia can make best and most responsible use of the vast amount of genomic data being generated globally, for the ben- efit of our communities, science, healthcare and the economy. We recognise that achieving this aim requires strategies to ensure fundamental human rights, public trust and freedom of research are protected and inno- vation is facilitated. The overarching aim of this pro- ject is to provide best-practice guidance for the design of regulatory and governance strategies to achieve these ends in Australia. The project includes the crea- tion and use of a series of GDS scenarios to map legal and quasi-legal facilitators and barriers to sharing, and to assess their roles in promoting public trust, using Warren et al. BMC Medical Genomics (2022) 15:275 Page 3 of 16 evidence-based processes and law reform methodology [19]. of data, storage and data security, return of results, con- sent procedures and monitoring and oversight. The aspect of the project reported in this article involved the use of simplified versions of the GDS sce- narios in semi-structured questionnaires. The aim was to determine if and/or how public attitudes towards GDS may be influenced by the context in which the genomic data is generated and shared. Method Participants An online semi-structured questionnaire was designed to obtain qualitative reactions to seven scenarios. Respond- ents were recruited by Qualtrics [20], a multinational survey and analytics company, over the period 12–20 February 2020. Qualtrics outsources recruitment to various companies that provide online panel members who have consented to participate in surveys for a small incentive, typically points that can be redeemed for prod- ucts or services. Since companies use different methods to promote survey completion response rates are diffi- cult to obtain. Recruitment employed a quota system to ensure a diversity of views were captured, using catego- ries for stratification that were either previously found or could reasonably be expected to influence attitudes to genomic data sharing in the scenarios provided [21]. All participation was anonymous. Data collection A survey instrument was constructed by the research team to obtain reactions to seven externally validated prototypical scenarios that categorize current Austral- ian GDS practices [19]. The scenarios ranged from the sharing of genomic data in clinical, research, biobank, and data repository settings, through to citizens shar- ing information generated through a direct-to-consumer company. Some previous studies have employed hypo- thetical vignettes in exploring public attitudes to GDS [21–23]; it is anticipated that the use of detailed, validated scenarios explicitly grounded in real-world data sharing practices will provide methodological rigour and strate- gic relevance to regulatory recommendations [19]. While these scenarios do not represent all current and emergent contexts in which GDS takes place, external validation confirmed that they captured a meaningful cross-section of prototypical genomic data sharing settings in Australia at the time of development [19]. The scenarios included a range of variables reflecting potential ethical, legal and social challenges identified through extensive qualitative interviews with stakeholders engaged with GDS in these contexts. These variables included the data source, pro- vider, intermediary and user, the purpose of sharing, type McWhirter et al. originally developed six scenarios [19], which were subsequently adapted for this compo- nent of the project to ensure they were accessible to a lay public with an average education level of Year 10. One scenario depicting GDS in a clinical setting for diagnos- tic purposes was repeated to compare an adult to infant patient to gauge reactions associated with making deci- sions on behalf of a minor. A summary of the seven sce- narios (S1-S7) used in this study is provided in Table 1. The full description of each scenario as read by partici- pants can be found in supplemental materials (see Addi- tional file 1). Where possible, language was simplified to be acces- sible to a lay audience, including referring to genomic data as ‘genomic information’. In addition, each scenario included definitions of concepts (e.g., accredited, whole genome sequence, medical information), entities (e.g. university ethics committee, consortium, data access committee), and systems (e.g., cloud based platform, clinical trial, data transfer agreement) that we anticipated could be unfamiliar. The terms were bolded, with par- ticipants being instructed to “hover your mouse or cur- sor over the highlighted word in the scenario” to obtain a definition. Five open-ended questions were developed to capture respondents’ views on whether or not respondents would share their genomic data in the given scenario, why they would or would not be happy to share, what their deci- sion to share would depend on, the perceived benefits and risks of sharing, which risks they would be willing to accept if sharing was certain to result in benefits, and views on what would increase comfort about sharing and any potential risks associated with sharing genomic infor- mation. This approach was employed to explore if, and/or how, participant responses varied between different shar- ing scenarios, enabling the identification of themes for further investigation with a view to developing empiri- cally supported recommendations for regulatory reform in Australia. Participants had an unlimited character count within the platform default of 20,000 characters in which to give their responses to open-ended questions. The whole survey instrument is provided in supplemen- tal materials (see Additional file 2). Before respondents received the scenario and ques- tions, they were instructed to view a two-minute video to ensure familiarity with concepts such as genomes, genes and sequences. To check that respondents were familiar with the key aspects covered in the video, six true or false questions were asked (see Additional file 2). A between groups design was employed, where respondents were randomly presented with one of the Warren et al. BMC Medical Genomics (2022) 15:275 Page 4 of 16 Table 1 Summary descriptions of the prototypical genomic data sharing scenarios Scenario number Scenario Short In-Text Title Scenario summary Number of participants S1 S2 S3 Diagnosis of infant Diagnosis of adult Association study S4 Waived consent S5 S6 Private clinical trial Biobank S7 Direct-to-consumer Clinician-led sharing of de-identified genomic and medical (phenotype) infant patient records for diagnosis of a rare condition with trusted doctors face to face and via a hospital computer system. Clinician-led sharing of de-identified genomic and medical (phenotype) adult patient records for diagnosis of a rare condition with trusted doctors face to face and via a hospital computer system. Clinician-led sharing of de-identified genomic, leftover tissue from a surgical procedure and medical records to an international consortium led genome wide association study approved by a university ethics committee. Data can only be accessed and analysed by a cloud-based platform. General though not individual research results are shared with donors via regular newsletters. A pancreatic cancer researcher who has sequenced the genomes from tissue obtained from an Australian tissue bank comprising samples from Aboriginal and Torres Strait Islander participants needs to submit the de-identified results to an international data repository for publication purposes via a cloud-based platform. A consent waiver is obtained from a univer- sity ethics committee, as consent has only been obtained for further use in pancreatic cancer research. Individual results cannot be returned despite international third-party researchers, approved by a data access committee, detecting preventable risk. A privately funded clinical trial to develop a new treatment genotypes blood samples (linked to medical records) obtained from a privately controlled biobank. Donors are able to with- draw from the trial, but not from the biobank and future use as their sample is de-identified. A data access committee approved team of Australian researchers access de-identified genome sequences from four international biobanks. A data transfer agreement stipulates that the data can only be used for specific purposes that are within the scope of the original consent provided, however most have provided broad consent. Re-identification of data is made difficult by the biobank’s use of data encryption, meaning participants cannot be contacted and will not know how their data will be used. A direct-to-consumer genetic testing customer uploads her genetic health risk report online to investigate a diagnosis and seek similar others. Her identity is protected by a private mes- saging system but is linked to actual email addresses, allowing private contact. Her data is used by the company for unknown research purposes and without acknowledgement of intellectual property rights. 40 30 41 30 32 32 40 seven scenarios. An extensive quota system was used to obtain a range of views rather than statistical rep- resentation of the Australian population. Quotas were designed to recruit a minimum of 1 male and 1 female respondent per scenario in each of the following cat- egories: age (< 50 years, ≥ 50 years), education (uni- versity educated, not university educated), indigeneity (Aboriginal and/or Torres Strait Islander, not Aborigi- nal and/or Torres Strait Islander), country of birth (Australian born, not Australian born), location (urban, rural or remote), parental status (have children, do not have children), experience of a diagnosis of a serious health condition personally (yes, no) or with an imme- diate family member (yes, no). With one exception, the quota was achieved for all scenarios and demographic characteristics. The excep- tion was that Scenario 2 and Scenario 5 were not viewed by an Aboriginal and/or Torres Strait Islander respondent (see Additional file 3). Although the num- ber of respondents receiving each scenario was not equal (range = 30–41), chi-square analyses revealed that there were no significant (at p < 0.05) differences Table 2 Pearson chi-square tests for scenario assignment by quota demographics Variable Age group Education Disease—Personal Disease—Family Cultural background Australian born χ2 4.92 3.17 7.24 9.45 9.07 6.02 df 6 6 6 6 6 6 p 0.554 0.788 0.300 0.150 0.170 0.421 Due to insufficient sample size, the test could not be performed for the Aboriginal and/or Torres Strait Islander variable between the scenario received and all demographic var- iables (see Table 2). Data analysis In seeking to identify themes and trends in partici- pant responses to the different scenarios as the starting point for further investigation, the analysis of this survey reflects a broadly post-positivist orientation. A thematic Warren et al. BMC Medical Genomics (2022) 15:275 Page 5 of 16 analysis approach [24] was used to examine the responses to the five questions initially across all scenarios and then within each. Initially two independent coders, blinded to scenarios, read all responses to extract mean- ing or themes from their content and sentence structure. Themes and their meaning were then compared by the two coders, which resulted in additional themes being created and some being merged or omitted. This coding structure was then presented to all authors for validation and comment, which resulted in no changes. The final structure was then used by both original coders to recode all comments independently. Disagreements were found for 42 responses (4.03%) and all were resolved by discus- sion between the coders. Finally, themes were examined across the seven scenarios. The final coding scheme with definitions and example quotes for each question can be found in supplemental materials (see Additional file 4). SPSS Version 26 was used to conduct descriptive statis- tics and logistic regression results. The latter involved the scenario received predicting the likelihood of sharing (yes/no/depends), perceived benefits (yes/no) and risks (yes/no). Results Participants A total of 243 members of the public took part in this survey. All but one quota was achieved, facilitating the inclusion of a range of views and attitudes (see Addi- tional file 5). All were Australian residing in all states and territories except the Northern Territory (NSW = 61, VIC = 71, QLD = 55, WA = 26, SA = 21, TAS = 6, ACT = 3). Approximately half were female (49.2%) and under the age of 50 years (46.5%), with 30% indicating that they had either an undergraduate or postgradu- ate university qualification. Most were Australian born (76.1%) and lived in an urban area (81.9%), with 2.5% identifying as Aboriginal and/or Torres Strait Islander. Over half had children (61.3%), and a minority had expe- rienced a diagnosis of a serious health condition person- ally (25.9%) or via an immediate family member (39.9%). Viewing time for video The time between clicking on the video link to mov- ing to the next page of the survey suggested that five respondents did not watch it in its entirety (i.e. seconds were < 120; range = 81.33–105.88). The mean viewing time was 203.08 s (SD = 174.71; range = 81.33–1612.05 after removing 3 extreme outliers). The majority of respondents answered all six questions correctly (74.9%), and the mean number of questions correct was 5.68 (SD = 0.65; range = 1–6). Responses Question 1: If you were the (patient/parent/research participant) would you be happy for the (doctor/ researcher) to share the results? Respondents reported a high intention to share their genomic data, with an overall mean of 71.2% across all the scenarios (Table 3). Within this result however, inten- tion to share varied dramatically between scenarios, with the highest intention to share at 90.0% in Scenario 2 (diagnosis of an adult) and the lowest at only 30.0% in Scenario 7 (direct-to-consumer). The highest inten- tion to share, found among the three clinical scenarios (S1, S2 and S3) was closely followed by the biobank and repository scenarios (S4 and S6), which were not signifi- cantly different from the three clinical scenarios in terms of intending to share versus not intending or selecting depends. Intention to share was significantly lower for both the direct-to-consumer (S7) and privately run clini- cal trial (S5) scenarios compared to all others, with one exception. That is, the tendency to respond ‘yes’ to shar- ing compared to ‘no’ for S5 was not significantly different from the researcher-led repository (S4). The privately run clinical trial (S5) was not significantly different from S7 in terms of tendency to share (or not), but the likelihood of answering “depends” (relative to “yes”) was significantly higher for the direct-to-consumer scenario (S7) com- pared to the private clinical trial (S5). Findings from the thematic analysis are characterised by a largely homogeneous perception of benefits and positive consequences, and a heterogeneous response to perceived risk and negative consequences across the dif- ferent scenarios. If yes, why would you be happy to share your genomic information in this scenario? The expectation of benefits was the dominant explana- tion for willingness to share data across all the scenarios (Fig. 1). Although some respondents, particularly in the two clinical diagnostic scenarios (S1 and S2), linked shar- ing to personal benefit, for example, “to ensure that I get the best care” (Younger Woman, S2), most framed benefit in terms of helping others, and in the researcher-led sce- narios particularly, scientific advancement “to help other people with the same issues its really great” (Younger Man, S6), “I am all for furthering the advancement of research” (Older Woman, S2). Others indicated that their willingness to share was based on a belief that “…the possible benefit of sharing the information would outweigh the risks in my estimation” (Older Woman, S2), on the basis of their trust in medical professionals and data security in the diagnostic scenar- ios (S1 and S2) “I would like to think that a doctor has my (or others) best interests at heart” (Younger Man, S2), and Warren et al. BMC Medical Genomics (2022) 15:275 Page 6 of 16 Table 3 Percentage of responses to sharing, perceived benefits and risks across scenario Scenario Clinician-led sharing of clinical genomic data for diagnosis (infant patient) Clinician-led sharing of clinical genomic data for diagnosis (adult patient) Clinician researcher-led sharing of genomic data for genome wide association study Researcher-led sharing of pre-existing genomic data for research based on waiver of consent, indigenous findings and return of results Sharing of genomic data obtained by a company-sponsored clinical trial based on participant consent Researcher-led sharing of genomic data for research from multiple sources Citizen-led sharing of genetic data from direct-to-consumer testing S1 S2 S3 S4 S5 S6 S7 Total If you were the patient/parent of the patient/research participant in this scenario, would you be happy for the doctor/researcher to share your/their results? If you did decide to share your/your child’s genomic information in this situation, do you think there would be any benefits or positive consequences? If you did decide to share your [your child’s] genomic information in this situation do you think that there would be any risks or negative consequences? Yes No Depends Yes No Yes No 82.5 2.5Y5,Y7 15.0Y7 90.0 3.3Y5,Y7 6.7Y7 87.8 4.9Y5,Y7 7.3Y7 92.5 96.7 82.9 7.5Y5 3.3Y5 17.1 30.0 70.0 30.0 70.0 31.7 68.3 n 40 30 41 73.3 6.7Y7 20.0Y7 90.0 10.0 26.7 73.3 30 62.5 21.9 15.6Y7 71.9 28.1 46.9 53.1 32 76.7 10.0Y7 30.0 30.0 13.3Y7 40.0 71.2 11.5 17.3 86.7 90.0 87.2 13.3 10.0 12.8 23.3 76.7Y7 30 40 47.5 52.5 34.2 65.8 243 Superscripts denote significant differences between a scenario and another scenario. Y(number)indicates a significant difference (at least at p < 0.05) between the category relative to a Yes response across scenarios the understanding that data would not be shared without their consent “I wouldn’t have agreed to the study in the first place if I were not happy with the terms” (Younger Woman, S5). If no, why would you not be happy to share your genomic information in this scenario? Privacy was an important explanation for not sharing data, mentioned in all scenarios except S4 (waiver of con- sent) for this question (Fig. 2). Participants expressed a desire to “…keep my details private” (Younger Woman, S3), along with concerns about re-identification and data security, particularly in the private clinical trial (S5) and biobank (S6) scenarios. Concerns about future use were also important to respondents, particularly in the direct-to-consumer scenario (S7), with respondents unwilling to share data “without further clarification on how the information will be shared into the future”(Younger Woman, S3). In a pattern that follows throughout the other ques- tions, commercialisation was a recurrent theme among respondents in the two scenarios involving private companies (S5 and S7) with respondents offering expla- nations such as “I don’t believe that my DNA should be used for commercial purposes that I am unaware of” (Younger Person, S7), and “its my information. What is stopping the company from selling that information” (Younger Man, S5). If depends, what would sharing depend on? For some respondents, the willingness to share genomic information was dependent on certain cri- teria being met or concerns being addressed (Fig. 3). While future use was a factor in determining willing- ness to share across all scenarios, it was particularly prevalent in the two associated with private industry (S5 and S7) and the researcher-led project with a con- sent waiver (S4). Respondents indicated that agreeing to share “…depends on who it is being shared with, their reason for needing it and what they will actually do with the outcome” (Older Man, S4), while others high- lighted concerns with further sharing outside the pro- posed scenario “how far does this information go? Or is it possible that company may share it further”? (Older Woman, S7). Other themes appearing in small numbers include con- cerns about commercialisation in S4 (waiver of consent), S5 (private clinical trial) and S7 (direct-to-consumer); the impact and urgency of benefits in S1 (diagnosis of infant), S4 (waiver of consent) and S7 (direct-to-consumer); and the need for explicit consent in S4 (waiver of consent), S6 (biobank) and S7 (direct-to-consumer). Warren et al. BMC Medical Genomics (2022) 15:275 Page 7 of 16 40 35 30 25 20 15 10 5 0 s e s n o p s e r f o y c n e u q e r F S1 (diagnosis of infant) S2 (diagnosis of adult) S3 (associa(cid:9)on study) S4 (consent waiver) S5 (clinical trial) S6 (biobank) S7 (DTC) Fig. 1 Distribution of themes for responses to Q1. Would be happy to share by scenario Unclear Consent Trust Privacy Collabora(cid:9)on Personal beneﬁt Beneﬁts No concerns Response themes by scenario 8 7 6 5 4 3 2 1 0 s e s n o p s e r f o y c n e u q e r F S1 (diagnosis of infant) S2 (diagnosis of adult) S3 (associa(cid:6)on study) S4 (consent waiver) S5 (clinical trial) S6 (biobank) S7 (DTC) Response themes by scenario Fig. 2 Distribution of themes for responses to Q1. Would not be happy to share by scenario General concern/fear Commercialisa(cid:6)on Consent Privacy Future use Warren et al. BMC Medical Genomics (2022) 15:275 Page 8 of 16 s e s n o p s e r f o y c n e u q e r F 6 5 4 3 2 1 0 Scenario 1 (diagnosis of infant) Scenario 2 (diagnosis of adult) Scenario 3 (associa(cid:6)on study) Scenario 4 (consent waiver) Scenario 5 (clinical trial) Scenario 6 (biobank) Scenario 7 (DTC) Response themes by scenario Consent Commercialisa(cid:6)on Beneﬁts Awareness/knowledge Type of results to be shared Governance Security of data Harm Trust Future use Privacy Unsure Fig. 3 Distribution of themes for responses to Q1. Sharing would depend by scenario Question 2: If your genomic information was shared in this situation, do you think there would be any benefits or positive consequences? Respondents saw all scenarios as likely to result in ben- efits regardless of intention to share (see Table 3); 90.0% of respondents in Scenario 7 (DTC) reported perceived benefits for sharing in their scenario, for example, despite only 30.0% of these respondents indicating an outright intention to share their genomic data. However, signifi- cantly more respondents indicated that benefits would result from the two clinical diagnostic scenarios (S1 and 2) compared to the privately run clinical trial scenario (S5). What do you think those benefits or positive conse- quences could be? Among those who believed there would be benefits or positive consequences to sharing their genomic informa- tion the strongest theme (Fig. 4) was an anticipation of cures and treatments, which participants linked to per- sonal benefits, general benefits, and in many cases both “it would definitely help to improve my health now and into the future. It also helps to identify those at risk to cer- tain conditions and enables labs to improve treatments for those issues” (Older Woman, S5). Many respondents across the scenarios showed a clear expectation that sharing data would have benefits for knowledge “I expect they would be able to learn new information or confirm previously gathered informa- tion” (Younger Woman, S5), and benefits to scientific methods and/or medical procedures, through “bet- ter use of time and resources to target the condition” (Younger Man, S2). This theme appeared in all scenar- ios despite the different purposes and methods of shar- ing genomic data described in each scenario, and was particularly strong in the clinical diagnosis scenarios (S1 and 2). Why do you think that there would be no benefits or positive consequences? Each scenario included respondents who answered that they thought there would not be any benefits or positive consequences to sharing their genomic infor- mation, with S5 (private clinical trial) returning the greatest number of ‘no’ responses (Fig. 5). Though responses were few and somewhat fragmented the- matically for this question, they generally fell under a scepticism of the benefits of sharing their genomic information or concerns regarding specific practices and potential consequences. In the former, respond- ents explained that they couldn’t imagine any benefits Warren et al. BMC Medical Genomics (2022) 15:275 Page 9 of 16 s e s n o p s e r f o y c n e u q e r F 30 25 20 15 10 5 0 S1 (diagnosis of infant) S2 (diagnosis of adult) S3 (associa(cid:10)on study) S4 (consent waiver) S5 (clinical trial) S6 (biobank) S7 (DTC) Response themes by scenario Cures or Treatments Diagnosis Awareness/Knowledge Predic(cid:10)on/Preven(cid:10)on Scien(cid:10)ﬁc Method/Medical procedures Communica(cid:10)on Unsure Unclear Fig. 4 Distribution of themes for responses to Q2. There would be benefits or positive consequences by scenario s e s n o p s e r f o y c n e u q e r F 6 5 4 3 2 1 0 S1 (diagnosis of infant) S2 (diagnosis of adult) S3 (associa(cid:8)on study) S4 (consent waiver) S5 (clinical trial) S6 (biobank) S7 (DTC) No personal beneﬁt, but others will beneﬁt Commercialisa(cid:8)on No beneﬁt Response themes by scenario Illegal Poten(cid:8)al for misuse Unclear Lack of transparency Accuracy/clinical u(cid:8)lity Privacy Unsure Fig. 5 Distribution of themes for responses to Q2. There would not be benefits or positive consequences by scenario Warren et al. BMC Medical Genomics (2022) 15:275 Page 10 of 16 s e s n o p s e r f o y c n e u q e r F 16 14 12 10 8 6 4 2 0 S1 (diagnosis of infant) S2 (diagnosis of adult) S3 (associa(cid:8)on study) S4 (consent waiver) S5 (clinical trial) S6 (biobank) S7 (DTC) Response themes by scenario Fig. 6 Distribution of themes for responses to Q3. There would not be risks or negative consequences by scenario Unclear Unsure Beneﬁts There are risks Safeguards No or negligible risk personally “I don’t have any health problems, nor do my family so what benefit would there be?” (Older Man, S5), or more broadly “what’s the point? Why are they study- ing random DNA?” (Younger Man, S4). Other responses indicated concerns about privacy “I don’t want anything shared that could identify me…” (Younger Woman, S7), lack of transparency conceal- ing undesirable motivations or outcomes “I just think if someone can’t disclose what this would be used for then there must be something not too good behind it, maybe not but this info is pretty important” (Younger Woman, S5), and suspicion of commercialisation, including potential implications for health insurance “not for me there wouldn’t be but the person who is selling it is who would benefit by making money” (Younger Woman S7), “it might not be beneficial to me as I might be charged higher private healthcare in the future” (Younger Man, S7). However, each of these themes were reported less frequently than the general theme of no benefit, which was the only response theme to appear in all scenarios. Question 3a: If your genomic information was shared in this situation, do you think there would be any risks or negative consequences? While respondents in each of the scenarios perceived risks or negative consequences they appeared far more frequently in the scenarios involving private industry (S5 and S7), although these differences were not significant. In these two scenarios close to half of the respondents (46.90% and 47.50% respectively) answered that they thought there would be risks or negative consequences to sharing their genomic data; in all the other scenarios a far greater proportion of respondents said that they perceived no risks or negative consequences, with a sig- nificant difference observed between the scenario with the fewest ‘yes’ responses, S6 (biobank), compared to S7 (direct-to-consumer) (see Table 3). Why do you think that there would be no risks or neg- ative consequences? Of those respondents who answered ‘no’ to this ques- tion, the dominant explanation was simply that they perceived no or negligible risks in their scenario “I can- not see any situation where there would be risks or conse- quences. The whole scenario looks positive to me.” (Older Woman, S4). A smaller subset of responses in this theme linked their interpretation of negligible risk to their understanding of the nature of the information being shared “genetic information doesn’t currently have the same risks as other forms of identification, such as bank details” (Younger Man, S4), or to the likelihood of risks manifesting “although it is possible to re-identify someone if you have their whole genetic sequence/information, it is far too time consuming and tedious to do so. I assume Warren et al. BMC Medical Genomics (2022) 15:275 Page 11 of 16 s e s n o p s e r f o y c n e u q e r F 8 7 6 5 4 3 2 1 0 S1 (diagnosis of infant) S2 (diagnosis of adult) S3 (associa(cid:6)on study) S4 (consent waiver) S5 (clinical trial) S6 (biobank) S7 (DTC) Response themes by scenario Unclear Unsure Risks are not known No or minimal risk Incidental ﬁndings Harm/consequences Unauthorised access Consent Privacy Future use Commercialisa(cid:6)on Fig. 7 Distribution of themes for responses to Q3. There would be risks or negative consequences by scenario most people simply would not bother to go to that effort to do so.” (Younger Woman, S2). The no or negligible risks theme was the most frequent response in all scenarios except S6 (biobank), in which responses around safeguards were returned more fre- quently (Fig. 6). The safeguards theme included a number of sub-categories, primarily confidence in de-identifica- tion protocols “I don’t see any personal risks, there are no personal identifiers” (Younger Woman, S5), but also including trust in the professionals involved in data shar- ing, confidence in existing regulatory safeguards, data security, and confidentiality. What do you think those risks or negative conse- quences could be? Privacy and future use were the most frequently iden- tified areas of risk or negative consequences across the scenarios (Fig. 7). Respondents reported concerns about future use in all scenarios except S1 (diagnosis of infant), and it was the most common theme in S6 (biobank), S7 (direct-to-consumer), and particularly S3 (association study). Risks associated with future use included sharing with unknown actors “…you never know who’s [sic] hands it could end up in” (Younger Man, S7), and for reasons outside the original purpose, which could lead to misuse “genetic information could be inappropriately released for nefarious purposes” (Older Man, S5). Privacy, which appeared in all scenarios, included con- cern both that “there is the chance of being identified…” (Younger Woman, S3) and that specific harms to the individual may result from being identified, or identifi- able, such as “negative impact on employment, insurance, medical decisions etc.” (Older Man, S4). Concern about unauthorised access was strongest in S1 (diagnosis of infant) and S5 (private clinical trial). Responses in this theme described the possibility that data may end up in the hands of unauthorised, unscru- pulous or inappropriate actors “I suppose these is always the risk that it gets mixed up with someone else’s DNA or it gets into the wrong the hands. It would need to be secure and only shared with trusted doctors and clearly labelled.” (Younger Man, S1). This was also associated with the possibility of hacking or data theft and, in direct con- trast to respondents in the preceding section, the belief that technological safeguards are not robust enough to mitigate these risks “online databases can be reasonably easily hacked, as such my information could be stolen and shared without my consent.” (Younger Person, S7). Other minor themes include commercialisation (equal first in S2 (diagnosis of adult) and S4 (waiver of consent), the anticipation of societal harms and consequences such as discrimination, the implications of confronting inci- dental findings, and concern about a lack of control with- out specific consent. Warren et al. BMC Medical Genomics (2022) 15:275 Page 12 of 16 s e s n o p s e r f o y c n e u q e r F 7 6 5 4 3 2 1 0 S1 (diagnosis of infant) S2 (diagnosis of adult) S3 (associa(cid:7)on study) S4 (consent waiver) S5 (clinical trial) S6 (biobank) S7 (DTC) Response themes by scenario Unclear Unsure Sharing to various researchers Selling data Condi(cid:7)onal Privacy Uncomfortable ﬁnding Beneﬁts Need more informa(cid:7)on Any or all risks Would not take risks There is no risk Fig. 8 Distribution of themes for responses to Q3. What risks would you be willing to take by scenario Question 3b (If YES TO 3a): If sharing your genomic information in this situation was certain to result in benefits, which risks, if any, would you be willing to take? Themes indicating that respondents would not take risks, or would only take conditional risks appeared most frequently across the scenarios (Fig. 8). ‘Would not take risks’ was the strongest theme for S3 (association study) and S5 (private clinical trial), with respondents across several scenarios simply stating that “I would never par- ticipate” (Younger Man, S4), and that they would take “no risk whatsoever” (Younger Man, S6). The acceptance of conditional risks, strongest in S4 (waiver of consent) and S7 (direct-to-consumer), was the most diverse of all the response categories in terms of its subthemes, which included conditions relating to consent “I would generally agree to share it but would need a writ- ten contract setting out what is being shared for and how my identity would be protected” (Older Man, S4), future use “I would only risk sharing my genetic information if I was able to know where it was going and who would have access to it…” (Younger Woman, S7), compensation “I would need to be fairly compensated for me to give up this sort of information” (Younger Man, S3), trust “I will try to minimise the risk by carefully share[sic] the information only on trustworthy site” (Younger Woman, S7), and reg- ulation or protocols “I would want to be across the terms and conditions” (Younger Woman, S7). An unusually high number of respondents indicated that they were unsure what risks they would accept; this theme was equal in total to those indicating accept- ance of conditional risks. ‘Unsure’ was the most frequent response theme in both S6 (biobank) and S5 (private clinical trial) along with would not take risks in the lat- ter. Responses for this theme were generally a straight- forward ‘I don’t know” (Older Woman, S2), while others indicated a need for more contextual information “not sure would have to know the risk first” (Older Woman, S5). Question 4: Please describe the most important things that could be done in this situation to make you feel more comfortable about sharing if you were the patient/parent/research participant The prospect of benefits was consistently strong across all scenarios (Fig. 9), and the most important measure to increase comfort in sharing for participants in S3 (asso- ciation study) and S6 (biobank). Most comments empha- sised the scientific and clinical benefits that may arise from sharing genomic information, for example, “your DNA is being used for the good benefit of humanity. It Warren et al. BMC Medical Genomics (2022) 15:275 Page 13 of 16 s e s n o p s e r f o y c n e u q e r F 16 14 12 10 8 6 4 2 0 S1 (diagnosis of infant) S2 (diagnosis of adult) S3 (associa(cid:8)on study) S4 (consent waiver) S5 (clinical trial) S6 (biobank) S7 (DTC) Unclear Privacy Unsure Consent Governance Transparency/communica(cid:8)on Beneﬁts Response themes by scenario Risks are inevitable Data security Return results Commercialisa(cid:8)on Future use Nothing Trust Fig. 9 Distribution of themes for responses to Q4. What measures would make you more comfortable sharing data by scenario will help improve everyone’s quality of life and may solve diseases that have lingered for years.” (Younger Man, S7). Personal benefits were mentioned in the clinician- led scenarios (S1, S2, S3) and citizen-led sharing in the direct-to-consumer scenario (S7), for example that com- fort in sharing would increase “if there are any benefits for me in health” (Younger Man, S3), however these were less frequent than benefits for all. Transparency and communication was the most impor- tant factor to respondents in S1 (diagnosis of infant), S2 (diagnosis of adult), and S4 (waiver of consent), while in S7 (direct-to-consumer) this factor was equal to privacy as a measure to increase comfort. Respondents empha- sised the need for “transparency at all stages, including information on the risks (including likelihood and con- sequences) and benefits” (Older Man, S1), with several respondents detailing the desire for clear communication around specific practices and protocols in data security, future use and the right to withdraw, among others. Assurance of privacy was a strong precursor to feel- ing more comfortable sharing genomic information in all the scenarios, and was the most important factor for respondents in S5 (private clinical trial) and S7 (direct-to- consumer), equal with transparency/communication in the latter. Comments in this theme generally emphasised the importance of anonymity: “…I do feel that if people could be reassured that their personal details would be secure that would be sufficient to ease any misgivings they may have” (Older Woman, S4), and confidentiality meas- ures providing “reassurance that personally identifying information is not what the researchers are focussed on” (Older Woman, S3). Themes around governance, future use, and data secu- rity, appeared across all scenarios though with less fre- quency, as did the theme ‘nothing’, which was particularly strong in S3 (association study) and S6 (biobank). In this theme responses largely indicated that no further meas- ures were needed as the respondents: “feel comfortable enough about it already” (Older Woman, S3), although for a small number the inverse was true in their scenario: “I honestly don’t think there is anything that could make me feel comfortable” (Younger Woman, S6). Discussion In this study, the high rate of intention to share genomic data is striking in comparison to findings on intention to share in other recent investigations; for example, less than half of the Australian sample in the Your DNA Your Say study reported a willingness to share genomic data [25]. It is not clear at this stage how this difference can be accounted for; it is possible that here the use of proto- typical scenarios provided a level of real-world grounding Warren et al. BMC Medical Genomics (2022) 15:275 Page 14 of 16 and detail that allowed respondents to contextualise their participation with greater confidence. This research reinforces previous findings that mem- bers of the public tend to value broad social benefits more than specific personal benefits [26]. The impor- tance of benefits to participants in genomic research has been reported in a number of empirical studies, particu- larly in the context of attitudes towards participation in biobanking, which involves storage and sharing of tissue, genomic information and other health information [26– 28]. It has been noted, for example, that in this context people tend to be less concerned about privacy and con- fidentiality and more about who might benefit [27, 28]. Findings on intention to share also highlight the limi- tations of investigating public responses to genomic data sharing without accounting for the setting in which sharing occurs; while the overall mean intention to share is high in this study, it masks the dramatic varia- tion between the lowest and highest rates of intention to share between the individual scenarios. The high degree of convergence both in the importance of benefits, and in the main types of benefits identified by respondents regardless of scenario suggests that the differentiation in intention to share between the scenarios may lie in the perceived risks, and the extent to which measures reported to increase participant comfort are addressed. The diverse responses to risk in this study indicate that not only does intention to share vary across different sce- narios, but that the factors that influence this intention vary in response to the context in which sharing takes place, highlighting the importance of further examination into how regulatory responses to genomic data sharing can account for public understandings and assessment of risk. Responding to generalized assumptions about risk and benefit without understanding the concerns specific to the sharing context may entrench the development or perpetuation of regulatory responses that are inappro- priate or ineffective in fostering participant confidence, and thus successful data sharing practices. For example, in Australian research review procedures, risk assess- ment is almost wholly concentrated on personal risks to the individual participant; almost no account is taken of perceived societal harms (such as racial discrimination, commercial exploitation), and the impact this may have on participant consent in specific scenarios. This has the potential to be particularly problematic for the secondary use of data through waivers of consent, where research ethics bodies make decisions about the secondary use of genomic data on behalf of participants within this limited framework of assessment. Similarly, in the two scenarios involving private indus- try (S5 and S7), which reported significantly lower inten- tion to share and more frequent perceptions of risk compared to the other scenarios (see Table 3), some respondents indicated that it was not the act of sharing itself that troubled them necessarily, but a suspicion of the motivations of commercial actors and the future con- sequences, both broad and specific, that may eventuate from sharing their data. For these respondents, participa- tion in scenarios involving private industry was perceived as a zero-sum game in which perceived commercial exploitation inherently undermined potential personal or public benefits. By contrast, fears of commercial exploita- tion and societal harms are less prevalent in Scenarios 3 and 6, which largely fit into more widely accepted con- ceptions of genomic research that are more likely to be perceived as altruistically motivated and serving the public good [17]. This finding is consistent with previ- ous research around public perceptions of commercial involvement in genomic data sharing and biobanking, which have similarly identified lower intention to share with, and greater suspicion of, commercial stakeholders [16, 17, 25]. The persistence of this finding both locally and internationally suggests an intrinsic discomfort or “natural prejudice” [29] regarding private enterprise in genomic data sharing that presents ongoing challenges for a research environment increasingly intertwined with commercial interests and involvement [30, 31]. We are particularly interested in further investigat- ing the consistent presence of future use in respondent risk perception. Though it appeared in differing levels across the scenarios, it was the most frequently reported risk overall (equal to privacy), indicating that concern about future use is common to members of the public regardless of the sharing context. In this study concerns about future use were not so related to malicious mis- use—though this does appear—but the unknown quality of future uses and actors that, while possibly legitimate within the ethical and regulatory frameworks that govern GDS, may nonetheless conflict with the respondents’ val- ues, priorities and thresholds for comfort. This presents a particularly complex regulatory challenge, given that it this same unknown quality of secondary use that informs current arguments towards open and broad consent practices in genomic data sharing [32]. It appears from this preliminary investigation that greater transparency around future use, along with clear communication of both benefits and regulatory safeguards are likely to con- tribute to ameliorating this concern; the character and relationship of these factors is likely to be a significant feature of subsequent investigation in this project. Limitations As an exploratory investigation with a relatively small sample size the study has some inherent limitations. For example, the sampling characteristics were not Warren et al. BMC Medical Genomics (2022) 15:275 Page 15 of 16 exhaustive and do not represent every population sub- group that may have unique perspectives on GDS, nor does the analysis attempt to examine the intersection of both the participants’ personal and social context and the context in which data sharing occurs. Further, despite the affordances of recruitment efficiency, the degree of opac- ity in participant selection through paid research panel operations can make it difficult to account for potential sampling biases. In this instance, a sample drawn from a paid research panel cohort may reasonably be expected to display a greater comfort in participating in other types of research; participants in this study did show a greater willingness to share genomic data compared to similar studies, though it is not clear if this finding can be explained by the recruitment method. While the between-groups design of the survey facili- tates a comparison of responses between the scenarios, it does potentially limit analysis compared to a within-groups design in which each participant would respond to all seven scenarios. However, given the technical complexity and length of the scenarios we determined that a between- groups design would minimise the likelihood of attrition and participant fatigue while still facilitating useful com- parative data. That said, the depth of the written survey responses are likely to be limited in comparison to more robust qualitative methods such as semi-structured inter- views. This is reflected in the brevity of some responses in the open-ended questions, which despite the maxi- mal character allowance had a mean count of 19 words per response. However, as a preliminary study, the initial findings do provide insight and justification for further scenario-based investigations that are likely to generate a richer dataset for deeper analyses and recommendations. Conclusion In this study we sought to initiate an exploratory quali- tative investigation into how members of the Austral- ian public respond to genomic data sharing in different sharing contexts, using validated prototypical scenarios. Our findings indicate observable patterns and diversity among the themes found within and between different sharing scenarios, as well as key themes that converge across the scenarios. These findings suggest that public responses to genomic data sharing are more complex than might be indicated by broad empirical research that does not take context of sharing into account; in particular, that the landscape of perceived risk is more diverse and contextual than might otherwise be understood without the comparative scenario-based analyses. These initial findings indicate the need for further investigation to explore public responses to and expectations of different GDS contexts in more depth, particularly if such analyses may contribute to the development of contextually appropriate recommenda- tions for regulatory approaches. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12920- 023- 01452-8. Additional file 1. Title: Full text of validated scenarios used in survey. Description: Full descriptive text of stakeholder validated prototypical surveys that were presented to survey respondents. Additional file 2. Title: Full survey instrument. Description: Full text of the survey instrument presented to respondents. Additional file 3. Title: Demographic distribution by scenario. Description: Presents demographic characteristics of the participant sample for each scenario. Additional file 4. Title: Glossary of results – Themes and sub themes. Description: Glossary of themes and subthemes including example quotes. Additional file 5. Title: Participant characteristics. Description: Full breakdown of participant characteristics as (n) and (%) within the whole sample. Acknowledgements We thank anonymous survey participants and other members of our project team. Authors’ information CC and DN were the Australian contributors to the Your DNA Your Say global survey. About this supplement This article has been published as part of BMC Medical Genomics Volume 15 Supplement 3, 2022: Personal Genomes: Accessing, Sharing and Interpretation During Pandemic Times. The full contents of the supplement are available online at https:// bmcme dgeno mics. biome dcent ral. com/ artic les/ suppl ements/ volume- 15- suppl ement-3. Author contributions Conceptualization: CC, RM, DN; Formal Analysis: VW, CC, JW; Funding acquisition: DN, CC; Methodology: CC, JW, VW; Project administration: DN; Writing—original draft: VW, RM, DN; Writing—review & editing: JW This work is dedicated to the memory of Prof Christine Critchley. All other authors read and approved the final manuscript. Funding This project was funded by the Australian Research Council Discovery Grants Scheme, DP180100269. Availability of data and materials The datasets generated and analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This study was approved as low risk by the Human Research Ethics Commit- tees at Swinburne University of Technology (reference number 20191425- 3113) and the University of Tasmania (reference number H0018566). All partici- pants provided written consent before accessing the survey. All participation was anonymous. Consent for publication Not applicable. Warren et al. BMC Medical Genomics (2022) 15:275 Page 16 of 16 Competing interests The authors declare that they have no competing interests. Author details 1 School of Law, University of Tasmania, Sandy Bay, TAS, Australia. 2 School of Health Science, Swinburne University of Technology, Hawthorn, VIC, Aus- tralia. 3 School of Medicine, Deakin University, Waurn Ponds, VIC, Australia. Received: 15 October 2021 Accepted: 1 February 2023 References 1. Middleton A. Society and personal genome data. Hum Mol Genet. 2018;1(27):R8–13. 2. Middleton A, Milne R, Thorogood A, Kleiderman E, Niemiec E, Prainsack B, et al. Attitudes of publics who are unwilling to donate DNA data for research. Eur J Med Genet. 2019;62(5):316–23. 3. Middleton A, Your DNA. Your DNA, Your Say. New Bioeth. 2017;23(1):74–80. 19. McWhirter R, Eckstein L, Chalmers D, Critchley C, Nielsen J, Otlowski M, et al. A scenario-based methodology for analyzing the ethical, legal, and social issues in genomic data sharing. J Empir Res Hum Res Ethics. 2020;15(4):355–64. 20. Qualtrics. Qualtrics. 2020 ed. Provo, Utah, USA2005. 21. Milne R, Morley KI, Howard H, et al. Trust in genomic data sharing among members of the general public in the UK, USA, Canada and Australia. Hum Genet. 2019;138:1237–46. 22. McCormack P, Kole A, Gainotti S, et al. ‘You should at least ask’. The expectations, hopes and fears of rare disease patients on large-scale data and biomaterial sharing for genomics research. Eur J Hum Genet. 2016;24:1403–8. 23. Haga SB, O’Daniel J. Public perspectives regarding data-sharing practices in genomics research. Public Health Genomics. 2011;14:319–24. 24. Braun V, Clarke V. Using thematic analysis in psychology. Qual Res Psychol. 2006;3(2):77–101. 25. Middleton A, Milne R, Almarri M, Anwer S, Atutornu J, Baranova E, et al. Global public perceptions of genomic data sharing: what shapes the willingness to donate DNA and health data? Am J Hum Genet. 2020;107(4):743–52. 26. Nicol D, Critchley C. Benefit sharing and biobanking in Australia. Public 4. Milne R, Morley KI, Howard H, Niemiec E, Nicol D, Critchley C, et al. Trust in Underst Sci. 2012;21(5):534–55. 27. Pullman D, Etchegary H, Gallagher K, Hodgkinson K, Keough M, Morgan D, et al. Personal privacy, public benefits, and biobanks: a conjoint analysis of policy priorities and public perceptions. Genet Med. 2012;14(2):229–35. 28. Critchley CR, Nicol D, McWhirter RE. Identifying public expectations of genetic biobanks. Public Underst Sci. 2017;26:671–87. 29. Nicol D, Critchley C, McWhirter R, Whitton T. Understanding public reac- tions to commercialization of biobanks and use of biobank resources. Soc Sci Med. 2016;162:79–87. 30. Critchley C, Nicol D. Commercialisation of genomic research: the issue of public trust. In: Freckleton I, Petersen K, editors. Tensions and traumas in health law. Australia: The Federation Press; 2017. p. 350–66. 31. Eckstein L, Chalmers D, Critchley C, Jeanneret R, McWhirter R, Nielsen J, et al. Australia: regulating genomic data sharing to promote public trust. Hum Genet. 2018;137(8):583–91. 32. Grady C, Eckstein L, Berkman B, Brock D, Cook-Deegan R, Fullerton SM, et al. Broad consent for research with biological samples: workshop conclusions. Am J Bioeth. 2015;15(9):34–42. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. genomic data sharing among members of the general public in the UK, USA, Canada and Australia. Hum Genet. 2019;138(11):1237–46. 5. Middleton A, Milne R, Howard H, Niemiec E, Robarts L, Critchley C, et al. Members of the public in the USA, UK, Canada and Australia expressing genetic exceptionalism say they are more willing to donate genomic data. Eur J Hum Genet. 2020;28(4):424–34. 6. Goodman D, Johnson CO, Bowen D, Smith M, Wenzel L, Edwards K. De- identified genomic data sharing: the research participant perspective. J Commun Genet. 2017;8(3):171–81. 7. Hassan L, Dalton A, Hammond C, Tully MP. A deliberative study of public attitudes towards sharing genomic data within NHS genomic medicine services in England. Public Underst Sci. 2020;29(7):702–17. Takashima K, Maru Y, Mori S, Mano H, Noda T, Muto K. Ethical concerns on sharing genomic data including patients’ family members. BMC Med Ethics. 2018;19(1):1–6. 8. 9. Hudson M, Nanibaa’A G, Sterling R, Caron NR, Fox K, Yracheta J, et al. Rights, interests and expectations: Indigenous perspectives on unre- stricted access to genomic data. Nat Rev Genet. 2020;21(6):377–84. 10. Nanibaa’A G, Barton KS, Porter KM, Mai T, Burke W, Carroll SR. Access and management: indigenous perspectives on genomic data sharing. Ethn Dis. 2019;29(Suppl 3):659. 11. Mulrine S, Blell M, Murtagh M. Beyond trust: amplifying unheard voices on concerns about harm resulting from health data-sharing. Med Access Point Care. 2021. https:// doi. org/ 10. 1177/ 23992 02621 10484 21. 12. Merson L, Phong TV, Nhan LNT, Dung NT, Ngan TTD, Kinh NV, et al. Trust, respect, and reciprocity: informing culturally appropriate data-sharing practice in Vietnam. J Empir Res Hum Res Ethics. 2015;10(3):251–63. 13. Briscoe F, Ajunwa I, Gaddis A, McCormick J. Evolving public views on the value of one’s DNA and expectations for genomic database governance: results from a national survey. PLoS ONE. 2020;15(3):e0229044. 14. Majumder MA, Cook-Deegan R, McGuire AL. Beyond our bor- ders? Public resistance to global genomic data sharing. PLoS Biol. 2016;14(11):e2000206. 15. Shabani M, Bezuidenhout L, Borry P. Attitudes of research participants and the general public towards genomic data sharing: a systematic literature review. Expert Rev Mol Diagn. 2014;14(8):1053–65. 16. Caulfield T, Burningham S, Joly Y, Master Z, Shabani M, Borry P, et al. A review of the key issues associated with the commercialization of biobanks. J Law Biosci. 2014;1(1):94–110. 17. Critchley CR, Nicol D, Otlowski MFA. The oimpact of commercialisa- tion and genetic data sharing arrangements on public trust and the intention to participate in biobank research. Public Health Genomics. 2015;18:160–72. 18. McWhirter R, Eckstein L, Chalmers D, Kaye J, Nielsen J, Otlowski, M et al. Essentially Ours: Assessing the Regulation of the Collection and Use of Genomic Information (Hobart: Centre for Law and Genetics Occasional Paper No 11, 2022), <https:// www. utas. edu. au/ law- and- genet ics/ publi catio ns/ occas ional- papers>. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12977-021-00568-y	Md Zahid et al. Retrovirology (2021) 18:23 https://doi.org/10.1186/s12977-021-00568-y RESEARCH Retrovirology Open Access Functional analysis of a monoclonal antibody reactive against the C1C2 of Env obtained from a patient infected with HIV-1 CRF02_AG Hasan Md Zahid1, Takeo Kuwata1, Shokichi Takahama1,2, Yu Kaku1, Shashwata Biswas1, Kaho Matsumoto1, Hirokazu Tamamura3 and Shuzo Matsushita1* Abstract Background: Recent data suggest the importance of non-neutralizing antibodies (nnAbs) in the development of vaccines against HIV-1 because two types of nnAbs that recognize the coreceptor binding site (CoRBS) and the C1C2 region mediate antibody-dependent cellular-cytotoxicity (ADCC) against HIV-1-infected cells. However, many studies have been conducted with nnAbs obtained from subtype B-infected individuals, with few studies in patients with non-subtype B infections. Results: We isolated a monoclonal antibody 1E5 from a CRF02_AG-infected individual and constructed two forms of antibody with constant regions of IgG1 or IgG3. The epitope of 1E5 belongs to the C1C2 of gp120, and 1E5 binds to 27 out of 35 strains (77 %) across the subtypes. The 1E5 showed strong ADCC activity, especially in the form of IgG3 in the presence of small CD4-mimetic compounds (CD4mc) and 4E9C (anti-CoRBS antibody), but did not show any neutralizing activity even against the isolates with strong binding activities. The enhancement in the binding of A32, anti-C1C2 antibody isolated from a patient with subtype B infection, was observed in the presence of 1E5 and the combination of 1E5, A32 and 4E9C mediated a strong ADCC activity. Conclusions: These results suggest that anti-C1C2 antibodies that are induced in patients with different HIV-1 subtype infections have common functional modality and may have unexpected interactions. These data may have implications for vaccine development against HIV-1. Keywords: HIV-1, Non-neutralizing antibody, C1C2 antibody, ADCC, Non-subtype B, CRF02_AG Background The human immunodeficiency virus (HIV-1) envelope glycoprotein trimer (Env) is exposed on the surface of both virions and infected cells. Thus, Env is the princi- pal target for neutralizing antibodies and antibodies Correspondence: shuzo@kumamoto-u.ac.jp 1 Division of Clinical Retrovirology, Joint Research Center for Human Retrovirus infection, Kumamoto University, 2-2-1 Honjo, Chuo-ku, Kumamoto 860-0811, Japan Full list of author information is available at the end of the article able to mediate antibody-dependent cellular cytotoxic- ity (ADCC). HIV-1 Env is a flexible molecule that exists in at least three different conformational states: states 1, 2 and 3 [1]. Before interacting with the primary recep- tor, CD4, Env preferentially adopts a compact, “closed” conformation (state 1) that is largely antibody-resistant. CD4 binding “opens” Env, increasing the vulnerability of infected cells to ADCC mediated by non-neutralizing antibodies (nnAbs), as these easily-elicited antibodies preferentially recognize epitopes exposed in the open conformational states (states 2/3). These antibodies © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecom- mons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Md Zahid et al. Retrovirology (2021) 18:23 Page 2 of 15 include the anti-coreceptor binding site (CoRBS) and the anti-C1C2 families of antibodies that, in combination with a small molecule that mimics CD4 (CD4mc), stabi- lize a new asymmetric Env conformation (state 2A) that is vulnerable to ADCC [1]. Approaches aimed at stabi- lizing this “open” conformation represent new interven- tional approaches to fight HIV-1 infection. ADCC can play a major role in limiting the infection and replication of HIV-1 [2–5]. Data on the correlates of protection in the RV144 vaccine trial suggested that in protected vaccinees, increased ADCC activity resulted into decreased HIV-1 acquisition [6]. Antibodies bind- ing to epitopes in the C1C2 region and mediating potent ADCC were isolated from some RV144 vaccinees [7]. Most of the nnAbs mediating ADCC require Env in a CD4-bound conformation [8] and target epitopes that overlap epitopes recognized by the anti-C1C2 antibody, such as A32 [9–11]. These CD4-induced (CD4i) immu- noglobulins (IgGs) are present in the sera, breast milk and cervicovaginal lavages of HIV-1-infected patients [12, 13]. CD4mc and anti-CoRBS antibodies (Abs) bind sequen- tially to Env trimer, opening its conformation and allow- ing recognition by anti-C1C2 antibodies whose epitopes are located in the gp120 inner domain and remain occluded in the native trimer [10, 14–17]. It is not only Fab fragments that interact with the same Env trimer, but Fc fragments of these two families of Abs also bind syn- ergistically with FcγRIIIa [18]. Fc-dependent mechanisms can impact on the viral load [19, 20] and slow disease progression by controlling HIV-1 infection [21, 22]. Recent reports have described the superior nature of the IgG3 class of antibodies over the IgG1 class, not only for Fc-receptor-mediated functions, such as ADCC and antibody-dependent cellular phagocytosis (ADCP) [21, 23–27], but also for their neutralization capability [25]. The level of IgG3 correlated with anti-HIV-1 function in the RV-144 trial [24, 27] and was involved in the control of disease in different cohorts [21, 28]. The hinge region of IgG3, which links the Fab and the Fc regions, is two to four times longer than that of IgG1. This increased hinge length may have an effect on the flexibility of the antibody and the recognition of antigen, which would ultimately result into differences in protection [29, 30]. Although IgG3 has higher functional potential, it has not been advanced clinically partly because of its shorter half-life compared with IgG1 [31, 32]. CRF02_AG comprises 46% of the circulating strains in West Africa and is the fourth most abundant sub- type in the world [33]. It is also considered as the most frequent non-B subtype spreading among European natives [34]. In this study we isolated a monoclonal antibody (mAb) against the C1C2 epitope, 1E5, from a CRF02_AG-infected patient. IgG1 and IgG3 forms of 1E5 were constructed and examined for their functional characteristics. Previous studies on ADCC have focused from subtype B-infected on anti-C1C2 antibodies patients, such as A32 and C11, which are also known as anti-cluster A antibodies [11, 12, 35]. We observed the ADCC activity of 1E5 against cells expressing HIV-1 Env, although no neutralizing activity of 1E5 was detected against HIV-1. Moreover, enhancement in the binding and ADCC activities of A32 were observed in the pres- ence of 1E5. The combination of anti-C1C2 antibodies induced by different subtypes may have implications for vaccine development against HIV-1. Results Isolation of monoclonal antibody 1E5 from a donor infected with CRF02_AG B cells from a donor infected with CRF02_AG were transformed by Epstein–Barr virus (EBV) and the super- natants were screened for reactivity to the Env of HIV-1 93TH966.8 (CRF01_AE) strain. Recombinant mAb, 1E5, was isolated from the single cell-sorted Env-reactive cul- ture by RT-PCR of the immunoglobulin heavy and light chain genes. Genetic analysis revealed that 1E5 used IGHV1-6909 and IGKV3-2001 as germline genes of the heavy and light chain genes, respectively (Additional file 1: Fig. S1). The binding activity of 1E5 to Env pro- teins from various HIV-1 strains was examined by flow cytometry (Fig. 1a). This mAb bound well to the Env pro- tein of subtype A (92UG037.8), CRF01_AE (93TH976.17 and 93TH966.8), subtype C (ZM233M.PB6) and subtype B (WITO4160.33 and RHPA4259.7). The results sug- gested that 1E5 cross-reacts to HIV-1 strains belonging to various subtypes, although the Env proteins of sub- type C (ZM109F.PB4) and subtype B (REJO4541.67 and JR-FL) were not recognized by 1E5. The binding activity of 1E5 to intra-subtype strains was tested using a panel of CRF02_AG Env proteins (Fig. 1b). The binding of 1E5 was observed in most CRF02_AG strains (13 out of 15 strains), suggesting that the 1E5 epitope is conserved among most CRF02_AG strains. However, the reactivity of 1E5 was moderate for AG-250, AG-258, AG-278 and AG-263, and was not detected for AG-235 and AG-928. IgG3 form of 1E5 showed better reactivity than the IgG1 form Recent reports described the superior nature of the IgG3 class over IgG1 not only for Fc-receptor-mediated func- tions, such as ADCC and ADCP, but also for neutraliza- tion capability [21, 23–27]. To obtain 1E5 with superior activities, we constructed an IgG3 form of 1E5 in addi- tion to the IgG1 form. Analysis of cross-reactivity using a global panel of HIV-1 [36] showed that both IgG1 and Md Zahid et al. Retrovirology (2021) 18:23 Page 3 of 15 a b t n u o C t n u o C NHG VRC01 1E5 92UG037.8 A 93TH966.8 CRF01_AE 93TH976.17 CRF01_AE ZM233M.PB6 C ZM109F.PB C JR-FL B REJO4541.67 B RHPA4259.7 B WITO4160.3 B Anti-human IgG-APC AG-33 AG-242 AG-252 AG-253 AG-255 AG-257 AG-266 AG-271 AG-280 AG-250 AG-258 AG-278 AG-263 AG-235 AG-928 Anti-human IgG-APC Fig. 1 Recognition of HIV-1 Env proteins by 1E5. HEK 293T cells were transfected with plasmid expressing EGFP and Env proteins of subtypes A, AE, B and C (a) and those of a panel of CRF02_AG (b). At 48 h post-transfection, cells were stained with primary antibody. Then, cells were fluorescently labeled with an allophycocyanin (APC)-conjugated anti-human IgG secondary Ab. Histograms of APC in EGFP+ cells stained with 1E5, normal human globulin (NHG) as a negative control, and VRC01 as a positive control, are shown by a blue line, gray shading and a red line, respectively IgG3 forms of 1E5 reacted to 8 out of 11 strains (Fig. 2). The reactivity of 1E5 appeared high for subtype A and AE strains, such as p398F1 and pCNE8, as well as subtype C and BC strains, such as p25710, pCE1176 and pCH119. The 1E5 reacted moderately to pCNE55, pTRO11, pX1632 and pCH119, but showed no detectable reactivity Md Zahid et al. Retrovirology (2021) 18:23 Page 4 of 15 to pX2278, pCE0217 and pBJOX2000. The reactivity of the IgG3 form of 1E5 was significantly higher than that of the IgG1 form (Additional file 2: Fig. S2). Taken together, the binding data revealed that 1E5 reacted to 27 out of 35 strains tested (77%). Determination of the epitope recognized by 1E5 To determine the 1E5 epitope, a panel of Env mutants, which were affected in their reactivity of potent anti- HIV-1 antibodies [37–39], were used to examine the reactivity to 1E5. However, the results revealed that point mutations in V2 (N160K, I165A, L165A, K169E, L175P and L179P), the CD4 binding site (CD4bs, D368R) and the CD4-induced epitope (CD4i, I420R) did not change the reactivity of 1E5 (Additional file 3: Fig. S3). Further- more, the addition of sCD4 or CD4mc, YIR-821 [40], did not change the reactivity of 1E5 (Additional file 4: Fig. S4). These results suggested that the V2, CD4bs, CD4i and V3 epitopes are not the target for 1E5. Next, we compared the binding activity of 1E5 to chimeric Env proteins from 93TH976.17, a 1E5-reactive strain, and REJO4541.67, a 1E5-non-reactive strain (Fig. 3). Chimera A, which possesses gp120 from 93TH976.17 and gp41 from REJO4541.67, retained reactivity to 1E5. Chimeric Env containing the C1-C2 domains from 93TH976.17 (Chimera B) showed reactivity, but that containing the V3-C5 domains (Chimera C) resulted in no reactivity, indicating that the 1E5 epitope is in the C1-C2 domain of gp120. Strong binding was observed when the C1 and C2 domains originated from 93TH976.17 (Chimera F), although all of the chimeric Env constructs possess- ing the C1-C2 domain of 93TH976.17 (Chimeras D, E, G, H and I) showed marginal reactivity to 1E5. Moreo- ver, strong binding of 1E5 to the V1/V2 deletion mutants clearly demonstrated that the epitope of 1E5 is not in the V1V2 region (Chimeras L, M and N). This suggested that the C1 and C2 domains constitute the 1E5 epitope, but that the V1 and V2 domains also affect the binding of 1E5. Consistent with this, chimera J, which contains the V1-V2 domain from REJO4541.67 in a 93TH976.17 back- bone, showed good reactivity to 1E5. The data also sug- gested that the epitope recognized by 1E5, consisting of C1 and C2, may be different from that of A32 because the W69G mutation had no effect on binding [11, 41]. The conformational epitope consisting of the C1 and C2 domains of gp120 contains the cluster A region [10, 42]. The binding of antibodies to cluster A, such as A32 and C11, was reported to be enhanced by a combina- tion of CD4mc and anti-CoRBS antibodies [8, 11, 35]. To investigate the activity of anti-cluster A antibody, the binding activity of 1E5 was analyzed in the presence of CD4mc and anti-CoRBS antibodies (Fig. 4). The addition of CD4mc (YIR-821) and an anti-CoRBS antibody (17b or 4E9C) [43, 44] markedly enhanced the binding activ- ity of 1E5. This enhancement required both CD4mc and anti-CoRBS antibody, and the addition of either CD4mc or anti-CoRBS antibody alone did not affect 1E5 bind- ing significantly. This enhancement effect of CD4mc and anti-CoRBS antibody on 1E5 binding was even more apparent than that on A32 binding. To further investigate whether 1E5 binds to an epitope that overlaps with the epitope for A32, we performed a binding inhibition assay using Env from CRF02_AG- 257-transfected cells as the target and biotinylated 1E5 or A32 as the probe (Additional file 5: Fig. S5). The findings revealed that 1E5 did not compete with A32 for bind- ing (Additional file 5: Fig. S5a), but significant enhance- ment of A32 binding was observed in the presence of 1E5 (Additional file 5: Fig. S5b). These data suggested that 1E5 binds to the C1C2 region that does not overlap with the A32 epitope, and further that binding of 1E5 can enhance the binding of A32. Neutralization and ADCC activities of 1E5 The neutralization activity of 1E5-IgG1 was tested by a standard single-round neutralization assay for HIV-1 strains belonging to subtype A, CRF01_AE and CRF02_ AG (Additional file 6: Fig. S6a). The neutralization activ- ity of 1E5-IgG3 was examined by 1E5 alone and 1E5 with anti-CoRBS antibody and CD4mc against CRF02_ AG-257 virus (Additional file 6: Fig. S6b). Neither the IgG1 nor the IgG3 forms of 1E5 showed any neutraliza- tion activity, similar to the other C1C2 antibodies [9, 11, 35]. The ADCC activity of 1E5 was examined against nine CRF02_AG strains that showed strong binding of 1E5 (Fig. 1b) by the detection of FcγRIIIa signaling (Fig. 5a). Both IgG1 and IgG3 forms of 1E5 showed low ADCC activity against most of the strains, although 1E5-IgG3 showed higher ADCC activity than 1E5-IgG1 against sev- eral strains, such as AG-242, AG-257 and AG-280. The combination of 1E5, both IgG1 and IgG3 forms, with (See figure on next page.) Fig. 2 Reactivity of the IgG1 and IgG3 forms of 1E5 to global panel of HIV-1 Envs. HEK 293T cells were transiently transfected with plasmids expressing genes for both Env and EGFP. At 48 h post-transfection, cells were stained with the test IgG 1E5 (IgG1 and IgG3 forms) and three control IgGs: A32 (cluster A), VRC01 (CD4bs) and 4E9C (CoRBS). Cells were then fluorescently labeled with an APC-conjugated anti-human IgG secondary Ab. Histograms of APC in EGFP+ cells stained with test IgG and NHG as a negative control are shown by a red line and gray shading, respectively Md Zahid et al. Retrovirology (2021) 18:23 Page 5 of 15 NHG Test IgG A32 4E9C VRC01 1E5-IgG1 1E5-IgG3 p398F1 A pCNE8 CRF01_ AE pCNE55 CRF01_ AE pTRO11 B pX2278 B pX1632 G t n u o C p25710 C pCE1176 C pCE0217 C pCH119 CRF07_ BC pBJOX 2000 CRF07_ BC Fig. 2 (See legend on previous page.) Anti-human IgG-APC Md Zahid et al. Retrovirology (2021) 18:23 Page 6 of 15 a V1 V2 gp120 V3 V4 V5 gp41 C1 C2 C3 C4 C5 REJO4541.67 93TH976.17 Chimera A Chimera B Chimera C Chimera D Chimera E Chimera F Chimera G Chimera H Chimera I Chimera J Mutant K Mutant L Mutant M Mutant N b 1 0 0 0 0 I F M g o L 1 0 0 0 1 00 P G N V R C 0 1 1 E 5 B A E A B C D E F G H I Chimera J K L M N Fig. 3 Determination of the Env regions required for 1E5 binding. a Schematic presentation of the Env recombinants constructed between REJO4541.67 (SVPB16), which is not recognized by 1E5, and 93TH976.17, which is strongly bound by 1E5. The regions from REJO4541.67 and 93TH976.17 are shown in light-blue and red, respectively. The W69G mutation is shown in yellow. Deletion in the V1, V2 and V1V2 region was shown by a dotted line. b The reactivity of Env recombinants to NHG, VRC01 and 1E5 was determined by flow cytometry analysis using cells expressing each Env recombinant. The reactivity was detected by APC-conjugated anti-human IgG secondary Ab, and the mean fluorescence intensity (MFI) of APC is shown. Experiments were performed three times, and the representative result is shown 4E9C and YIR-821 increased ADCC activity against all of the strains tested except for AG-252. This lack of ADCC enhancement against AG-252 was consistent with the lack of enhancement of binding activity to AG-252 by 4E9C and YIR-821 (Additional file 7: Fig. S7). The com- bination effect of ADCC activity was statistically signifi- cant, and the combination of 1E5-IgG3 with 4E9C and YIR-821 showed a significantly higher level of ADCC activity than the other combinations (Fig. 5b). This was consistent with the enhancement of binding activity of 1E5 with CD4mc and anti-CoRBS antibodies (Fig. 4) and previous reports describing the enhancement of ADCC activity by the combination of C1C2 antibody, CD4mc and anti-CoRBS antibodies [8, 11, 35]. However, addi- tional analysis of the combination effect revealed that CD4mc and sCD4 were not required for ADCC enhance- ment in the ADCC assay detecting FcγRIIIa signaling (Fig. 6a). The addition of 4E9C alone increased ADCC activity more than the combination with CD4mc or sCD4. The lack of an effect with CD4mc and sCD4 may be due to the CD4 molecules expressed on the surface of effector cells, which possibly change the Env confor- mation accessible to anti-CoRBS antibodies. The level of dose-dependent ADCC activity of 1E5 in the presence of 4E9C was the same in both the presence and absence of YIR-821 (Additional file 8: Fig. S8). As reported previ- ously [10, 11], anti-CoRBS antibody alone did not medi- ate ADCC despite its strong recognition of the target cells (Additional file 9: Fig. S9). These results suggested that 1E5 mediates ADCC in combination with CD4 and anti-CoRBS antibody. Enhancement of ADCC by the dual and triple combination of anti‑C1C2 antibodies and an anti‑CoRBS antibody Although 1E5 recognized the C1C2 region of gp120, the epitope recognized by 1E5 did not overlap with that of A32, the representative anti-cluster A antibody (Addi- tional file 5: Fig. S5a). The binding of A32 was even enhanced in the presence of 1E5 (Additional file 5: Fig. S5b), and enhancement of FcγRIIIa signal detecting ADCC activity was examined using three antibodies, 1E5, A32 and 4E9C (Fig. 6b). These antibodies medi- ated a two- to three-fold change in ADCC when used alone, but mediated a six to eight-fold change when used in combination. Moreover, a triple combination of anti- bodies showed the highest ADCC activity, although the difference of ADCC activities was not statistically signifi- cant. Taken together, these data suggested that not only the combination of anti-C1C2 antibody and anti-CoRBS antibody, but also two anti-cluster A antibodies coordi- nately, can mediate strong ADCC activity, the phenom- enon that has not been reported previously. Strong ADCC activity against HIV‑1‑infected cells by combination of anti‑C1C2 antibodies, an anti‑CoRBS antibody and CD4mc We have examined ADCC activity of 1E5 by detecting FcγRIIIa signaling of the effector cells. However, this method does not measure a decrease of HIV-1-infected target cells, and expression of CD4 on the effector cell line may affect ADCC activity by changing Env con- formation (Fig. 6a). To examine ADCC activity of 1E5 against HIV-1-infected cells without the effect of CD4 expression on the effector cells, we performed ADCC Md Zahid et al. Retrovirology (2021) 18:23 Page 7 of 15 Biotinylated NHG Biotinylated Test IgG + anti-CoRBS Biotinylated Test IgG Biotinylated Test IgG + YIR-821 Biotinylated Test IgG + YIR-821 + anti-CoRBS 17b 4E9C 1E5 - IgG1 - Biotin 1E5 - IgG3 - Biotin t n u o C A32 - Biotin Streptavidin-APC Fig. 4 Anti-CoRBS antibody and CD4mc increased binding of 1E5 to CRF02_AG Env. HEK 293T cells were transfected with plasmid expressing both EGFP and CRF02_AG-257 Env. At 48 h post-transfection, cells were stained with biotinylated test IgG (1E5-IgG1 and 1E5-IgG3, 10 µg/ml) alone or with anti-CoRBS antibody (17B or 4E9C, 5 µg/ml) in the presence or absence of CD4mc YIR-821 (20 µM). Biotinylated A32 and biotinylated NHG were used as a control. Then, cells were fluorescently labeled with APC-conjugated streptavidin. Histograms of APC in EGFP+ cells are shown; Biotinylated NHG (gray shading), test IgG alone (orange), test IgG and YIR-821 (green), test IgG and anti-CoRBS antibody (blue), and IgG, YIR-821 and anti-CoRBS antibody (red). The representative result from three independent experiments is shown. Each independent experiment was done at once with the same transfected cells activity using NKR24 (CEM.NKR-CCR5 cells with LTR- Luc) cells infected with HIV-1 as target cells and N6 (human NK cell line KHYG-1 expressing FcγRIIIa) cells as effector cells. Luciferase activity, which increases by the LTR-driven luciferase gene in HIV-1-infected CEM. NKR-CCR5 cells, was measured, and the cell killing was calculated by a decrease of luciferase activity in the presence of antibodies. Decrease of infected cells was observed slightly in the presence of 1E5, A32 or 4E9C, considerably in the presence of two antibodies, and the triple combination of 1E5, A32 and 4E9C showed the strongest ADCC activity (Fig. 7). This is consistent with the results obtained by ADCC assay detecting FcγRIIIa signaling (Fig. 6b). Furthermore, YIR-821 was required for the ADCC activity mediated with these antibodies, suggesting that conformational change by CD4 is nec- essary for ADCC mediated with anti-C1C2 and anti- CoRBS antibodies. Discussion We isolated a monoclonal antibody 1E5 belonging to the anti-C1C2 antibody family, which targets the C1C2 of HIV-1 gp120, from a patient infected with CRF02_AG. Genetic analysis of the immunoglobulin heavy (VH) and Md Zahid et al. Retrovirology (2021) 18:23 Page 8 of 15 a b Fig. 5 1E5-IgG3 induces higher ADCC than 1E5-IgG1 in the presence of anti-CoRBS antibody and CD4mc. The ability of IgG1 and IgG3 forms of 1E5 to mediate signaling through FcγRIIIa when bound to cells expressing Env is shown. HEK 293T cells transfected with HIV-1 Env (nine binding assay positive Env proteins from a CRF02_AG panel) were incubated with the indicated antibodies and the ADCC indicator cell line expressing FcγRIIIa. Simultaneous binding of antigen and FcγRIIIa results in activation of the NFAT transcription factor, which induces luciferase in indicator cells. IE5 was used (10 µg/ml) alone or in combination with 4E9C (5 µg/ml) and CD4mc YIR-821 (20 µM). The fold change was calculated by dividing the luminescence units in the presence of Ab with those in the absence of Ab. NHG was used as a control. a ADCC activity against each target cell is shown. Experiments were performed in triplicate, and the means standard errors of the means are shown. b ADCC activity against target cells expressing Env from nine CRF02_AG strains was plotted, and statistically analyzed. The means significance was tested using a paired t test (, P < 0.05; , P < 0.01; , P < 0.001; **, P < 0.0001; ns, non-significant) standard errors of the means are shown. Statistical ± ± light (VL) chain variable domain gene segments revealed that VH was derived from VH1-69 and VL from the VK3- 20 germline. A recent report of germline VH1-69-de- rived antibodies demonstrated the defining features of VH1-69-utilizing antibodies against gp120, namely: the hydrophobic nature of the complementarity determining region-2 (CDRH2) regions with grand average hydropa- thy (GRAVY) scores ranging from 0.34 to 2.7, shorter complementarity-determining region-3 (CDRH3) with a median CDRH3 length of 14 amino acids and a higher isoelectric point (pI) with a median value of 6.15 [45]. The characteristics of these antibodies also apply to 1E5, which has a high GRAVY score of 1.85 for the CDRH2 region, a short CDRH3 region involving 12 amino acids and a high pI value of 8.59. It has been reported that the interactions between VH1-69 CDRH2 and the cavities within HIV-1 gp120 are hydrophobic [46]. High CDRH2 hydrophobicity was detected as a unique and universal feature of the VH1-69-utilizing antibodies [45]. Binding analysis with chimeric envelope constructs indicated that 1E5 recognizes a conformational epitope involving the C1 and C2 regions of gp120 (Fig. 3). A pre- vious detailed study mapped three unique clusters (A, B and C) of CD4i antibodies based on a cross competi- tion assay [10]. Usually occluded Cluster A epitopes can be exposed by conformational changes mediated by cel- lular CD4 binding to Env trimer during the viral entry process or co-expression of CD4 and the viral envelope on the same cell surface [14, 15]. However, the Env pro- tein expressed on the cell surface was recognized by A32 and 1E5 (Fig. 2). The reactivity of cluster A antibodies to Env trimers may be due to the HIV-1 strains used in this study. HIV-1 strains including non-B subtypes may be structurally different from the representative subtype B strains. It is also possible that the unprocessed Env pro- tein may be the target for A32 and 1E5. A32 and C11 are the major examples of anti-cluster A antibodies possess- ing non-overlapping epitopes involving the C1 and C2 domains [10, 42]. A32-like antibodies were found to be associated with the majority of ADCC activity in chron- ically-infected patients [9]. A flow cytometry-based inhi- bition assay demonstrated that the epitope of 1E5 did not overlap with that of A32, and that 1E5 even enhanced A32 binding (Additional file 5: Fig. S5). Taken together, these data suggested that 1E5 binds to a part of the C1C2 region that does not overlap with the A32 epitope, but that binding of 1E5 can enhance the accessibility of A32 binding. Despite having high polymorphism [25] and a shorter half-life than IgG1 [32], IgG3 is the most poly-functional IgG subclass, having the most potent Fc effector func- tion covering the widest range [23]. In the RV144 HIV vaccine trial, IgG3-mediated Fc effector functions, such as ADCC, ADCP and complement deposition, corre- lated with protection [24, 27]. Considering these facts, the IgG3 form of 1E5 was constructed and used in Md Zahid et al. Retrovirology (2021) 18:23 Page 9 of 15 a e g n a h c d l o F 5 4 3 2 1 0 4E9C only 4E9C +YIR-821 - + + + D4 +sC - - + - 4E9C - - - + 4E9C only 4E9C +YIR-821 - + + + D4 +sC - - + - 4E9C - - - + 4E9C only 4E9C +YIR-821 - + + + D4 +sC - - + - 4E9C - - - + 4E9C YIR-821 sCD4 1E5-IgG1 1E5-IgG3 Control b 1 0 e g n a h c d l o F 8 6 4 2 0 3+4E9C G H N 3 A32 1E5-IgG 4E9C A32+1E5-IgG 3 A32+4E9C 3+4E9C 1E5-IgG A32+1E5-IgG I g G c o m b in a t io n Fig. 6 Combination of anti-C1C2 antibodies and anti-CoRBS antibody enhanced ADCC activity. The ability of 1E5 to mediate ADCC activity was analyzed by measuring the signal through FcγRIIIa. HEK 293T cells expressing AG-257 Env were incubated with the indicated antibodies, YIR-821 and sCD4, and co-cultivated with the ADCC indicator cell line expressing FcγRIIIa. Fold change was calculated by dividing the luminescence units in the presence of Ab by those in the absence of Ab. NHG was used as a control. a 1E5-IgG1 or 1E5-IgG3 was used (10 µg/ml) alone or in combination with 4E9C (5 µg/ml), sCD4 (2 µg/ml) or CD4mc YIR-821 (20 µM). The effect of 4E9C, sCD4 and YIR-821 in the absence of test antibody is shown as a control. b Anti-C1C2 antibodies, A32 and 1E5-IgG3, and anti-CoRBS antibody, 4E9C, were used at a concentration of 10 µg/ml. The combination of anti-C1C2 antibody and anti-CoRBS antibody, as well as the combination of two anti-C1C2 antibodies, mediated stronger ADCC activity than each antibody alone. The combination of two anti-C1C2 antibodies and one anti-CoRBS antibody reached maximum enhancement at an eight-fold increase AG-257 AG-280 30 g n i l l i K % 20 10 0 30 20 10 0 g n i l l i K % NHG 4E9C A32 1E5-IgG3 1E5-IgG3+A32 A32+4E9C 1E5-IgG3+4E9C 1E5-IgG3+A32+4E9C NHG 4E9C A32 1E5-IgG3 1E5-IgG3+A32 A32+4E9C 1E5-IgG3+4E9C 1E5-IgG3+A32+4E9C YIR-821 + - Fig. 7 ADCC activity against HIV-1-infected cells mediated by combination of anti-C1C2 antibodies, an anti-CoRBS antibody and CD4mc. The ability of 1E5 to mediate ADCC activity was analyzed by measuring the decrease of HIV-1-infected cells. NKR24 cells infected with HIV-1, pNL-AG-257.ecto and pNL-AG-280.ecto, and N6 cells were used as target and effector cells, respectively. Decrease of HIV-1-infected cells was calculated from luciferase activity, which was increased by the LTR-driven luciferase gene in HIV-1-infected CEM. NKR-CCR5 cells. Antibodies, 1E5-IgG3 (5 µg/ml), A32 (1.25 µg/ml) and 4E9C (5 µg/ml) were used solely or in combination in the presence or absence of YIR-821 (20 µM) different assays in parallel with IgG1. When comparing the binding activity to Env proteins from a global panel of HIV-1, the IgG3 form showed significantly stronger binding than IgG1 (Additional file 2: Fig. S2). As shown in Figs. 5 and 6, the IgG3 form of 1E5 exhibited signifi- cantly higher ADCC activity than IgG1 in any combina- tion with CD4mc and/or anti-CoRBS antibody. Factors other than the epitope, such as the angle of binding, can influence the Fc function of an antibody [45]. Increased hinge length can allow more flexibility and therefore may increase the Fc-mediated effector functions of IgG3 [25]. Most importantly, IgG3 has the highest affin- ity to FcγRIIIa [26, 47]. Stronger binding of FcγRIIIa at an appropriate angle can favor the IgG3 form to mediate better ADCC than IgG1. The binding of biotinylated 1E5 with CRF02_AG Env- expressing target cells was markedly increased in the presence of CD4mc (YIR-821) and anti-CoRBS antibody (4E9C). CD4mc and anti-CoRBS antibody alone could Md Zahid et al. Retrovirology (2021) 18:23 Page 10 of 15 not mediate noticeable binding enhancement (Fig. 4). A similar pattern of binding was observed with A32. This indicates that 1E5 reactivity was the same as that of A32, requiring CD4mc and anti-CoRBS antibodies for enhanced binding and the stabilization of state 2 A in the presence of CRF02_AG-257 Env. Most of the previ- ous studies analyzing anti-C1C2 antibody binding used Env belonging to subtype B viruses [1, 35, 48]. Here, we observed the same phenomena for CRF02_AG Env by means of anti-C1C2 antibodies 1E5 and A32. The 1E5 did not demonstrate any neutralization activ- ity (Additional file 6: Fig. S6) when analyzing its ability to reduce the infectivity of the subtype-A, CRF01_AE and CRF02_AG Env pseudotype viruses. Being derived from germline VH1-69, this observation indicates the ADCC potential of 1E5, as described by previous research [45]. A model was described for the sequential opening of tri- meric Env that required anti-CoRBS antibodies to reveal the occluded epitope recognized by anti-C1C2 antibod- ies [35]. Engagement of CD4mc with the Phe43 cavity of the CD4 binding site causes a partial opening of trimeric Env, which enable anti-CoRBS antibodies to bind to Env but does not expose the inner region consisting of C1 and C2 regions. Binding of anti-CoRBS antibodies with two gp120 subunits possibly exposes epitopes recognized by anti-C1C2 antibodies resulting in state 2 A stabiliza- tion [1]. This recognition translated into efficient ADCC by anti-C1C2 antibodies [35] and may be involved in the ADCC exhibited by anti-C1C2 antibodies in HIV + sera [10, 11]. Figures 5 and 6a indicate that the highest level of ADCC was exhibited by the combination of IgG3 form 1E5 and anti-CoRBS antibody 4E9C. CD4mc (YIR-821) did not contribute to the enhancement of ADCC activ- ity in the FcγRIIIa signaling assay, but was mandatory in the cell killing assay using HIV-1-infected cells. The lack of requirement for CD4mc in the FcγRIIIa signaling assay is explained by the expression of CD4 on the surface of effector cells used for the ADCC assay. Several studies have suggested that ADCC may play a role in controlling HIV-1 infection [22, 49]. In the RV144 trial, ADCC was mainly found to be responsi- ble for conferring protection [6]. The C1C2 region is immunodominant in the case of both natural infection and vaccination. The majority of the ALVAC-HIV/AID- SVAX B/E vaccine recipients developed ADCC-mediat- ing antibodies with the C1C2 region specific A32-like antibodies comprising the significant portion [7]. This study demonstrated that the binding of A32 increased in the presence of 1E5 (Additional file 5: Fig. S5b). This observation raised the possibility of enhancement of ADCC using a combination of two anti-C1C2 antibod- ies. When used in an ADCC assay using FcγRIIIa sign- aling, the combination of 1E5-IgG3 and A32 showed higher level of fold change than their individual com- binations with anti-CoRBS IgG (Fig. 6b). This combi- nation effect by two anti-C1C2 antibodies were also confirmed in an ADCC assay using infected cells as target cells (Fig. 7). As anti-C1C2 antibodies are the mediators of ADCC activity exhibited by HIV-1 + sera [9–11], and these antibodies can be elicited by vaccina- tion [7], targeting this combination of anti-C1C2 anti- bodies may be a major tool for the protection against HIV-1. Moreover, a recent study on the elicitation of anti-C1C2 and anti-CoRBS antibodies observed higher and more efficient induction of anti-C1C2 antibodies in immunized guinea pigs [50]. Our results suggested that some of the anti-C1C2 antibodies, such as 1E5, can sta- bilize the Env conformation at state 2a in the presence of CD4mc resulting in the enhancement of binding of the other anti-C1C2 antibodies, such as A32, to exert higher ADCC activities. This finding may have implica- tions in terms of vaccine strategies to induce appropri- ate combinations of antibodies for improved outcomes. Conclusions Our findings indicate that the IgG3 form of anti-C1C2 antibody 1E5 isolated from a CRF02_AG-infected indi- vidual can mediate higher ADCC than the IgG1 form. The combination of two anti-cluster A antibodies, together with an anti-CoRBS antibody, mediated the highest level of ADCC. Methods Cells and reagents The 293T, 293 A and TZM-bl cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Nacalai Tesque, Inc., Kyoto, Japan) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Mediatech Inc., Corning, Manassas, VA, USA). Mab 4E9C [43, 44] and CD4mc YIR-821 [40] have previously been reported. Soluble CD4 was purchased commercially (sCD4, R&D systems, Inc., Minneapolis, MN, USA). Heavy and light chain gene-expressing plasmids of VRC01 [51], Env clones from a global panel [36], sub- type B [52], A [52], C [53] and CRF02_AG [54] were obtained through the NIH AIDS Reagent Program. NKR24 cells, CEM.NKR-CCR5 cells with LTR-Luc, were maintained in RPMI1640 (Fujifilm, Osaka, Japan) supplemented with 10% FBS, 2mM Glutamax, 0.1 mg/ ml Primocin (R10). N6 cells, human NK cell line KHYG-1 expressing human CD16, were maintained in R10 medium supplemented with 1 µg/ml Cs-A and 5 U/ml IL-2. NKR24 cells and N6 cells were kindly pro- vided by Dr. Evans [55]. Md Zahid et al. Retrovirology (2021) 18:23 Page 11 of 15 Isolation of IgG‑producing single B cells by fluorescence activated cell sorting A blood sample was obtained from patient KMCB2 of Kyushu Medical Center, who was infected with the CRF02_AG subtype of HIV-1. B cells were transformed by EBV and cultured at a concentration of 103 cells/ well for 10 days, as previously reported [45]. Single cells were sorted from the wells of an EBV-transformed B cell culture that scored positive for binding to Env (HIV-1 93TH966.8)-expressing cells using FACSAria II (BD Bio- sciences, San Jose, CA, USA). The cells were stained with anti-human IgG-BV421 and anti-human IgM-APC/Cy7 ˗ (BioLegend, San Diego, CA, USA), and IgG+IgM cells were sorted at single cell density into 4 µl/well of ice- cold 0.5× phosphate-buffered saline (PBS) containing 10 mM DTT, 8 U RNAsin® (Promega, WI, USA), 0.4 U 5′-3′ Prime RNAse Inhibitor™ (Eppendorf ) as previously described [56]. Cloning and analysis of 1E5 immunoglobulin variable genes cDNA was synthesized as previously described [56] in a total volume of 14 µl/well in a 96-well sorting plate. Total RNA from single cells was reverse transcribed in nucle- ase-free water (Eppendorf ) using 150 ng random hex- amer primer (pd(N) 6, GE Healthcare, Buckinghamshire, UK), 0.5 µl (10 mM) of each nucleotide dNTP-Mix (Inv- itrogen, Carlsbad CA, USA), 1 µl (0.1 M) of DTT (Inv- itrogen), 0.5% v/v Igepal CA-630 (Sigma), 4 U RNAsin® (Promega), 6 U Prime RNAse Inhibitor™ (Eppendorf ) and 50 U Superscript® III reverse transcriptase (Thermo Fisher Scientific, MA USA). The reverse transcription (RT) was performed as follows: 42 °C for 10 min, 25 °C for 10 min, 50 °C for 60 min and 94 °C for 5 min. For cloning of 1E5 immunoglobulin variable genes, the first round of nested PCR was performed according to the methods described by Tiller et al. [56] using the same primer pairs, while second-round primers were modified to have a 15 base overlap at the 5ʹ end with the specific vectors. The second PCR primer sequences are listed in Additional file 10: Table S1. The IgG heavy and light chain expression plasmids were constructed by recombination of the designated second PCR product with pIgGH and pKVA2, respectively [43], using the GeneArt Seamless Cloning and Assembly kit (Invitrogen). The nucleotide sequences of the immuno- globulin variable regions were aligned and compared to avoid possible PCR error. The sequences were analyzed for germline gene verification, framework and CDR map- ping, quantification of percent identity to germline, CDR amino acid length and pI using IMGT vquest (http:// imgt. org/ IMGT_ vquest/ vquest). CDRH2 grand average of hydropathy (GRAVY) scores were calculated using an online tool (http:// www. gravy- calcu lator. de/). Construction of IgG3 heavy chain‑expressing plasmid The region from CH1 to CH3 of IgG1 heavy chain- expressing vector pIgGH was exchanged with the corresponding region of IgG3, and IgG3 heavy chain- expressing vector pIgG3H was constructed. Briefly, the CH1-CH3 region of IgG3 was amplified using primers, CHApa-F (AGC CTC CAC CAA GGG CCC ATC GG), IgG3-R (TCA CCA AGT GGG GTT TTG AGC TCA), CHPme-R (CTG ATC AGC GGG TTT AAA CTA TCA TTT ACC CGG AGA CAG GG) and IgG3-F (ACA AGA GAG TTG AGC TCA AAA CCC C) from cDNA, which was synthesized from the RNA of healthy donor periph- eral blood mononuclear cells. The CH1-CH3 region of pIgGH was excluded by digestion with ApaI and PmeI, and the IgG3 fragments were inserted into the vector using the GeneArt Seamless Cloning and Assembly kit (Invitrogen). The variable region of 1E5 was inserted into pIgG3H to obtain 1E5-IgG3. Production and purification of recombinant IgG Recombinant IgG was produced and purified as pre- viously described [43]. Briefly, heavy and light chain plasmids were transfected into 293 A cells using TransIT®-LT1 Transfection Reagent (Mirus Bio LLC, WI, USA), and the cells stably expressing IgG were selected with G418 (1000 µg/ml) and hygromycin (150 µg/ml). IgG1 and IgG3 proteins were purified using a HiTrap™ rProtein A FF Column and a HiTrap™ Protein G HP col- umn, respectively (GE Healthcare). Analysis of the binding activity of antibodies by flow cytometry The binding activity of antibodies was analyzed as previ- ously described [57]. Briefly, 293T cells were transfected with a plasmid expressing both HIV-1 Env and enhanced green fluorescent protein (EGFP). After 48 h of transfec- tion, the cells were stained with primary antibody for 15 min at room temperature (RT). The cells were washed twice with PBS containing 0.2% BSA, and incubated with allophycocyanin-conjugated AffiniPure F(ab’)2 Fragment Goat Anti-Human IgG (H + L) (Jackson ImmunoRe- search, West Grove, PA, USA) for 15 min at RT. Cells were fixed with PBS containing 10% formalin and ana- lyzed using the FACSCanto II (BD Biosciences, San Jose, CA, USA). The reactivity of the antibodies was analyzed after gating the EGFP + cells using FlowJo (TreeStar, San Carlos, CA, USA). All experiments were performed at least twice independently, and the representative results are shown. Md Zahid et al. Retrovirology (2021) 18:23 Page 12 of 15 Neutralization assay using pseudovirus The neutralization activity of antibodies was determined as previously described [43, 58]. In brief, 293T cells were transfected with pSG3ΔEnv and Env expression vec- tor, and the supernatant after 48 h of transfection was stored at ˗80 °C. The median tissue culture infectious dose (TCID50) of each pseudovirus was determined using TZM-bl cells. Serially diluted antibody and virus (400 TCID50) were incubated for 1 h, and TZM-bl cells were added. After incubation for 48 h, the galactosidase activity was measured using galactosidase substrate (Tropix Gal-Screen substrate, Applied Biosystems) and an EnSpire Multimode Plate Reader (PerkinElmer, MA, USA). The relative light units (RLU) were compared to calculate the reduction in infectivity and 50% of the max- imal inhibitory concentration (IC50) was calculated using nonlinear regression. ADCC reporter assay to detect FcγRIIIa‑mediated signaling The detection of FcγRIIIa-mediated signaling was per- formed using a Jurkat NFAT-luc FCγRIIIa cell line (BPS Bioscience, CA, USA), as described previously [59]. The target cells were 293T cells expressing Env, which were transfected with Env-expressing plasmid 48 h before the ADCC assay. The target cells were washed with PBS, treated with 0.05% trypsin, and resuspended in RPMI- 1640 with 4% FBS at a concentration of 3 × 106 cells/ml. Then, 25 µl of the target cells were incubated with anti- bodies for 15 min, after which 25 µl of effector Jurkat cells were added at a ratio of 1:1 and were co-cultured for 6 h. The cells were lysed and the firefly luciferase activ- ity was determined with a luciferase assay kit (Promega) and EnSpire® Multimode Plate Reader. The co-culture in the absence of antibody provided background (antibody- independent) luciferase activity. The RLU obtained in the presence of antibody were divided by the background level to calculate the fold change. Infectious molecular clones for ADCC assay Replication competent infectious molecular clones (IMC) were designed to encode AG-257 and 280 env genes in pNL4-3 (GenBank: AF324493). The region of the env gene that encode the ectodomain were amplified by primers, NL-4-AG-257-F (AAG CAG TAA GTA GTA AAT GCA ACC TTT AGC ) and NL-257-ECTO-R (TAT CAT TAT GAA TAA TTT TAT ATA CCA CAG ) for AG-257, and NL-4-AG-280-F (AAG CAG TAA GTA GTA GAT GCA ATC TTT AAT ) and NL-280-ECTO-R (TAT CAT TAT GAA TAA TTT AAT ATA CCA CAG ) for AG-280. The amplified fragment was replaced with the correspond- ing region of pNL4-3, generating pNL-AG-257.ecto and pNL-AG-280.ecto [60] using GeneArt Seamless Cloning and Assembly enzyme mix (Invitrogen, Carlsbad, CA, USA). ADCC assay using HIV‑infected cells and N6 cells ADCC assay was performed according to the method described before with some modifications [55]. In brief, NKR24 cells were infected by spinoculation in round bottom tubes. 5 × 105 target cells and infectious viral inoculum were subjected to centrifugation at 1200×g for 2 h at 25 °C. Then the viral inoculum was removed, and the target cells were cultured in R10 medium for 2 days. Target cells were washed 3 times in R10 medium and suspended in R10 medium containing 10 U/ml IL-2 without CsA. In round-bottom, tissue culture-treated polystyrene 96-well plates, 40 µl each of target and effec- tor cells were added at 2.5 × 105 cells/ml and 2.5 × 106 cells/ml, respectively. N6 effector cells and uninfected target cells were a control to define 0% RLU. N6 cells and infected targets without antibody were a control to define 100 % RLU. Antibodies (20 µl) were added in triplicate. After 8 h incubation, a 40 µl of cells was resuspended and mixed with 40 µl of BriteLite Plus (Perkin Elmer) in 96-well white 1/2 area microplate. Luciferase activity was measured using an EnSpire Multimode Plate Reader, and the %killing of HIV-infected cells was calculated from the decrease of RLU. Abbreviations CoRBS: Coreceptor binding site; ADCC: Antibody-dependent cellular-cytotox- icity; CD4mc: CD4-mimetic compounds; HIV-1: Human immunodeficiency virus type 1; Env: Envelope glycoprotein; nnAbs: Non-neutralizing antibodies; CD4i: CD4-induced; IgG: Immunoglobulins; Ab: Antibody; ADCP: Antibody- dependent cellular phagocytosis; mAb: Monoclonal antibody; EBV: Epstein– Barr virus; CD4bs: CD4 binding site; VH: Immunoglobulin heavy chain variable domain; VL: Immunoglobulin light chain variable domain; CDRH2: Heavy chain complementarity determining region-2; CDRH3: Heavy chain complementa- rity determining region-3; GRAVY: Grand average hydropathy; pI: Isoelectric point; sCD4: Soluble CD4; EGFP: Enhanced green fluorescent protein; TCID50: Median tissue culture infectious dose; RLU: Relative light units; IC50: 50% of the maximal inhibitory concentration. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12977- 021- 00568-y. Additional file 1: Fig. S1. Amino acid sequences of 1E5 heavy and light chains Additional file 2: Fig. S2. gG3 exhibits stronger binding affinity to a panel of global HIV-1 Env clones. HEK 293T cells were transfected with plasmids expressing 11 Env proteins from a global panel of HIV-1. At 48 h post-transfection, cells were stained with primary antibody (1E5). Then, cells were fluorescently labeled with an APC-conjugated anti-human IgG secondary Ab. The plasmids used for transfection express HIV-1 Env and EGFP. Means standard errors of the means of APC intensity in EGFP+ ± Md Zahid et al. Retrovirology (2021) 18:23 Page 13 of 15 cells are shown. Statistical significance was tested using a paired t test (, P <0.05). Additional file 3: Fig. S3. 1E5 did not recognize a previously well-defined epitope. HEK 293T cells were transfected with plasmids encoding both EGFP and HIV-1 Env (subtype A and CRF01_AE) with point mutations at previously described positions in V2, CD4i, CD4bs and glycosylation sites. Cells were stained with 1E5-IgG or VRC01, and binding was detected by an APC-conjugated anti-human IgG secondary Ab. The MFI values of APC in EGFP cells are shown. + Additional file 4: Fig. S4. Binding of 1E5 was not enhanced by sCD4 or CD4mc. HEK 293T cells expressing EGFP and HIV-1 Env of CRF01_AG, CRF02_AE and subtype B were incubated with sCD4 (2 µg/ml) or YIR-821 (10 µM), and stained with primary antibody (5 µg/ml). Then, IgG binding was detected by APC-conjugated anti-human IgG secondary Ab. The MFI values of APC in EGFP cells are shown.. + Additional file 5: Fig. S5. Analysis of the competitive binding between 1E5 and A32. HEK 293T cells expressing EGFP and AG-257 Env were stained with 3 µg/ml of biotinylated 1E5 (a) or A32 (b) in the presence of serially-diluted non-biotinylated competitors, A32 and 1E5, respectively. Biotinylated antibodies were detected by APC-conjugated streptavidin. The MFI values of APC in EGFP absence of competitor is shown by dotted line. cells are shown. The MFI value in the + Additional file 6: Fig. S6. 1E5 is a non-neutralizing antibody. (a) The neutralization activity of 1E5-IgG1 was tested in TZM-bl assays against Env-pseudotyped viruses from subtype A, CRF01_AE and CRF02_AG of HIV-1. (b) Neutralization activity of 1E5-IgG3 was examined alone or with anti-CoRBS Ab (4E9C) and CD4mc (YIR-821) against pseudotyped virus with AG-257 Env. Additional file 7: Fig. S7. Anti-CoRBS antibodies failed to increase the binding of anti-cluster A antibodies to AG-252 Env. HEK 293T cells expressing both EGFP and AG-252 Env were stained with biotinylated test IgG (1E5-IgG1, 1E5-IgG3 and A32 at 10 µg/ml) alone or with anti-CoRBS antibody (17b or 4E9C, 5 µg/ml) and CD4mc YIR-821 (20 µM), and the binding was detected by APC-conjugated streptavidin. Histograms of APC in EGFP cells are shown. + Additional file 8: Fig. S8. IgG3 form of 1E5 shows stronger ADCC in the presence of anti-CoRBS antibody. The effect of anti-CoRBS antibody and CD4mc on ADCC activity, which was measured by signaling through FcγRIIIa, was analyzed using IgG3 forms of 1E5. HEK 293T cells transfected with plasmid expressing HIV-1 Env (nine binding assay-positive Env proteins from a CRF02_AG panel and one subtype A) were incubated with the indicated antibodies and co-cultivated with the ADCC indicator cell line expressing FcγRIIIa. ADCC activities obtained by serial dilution of 1E5 (0.1, 1 and 10 µg/ml) alone (RPMI) or in combination with 4E9C (5 μg/ml) and YIR-821 (20 µM) are shown. ‘0’ on the X axis indicates the absence of 1E5. The fold change was calculated by dividing the luminescence units in the presence of Ab by those in the absence of Ab. Experiments were performed in triplicate, and the means shown. standard errors of the means are ± Additional file 9: Fig. S9. Recognition of target cells by anti-CoRBS anti- bodies. HEK 293T cells expressing both EGFP and AG-257 Env were stained with test IgG (4E9C and 17b, 10 µg/ml) alone or in the presence of CD4mc YIR-821 (20 µM). Antibody binding was detected by APC-conjugated anti- human IgG secondary Ab. Histograms of APC in EGFP cells are shown + Additional file 10: Table S1. Second-round PCR primers. Acknowledgements We are grateful to Dr. J. Robinson, Dr. (A) Finzi and Dr. D. Evans for their gener- ous gifts of 17b, A32 and cells for ADCC. The following reagent was obtained through the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH from Drs. Montefiori, F. Gao, M. Li, (B) H. Hahn, J. F. Salazar-Gonzalez, X. Wei, G. M. Shaw, D. L. Kothe, S. Abdool Karim, G. Ramjee, (C) Williamson, Y. Li, C. (A) Derdeyn, E. Hunter, L. Morris, K. Mlisana and Dr. Julie Overbaugh, J. C. Kappes, D. X. Wu and Tranzyme Inc: Panel of HIV-1 Subtype B Env Clones (cat#11227), Panel of HIV-1 Subtype C Env Clones (cat#11326), and Panel of Global HIV-1 Env Clones, from Dr. Julie Overbaugh: Panel of HIV-1 Env Clones – Subtypes A, A/D, A2/D, C, and D (cat# 11947), from Drs. D. Ellenberger, (B) Li, M. Callahan, and S. Buterra: Panel of HIV-1 Subtype A/G Env Clones (cat# 11673), from Drs. John (C) Kappes and Xiaoyun Wu: pSG3ΔEnv (cat# 11051) and TZM-bl cells (cat# 8129), from Dr. John Mascola: VRC01 mAb Heavy and Light Chain Expression Vectors (cat# 12035 and 12036). We thank Kate Fox, DPhil, from Edanz Group (https:// en- author- servi ces. edanz group. com/ ac) for editing a draft of this manuscript. We thank Yoko Kawanami and Mikiko Shimizu for their assistance in the experi- ments, and Miki Tsukiashi for her kind administrative assistance. Authors’ contributions HMZ, TK and SM conceived and designed the study. HMZ, ST, TK and SM per- formed experiments. KY, SB, and KM prepared reagents including pseudovi- ruses for the experiments. HT prepared the CD4mc YIR-821. HMZ, TK and SM prepared the manuscript. All authors read and approved the final manuscript. Funding This work was supported in part by Global Education and Research Center Aiming at the Control of AIDS, by JSPS KAKENHI Grant Number18H0285400 and by a grant for Research Program on HIV/AIDS from the Japan Agency for Medical Research and Development (JP19fk0410025h0001, JP20fk0410025h0002, JP21fk0410025h0003). Availability of data and materials Not applicable. Declarations Ethics approval and consent to participate Human blood samples were collected after signed informed consent was obtained in accordance with the study protocol approved by the Ethics Committee for Clinical Research and Advanced Medical Technology at the Kumamoto University School (Approval No. 1637). Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1 Division of Clinical Retrovirology, Joint Research Center for Human Retrovirus infection, Kumamoto University, 2-2-1 Honjo, Chuo-ku, Kumamoto 860-0811, Japan. 2 Laboratory of Immunosenescence, National Institutes of Biomedi- cal Innovation, Health and Nutrition, Osaka, Japan. 3 Institute of Biomaterials and Bioengineering, Tokyo Medical and Dental University, Tokyo, Japan. Received: 9 October 2020 Accepted: 9 August 2021 References 1. Alsahafi N, Bakouche N, Kazemi M, Richard J, Ding S, Bhattacharyya S, et al. An asymmetric opening of HIV-1 envelope mediates antibody- dependent cellular cytotoxicity. Cell Host Microbe. 2019;25(4):578 – 87.e5. 2. Alpert MD, Harvey JD, Lauer WA, Reeves RK, Piatak M Jr, Carville A, et al. ADCC develops over time during persistent infection with live-attenu- ated SIV and is associated with complete protection against SIV(mac)251 challenge. PLoS pathogens. 2012;8(8):e1002890. 3. Chung AW, Isitman G, Navis M, Kramski M, Center RJ, Kent SJ, et al. Immune escape from HIV-specific antibody-dependent cellular cytotox- icity (ADCC) pressure. Proc Natl Acad Sci U S A. 2011;108(18):7505–10. 4. Mabuka J, Nduati R, Odem-Davis K, Peterson D, Overbaugh J. HIV-specific antibodies capable of ADCC are common in breastmilk and are associ- ated with reduced risk of transmission in women with high viral loads. PLoS pathogens. 2012;8(6):e1002739. 5. Williams KL, Cortez V, Dingens AS, Gach JS, Rainwater S, Weis JF, et al. HIV-specific CD4-induced antibodies mediate broad and potent Md Zahid et al. Retrovirology (2021) 18:23 Page 14 of 15 antibody-dependent cellular cytotoxicity activity and are com- monly detected in plasma from HIV-infected humans. EBioMedicine. 2015;2(10):1464–77. 6. Haynes BF, Gilbert PB, McElrath MJ, Zolla-Pazner S, Tomaras GD, Alam SM, et al. Immune-correlates analysis of an HIV-1 vaccine efficacy trial. N Engl J Med. 2012;366(14):1275–86. 7. Bonsignori M, Pollara J, Moody MA, Alpert MD, Chen X, Hwang K-K, et al. Antibody-dependent cellular cytotoxicity-mediating antibodies from an HIV-1 vaccine efficacy trial target multiple epitopes and preferentially use the VH1 gene family. J Virol. 2012;86(21):11521–32. 8. Veillette M, Coutu M, Richard J, Batraville L-A, Dagher O, Bernard N, et al. The HIV-1 gp120 CD4-bound conformation is preferentially targeted by antibody-dependent cellular cytotoxicity-mediating antibodies in Sera from HIV-1-infected individuals. J Virol. 2015;89(1):545–51. Ferrari G, Pollara J, Kozink D, Harms T, Drinker M, Freel S, et al. An HIV-1 gp120 envelope human monoclonal antibody that recognizes a C1 conformational epitope mediates potent antibody-dependent cellular cytotoxicity (ADCC) activity and defines a common ADCC epitope in human HIV-1 serum. J Virol. 2011;85(14):7029–36. 9. 10. Guan Y, Pazgier M, Sajadi MM, Kamin-Lewis R, Al-Darmarki S, Flinko R, et al. Diverse specificity and effector function among human antibodies to HIV-1 envelope glycoprotein epitopes exposed by CD4 binding. Proc Natl Acad Sci U S A. 2013;110(1):E69–78. 11. Ding S, Veillette M, Coutu M, Prévost J, Scharf L, Bjorkman PJ, et al. A highly conserved residue of the HIV-1 gp120 inner domain is important for antibody-dependent cellular cytotoxicity responses mediated by anti- cluster A antibodies. J Virol. 2016;90(4):2127–34. 12. Richard J, Veillette M, Brassard N, Iyer SS, Roger M, Martin L, et al. CD4 mimetics sensitize HIV-1-infected cells to ADCC. Proc Natl Acad Sci U S A. 2015;112(20):E2687-94. 13. Batraville LA, Richard J, Veillette M, Labbe AC, Alary M, Guedou F, et al. Short communication: Anti-HIV-1 envelope immunoglobulin Gs in blood and cervicovaginal samples of Beninese commercial sex workers. AIDS Res Hum Retroviruses. 2014;30(11):1145–9. 14. Acharya P, Tolbert WD, Gohain N, Wu X, Yu L, Liu T, et al. Structural defini- tion of an antibody-dependent cellular cytotoxicity response implicated in reduced risk for HIV-1 infection. J Virol. 2014;88(21):12895–906. 15. Veillette M, Désormeaux A, Medjahed H, Gharsallah N-E, Coutu M, Baalwa J, et al. Interaction with cellular CD4 exposes HIV-1 envelope epitopes targeted by antibody-dependent cell-mediated cytotoxicity. J Virol. 2014;88(5):2633–44. 16. Tolbert WD, Gohain N, Veillette M, Chapleau JP, Orlandi C, Visciano ML, et al. Paring down HIV env: design and crystal structure of a stabilized inner domain of HIV-1 gp120 displaying a major ADCC target of the A32 region. Structure. 2016;24(5):697–709. 17. Tolbert WD, Gohain N, Alsahafi N, Van V, Orlandi C, Ding S, et al. Targeting the late stage of HIV-1 entry for antibody-dependent cellular cyto- toxicity: structural basis for Env epitopes in the C11 region. Structure. 2017;25(11):1719–31. e4. 18. Anand SP, Prévost J, Baril S, Richard J, Medjahed H, Chapleau J-P, et al. Two families of Env antibodies efficiently engage Fc-Gamma receptors and eliminate HIV-1-infected cells. J Virol. 2019;93(3):e01823-18. 25. Richardson SI, Lambson BE, Crowley AR, Bashirova A, Scheepers C, Garrett N, et al. IgG3 enhances neutralization potency and Fc effector function of an HIV V2-specific broadly neutralizing antibody. PLoS pathogens. 2019;15(12):e1008064. 26. Vidarsson G, Dekkers G, Rispens T. IgG subclasses and allotypes: from structure to effector functions. Front Immunol. 2014;5(520):520. 27. Yates NL, Liao H-X, Fong Y, deCamp A, Vandergrift NA, Williams WT, et al. Vaccine-induced Env V1-V2 IgG3 correlates with lower HIV-1 infection risk and declines soon after vaccination. Sci Transl Med. 2014;6(228):228ra39-ra39. 28. Sadanand S, Das J, Chung AW, Schoen MK, Lane S, Suscovich TJ, et al. Temporal variation in HIV-specific IgG subclass antibodies during acute infection differentiates spontaneous controllers from chronic progressors. AIDS. 2018;32(4):443–50. 29. Liu H, May K. Disulfide bond structures of IgG molecules. mAbs. 2012;4(1):17–23. 30. Roux KH, Strelets L, Michaelsen TE. Flexibility of human IgG subclasses. J Immunol. 1997;159(7):3372–82. 31. Braster R, Grewal S, Visser R, Einarsdottir HK, van Egmond M, Vidarsson G, et al. Human IgG3 with extended half-life does not improve Fc- gamma receptor-mediated cancer antibody therapies in mice. PloS one. 2017;12(5):e0177736-e. 32. Yates NL, Lucas JT, Nolen TL, Vandergrift NA, Soderberg KA, Seaton KE, et al. Multiple HIV-1-specific IgG3 responses decline during acute HIV-1: implications for detection of incident HIV infection. AIDS. 2011;25(17):2089–97. 33. Hemelaar J, Elangovan R, Yun J, Dickson-Tetteh L, Fleminger I, Kirtley S, et al. Global and regional molecular epidemiology of HIV-1, 1990–2015: a systematic review, global survey, and trend analysis. Lancet Infect Dis. 2019;19(2):143–55. 34. Dauwe K, Mortier V, Schauvliege M, Van Den Heuvel A, Fransen K, Servais J-Y, et al. Characteristics and spread to the native population of HIV-1 non-B subtypes in two European countries with high migration rate. BMC Infect Dis. 2015;15(1):524. 35. Richard J, Pacheco B, Gohain N, Veillette M, Ding S, Alsahafi N, et al. Co-receptor binding site antibodies enable CD4-mimetics to expose con- served anti-cluster A ADCC epitopes on HIV-1 envelope glycoproteins. EBioMedicine. 2016;12:208–18. 36. deCamp A, Hraber P, Bailer RT, Seaman MS, Ochsenbauer C, Kap- pes J, et al. Global panel of HIV-1 Env reference strains for standard- ized assessments of vaccine-elicited neutralizing antibodies. J Virol. 2014;88(5):2489–507. 37. Dingens AS, Arenz D, Weight H, Overbaugh J, Bloom JD. An antigenic atlas of HIV-1 escape from broadly neutralizing antibodies distinguishes functional and structural epitopes. Immunity. 2019;50(2):520 – 32.e3. 38. Scheid JF, Mouquet H, Feldhahn N, Seaman MS, Velinzon K, Pietzsch J, et al. Broad diversity of neutralizing antibodies isolated from memory B cells in HIV-infected individuals. Nature. 2009;458(7238):636–40. 39. Shibata J, Yoshimura K, Honda A, Koito A, Murakami T, Matsushita S. Impact of V2 mutations on escape from a potent neutralizing anti-V3 monoclonal antibody during in vitro selection of a primary human immunodeficiency virus type 1 isolate. J Virol. 2007;81(8):3757–68. 19. Horwitz JA, Bar-On Y, Lu CL, Fera D, Lockhart AAK, Lorenzi JCC, et al. 40. Ohashi N, Harada S, Mizuguchi T, Irahara Y, Yamada Y, Kotani M, et al. Non-neutralizing atibodies alter the course of HIV-1 infection in vivo. Cell. 2017;170(4):637–48. e10. Small-molecule CD4 mimics containing mono-cyclohexyl moieties as HIV entry inhibitors. ChemMedChem. 2016;11(8):940–6. 20. Santra S, Tomaras GD, Warrier R, Nicely NI, Liao HX, Pollara J, et al. Human 41. Finzi A, Xiang SH, Pacheco B, Wang L, Haight J, Kassa A, et al. Topological non-neutralizing HIV-1 envelope monoclonal antibodies limit the num- ber of founder viruses during SHIV mucosal infection in rhesus macaques. PLoS pathogens. 2015;11(8):e1005042. 21. Ackerman ME, Mikhailova A, Brown EP, Dowell KG, Walker BD, Bailey- Kellogg C, et al. Polyfunctional HIV-specific antibody responses are associated with spontaneous HIV control. PLoS pathogens. 2016;12(1):e1005315. 22. Lambotte O, Ferrari G, Moog C, Yates NL, Liao H-X, Parks RJ, et al. Hetero- geneous neutralizing antibody and antibody-dependent cell cytotoxicity responses in HIV-1 elite controllers. AIDS. 2009;23(8):897–906. 23. Damelang T, Rogerson SJ, Kent SJ, Chung AW. Role of IgG3 in infectious diseases. Trends Immunol. 2019;40(3):197–211. 24. Chung AW, Ghebremichael M, Robinson H, Brown E, Choi I, Lane S, et al. layers in the HIV-1 gp120 inner domain regulate gp41 interaction and CD4-triggered conformational transitions. Mol Cell. 2010;37(5):656–67. 42. Tolbert WD, Sherburn RT, Van V, Pazgier M. Structural basis for epitopes in the gp120 cluster A region that invokes potent effector cell activity. Viruses. 2019;11(1):69. 43. Ramirez Valdez KP, Kuwata T, Maruta Y, Tanaka K, Alam M, Yoshimura K, et al. Complementary and synergistic activities of anti-V3, CD4bs and CD4i antibodies derived from a single individual can cover a wide range of HIV-1 strains. Virology. 2015;475:187–203. 44. Tanaka K, Kuwata T, Alam M, Kaplan G, Takahama S, Valdez KPR, et al. Unique binding modes for the broad neutralizing activity of single-chain variable fragments (scFv) targeting CD4-induced epitopes. Retrovirology. 2017;14(1):44. Polyfunctional Fc-effector profiles mediated by IgG subclass selection dis- tinguish RV144 and VAX003 vaccines. Sci Transl Med. 2014;6(228):228ra38. 45. Smith SA, Burton SL, Kilembe W, Lakhi S, Karita E, Price M, et al. VH1-69 uti- lizing antibodies are capable of mediating non-neutralizing Fc-mediated Md Zahid et al. Retrovirology (2021) 18:23 Page 15 of 15 effector functions against the transmitted/founder gp120. Front Immu- nol. 2019;9(3163):3163. 46. Huang CC, Venturi M, Majeed S, Moore MJ, Phogat S, Zhang MY, et al. Structural basis of tyrosine sulfation and VH-gene usage in antibodies that recognize the HIV type 1 coreceptor-binding site on gp120. Proc Natl Acad Sci U S A. 2004;101(9):2706–11. 47. Bruhns P, Iannascoli B, England P, Mancardi DA, Fernandez N, Jorieux S, et al. Specificity and affinity of human Fcγ receptors and their polymor- phic variants for human IgG subclasses. Blood. 2009;113(16):3716–25. 48. Munro JB, Gorman J, Ma X, Zhou Z, Arthos J, Burton DR, et al. Conforma- tional dynamics of single HIV-1 envelope trimers on the surface of native virions. Science. 2014;346(6210):759–63. 49. Barouch DH, Liu J, Li H, Maxfield LF, Abbink P, Lynch DM, et al. Vaccine protection against acquisition of neutralization-resistant SIV challenges in rhesus monkeys. Nature. 2012;482(7383):89–93. 50. Beaudoin-Bussières G, Prévost J, Gendron-Lepage G, Melillo B, Chen J, Smith Iii AB, et al. Elicitation of cluster A and Co-receptor binding site antibodies are required to eliminate HIV-1 infected cells. Microorganisms. 2020;8(5):710. 51. Lynch RM, Boritz E, Coates EE, DeZure A, Madden P, Costner P, et al. Virologic effects of broadly neutralizing antibody VRC01 administration during chronic HIV-1 infection. Sci Transl Med. 2015;7(319):319ra206–6. 52. Gao F, Yue L, Craig S, Thornton CL, Robertson DL, McCutchan FE, et al. Genetic variation of HIV type 1 in four World Health Organization- sponsored vaccine evaluation sites: generation of functional envelope (glycoprotein 160) clones representative of sequence subtypes A, B, C, and E. WHO Network for HIV Isolation and Characterization. AIDS Res Hum Retroviruses. 1994;10(11):1359–68. 53. Li M, Salazar-Gonzalez JF, Derdeyn CA, Morris L, Williamson C, Robinson JE, et al. Genetic and neutralization properties of subtype C human immunodeficiency virus type 1 molecular env clones from acute and early heterosexually acquired infections in Southern Africa. J Virol. 2006;80(23):11776–90. 54. Kulkarni SS, Lapedes A, Tang H, Gnanakaran S, Daniels MG, Zhang M, et al. Highly complex neutralization determinants on a monophyletic lineage of newly transmitted subtype C HIV-1 Env clones from India. Virology. 2009;385(2):505–20. 55. Alpert MD, Heyer LN, Williams DE, Harvey JD, Greenough T, Allhorn M, et al. A novel assay for antibody-dependent cell-mediated cytotoxic- ity against HIV-1- or SIV-infected cells reveals incomplete overlap with antibodies measured by neutralization and binding assays. J Virol. 2012;86(22):12039–52. 56. Tiller T, Meffre E, Yurasov S, Tsuiji M, Nussenzweig MC, Wardemann H. Efficient generation of monoclonal antibodies from single human B cells by single cell RT-PCR and expression vector cloning. J Immunol Methods. 2008;329(1–2):112–24. 57. Alam M, Kuwata T, Shimura K, Yokoyama M, Ramirez Valdez KP, Tanaka K, et al. Enhanced antibody-mediated neutralization of HIV-1 variants that are resistant to fusion inhibitors. Retrovirology. 2016;13(1):70. 58. Montefiori DC. Measuring HIV Neutralization in a Luciferase Reporter Gene Assay. In: Prasad VR, Kalpana GV, editors. HIV Protocols. Totowa: Humana Press; 2009. pp. 395–405. 59. Thida W, Kuwata T, Maeda Y, Yamashiro T, Tran GV, Nguyen KV, et al. The role of conventional antibodies targeting the CD4 binding site and CD4-induced epitopes in the control of HIV-1 CRF01_AE viruses. Biochem Biophys Res Commun. 2019;508(1):46–51. 60. Edmonds TG, Ding H, Yuan X, Wei Q, Smith KS, Conway JA, et al. Replication competent molecular clones of HIV-1 expressing Renilla luciferase facilitate the analysis of antibody inhibition in PBMC. Virology. 2010;408(1):1–13. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12954-023-00771-4	Lowrie et al. Harm Reduction Journal (2023) 20:46 https://doi.org/10.1186/s12954-023-00771-4 Harm Reduction Journal RESEARCH Open Access Baseline characteristics of people experiencing homelessness with a recent drug overdose in the PHOENIx pilot randomised controlled trial Richard Lowrie1, Andrew McPherson1, Frances S. Mair2, Kate Stock1, Caitlin Jones2, Donogh Maguire3, Vibhu Paudyal4, Clare Duncan5, Becky Blair1, Cian Lombard1, Steven Ross6, Fiona Hughes1, Jane Moir1, Ailsa Scott6, Frank Reilly6, Laura Sills7, Jennifer Hislop8, Natalia Farmer9, Sharon Lucey1, Stephen Wishart11, George Provan6, Roy Robertson10 and Andrea Williamson2 Abstract Background Drug-related deaths in Scotland are the highest in Europe. Half of all deaths in people experiencing homelessness are drug related, yet we know little about the unmet health needs of people experiencing homeless- ness with recent non-fatal overdose, limiting a tailored practice and policy response to a public health crisis. Methods People experiencing homelessness with at least one non-fatal street drug overdose in the previous 6 months were recruited from 20 venues in Glasgow, Scotland, and randomised into PHOENIx plus usual care, or usual care. PHOENIx is a collaborative assertive outreach intervention by independent prescriber NHS Pharmacists and third sector homelessness workers, offering repeated integrated, holistic physical, mental and addictions health and social care support including prescribing. We describe comprehensive baseline characteristics of randomised participants. Results One hundred and twenty-eight participants had a mean age of 42 years (SD 8.4); 71% male, homelessness for a median of 24 years (IQR 12–30). One hundred and eighteen (92%) lived in large, congregate city centre tempo- rary accommodation. A quarter (25%) were not registered with a General Practitioner. Participants had overdosed a mean of 3.2 (SD 3.2) times in the preceding 6 months, using a median of 3 (IQR 2–4) non-prescription drugs concur- rently: 112 (87.5%) street valium (benzodiazepine-type new psychoactive substances); 77 (60%) heroin; and 76 (59%) cocaine. Half (50%) were injecting, 50% into their groins. 90% were receiving care from Alcohol and Drug Recovery Services (ADRS), and in addition to using street drugs, 90% received opioid substitution therapy (OST), 10% diazepam for street valium use and one participant received heroin-assisted treatment. Participants had a mean of 2.2 (SD 1.3) mental health problems and 5.4 (SD 2.5) physical health problems; 50% received treatment for physical or mental health problems. Ninety-one per cent had at least one mental health problem; 66% had no specialist mental health support. Participants were frail (70%) or pre-frail (28%), with maximal levels of psychological distress, 44% received one or no daily meal, and 58% had previously attempted suicide. Correspondence: Richard Lowrie Richard.Lowrie@ggc.scot.nhs.uk Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 2 of 18 Conclusions People at high risk of drug-related death continue to overdose repeatedly despite receiving OST. High levels of frailty, multimorbidity, unsuitable accommodation and unmet mental and physical health care needs require a reorientation of services informed by evidence of effectiveness and cost-effectiveness. Trial registration UK Clinical Trials Registry identifier: ISRCTN 10585019. Keywords Homelessness, Chronic disease, Opioid addiction, Primary health care, Randomised controlled trial, Public health, Polydrug use, Drug-related death Introduction The individual, societal, health and economic burdens of homelessness and drug-related deaths are undisputed, overlap and are increasing [1–4]. People experiencing homelessness with problem drug use including opioid use disorder are at higher risk of fatal overdose than peo- ple with opioid use disorder living in mainstream society [5–7]. Polydrug use includes street benzodiazepines, and cocaine, for which there are no evidence-based substitute prescriptions known to prevent overdose [5, 8]. Almost half of all deaths among people experiencing homeless- ness are caused by polydrug overdose [2–4], suggesting an urgent need to investigate the characteristics includ- ing unmet needs of people experiencing homelessness with recent overdose, and test innovative, additional approaches to improving outcomes. Experiencing a non-fatal overdose increases the odds of a subsequent fatal overdose [6, 11] and is associated with multiple physical and mental health problems [9, 10]. Multiple severe disadvantages including unmet mental health needs act synergistically to increase the risk of pre- mature mortality from overdose and other causes [11– 14]. Timely prevention and treatment of wide-ranging health problems in people with problem opiate use has been suggested as a way to prevent drug-related deaths [7, 12], but gold-standard randomised controlled trial evidence of health and housing interventions improving health outcomes is lacking [15–17]. In practice, access- ing care for multiple problems requires attendance at different parts of a fragmented healthcare system where specialists cater separately for: problem drug use; men- tal health; physical health; housing; benefits; and social prescribing [18]. This suggests merit in testing accessible, holistic interventions [1, 19]. Helping people experiencing homelessness who have had an overdose requires a prior understanding of their detailed unmet health and social care needs. Previously, this understanding has come from studies describing sec- ondary data e.g. using data linkage enabling inferences about populations [7, 12]. However, more nuanced data that capture information about non-prescribed and pre- scribed drug use, health service engagement, housing and other variables are also needed to understand unmet health and social care need at the individual level to inform targeted interventions [1, 20]. To date, published randomised controlled interven- tion trials targeting people experiencing homelessness with or without previous overdose (Additional file 1 provides a summary of recent randomised controlled trials (RCTs) lack sufficient detail on participants’ com- bined health and social care problems, treatments and management [15]. Multifaceted interventions aiming to improve health include peer health advisers, cash incen- tives or enhanced nurse led management of specific dis- eases, housing interventions and/or enhanced addictions management (Additional file 1) [15]. Randomised con- trolled trials in people with opioid use disorder (includ- ing people experiencing homelessness) have focussed on pharmacological interventions for opiate dependence in a younger cohort [21] than those experiencing homeless- ness with polydrug use [22]. To our knowledge, people experiencing homelessness post-overdose, despite their elevated risk of death, have not formed the target group of any intervention study (Additional file 1) [15]. In Scotland, 2021 was the first year since 2013 where drug misuse deaths have not increased (1330 in 2021 vs 1339 in 2020) [4]. This makes it the second highest annual total number of drug misuse deaths on record, 3.5 times higher than in the rest of the UK and many times higher than reports from European countries and per head of population than the USA [3, 4]. Deaths are caused, at least in part, by drugs other than opioids [4]. Strategic policy responses have prioritised uptake, access and patient choice in, substitute prescribing for opioid use disorder, provision of naloxone for emergency rever- sal of opioid overdose and heroin-assisted treatment for street heroin use [23]. In relation to problem street ben- zodiazepine use, clinical practice is informed by emerg- ing evidence from a retrospective cohort study plus local guidance [24–27]. Detailed assessment and management of problem benzodiazepine use may or may not involve benzodiazepine prescribing. It is not clear whether current approaches reduce over- doses or drug-related deaths in people with poly problem drug use. Management approaches focussed on addressing problem drug use may not address patients’ competing Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 3 of 18 priorities. These include unmet mental and physical health needs or a need for stable housing, which are associated with worse outcomes [14, 16, 17, 28, 29]. This suggests merit in offering holistic patient-centred care for those at highest risk of overdose, by addressing com- peting physical health problems, trauma and associated adverse childhood experiences [11, 30–32], housing and physical health problems. PHOENIx (Pharmacist and Homeless Outreach Engagement and Non-medical Independent prescribing Rx) is a collaborative NHS independent prescribing phar- macist and third sector homeless charity (Simon Com- munity Scotland and Marie Trust) outreach intervention offering holistic health and social care support for people experiencing homelessness post overdose [33, 34]. We hypothesise that identifying and holistically addressing multiple health and social care problems in people expe- riencing homelessness may offer an alternative, success- ful route to reducing non-fatal and fatal overdoses. This pilot study describes baseline findings from an ongoing pilot RCT. It fills gaps in our understanding of contemporary, comprehensive patient level health and social care needs, and tailored interventions aiming to improve outcomes in people experiencing homeless- ness. As a pilot RCT of a complex intervention, it follows a previous feasibility study [33] and precedes a planned definitive RCT conditional on achievement of progres- sion criteria and a signal of improved patient outcomes [35]. We report baseline findings from the ongoing PHOENIx after overdose pilot randomised controlled trial, the results of which will be available in April 2023. Methods This was a prospective, parallel group, randomised con- trolled pilot study. Participants Participant eligibility criteria are described in Table 1 and have been described in detail previously [36]. Setting The study setting is Glasgow, Scotland (drug deaths account for 33.7 per 100,000 population and over half of all deaths in people experiencing homelessness (59%, 151 deaths)) [4]. Due to the risks associated with co-prescribing OST, diazepam and gabapentinoids together, specialist alcohol and drug recovery teams take responsibility for combina- tion prescribing in people experiencing homelessness in Glasgow. Patients receiving these combinations tend to have their medicines dispensed daily, with consumption supervised in community pharmacies. For these reasons, overdose with prescribed medicines is less likely. The study therefore targeted people experiencing homeless- ness who had overdosed with non-prescribed (street) drugs. In the UK, facilities for people experiencing homeless- ness who also have problem drug use, include residen- tial rehabilitation units. These provide in-house short to medium term detoxification or stabilisation for people who have needs that cannot be met, although there are a shortage of rehabilitation beds. This level of respite care requires specialist addiction team input, and Glasgow is no different in this respect. Because of the level of spe- cialist care needed to oversee stabilisation or detoxifi- cation, these units have qualified medical and nursing staff in-house. Clinical information relating to episodes of patient care in rehabilitation units include treatments is shared with the patient’s NHS primary and secondary care providers, to enable continuity of care after patients leave rehabilitation units. There are no barriers to infor- mation sharing with NHS practitioners including NHS employee PHOENIx Pharmacists. The PHOENIx team often refer patients into these residential rehabilitation units because their care needs cannot be met elsewhere. In Glasgow, the care of problem drug or alcohol use in people experiencing homelessness with problem drug use is provided by specialist Alcohol and Drug Recovery Services (ADRS). Mental health services are provided by specialist homeless mental health teams, specialist com- munity mental health teams or via ADRS. Mainstream Table 1 Trial inclusion and exclusion criteria Inclusion criteria Exclusion criteria Homeless (living in temporary accommodation, no fixed abode or rough sleeping) [37] and Aged 18 years or over and One or more non-prescribed drug overdoses in past 6 months confirmed by self-report and witnessed overdose/ambulance call out/emergency depart- ment (ED) visit/naloxone use Living in residential or community-based rehabilitation facility which has direct access to in-house medical and nursing care or Unable to give written informed consent Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 4 of 18 General Practices, or the Specialist Homeless Health Service General Practitioner led service provide gener- alist care for physical and mental health needs, referring patients for specialist ADRS, mental health services and specialist hospital care as required. Community pharma- cies dispense medicines but have no access to patients’ clinical information. Instead, community pharmacy staff retain their own records and share information about patients with ADRS staff as and when required. All health care (including medicines) is available free of charge in NHS Scotland. Social care and third sector charity ser- vices records are not routinely shared with other services. Prescribing for problem substance use is undertaken by ADRS; Glasgow also has Scotland’s first Heroin-Assisted Treatment Unit, with capacity for approximately 20 patients. Acute and long-term prescribing of most other medicines except antiretrovirals and some other spe- cialised medication such as cancer chemotherapy is undertaken by General Practitioners and specialists in secondary care. Mental health teams (specifically psy- chiatrists) take responsibility for initiating antipsychotic medication. To have their health care needs met, patients with multimorbidity (two or more long-term conditions) [36, 38, 39] are therefore linked with at least two clinical services which are rarely collocated. The PHOENIx intervention is described below, using the TIDieR checklist [40]. Full intervention details are described previously [36]. assessment and immediate support for holistic health and social care needs, e.g. unmet mental, physical and social care needs [33, 34, 41]. This aims to improve access and continuity of care while reducing the number of services patients need to attend, and facilitating attend- ance at others, e.g. ADRS. PHOENIx aims to improve self-care and prevent deterioration in health through timely, immediate health and social care intervention on outreach. Where The PHOENIx team assertively outreach and deliver the intervention in various locations in Glasgow where people experiencing homelessness gather. This includes homeless congregate accommodation (large buildings with individual rooms, which do not have cooking facili- ties) housing multiple people experiencing homelessness. How The PHOENIx team always work in pairs. They access the patient’s existing NHS clinical records on a laptop with remote connection, while asking the patient to describe the health and social care problems that are important to them. These are recorded on paper forms and the patient’s clinical records. Through weekly conversations, PHOENIx build trusting therapeutic relationships with people experiencing homelessness, tackling problems in turn. Background What Over 7500 (13%) of UK-based registered pharmacists have undergone additional subsequent training in thera- peutics and completed a period of additional supervised clinical training, to gain an independent prescribing qual- ification enabling diagnosis and prescribing of common conditions. Independent prescribing pharmacists work in tandem with staff from Glasgow’s third sector homeless charities (the Simon Community Scotland and the Marie Trust) to offer the PHOENIx intervention. Previous qual- itative work suggests benefit to patients [41], and a feasi- bility study describes the pharmacist assessing, treating, prescribing for acute and chronic health problems and referring for initiation of opioid substitution treatment (OST), while the homeless charity link worker addresses benefits, housing and social prescribing [33, 34]. Why PHOENIx is a complex secondary prevention inter- vention. It is offered in addition to usual care, targeting people experiencing homelessness with recent overdose. It seeks to address overdose risk directly through con- ventional harm reduction (naloxone, same day access to ADRS for opioid substitution therapy) and offers Working within the clinical governance framework pro- vided by the patient’s General Practitioner and the local emergency department, the pharmacist leads on a full health assessment including measurements of weight, height, respiratory function and blood pressure, using routinely available NHS equipment. During consulta- tions, the pharmacist, homeless worker and patient may decide to use standardised questionnaires as screening tools for common conditions: anxiety/depression (PHQ- 4); modified Medical Research Council breathlessness score (mMRC) [42]; cardiovascular disease (ASSIGN); Malnutrition Universal Screening Tool (MUST); and alcohol screening (CAGE). Objective measures and sub- jective assessment scores help confirm diagnoses or severity of conditions. In some cases, depending on the patient’s clinical situation and priorities, these measures form an important part of the intervention when phar- macists chose to use them. They aid diagnoses and clini- cal decision-making in the clinical setting for pharmacists in the PHOENIx teams. Patients are routinely asked about common conditions including: hepatitis; HIV; den- tal problems; and injection site wounds; however, consul- tations follow the patient’s priorities and are personalised Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 5 of 18 to their needs. The pharmacist listens, assesses and treats accordingly which may include a handwritten prescrip- tion (for any health condition), de-prescribing, and refers to a range of different health services as needed. The third sector worker manages the patient’s benefits claim support, offers social prescribing, advocacy, liaises with the patient’s existing support workers to optimise their accommodation and attends appointments with the patient if needed. hospitalisations; treatment uptake for physical health, mental health and problem drug use; health-related qual- ity of life; and patient experience of treatment burden. Based on available guidance and data on recruitment and mortality from our feasibility study, we aimed to recruit at least 100 participants by inviting approximately 160, anticipating at least 64 with follow-up data to inform a sample size for a subsequent definitive RCT [33, 46]. When and how much Baseline assessments PHOENIx aims to visit patients once weekly, and with consultations lasting an hour on average, the team follow patients wherever possible. Some patients require addi- tional support and others require less, depending on the urgency, number of needs and patient preferences. Who provided PHOENIx staff are recruited based on their clinical independent prescribing (NHS Advanced Clinical Phar- macist) and housing (third sector worker) knowledge and skills, but also because of their street sense, empa- thy, active listening skills and non-judgemental attitude. These attributes were felt to be important to maxim- ise the chances of immediately building rapport. Peo- ple experiencing homelessness may have had difficulties forming and maintaining relationships because of past traumas [43], value receiving care directly rather than brokering [41, 44] and consider treatment for problem drug use to be effective when the care provider is com- passionate and non-judgemental, taking time to under- stand the complexity of their lives [45]. Seven NHS employee pharmacists are available to deliver outreach visits, all working part time and three part time third sector outreach workers. Pairs attempt to retain contact with the same patients continuously. Intervention fidelity Assessed by the study lead, visiting PHOENIx teams on outreach every month, sitting in during consultations to check that the patient’s expressed needs were identi- fied, and the team were supporting the patient with these needs and recording relevant information. Full methods for the pilot RCT are described previ- ously (https:// doi. org/ 10. 1186/ ISRCT N1058 5019) [36]. Briefly, the main outcome is whether to progress to a subsequent definitive randomised controlled trial based on progression criteria: recruitment of ≥ 100 participants within 4 months; ≥ 60% patients remaining in the study at 6 -and 9-month follow-up; ≥ 60% participants in the PHOENIx group receiving the intervention; and ≥ 80% participants with data collected. Secondary outcomes include: rates and time to overdose; rates and time to In-person baseline assessments and subsequent access by researchers to clinical and administrative records ena- bled gathering of comprehensive information (Additional file 2). Diagnoses data were collected through a combina- tion of self-report and confirmed from medical records (Hospital, General Practice, or Alcohol and Drug Recov- ery Service clinical records). This is a necessary approach when collecting data in people experiencing homeless- ness because their lack of repeated engagement with primary medical care leads to low levels of General Prac- tice registration. In turn, this means low levels of diag- noses recorded in General Practice clinical records and a requirement to access multiple clinical records to obtain diagnoses information. In addition, patient-reported information is important because people experiencing homelessness are itinerant, and may have registered with multiple General Practitioners, leading to missed infor- mation during transfers between care providers. Data items relate to the date of baseline data collection with the exception of questions about any overdoses, previ- ous assaults and healthcare contacts which related to the period from 6 months prior to and including the date of baseline data collection, and blood results which were included if reported within a year of the date of baseline data collection. Trial schema is summarised in Fig. 1. The research team cross-checked a 10% sample of data entries for accuracy and completeness. The Patient Experience with Treatment and Self-Management (PETS) questionnaire [47] assesses treatment burden in patients with chronic health conditions requiring self- management [47]. PETS had not been used previously in research with people experiencing homelessness; therefore, the research and clinical team worked with the developer (Dr David Eton) to adapt PETS version 2.0, to better suit the target group. The PETS including all modi- fied, translated and adapted versions of it is protected by copyright, ©2020 Mayo Foundation for Medical Educa- tion and Research. All rights reserved. Permission to use the PETS can be sought from Dr Eton. At 6- and 9-month follow-up, the research team will make repeated attempts to re-engage patients, as will the PHOENIx team during the intervention phase. Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 6 of 18 Fig. 1 Trial schema Descriptive outcomes will be conducted by independ- ent statisticians after collection of 9-month follow-up data, using MINITAB statistical software (version 21) [48]. An embedded economic evaluation will examine the feasibility of determining the cost-effectiveness of the PHOENIx intervention in a subsequent definitive trial. The main analysis will consider a health and social care service perspective whereby unit costs are applied to each item of health (e.g. hospitalisation) and social care service use data. Unit costs will be taken from routine sources where possible including missed appointments [49–51]. The effectiveness of the intervention will be explored in terms of health state utilities (for a future cost utility analysis), as measured using the EQ-5D-5L to generate quality-adjusted life years (QALYs) to be used alongside the cost data to give an indicative picture of cost-effec- tiveness [52–54]. Qualitative components are embedded in this study, to enable an understanding of how participants respond to the intervention alongside an exploration of the contex- tual issues in which the RCT occurs [55]. This includes a process evaluation of the PHOENIx intervention and an exploration of participants’ perspectives of their drug use and overdoses, using normalisation process theory (NPT) to inform conceptualisation of the process evalu- ation data [56]. Coding will be conducted independently by NF and checked by the research team to reduce the risk of bias, ensure consistency and rigour. Data will be analysed using NVivo V.12 software [57]. Baseline results Recruitment Visiting 20 different temporary accommodation ven- ues across Glasgow city centre, researchers passed study information to staff [36]. Researchers (AMcP and JM) each had over 20 years of experience working with people experiencing homelessness and those with drug and/or mental health problems. Staff in homeless accommodations identified and approached patients they knew had overdosed in the preceding 6 months. Study information was received by people experienc- ing homelessness in low-threshold city centre venues, temporary accommodation or the street. Researchers offered a £10 shopping voucher to each participant on completion of baseline assessment. One hundred and thirty eligible participants with at least one overdose in the past 6 months were offered recruitment across 20 different sites in Glasgow between 11 May and 1 September 2021. Two patients declined, leaving 128 who provided informed consent (Fig. 1). Baseline interviews lasted approximately one hour. Following baseline interviews, which were conducted in person in the patient’s choice of venue, researchers accessed each patient’s multiple clinical and social care records: General Practitioner; hospital; ADRS/men- tal health; prescribing; social work; and third sector homelessness charity. Records were sought from NHS health board areas outside Greater Glasgow and Clyde if required. This enabled capture of complete data on diagnoses, laboratory tests, prescribing, healthcare contacts, housing and registration with services. Baseline demographic, physical and mental health characteristics (Table 2) Data presented in Tables 2, 3, 4, 5 and 6 were obtained from self-report during in person assessments at base- line, and/or from case notes. Participants were on aver- age 42.2 years old (SD 8.4), 91 (71.1%) male, 127 (99.2%) described their ethnicity as white, and had been expe- riencing homelessness for a median of 23.5 years (IQR Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 7 of 18 12–29.8). Participants lived in congregate temporary accommodation with half residing in city centre hostels or hotels, staffed by third sector homeless organisations. Others lived in low-cost congregate, emergency accom- modation without any on-site support from dedicated homelessness workers, were sleeping rough, or had no fixed abode. Thirty-two (25%) participants were not reg- istered with a General Practice. A total of 124 (96.9%) had at least one physical health problem, and 117 (91%) had at least one mental health problem. Forty-three (33.8%) were under the care of a mental health team. One hun- dred and thirteen (89.7%) were registered with specialist ADRS. Approximately one-third of participants reported being isolated with no friends or family, 24 (18.8%) felt unsafe, and almost half had been assaulted in the preced- ing 6 months. Participants had a wide range of health conditions. The most common conditions were seizures, followed by den- tal problems, visual impairment, head injuries, wounds and respiratory conditions. Fifty (39%) of participants had blood borne viruses (Hepatitis C and/or HIV). The mean number of physical and mental health conditions per participant was 5.4 (SD 2.5) and 2.2 (SD 1.3), respec- tively. Eighty-five (66.4%) of participants had either self- reported and/or diagnosed depression, 56 (43.7%) had self-reported and/or diagnosed anxiety, 38(29.7%) had a history of suicide/self-harm and 25 (19.5%) had self- reported and/or diagnosed post-traumatic stress disor- der. As psychological pain is a predictor of overdose risk [14], and levels of non-engagement with mental health services, we included the PHQ-4 questionnaire, which determines levels of psychological pain/distress in base- line assessments [58]. One hundred and twenty-seven (99.2%) participants completed the PHQ-4. Scores of three or over are diagnostic of clinically relevant psy- chological pain and/or distress, and 12 is the maximum score. The median score was 12 (IQR 8–12), signifying maximum levels of anxiety and depression. Overdose and problem drug use (Table 3) Participants used a median of three different street drugs (IQR 2–4) in addition to OST, diazepam, and in one case, diamorphine from the Heroin-Assisted Treat- ment Unit. The mean (SD) number of overdoses in the past 6 months was 3.2 (3.2). A total of 81 (64%) partici- pants were able to recall the drugs taken at the time of overdose; 65 (80%) identified street valium (benzodi- azepine-type new psychoactive substance) [8] alone or with other substances as the main contributor. Half of participants described injecting drug use of whom half routinely accessed their femoral vein. Accessing either of the femoral veins constitutes risky injection practice because of the level of difficulty finding and accessing Table 2 Baseline demographic, physical and mental health characteristics (N% or mean (SD)/median (IQR) Characteristic Age (years) Sex (male) Body mass index (kg/m2)a Underweight (< 18.5 kg/m2) Overweight/Obese (> 25 kg/m2) Ethnicity (White) Number of years experienced homelessc Temporary accommodation Supported (hostel/low-cost hotel) Unsupported (hotel/bed and breakfast) Temporary Furnished Flat Rough Sleeping No fixed abode GP registered (patient reported) Mainstream GP Homeless Health Service GP Unknown Addictions Team registeredb Homeless Addictions team Mainstream addictions team Mental Health Team registeredb Homeless mental health team Mainstream mental health team Physical health conditions* Vascular Blood Borne Viruses Anaemia Skin Seizures Cardiovascular Chronic painful condition Fracture Alcohol related brain injury Respiratory Coronary heart disease Gastrointestinal (upper) Infection Epilepsy Alcohol related seizures Head/brain condition Neurological Chronic kidney disease Endocrine Genitourinary/pelvic Musculoskeletal Cachexia Rheumatic Wounds PHOENIx participants (n 128) = 42.2 (8.4) 91 (71.1%) 23.8 (5.1) 17 (14.7%) 45 (38.8%) 127 (99.2%) 23.5 (12–29.8) 72 (56.2%) 46 (35.9%) 3 (2.3%) 5 (3.9%) 2 (1.6%) 56 (43.8%) 42 (32.8%) 32 (25.2%) 113 (89.7%) 50 (39.1%) 63 (49.2%) 43 (33.8%) 13 (11.1%) 17 (14.5%) 30 (23.4%) 50 (39.0%) 3 (2.3%) 26 (20.3%) 98 (76.5%) 15 (11.7%) 20 (15.6%) 13 (10.1%) 1 (0.8%) 38 (29.6%) 9 (7.0%) 12 (9.4%) 11 (8.6%) 13 (10.2%) 2 (1.6%) 2 (1.6%) 11 (8.6%) 2 (1.6%) 16 (12.5%) 3 (2.3%) 4 (3.2%) 2 (1.6%) 3 (3.2%) 42 (32.8%) Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 8 of 18 Table 2 (continued) Characteristic Dental condition Dentures Hearing condition Eye condition Head Injury Other Number of physical health conditions/ patient PHOENIx participants (n 128) = 86 (67.2%) 48 (37.5%) 24 (18.8%) 72 (56.2%) 47 (36.7%) 7 (5.5%) 5.4 (2.5) Patients with any physical health condition 124 (96.9%) Mental health conditions Depression Anxiety Personality disorder Suicide attempt Mania/hypomania PTSD Complex trauma Childhood abuse/neglect Drug-induced psychosis Schizophrenia/psychosis Other mental health condition 85 (66.4%) 56 (43.7%) 11 (8.6%) 73 (57.9%) 1 (0.8%) 25 (19.5%) 10 (7.8%) 4 (3.1%) 8 (6.2%) 15 (11.7%) 26 (20.3%) Number of mental health conditions/patient 2.2 (1.3) Patients with any mental health condition 117 (91.4%) Psychological distresse (0–2 none; 3–5 mild; 6–8 moderate; 9–12 severe) PHQ-4 PHQ-4 3 ≥ Anxiety subscale Anxiety score 3 ≥ Depression subscale Depression score 3 ≥ Any long-term health condition 0–1 2–4 5–8 9–16 Charlson comorbidity scored Charlson 10-year survival percentaged Assaulted (past 6 months) Feels unsafe No reported next of kin 12 (8–12) 117 (92.1%) 6 (4–6) 108 (85.0%) 6 (4–6) 105 (82.7%) 0 (0%) 11 (8.6%) 59 (46.1%) 58 (45.3%) 2.8 (2.2) 67.7 (34.9) 58 (45.3%) 24 (18.8%) 39 (30.5%) = 1; cn 12. bn Ever diagnosed (from self-report or medical records) Missing data: an 80 d Charlson comorbidity index calculator assesses the 10-year survival in mainstream housed patients with several comorbidities based on the CCI scoring system e 3 threshold for screening (data missing n 1) = = ≥ = Table 3 Baseline overdose and problem drug use (N% or mean (SD)/median (IQR) Characteristic Number of overdoses in past 6 monthsa 1–2 3–5 6–10 > 11 Number of illicit drugs used concurrentlyb Problem drugs used concurrently One Two Three Four Five Six Main cause of overdoseb1 Unable to recall ‘Street Valium’ ‘Street Valium’ other drugs + Cocaine Heroin Suboxone Alcohol Main route of drug administrationb Injection Injection Sites Groin Groin and leg Groin sinuses Arms and groin Arms Legs Hands All over Feet/neck Thigh Not Sure Number of injection sitesc One to four Five to ten Too many to count Possesses naloxoned Knows how to use naloxoned Heroin Currentb Frequencye Once or more daily/most days Every few days/weekly Every 2 weeks/monthly Rarely PHOENIx participants n 128 = 3.2 (3.2) 70 (55.6%) 40 (31.7%) 14 (11.1%) 2 (1.6%) 3 (2–4) 21 (16.4%) 34 (26.6%) 41 (32.0%) 24 (18.8%) 6 (4.7%) 2 (1.6%) 46 (36.0%) 46 (56.8%) 19 (23.4%) 5 (6.2%) 6 (7.4%) 3 (3.7%) 2 (2.5%) 65 (50.8%) 23 (35.4%) 3 (4.6%) 6 (9.2%) 1 (1.5%) 14 (21.5%) 8 (12%) 2 (3.1%) 2 (3.1%) 1 (1.5%) 1 (1.5%) 1 (1.5%) 44 (34.4%) 15 (11.8%) 6 (4.7%) 80 (68.4%) 103 (88.0%) 77 (60.1%) 36 (49.3%) 9 (12.3%) 10 (13.7%) 18 (24.7%) Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 9 of 18 Table 3 (continued) Characteristic Dosef < 0.4 g ( £10) ≤ > 0.4 g but 2 g (£11–£50) ≤ > 2 g but up to 4 g (£51–£100) Routeg Intravenous Snort Smoke Age started (years) Cocaine Current Frequencyh Once or more daily/most days Every few days/weekly Every 2 weeks/monthly Rarely Dosei < one bag (0.4 g; 2 lines); ( > 1 bag—2 bags (£10—£20) £10) ≤ > 2 bags—1 g (£21—£25) > 1 g (2.5 bags; > £25) Routej Intravenous Snort Smoke Age started (years) Street Valium Current Frequencyj Once or more daily/most days Every few days/weekly Every 2 weeks/monthly Rarely Dosek 1–10 tablets 11–25 tablets 26–50 tablets 51–100 tablets > 100 tablets Routej Oral Age started (years)l Spice Current Only when in prison Street gabapentinoids Current Cannabis Current PHOENIx participants n 128 = 16 (35.6%) 26 (57.8%) 3 (6.7%) 36 (55.4%) 3 (4.6%) 26 (40.0%) Table 3 (continued) Characteristic Frequencyn Once or more daily/most days Every few days/weekly Every 2 weeks/monthly Rarely Age started (years) Tobacco Current 20.0 (16.3–26.0) Cigarettes/roll-ups/day PHOENIx participants n 128 = 23 (50.0%) 7 (15.2%) 2 (4.3%) 16 (34.8%) 15 (14–20) 113 (88.3%) 15.7 (15.2) 13.0 (5–15) 46 (36.0%) 200.7 (151.1) 13 (11.5–15.0) 38 (82.6%) 30 (65.2%) Age started (years) Alcohol Daily drinking Units/week (recommended 4u maximum) Age of first drink (years)m Previous withdrawals/DTs Previous detox/rehab for alcohol a Self-reported; adata missing n cFrom 65 respondents reporting injecting drug use. dData missing n fn 38; j1Street Valium 32; gn mData from n 1; in 105; missing n = 33; jn = 7 2. bIn past 6 months. b1Data missing n 12; hn 21; ln 8; kn = + = = = = = = 1. = 11. en hash; 4; = = = the vein and its proximity to the femoral artery. Mul- tiple injection sites were common. Most participants (80 (68.4%) possessed naloxone. Table 3 also describes detailed patterns of use (frequency, dose and route) for each of the main street drugs. Most (112 (87.5%) par- ticipants took large amounts of street valium and 60% of participants used heroin and/or cocaine, mostly by injection. The majority (88%) smoked tobacco and 41% smoked cannabis. Forty-six (36%) reported daily alco- hol consumption. The long-term nature of problem drug use was reflected by the ages at starting different drugs: on average, participants had their first cigarette and alcoholic drink aged 13 years, moving onto can- nabis age 15 years, heroin age 20 years, cocaine age 30 years with street valium one and a half years later. Significant numbers (almost 20%) also bought and took street pregabalin or gabapentin. Prescribed medicines (Table 4) More participants took daily prescribed OST (115 (89.8%) than smoked or injected heroin (77 (60.1%). In contrast, fewer people (13 (10.2%) received prescribed diazepam, than reported problem street valium use (112 (87.5%). None of the participants reported receiving counselling or other psychological behavioural therapies. Ninety-one per cent of participants had at least one (self- reported and/or confirmed by case notes) mental health problem; however, only half (67 (52.3%) were receiving 76 (59.4%) 21 (28%) 14 (18.7%) 18 (24%) 22 (33.0%) 8 (18.6%) 12 (27.9%) 4 (9.3%) 19 (44.2%) 50 (73.5%) 10 (14.7%) 8 (11.8%) 30.0 (18.5–37.0) 112 (87.5%) 71 (68.3%) 15 (14.4%) 9 (8.7%) 9 (8.7%) 24 (26.5%) 35 (38.4%) 19 (20.9%) 10 (11.0%) 3 (3.3%) 105 (100%) 31.5 (20.0–43.3) 8 (6.2%) 8 (6.2%) 22 (17.2%) 53 (41.4%) Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 10 of 18 Table 4 Prescribed medicines (N% or mean (SD)/median (IQR) Medicine PHOENIx participants Opiate substitution treatment Methadone Daily dose (mg) Buprenorphine oral/sublingual/with naloxone Daily dose (mg)a Buprenorphine injection Weekly dose (mg) Diamorphine injection Daily dose (mg) Diazepam treatment Daily dose (mg) Number of medicines for problem drug use Any medicine for problem drug use Medicines for mental health problem Any mental health medicine 128 n = 115 (89.8%) 95 (74.2%) 86.9 (29.4) 17 (13.3%) 16 (10–19.5) 2 (1.6%) 96 (64–128) 1 (0.8%) 300 (–) 13 (10.2%) 21.3 (8.9) 1 (0–1) 115 (89.8%) 67 (52.3%) Number of medicines for mental health problem 1 (0–1) Type of medicine for mental health problem Antipsychotic Antidepressant Anxiolyticb 19 (14.8%) 53 (41.4%) 23 (17.9%) Medicines for physical health problem Any medicine for physical health problem 65 (50.8%) Number of medicines for physical health problem 1 (0–2) Type of medicine for physical health problem Nutrition and anaemia Analgesic Topical for skin condition Antiepileptic Nocturnal leg cramps Upper gastrointestinal Laxative Respiratory Diabetes Antiretroviral Antibacterial/antifungal Antiplatelet Diuretic Statin Sex hormone Antihypertensive Hormone Replacement Therapy Drug for movement disorder COVID-19 vaccinec None Declined to answer question 1st dose only 1st and 2nd doses 21 (16.4%) 32 (25.0%) 2 (1.6%) 7 (5.5%) 1 (0.8%) 11 (8.6%) 1 (0.8%) 14 (10.9%) 4 (3.1%) 10 (7.8%) 5 (3.9%) 2 (1.6%) 2 (1.6%) 3 (2.3%) 1 (0.8%) 5 (3.9%) 2 (1.6%) 1 (0.8%) 23 (18%) 54 (42.2%) 28 (21.9%) 23 (18.0%) *Current Data missing an = 1; bDiazepam. cBooster unavailable at time of recruitment any mental health treatment, most commonly antide- pressants followed by anxiolytics (including diazepam) and antipsychotics. Almost all (97%) of participants had multiple treatable current physical health problems (Table 2), but only 65 (50.8%; Table 4) were receiving any treatment, the most prevalent being analgesics, medicines to treat nutritional deficiencies or anaemias, respiratory problems, upper gastrointestinal problems and antiretrovirals for blood- borne viruses. The first and second COVID-19 vaccines had been administered to most of the Scottish population in the period May–September 2021 [59]; however, only 23 (18%) of participants reported receiving both, and 28 (21.9%) reported receiving their first dose only. Baseline function, quality of life and objective health measures (Table 5) Frailty is a syndrome of vulnerability conferring an increased risk for falls, disability, hospitalisation and mortality [60]. Frailty was examined because of our pre- vious finding that people experiencing homelessness in Glasgow and Edinburgh, despite being 43 years old on average, had high levels of multimorbidity comparable to people aged 85 years in mainstream society [22]. We used an adapted Fried’s frailty phenotype [60] (Table 5) which included five measures assessed through standard ques- tions (unintentional weight loss; self-reported exhaus- tion; low physical activity; and slow walking speed) and weakness (through a hand dynamometer). Participants with three or more scores above the relevant threshold for each measure are considered frail, and those with one or two criteria are pre-frail. Of the 71 participants with all five measures available, most (50/71 (70.4%) were frail and all but one of the remainder were pre-frail. Table 5 describes EQ-5D-5L data, which enabled participants to rate their health under five domains: mobility; self-care; usual activities; pain/discomfort; and anxiety/depression. Each domain had five possible answers ranging from the participant being unable to walk about, wash/dress or self-care, having extreme pain/ discomfort/anxiety or depression (all scored as 5), to hav- ing no problem with any of the domains (scoring 1). A separate visual analogue scale ranging from 0 (worst pos- sible health) to 100 (best possible health) enabled partici- pants to rate their health. Information was available for 125 participants (98%) although two of these participants only completed the visual analogue scale section and so indexed data (cross-walked to the EQ-5D-3L value set for the UK) [53] were available for 123 participants. Overall, reported domain scores were highest (indicating poorer quality of life) at 4 (IQR 3–5) for the “depression/anxiety” Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 11 of 18 Table 5 Baseline functional, quality of life (N%) or mean (SD)/median (IQR) PHOENIx participants (n 128) = Frailty criteriah Weight loss Exhaustion Low activity Slowness Weakness Positive for frailty phenotype Pre-frail Quality of life (EQ5D5L)b Mobility (1 no problem; 2 = Usual activities slight; 3 = = (1 no problem; 2 = Pain/discomfort slight; 3 = = (1 no problem; 2 Anxiety/depression = slight; 3 = = (1 no problem; 2 moderate; 4 severe; 5 unable to mobilise) slight; 3 = = = Self-care moderate; 4 severe; 5 unable to self-care) = = = = = = = = moderate; 4 severe; 5 unable to do usual activities) moderate; 4 severe; 5 extreme pain/discomfort) extreme) (1 no problem; 2 severe; 5 = = = Overall health number—Visual Analogue Scale (VAS)c moderate; 4 slight; 3 = = (0 worst health imaginable; 100 best health imaginable) = Index Score—crosswalk method to UK Value Set = ( 0.5 = − Mealsc1 lowest score on all five domains; 1 = highest score on all five domains) Breakfast, lunch and dinner One meal only per day Two meals only per day No daily meals Modified Medical Research Council breathlessness scaled1 Oxygen saturation (%)g Peak expiratory flow rate (% predicted, l/min)i Systolic blood pressure (mmHg)e Heart rate (beats per minute)f Sodium (133–146 mmol/l)k Mean (SD) Potassium (3.5–5.3 mmol/l)l Mean (SD) Creatinine (40–130umol/l)m Mean (SD) Estimated GFR (% > 60 ml/min)m Alanine aminotransferase (% < 50 U/L)n Mean (SD) Asparate aminotransferase (% < 40 U/L)0 Mean (SD) Alkaline phosphatase (< 130 U/L)0 Mean (SD) Albumin (> 35 g/l)m Mean (SD) Calcium (adj 2.2–2.6)t 67 (65%) 92 (77.3%) 88 (69.8%) 75 (66.4%) 28 (25.2%) 50 (70.4%) 20 (28.2%) 3 (1–4) 2 (1–3) 3 (1–4) 3 (2–4) 4 (3–5) 34.4 (23.9) 0.2 (0.3) 35 (28.5%) 34 (27.6%) 34 (27.6%) 20 (16.3%) 2 (1–3)d 96.4% (2.3%) 70.6 (21.5) 115.8 (18.5) 78.3 (14.0) 103 (95.4%) 138.5 (3.2) 104 (95.4) 4.2 (0.5) 108 (99.1%) 68.4 (15.1) 106 (97.2%) 86 (80.4%) 30.9 (28.6) 80 (75.5%) 40.2 (40.7) 89 (84.0%) 103.9 (56.1) 63 (57.8%) 36.5 (5.6) 77 (89.5%) Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 12 of 18 Table 5 (continued) Mean (SD) Magnesium (> 0.7 mmol/l)p Mean (SD) C-reactive protein (< 10 mg/l)q Mean (SD) B12 (200–883 ng/l)r Mean (SD) Folate (serum: 3.1–20.0 ng/ml)s Mean (SD) Red cell count (4.5–6.5)u Mean (SD) Platelets (150–410)u Mean (SD) PHOENIx participants (n 128) = 2.3 (0.1) 52 (81.2%) 0.8 (0.2) 53 (51.5%) 19.4 (28.7) 20 (90.9%) 508.3 (217.4) 11 (47.6%) 4.3 (2.4) 33 (31.1%) 2.3 (0.1) 96 (90.6%) 258 (85.6) 5; cn 3. c1n Data missing bn 0 (breathless only on hard exercise-1-2-3-4 (too breathless to leave accommodation). hFrieds frailty phenotype (adapted): criteria en intermediate or pre-frail. d“Walk slower than other people of same age or stop for breath when walking at own pace” data missing n = 42; un 5. d, e, f, g, h and i collected at interview. k through to s: collected from clinical records most recent in past year. d1 Options 12; Data missing kn 57; Imissing data n ≥ 106; sn = 108; tn positive; 1 or 2 3 criteria 20; l, mn 22; pn 64; qn 19; nn 21; on = 5; fn 25; rn 4; gn 3; hn 22 = = = 13; Data missing = = = = = = = = = = = = = = = domain. The “mobility” and “activities of daily living” and the “pain/discomfort” domains were rated as 3 (2–4). Table 5 describes meals received by participants. Sin- gle rooms in temporary accommodation had no cook- ing facilities and given the level of destitution associated with being homeless, participants relied on food hand- outs from their accommodation or soup kitchens. Most participants (except those with no fixed abode or in temporary furnished flats, where they have no immedi- ate access to ready meals) had in-house, ultra-processed ready to eat or heat meals, soft drinks, crisps, packaged snacks, commercial bread, cakes and biscuits (particu- larly shortbread), sweetened breakfast “cereals”, sugared milk-based and “fruit” drinks. Temporary accommoda- tion had one communal microwave in the reception area, for large numbers of residents. Approximately 60% of participants had fewer than three meals/day and 20 (16%) had no daily meals, living on snacks. The majority were either underweight (15%) or overweight/obese (40%). Objective health measures (Table 5) Objective measures of respiratory status at the time of baseline interviews included an assessment of the func- tional impact of breathlessness using the modified Medi- cal Research Council breathlessness scale [42]. The scale ranged from 0 (breathless only on hard exercise) to 4 (too breathless to leave accommodation), and most partici- pants scored 2 (on level ground, participants walk slower than people of the same age because of breathlessness, or have to stop for breath when walking at their own pace on the level). Twenty-two (17%) had oxygen saturation measurements less than 95% at rest. Mean systolic blood pressure was 115.8 (SD18.5). Biochemical values were collected from medical records, expressed as the propor- tion out with NHS Greater Glasgow and Clyde laboratory reference ranges and mean (SD). In most cases, samples were taken from visiting ED or during a hospital inpa- tient episode of care rather than for screening purposes. Healthcare utilisation in preceding 6 months (Table 6) One-third of participants had been in contact with a General Practitioner and fewer had received care from a General Practice-based nurse or other healthcare profes- sional. In contrast, two-thirds of participants (80 (62.5%) had received at least one consultation with a nurse from the ADRS, and similar numbers had received care from social care staff. Between 15 and 16% had received input from an addictions doctor or non-medical (ADRS) pre- scriber, respectively. One quarter of participants had received care from a mental health nurse during the pre- vious 6 months, and fewer than one in 20 had received input from a psychiatrist. Participants had a median of three ED visits in the past 6 months, 82% having visited ED at least once. Sev- enty per cent had spent time in the local general hospi- tal although unlike mental health admissions where the median length of stay was 11.5 (5–22) days, the median length of stay in the general hospital was 2 (1–4) days. Half of participants had attended, and 40% had not managed to attend at least one scheduled outpatient appointment. Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 13 of 18 Table 6 Healthcare contacts in past 6 months (N% or mean (SD)/median Characteristic Primary care General Practice (specialist homeless or mainstream) Contacts/patienta Patients with GP contact GP-based physical health nurse consultations/patient Patients with GP-based physical health nurse contacts Other primary healthcare staff consultations/patientb Patients with other primary healthcare contactsb Alcohol and Drug Recovery Service Nurse contacts/patient Patients with any ADRS nurse contacts Pharmacist contacts/patient Patients with any addictions pharmacist contacts Medic contacts/patient Patients with any addictions medic contacts Mental health (specialist homeless or mainstream) Mental health nurse contacts/patient Patients with any mental health nurse contacts Consultant psychiatrist contacts/patient Patients with any consultant psychiatrist contacts Social care Social care staff consultations/patient Patients with any social care contacts Hospital Mental health Mental health hospitalisations/patient Patients with any mental health hospitalisation Duration of mental health hospitalisations (days) General hospital Emergency department (ED) contacts/patient Patients with any ED contacts Hospitalisations/patient Patients with any general hospitalisations Duration of general hospitalisations (days) Outpatient clinic attendance/patient Patients with any outpatient attendances Outpatient clinic appointments not attended Patients with 1 non-attendance at outpatient clinic Rehabilitation for drug use (residential) ≥ Residential rehabilitation stays/patient Patients with any residential rehab stays Duration of rehab (days) a Homeless Health service GP/mainstream GP b Includes Occupational Therapist, Dietician, Podiatrist, sexual health nurse and others, excludes addiction and mental health team PHOENIx participants n 128 = 0 (0–1) 40 (31.2%) 1 (1–2) 20 (15.6%) 0 (0–0) 14 (10.9%) 1 (0–4) 80 (62.5%) 0 (0–0) 20 (15.6%) 0 (0–0) 19 (14.8%) 0 (0–1) 33 (25.8%) 0 (0–0) 5 (3.9%) 1 (0–4) 79 (61.7%) 0 (0–0) 8 (6.2%) 11.5 (5.0–22.0) 3 (1–5) 105 (82.0%) 1 (0–2) 89 (69.5%) 2 (1–4) 1 (0–2) 66 (51.6%) 0 (0–3) 51 (39.8%) 0 (0–0) 5 (3.9%) 21.5 (11.5–23.8) Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 14 of 18 Discussion Despite having a reputation for being hard to reach, 128 from 130 participants were engaged, provided consent and detailed baseline information during lengthy in per- son assessments in one of 20 different venues. The median duration of homelessness was 23.5 years, which is approximately half the life expectancy of a per- son experiencing homelessness in Scotland [3, 11, 61]. Participants had pervasive, high-risk polydrug use, using a median of three different street drugs in addition to prescribed OST and in some cases, prescribed diazepam. Participants had a mean of three non-fatal overdoses in the previous 6 months. Participants were frail and had a greater number of health conditions than people double their age in mainstream society [62], conferring a high level of susceptibility to, and impact from overdose. Most participants were known to and receiving OST from ADRS. However, participants continued to overdose, in most cases, with street valium, for which there is no strong evidence base for treatment [8, 24] although one in ten were prescribed maintenance diazepam. Heroin-assisted treatment of problem opiate use has some evidence of reduced street heroin use; however, the impact on overdose remains uncertain [63]. Patients with active significant medical or psychiatric conditions were excluded from the most recent, definitive RCT of her- oin-assisted treatment which included 127 participants [63]. One hundred and twenty-four (96.9%) participants had active significant physical health problems, and 117 (91.4%) had psychiatric conditions: the majority would have been excluded from the most recent trial of heroin- assisted treatment [63]. This makes it difficult to general- ise the utility of prescribed heroin to our cohort. The effectiveness of current care for problem drug use in people experiencing homelessness could be assessed by measures such as the number of drug-related deaths or overdoses or the number of participants present- ing to emergency department with drug-related prob- lems. Given one of the main outcomes from treatments such as OST and diazepam is to reduce harm and pre- vent overdose and deaths, the effectiveness of current care appears conditional on the effectiveness of these treatments for poly problem drug use. That participants repeatedly overdosed despite receiving OST and in some cases, diazepam, highlights an evidence and practice gap in the care of participants in this study, in terms of their problem drug use. In terms of established interventions, OST is proven to reduce all cause and overdose mortality in people dependent on street opioids [21]. However, it remains uncertain whether people currently experienc- ing homelessness with polydrug use including opioids, accrue these benefits because of exclusions from previous RCTs [21]. The extent of ongoing polydrug use including street heroin (60% of participants, 49% of whom use at least 0.4 g once daily (Table 3) suggests more studies are required to examine whether OST at optimal dose (mean 87 mg) (Table 4) prevents illicit drug use and overdose in our cohort. People experiencing homelessness were largely absent from trials of OST, and participants tak- ing three different street drugs were also absent, mean- ing there is a lack of evidence of benefit in the type of patients within our cohort [21]. In addition, now that the characteristics of people experiencing homelessness with recent overdose are known (Tables 2, 3, 4, 5, 6), a comparison with existing literature shows there are no established interventions known to reduce overdoses or emergency department visits in this type of cohort (Additional file 1). Difficulties associated with recruitment of people experiencing homelessness in trials may have previously limited collection of information about characteristics [64, 65]. Our recruitment rate (128 participants from 130 potential participants) was higher than expected. This may have been due to close collaboration between researchers and third sector homeless charity workers, who had established relationships with eligible partici- pants. Outreaching to participants in their own spaces enabled engagement. Participants were offered a shop- ping voucher on completion of baseline interviews. This and the non-judgemental, empathetic approach by the researchers may have contributed to high recruitment rates. A parallel process evaluation is underway and will capture information on barriers and facilitators to recruitment from the perspectives of participants and staff. “Housing First” offers permanent, self-contained hous- ing for people experiencing homelessness, alongside wrap-around health and social care support. It is an evi- denced approach to ending homelessness for people with complex needs including mental health problems [16, 17, 66]. In this sample, no participants were being considered for Housing First accommodation, at baseline, despite being eligible for Housing First. It is out with the scope of this pilot RCT to fully explore why this was the case. Participants had an average of two mental health problems. Medicines and other approaches have a role in the treatment of people with complex trauma and other mental health problems. Sixty-seven (50%) were in receipt of medicines for mental health problems, but none of the 128 participants were receiving any other form of specialist mental health input. While this is in keeping with previous work [67], restrictions due to COVID-19 may have contributed to participants experi- encing difficulties accessing mental health services, par- ticularly when digital options were not available to those in temporary accommodation. Participants had more Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 15 of 18 physical health problems than mental health problems. Only half of those with physical or mental health prob- lems were receiving any treatment, a finding consistent with the rule of halves [68]. Residential treatment services for drug use are scarce across Scotland [69]. Our findings confirmed less than one in 20 participants had received residential treat- ment in the previous 6 months. This may represent low uptake, but comparisons are not possible because previ- ous studies of people experiencing homelessness have not recruited a comparable sample in terms of range and chronicity of drug use with complex physical and mental health problems (Additional file 1) [15]. Inequities in the prescribing of diazepam may be due to clinical decision making in the light of uncertain evi- dence of benefit particularly in a high-risk cohort using multiple drugs who are frail. Given the high prevalence and importance of problem street diazepam use in our participants, and the congregate living conditions which bring participants into close proximity, it is likely that participants are aware of each other’s drug habits and treatments. Prescribed diazepam inequalities are unlikely to be lost on participants who already have a height- ened sense of discrimination and stigmatisation and live together. Other unexplained inequities shown by these data include: variable uptake of COVID-19 vaccination; irregular registration with General Practitioners; and low levels of registration with mental health services. Health-related quality of life is regarded to be the most relevant outcome for people experiencing homelessness; health outcomes are significantly associated with quality- of-life scores [70, 71]. Participants’ current situation, plus the cumulative long-term impact of severe and multiple disadvantage, was manifest in quality-of-life findings which were rated in the bottom third of the EQ-5D Vis- ual Analogue Scale. Patient responses to EQ5D5L scores are matched to a general population sample that has pre- viously rated every possible response combination to the questionnaire’s five domains, to estimate how much the population values being in (or avoiding) that particular health state [54]. These population values range from 1 (full health) to a minimum of − 0.224, beyond the zero score for death. This accounts for the possibility that there are some health states the public would prefer to avoid so much that they would rather be dead. Matching PHOENIx participants to these scores showed one-third of respondents at baseline were in health states consid- ered “worse than death”. Quality of life offers a possible primary outcome measure in the future randomised con- trolled trials of people experiencing homelessness. Our findings demonstrate extensive unmet health and social care needs of people experiencing homelessness post overdose. These needs are unlikely to be met by continuation of care as usual. Innovative models of care and new interventions are necessary to address the status quo, accompanied by robust, pragmatic research includ- ing qualitative research to understand the complexity and barriers and facilitators to real world implementa- tion [1, 72–74]. The PHOENIx intervention and RCT offers a novel, generalist approach instead of the current problem drug use oriented approach which characterises usual care. PHOENIx acknowledges patients’ priorities, and their multiple and competing relational, social care and health problems including maximum levels of frailty, anxiety and depression, which contribute to overdose risk [14]. People experiencing homelessness are known to have more difficulty using fragmented care systems, as compared with people without multiple health needs [62, 71, 75]. The existing evidence base for reducing drug- related deaths does not favour the current approach of tackling single morbidities, e.g. problem drug use, in iso- lation [1, 30, 76]. People experiencing homelessness do not favour the current approach either. [30, 76]. Together, the range and complexity of life threatening problems and under treatment characterising study participants makes a case for testing a transformational approach to offering and providing comprehensive, continuous and co-ordinated health and social care. The competing needs of finding safety, managing the impact of an accumulated treatment burden and self- medicating for anxiety and substance dependence may have diverted attention away from health seeking behav- iour until problems became overwhelming and required ED attendance [77]. The alliance–outcome relation- ship is one of the strongest predictors of treatment suc- cess [78]. We hypothesise that supportive relationships built through outreach may prevent or delay emergency department attendance if the skills and knowledge of those delivering outreach are sufficient to deal with most of the patient’s problems. Supportive relationships in conjunction with practical, immediate help with a range of health and social care problems are core features of the PHOENIx intervention [41]. Limitations to generalisability include most of the participants identifying as Caucasian, and recruitment from one Scottish city albeit across 20 different venues. Screening for other specific conditions, e.g. atrial fibril- lation through electrocardiography or blood samples for nutritional deficiencies, did not form part of baseline assessments which limits our understanding of these and other important needs. Worldwide, proportions of people experiencing homelessness using multiple street drugs and overdosing are unclear, making generalisations based on these data difficult. There were 26,166 home- lessness applications across Scotland in 2021/2022 [79]. The number of people experiencing homelessness with Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 16 of 18 problem drug use in Glasgow is in the region of 3500 [22]; however, numbers overdosing remain uncertain, making it difficult to know whether findings from 128 participants are generalizable. Characteristics of peo- ple dying drug-related deaths show the mean age has increased from 35 years in 2009, to 42 years in 2018, and the most commonly implicated substances were street benzodiazepines, methadone and heroin/morphine [4, 80, 81]. Our sample demographic is a close match to the characteristics of those experiencing homelessness and dying from drug-related causes in Scotland as a whole [3]. The number of participants in our pilot RCT is com- parable to the numbers recruited in previous (definitive) studies (Additional file 1) [15]. Conclusion People experiencing homelessness with recent over- dose can be recruited, and their characteristics can be described through comprehensive baseline data collected in the context of a pilot RCT. Complex drug use and frequent overdose combined with multiple unmet health needs. This suggests the cur- rent focus on stabilising street drug use and reducing harm from drugs without attention to wider health and social care needs including unstable housing, are failing to protect against non-fatal and by inference, fatal over- dose. Current models of care in Glasgow and worldwide (Additional file 1) [15] tend to focus on single conditions, an approach that does not seem sensible when multimor- bidity, re-traumatising living conditions [82] and frailty are the norms. This signals an urgent need for broaden- ing the scope of support offered on outreach, to include a health and social care partnership, to address wider determinants of non-fatal and fatal overdoses. If retention and intervention delivery targets are achieved, together with a signal of improvement in outcomes such as overdoses or quality of life, funding will be sought for a definitive RCT of the PHOENIx intervention. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12954- 023- 00771-4. Additional file 1. RCTs of interventions to improve health outcomes. Additional file 2. Baseline data collection form. Acknowledgements We acknowledge the efforts of every patient and member of staff who will- ingly gave their time to enable recruitment, and study pharmacists (KS, CL, RB, RR, AB), Simon Community (AS, SR) and Marie Trust staff (DB, FMcK). Author contributions Each author (RL, AMcP, FSM, DM, VP, BB, DB, JM, FH, CD, KS, NF, RR, CL, SR, AS, GP, LS, JH, FR and AEW) helped write and draft the protocol and baseline manuscript and gave final approval before submission for publication. Plan- ning, conduct and reporting were undertaken by RL, AMcP, FSM, DM, VP, FH, NF, JH, CJ, SL and AEW. The study was conceived by RL. Design input was received from RL, FSM, AEW and NF. Acquisition of data was done by AMcP, JM, NF, FH, AS, SR, DB, BB, CL, RR and RL. Analysis was performed by AMcP, FH and CJ. Interpretation was done by RL, AMcP, FSM, DM, VP, BB, DB, JM, FH, CD, KS, NF, RR, CL, SR, AS, GP, LS, CJ, SL, JH, FR and AEW. All authors read and approved the final manuscript. Funding This study was funded by the NHS Greater Glasgow and Clyde and the Scot- tish Government’s Drug Deaths Task Force. The study design, data collection, analysis, interpretation and governance are independent from commercial sponsorship of any kind. Availability of data and materials All data used during the study are available on reasonable request from the corresponding author. Declarations Ethics approval and consent to participate This study was approved by NHS South East Scotland Research Ethics Com- mittee 01. REC reference 21/SS/0004. Consent for publication All participants gave consent for publication of their aggregated, anonymised data. Competing interests The authors declare that they have no competing interests. Author details 1 Pharmacy Services, Homeless Health/Research and Development, NHS Greater Glasgow and Clyde, Glasgow G76 7AT, Scotland, UK. 2 General Practice and Primary Care, School of Health and Wellbeing, College of Medical, Veteri- nary and Life Sciences, University of Glasgow, Glasgow, Scotland, UK. 3 Emer- gency Medicine, Glasgow Royal Infirmary, Glasgow, Scotland, UK. 4 School of Pharmacy, University of Birmingham, Birmingham, England, UK. 5 Addictions Psychiatry, NHS Ayrshire and Arran, Crosshouse, Scotland, UK. 6 Simon Com- munity Scotland Street Team, Glasgow, Scotland, UK. 7 East End Addictions Services, Alcohol and Drug Recovery Service, Glasgow Health and Social Care Partnership, NHS Greater Glasgow and Clyde, Glasgow, UK. 8 Healthcare Improvement Scotland, Glasgow, Scotland, UK. 9 Department of Social work, School of Health and Life Sciences, Glasgow Caledonian University, Glasgow, UK. 10 Usher Institute, College of Medicine and Veterinary Medicine, The Uni- versity of Edinburgh, Edinburgh, UK. 11 Shelter Scotland, Glasgow, UK. Received: 25 November 2022 Accepted: 16 March 2023 References 1. NICE. Integrated health and social care for people experiencing home- lessness. NICE Guideline 214; 2022. https:// www. nice. org. uk/ guida nce/ ng214. 2. ONS. Deaths of homeless people in England and Wales: 2020 registra- tions. Deaths of homeless people in England and Wales - Office for National Statistics (ons.gov.uk); 2021. 3. National Records of Scotland. Homeless Deaths 2020. Edinburgh: Crown Copyright. Homeless Deaths 2020\|National Records of Scotland (nrscot- land.gov.uk); 2021. 4. National Records of Scotland. Drug-related deaths in Scotland in 2021. Edinburgh: Crown Copyright. Drug-related deaths in Scotland in 2021, Report (nrscotland.gov.uk); 2022. Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 17 of 18 5. Larney S, Tran LT, Leung J, et al. All cause and cause specific mortality among people using extramedical opioids: a systematic review and meta analysis. JAMA Psychiat. 2020;77:493–502. 6. Caudarella A, Dong H, Milloy MJ, et al. Non fatal overdose as a risk factor for subsequent fatal overdose among people who inject drugs. Drug Alcohol Depend. 2016;162:51–5. 7. Aldridge RW, Menezes D, Lewer D, Cornes M, Evans H, Blackburn RM, Byng R, Clark M, Denaxas S, Fuller J, Hewett N, Kilmister A, Luchenski S, Manthorpe J, McKee M, Neale J, Story A, Tinelli M, Whiteford M, Wurie F, Hayward A. Causes of death among homeless people: a population- based cross-sectional study of linked hospitalisation and mortality data in England. Wellcome Open Res. 2019;11(4):49. https:// doi. org/ 10. 12688/ wellc omeop enres. 15151.1. 8. McAuley A, Matheson C, Robertson JR. From the clinic to street: the changing role of benzodiazepines in the Scottish overdose epidemic. Int J Drug Policy. 2022;100:103512. 25. Anon. Working with people at harm from street benzodiazepine use— guidance for ADRS front line staff. NHS Greater Glasgow and Clyde; April 2021. 26. Anon. Guidance on the principles of benzodiazepine prescribing with concomitant opiate dependence. GG&C Alcohol and Drug Recovery Services; February 2019. 27. Anon. MAT standards informed response for benzodiazepine harm reduc- tion. Drugs Death Task Force/Public Health Scotland; August 2021. 28. Park JN, Rouhani S, Beletsky LEO, Vincent I, et al. Situating the continuum of overdose risk in the social determinants of health: a new conceptual framework. Millbank Q. 2020;98:700–46. 29. Joudrey PJ, Khan MR, Wang EA, et al. A conceptual model for under- standing post-release opioid related overdose risk. Addict Sci Clin Pract. 2019;14:17. 30. Hwang SW, Burns T. Health interventions for people who are homeless. The Lancet. 2014;384:1541–7. 9. Warner-Smith M, Darke S, Day C. Morbidity associated with non fatal 31. Campbell et al. Primary healthcare needs and barriers to care among heroin overdose. Addiction. 2002;97(8):963–7. 10. Schneider KE, Tomko C, et al. Conceptualising overdose trauma: the relationships between experiencing and witnessing overdoses with PTSD symptoms among street-recruited female sex workers in Baltimore, Maryland. Int J Drug Policy. 2020;92:102589. 11. Morrison DS. Homelessness as an independent risk factor for mortality: results from a retrospective cohort study. Int J Epidemiol. 2009;38(3):877– 83. https:// doi. org/ 10. 1093/ ije/ dyp160. 12. Tweed E, Leyland A, Morrison D, Katikireddi SV. Premature mortality in people affected by co-occurring homelessness, justice involvement, opioid dependence, and psychosis: a retrospective cohort study using linked administrative data. Lancet Public Health. 2022;7:e733–43. 13. Bramley G, Fitzpatrick S, Wood J, Sosenko F, Blenkinsopp J, Littlewood M, Frew C, Bashar T, McIntyre J, Johnsen S. Hard Edges Scotland: new conversations about severe and multiple disadvantage. London: Lankelly Chase Foundation; 2019. 14. Tomko C, Schneider KE, Rouhani S, et al. Identifying pathways to recent non fatal overdose among people who use opioids non-medially: How do psychological pain and unmet mental health need contribute to overdose risk? Addict Behav. 2022;127:107215. 15. Hanlon P, Yeoman L, Gibson L, Esiovwa R, Williamson AE, Mair FS, Lowrie R. A systematic review of interventions by healthcare professionals to improve management of non-communicable diseases and communica- ble diseases requiring long-term care in adults who are homeless. BMJ Open. 2018;8:e020161–e020161. Calgary’s homeless populations. BMC Fam Pract. 2015. 32. Adverse childhood experiences and their impact on health-harming behaviours in the Welsh adult population. 33. Lowrie R, Stock K, Lucey S, et al. Pharmacist led homeless outreach engagement and non-medical independent prescribing (Rx) (PHOENIx) intervention for people experiencing homelessness: a non- randomised feasibility study. Int J Equity Health. 2021;20:19. https:// doi. org/ 10. 1186/ s12939- 020- 01337-7. 34. Lowrie F, Gibson L, Towle I, Lowrie R. A descriptive study of a novel phar- macist led health outreach service for those experiencing homelessness. IJCP. 2019. https:// doi. org/ 10. 1111/ ijpp. 12520. 35. Skivington K, Matthews L, Simpson SA, et al. A new framework for devel- oping and evaluating complex interventions: update of medical research council guidance. BMJ. 2021;374:n2061. 36. Lowrie R, McPherson A, Mair FS, Maguire D, Paudyal V, Blair B, Brannan D, Moir J, Hughes F, Duncan C, Stock K, Farmer N, Ramage R, Lombard C, Ross S, Scott A, Provan G, Sills L, Hislop J, Reilly F, Williamson AE. Improv- ing the health of people experiencing homelessness with recent drug overdose: rationale for and design of the Pharmacist and Homeless Outreach worker Engagement Non-medical Independent prescribing Rx (PHOENIx) pilot randomised controlled trial. BMJ Open. 2022;12:e064792. https:// doi. org/ 10. 1136/ bmjop en- 2022- 064792. 37. ETHOS typology on Homelessness and Housing Exclusion https:// www. feant sa. org/ en/ toolk it/ 2005/ 04/ 01/ ethos- typol ogy- on- homel essne ss- and- housi ng- exclu sion. 16. Onapa H, Sharpley CF, Bitsika V, et al. The physical and mental health 38. Johnston M, Crilly M, Black C, Prescott GJ, Mercer SW. Defining and effects of housing homeless people: a systematic review. Health Soc Care Community. 2022;30:448–68. measuring multimorbidity: a systematic review of systematic reviews. Eur J Pub Health. 2019;29(1):182–9. 17. Aubry T, Bloch G, Brcic V, et al. Effectiveness of permanent supportive 39. Skou ST, Mair FS, Fortin M, et al. Multimorbidity. Nat Rev Dis Primers. housing and income assistance interventions for homeless individuals in high income countries: a systematic review. Lancet Public Health. 2020;5:e342–60. 18. Samet JH, Friedmann P, Saitz R. Benefts of linking primary medical care and substance abuse services: patient, provider, and societal perspec- tives. JAMA Intern Med. 2001;161(1):85–91. 19. Rae M, Howkins J, Holland A. Escalating drug deaths in the UK. A funda- mental reorientation in approach is needed. BMJ. 2022;378:o2005. 20. Priester MA, Browne T, et al. Treatment access barriers and disparities among individuals with co-occurring mental health and substance use disorders: an integrative literature review. J Subst Abuse Treat. 2016;61:47–59. 21. Sordo L, Barrio G, Bravo MJ, et al. Mortality risk during and after opioid substitution treatment: systematic review and meta-analysis of cohort studies. BMJ. 2017;357:j1550. https:// doi. org/ 10. 1136/ bmj. j1550. 22. Zeitler M, Williamson AE, Budd J, Spencer R, Queen A, Lowrie R. Compar- ing the impact of primary care practice design in two inner city UK homelessness services. J Prim Care Community Health. 2020;11:1–13. 23. Scottish Government. Medication assisted treatment standards: access, choice, support; 2021. https:// www. gov. scot/ publi catio ns/ medic ation- assis ted- treat ment- mat- stand ards- scotl and- access- choice- suppo rt/. 24. Matheson C et al. Benzort final report. University of Stirling; September 2022. 2022;8:48. https:// doi. org/ 10. 1038/ s41572- 022- 00376-4. 40. Hoffmann T, Glasziou P, Boutron I, et al. Better reporting of interventions: template for intervention description and replication (TIDieR) checklist and guide. BMJ. 2014;348:g1687. 41. Johnsen S, Cuthill F, Blenkinsopp J. Outreach-based clinical pharmacist prescribing input into the healthcare of people experiencing homeless- ness: a qualitative investigation. BMC Health Serv Res. 2021;21:7. https:// doi. org/ 10. 1186/ s12913- 020- 06013-8. 42. Jones PW, Adamek L, Nadeau G, et al. Comparisons of health status scores with MRC grades in COPD: implications for the GOLD 2011 clas- sification. Eur Respir J. 2013;42:647–54. 43. Lester S, Khatwa M, Sutcliffe K. What support do young people affected by adverse childhood experiences need? Child Youth Serv Rev. 2020;118:105429. 44. Coldwell CM, Bender WS. The effectiveness of assertive community treatment for homeless populations with severe mental illness: a meta analysis. Am J Psychiatry. 2007;164:393–9. 45. Carver H, Ring N, Miler J, Parkes T. What constitutes effective problem- atic substance use treatment from the perspective of people who are homeless? A systematic review and met-ethnography. Harm Reduct J. 2020;17:10. 46. Sim J, Lewis M. The size of a pilot study for a clinical trial should be calculated in relation to considerations of precision and efficiency. J Clin Epidemiol. 2012;65(3):301–8. Lowrie et al. Harm Reduction Journal (2023) 20:46 Page 18 of 18 47. Eton DT, Yost KJ, Lai J, Ridgeway JL, et al. Development and valida- 72. Morrison D. Extent, nature, and causes of homelessness in Glasgow. A tion of the patient experience with treatment and self-management (PETS): a patient-reported measure of treatment burden. Qual Life Res. 2017;26(2):489–503. needs assessment. Glasgow: Homeless Partnership, Greater Glasgow and Clyde Health Board; 2003. 73. Hwang SW, Tolomiczenko G, Kouyoumdjian FG, Garner RE. Interventions 48. MINITAB statistical software (Version 21). MINITAB LLC. (2021). Minitab. 49. Public Health Scotland Data and Intelligence. Scottish Health Service Costs. Edinburgh; 2019. https:// beta. isdsc otland. org/ find- publi catio ns- and- data/ healt hcare- resou rces/ finan ce/ scott ish- health- servi ce- costs/. 50. Curtis L, Burns A. Unit costs of health and social care 2019, Personal Social Services Research Unit, University of Kent, Canterbury; 2018. https:// www. pssru. ac. uk/ proje ct- pages/ unit- costs/ unit- costs- 2019/. 51. NHS Improvement. National Schedule of NHS Costs. London: NHS Improvement 2019. https:// impro vement. nhs. uk/ resou rces/ natio nal- cost- colle ction/. 52. Hernández Alava M, Pudney S, Wailoo A. Estimating the relationship between EQ-5D-5L and EQ-5D-3L: results from an English population study (PDF, 1.3MB). Policy Research Unit in Economic Evaluation of Health and Care Interventions. Universities of Sheffield and York. Report 063; 2020. to improve the health of the homeless: a systematic review. Am J Prev Med. 2005;29(4):311–9. https:// doi. org/ 10. 1016/j. amepre. 2005. 06. 017. 74. Fazel S, Geddes J, Kushel M. The health of homeless people in high income countries: descriptive epidemiology. Health consequences, and clinical and policy recommendations. The Lancet. 2014;384:1529–40. 75. Gunner E, Chandan SK, Marwick S, Saunders K, Burwood S, Yahyouche A, Paudyal V. Provision and accessibility of primary healthcare services for people who are homeless: a qualitative study of patient perspectives in the UK. Br J Gen Pract. 2019;69(685):e526–36. https:// doi. org/ 10. 3399/ bjgp1 9X704 633. 76. Johnsen S, Blenkinsopp J, Rayment M. Scotland’s Housing first pathfinder evaluation final report. Heriot-Watt University, Edinburgh (in press). 77. Gelberg L, Gallagher TC, Andersen RM, Koegel P. Competing priorities as a barrier to medical care among homeless adults in Los Angeles. Am J Pub Health. 1987;87(2):217–20. 53. EuroQoL. EQ-5D-5L\|Valuation\|Crosswalk Index Value Calculator. Rotter- 78. Horvath AO, Del Re AC. Alliance in individual psychotherapy. Psychotherapy. 79. Homelessness in Scotland: 2021/22. https:// www. gov. scot/ publi catio ns/ homel essne ss- scotl and- 2021- 22/ pages/4/. 80. Detailed insight into drug-related deaths in Scotland seeks to inform col- lective response - News - Public Health Scotland. Public Health Scotland; July 2022 81. Drug related deaths in Scotland 2021—additional analyses. NRS Scotland. 82. Farmer N, McPherson A, Lowrie R. What aspects of support are perceived as most significant in order to prevent drug overdose in people who experience homelessness? SAPC 2022. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. dam. 2020 https:// euroq ol. org/ eq- 5d- instr uments/ eq- 5d- 5l- about/ valua tion- stand ard- value- sets/ cross walk- index- value- calcu lator/. 54. Dolan P. Modelling valuations for EuroQol health states. med Care. 1997;35(11):1095–108. 55. O’Cathain A, Thomas KJ, Drabble SJ, Rudolph A, Hewison J. What can qualitative research do for randomised controlled trials? A systematic mapping review. BMJ Open. 2013;3(6):e002889. 56. May C, Finch T. Implementing, embedding and integrating practices: an outline of normalisation process theory. Sociology. 2009;43(3):535–54. 57. QSR International Pty Ltd. NVivo (Version 12); 2018. https:// www. qsrin terna tional. com/ nvivo- quali tative- data- analy sis- softw are/ home. 58. Kroenke K, Spitzer RL, Williams JB, Löwe B. An ultra-brief screening scale for anxiety and depression: the PHQ-4. Psychosomatics. 2009;50(6):613– 21. https:// doi. org/ 10. 1176/ appi. psy. 50.6. 613. 59. Coronavirus (COVID-19): data for Scotland. (COVID-19 Daily Dashboard). Scottish Government. Coronavirus (COVID-19): data for Scotland - gov. scot (www. gov. scot). 60. Fried LP, et al. Frailty in older adults: evidence for a phenotype. J Gerontol Ser A Biol Sci Med Sci. 2001. https:// doi. org/ 10. 1093/ gerona/ 56.3. M146. 61. National Records of Scotland. Homeless Deaths 2021. Edinburgh: Crown Copyright. National Records of Scotland (nrscotland.gov.uk); 2022. 62. Barnett K, Mercer S, Norbury M, Watt GCM, Wyke S, Guthrie B. Epidemiol- ogy of multimorbidity and implications for health care, research and medical education: a cross sectional study. Lancet. 2012;380:37–43. 63. Strang J, Metrebian N, Lintzeris N, et al. Supervised injectable heroin or injectable methadone versus optimised oral methadone as treatment for chronic heroin addicts in England after persistent failure in orthodox treatment (RIOTT): a randomised trial. Lancet. 2010;375:1885–95. 64. Hough RL, Tarke H, Renker V, et al. Recruitment and retention of homeless mentally ill participants in research. J Consult Clin Psychol. 1996;64(5):881–91. 65. Aldridge S, Enevoldsen C. Homelessness and Street Homelessness in Eng- land: trends, causes and what works. Eur J Homelessness. 2021;15:143–59. 66. Baxter AJ, Tweed EJ, Katikireddi SV, Thomson H. Effects of housing first approaches on health and well-being of adults who are homeless or at risk of homelessness: systematic review and meta-analysis of randomised controlled trials. J Epidemiol Community Health. 2019;73(5):379–87. 67. Hard Edges Scotland. https:// lanke llych ase. org. uk/ wp- conte nt/ uploa ds/ 2019/ 06/ Hard- Edges- Scotl and- full- report- June- 2019. pdf. 68. Tudor HJ. Rule of halves: implications of increasing diagnosis and reduc- ing dropout for future workload and prescribing costs in primary care. BJGP. 1992;42:116–9. 69. Scottish drugs Forum. Drug-related deaths in Scotland: MSP Briefing; 2021. 70. Henwood BF, Matejkowski J, Stefancic A, et al. Quality of life after housing first for adults with serious mental illness who have experienced chronic homelessness. Psychiatry Res. 2014;220(1–2):549–55. https:// doi. org/ 10. 1016/j. psych res. 2014. 07. 072. 71. Schoen C, Osborn R. 2010 Commonwealth Fund international health policy survey. http:// www. commo nweal thfund. org/. Surve ys/ 2010/ Nov/ 2010- Inter natio nal- Survey. aspx. Accessed 22 Dec 2011. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12931-023-02455-w	Warren et al. Respiratory Research (2023) 24:162 https://doi.org/10.1186/s12931-023-02455-w Respiratory Research Alveolar macrophages from EVALI patients and e-cigarette users: a story of shifting phenotype Kristi J. Warren1,2†, Emily M. Beck1,2†, Sean J. Callahan1,2†, My N. Helms1, Elizabeth Middleton1, Sean Maddock2,3, Jason R. Carr1,4, Dixie Harris4, Denitza P. Blagev4, Michael J. Lanspa4, Samuel M. Brown4 and Robert Paine III1,2 Abstract Exposure to e-cigarette vapors alters important biologic processes including phagocytosis, lipid metabolism, and cytokine activity in the airways and alveolar spaces. Little is known about the biologic mechanisms underpinning the conversion to e-cigarette, or vaping, product use-associated lung injury (EVALI) from normal e-cigarette use in otherwise healthy individuals. We compared cell populations and inflammatory immune populations from bronchoalveolar lavage fluid in individuals with EVALI to e-cigarette users without respiratory disease and healthy controls and found that e-cigarette users with EVALI demonstrate a neutrophilic inflammation with alveolar macrophages skewed towards inflammatory (M1) phenotype and cytokine profile. Comparatively, e-cigarette users without EVALI demonstrate lower inflammatory cytokine production and express features associated with a reparative (M2) phenotype. These data indicate macrophage-specific changes are occurring in e-cigarette users who develop EVALI. Keywords E-cigarette, Or vaping, Product use-associated lung injury (EVALI), Bronchoalveolar lavage (BAL), Alveolar macrophages (AM) Introduction E-cigarette, or vaping, product use-associated lung injury (EVALI) is an acute respiratory illness that inflicts sub- stantial morbidity and mortality [1, 2]. The United States experienced a major outbreak of EVALI in 2019 pre- dominantly in association with vitamin E acetate (VEA) adulteration of THC containing e-cigarettes [3]. This out- break led to the recognition of EVALI and to thousands of illnesses and over sixty confirmed deaths [4]. Its clini- cal features include an acute inflammatory state associ- ated with extensive ground glass opacities, neutrophilic alveolitis and increased numbers of lipid-laden macro- phages (LLM) in bronchoalveolar lavage (BAL). Many EVALI patients have a rapid and dramatic improvement with corticosteroid treatment [2, 4]. Unlike lung injury †Kristi J. Warren, Emily M. Beck and Sean J. Callahan contributed equally to this work. Correspondence: Kristi J. Warren kristi.warren@hsc.utah.edu 1Department of Internal Medicine, Division of Pulmonary & Critical Care Medicine, University of Utah, Salt Lake City, UT 84132, USA 2George E. Wahlen VA Medical Center, 500 Foothill Dr, Salt Lake City, UT 84148, USA 3Division of Pulmonary and Critical Care Medicine, National Jewish Health, Denver, CO 80206, USA 4Intermountain Healthcare, Department of Pulmonary & Critical Care Medicine, Murray, UT 84107, USA This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. RESEARCHOpen AccessPage 2 of 10 caused by traditional cigarette use, EVALI develops rap- idly in a young (< 35 years of age) healthy population [5]. and chemokine (CCL22), allowing for resolution of the inflammatory response [18]. At present, the pathobiology of EVALI is poorly characterized. It is probable there are multiple cellu- lar contributors to its development [6], but the alveolar macrophage (AM) is likely to play a central role. Health agencies identified vitamin E acetate (VEA), a diluent often used in tetrahydrocannabinol (THC) e-cigarette liquids, as the causative agent leading to EVALI [7]. Sub- sequent analyses supported this association, as investi- gators demonstrated LLM and markers of inflammation in mice exposed to vaporized VEA that was not seen in mice exposed to the nicotine-containing JUUL™ (a brand of e-cigarettes) aerosol [7]. Interestingly, even during the initial EVALI outbreak, a small but significant minority of patients had no preceding exposure to THC [8]. Federal authorities removed much of the illicit THC from circu- lation later in 2019, resulting in a marked reduction in case counts [3, 5, 9–11]. In the years since the recogni- tion of VEA as a major driver of the 2019 EVALI outbreak and through the COVID-19 pandemic, the number of patients with EVALI has decreased; nevertheless a steady number of patients continue to present with EVALI [12, 13]. Therefore, it remains unclear whether EVALI occurs because of a single vaping constituent or if it is a multi- factorial disorder involving host-specific conditions (i.e. genetic predisposition), perhaps in combination with dose effects of vaping products. Alveolar macrophages (AM) are lung resident cells that play a crucial role in host defense and repair in the lung. They have important roles in clearing pathogens and as sentinel cells that secrete inflammatory and anti-inflam- matory mediators to recruit and activate other immune and inflammatory cells, including circulating mononu- clear cells and neutrophils. Under normal circumstances, their activity is carefully regulated to optimize host defense while preventing impaired gas exchange and, ultimately, acute lung injury. Macrophages demonstrate a broad range of phenotypic characteristics, imperfectly captured along a spectrum characterized as M1 (clas- sically activated/inflammatory) and M2 (alternatively activated/reparative) phenotypes [14]. Resident AM in the healthy lung strike a balance in which they express a mixture of M1 and M2 features [15]. In the setting of acute insults such as pneumonia, AM become fully acti- vated, expressing abundant early response cytokines (TNF-α, IL-1β) and chemokines (IL-8 and CCL2), tak- ing on an M1 predominant phenotype [16]. In addition to activation of resident AM during acute injury, circu- lating mononuclear phagocytes are recruited to the lung and differentiate into tissue (alveolar) macrophages [17]. In later stages, macrophages express M2 characteristics, such as expression of the reparative cytokine (IL-10) The consistent evidence of severe alveolar inflamma- tion and morphologic changes found in AM of EVALI patients suggest that AM are a key target for investiga- tion. We hypothesized that a localized immune response induces alterations in macrophage activity, which results in an exuberant inflammatory response in e-Cig users who develop EVALI. Here, we postulate that EVALI is a consequence of an exuberant inflammatory response to vaped substances by AM, resulting in recruitment and activation of neutrophils and recruitment of circulating mononuclear phagocytes, which contribute to the devel- opment of extensive alveolar exudates and respiratory failure. Methods Human subjects Research subjects were recruited from the University of Utah and Intermountain Healthcare medical systems under IRB-approved protocols and provided informed consent. Eligible participants were between the ages of 18–50, had no history of chronic lung disease, and no recent respiratory symptoms/illness in the preceding four weeks. All EVALI subjects were hospitalized and met the CDC definition of “confirmed” or “probable” EVALI [19]. EVALI subjects were identified for enrollment by inpa- tient treating clinicians and recruited by study coordina- tors; none required mechanical ventilation or received corticosteroids prior to bronchoscopy. Subjects with EVALI were excluded if bronchoscopy presented excess risk for intubation by the treating clinician. Healthy sub- jects (healthy) and chronic users of e-cigarettes with- out EVALI (e-Cig controls) self-referred for enrollment. Healthy subjects reported no current or past use of cig- arettes or vaping. E-cigarette users without EVALI self- reported current vaping every day, or most days, with no current cigarette use (< 100 traditional cigarettes in their lifetime and quit ≥ 1 year prior to enrollment). All healthy volunteers and three of the e-Cig subjects were recruited for a prior bronchoscopy study that was not published; BAL procedure and sampling protocols were identical across groups. Bronchoscopy and BAL collection All subjects underwent bronchoscopy with BAL under moderate sedation. In addition to BAL, a standard e-cig- arette exposure history was completed for subjects who were vaping. We obtained BAL fluid from healthy vol- unteers, e-Cig controls, and subjects hospitalized with EVALI. Each BAL was performed using a standardized protocol from the SPIROMICS Bronchoscopy Substudy and in line with previously published research bronchos- copy protocols [20, 21]. Briefly, subjects were allowed Warren et al. Respiratory Research (2023) 24:162 Page 3 of 10 nothing by mouth for at least four hours prior to the procedure. Topical anesthesia was achieved using 1% lidocaine and moderate sedation was administered per hospital protocol. Subjects were continuously monitored with pulse oximetry, sphygmomanometry, and 3-lead ECG throughout the procedure. BAL was performed in the right middle lobe and lingula. Serial aliquots were instilled up to a total volume not to exceed 150 mL in each lung or 300 mL total. All samples were pooled, immediately stored, and transported on ice, and placed in a 4 °C refrigerator for short term (< 24 h) storage until further processing was completed. BAL cytospin preparation and quantitation BAL cell counts were determined and 2,000–4,000 cells were diluted in 500 µL of sterile saline. 250 µL were applied to glass slides using the Thermo Scientific Cyto- spin 4 centrifuge; 2 slides were prepared for each study participant. Cells were applied to slides using the slide adaptors and filters (Biorad) by gentle centrifugation at 100 X g for 10 min at room temperature. Slides were dried at room temperature for 5 min, then treated with cold methanol for 5 min, before staining with Giemsa stain (Sigma) according to the manufacturers protocol, also described in Misharin et al. (2017) [22]. All slides were examined under an inverted light microscope. Eosinophils (Eos), neutrophils (PMNs), lymphocytes (Lymphs), macrophages (Macs) and ciliated airway epi- thelial cells were easily differentiated. 20–50 cells were quantified per field, 4–6 fields were assessed per subject. Finally, cell numbers were normalized to 200 cells per cell type and graphed as mean cell count per standard error of the mean (SEM). These data are shown in Fig. 1. Flow cytometry Cells were separated from supernatants by gentle cen- trifugation at 300 X g for 10 min. Total BAL cells were counted using trypan blue exclusion. 1–3 × 106 cells were added into 5 mL polystyrene tubes (Falcon) containing Zombie Aqua (Biolegend; cat# 77,143). The following anti-human antibodies were acquired from Biolegend: APC-Fire 750 HLA-DR (clone: I.243; cat#307,657), CD14 conjugated to Brilliant Violet 711 (clone: M5E2; cat# 301,837), CD163 conjugated to PE-Cy7 (clone: GHI/61; cat#333,613), CD64 conjugated to Brilliant Violet 605 (clone: 10.1; cat# 305,033), CD11b conjugated to Brilliant Violet 750 (clone: M1/70; cat# 101,267). Anti-Human CD11c conjugated to AF700 was acquired from Invit- rogen (clone: 3.9; cat# 56-0116-41). The remaining anti- human antibodies were purchased from BD Biosciences; anti-human CD16 conjugated to BUV 496 (clone: 3G8; cat# 612,945), anti-human CD45 conjugated to BUV 805 (clone: HI30, cat# 612,891) and anti-human CD206 conjugated to BUV 395 (clone: 19.2; cat# 740,309). Com- pensation was completed using UltraComp eBeads from Invitrogen (cat# 01-2222-42), Zombie aqua only stained and unstained cellular events were collected for each sample including the healthy, e-Cig and EVALI groups to properly set flow cytometric gating for data acquisition on the Cytek Aurora. Individual samples were further Fig. 1 Increased neutrophils and MPO are detected in EVALI subjects. Healthy and e-cigarette controls were recruited as controls for EVALI subjects that had been admitted to the hospital for acute lung injury. Bronchoalveolar lavage fluids were collected following standard SPIROMICS procedures and cellular content was separated from supernatant. (A) Cytospins were prepared from the separate cells and eosinophils (EOS), lymphocytes (Lymphs), neutrophils (PMNs), Macrophages (MACs) and airway epithelial cells were quantified. At least 20–50 cells were counted field, 4–6 fields were counted per slide. Results as displayed as mean ± SEM. For healthy controls n = 7, e-cigarette controls (e-Cig) n = 13, and EVALI subjects n = 10. Statistical significance was determined by ordinary one-way ANOVA followed by Dunnett’s multiple comparison post-test to determine between groups differences Warren et al. Respiratory Research (2023) 24:162 Page 4 of 10 analyzed using FlowJo, v.10 software, using unstained cellular events from each study subject to set flow gating (shown in Fig. 2). ELISA A 1 milliliter aliquot of bronchoalveolar lavage fluid was pre-cleared by centrifugation (500 x g for 10 min) prior to analysis by ELISA. All ELISA were performed accord- ing to the manufacturer’s instructions without modifica- tions. A BCA (cat# 71285-3) protein assay was purchased from Millipore. Human IL-6 (cat# DY206), IL-8 (DY208), RAGE (DY1145), and CRP (Cat# QK1707) ELISA kits were acquired from R&D Systems (Minneapolis, MN). Human serum amyloid A, or SAA (cat# KHA0011), MPO (BMS2038INST), and sICAM ELISA kits (cat# BMS201) were acquired from Invitrogen (Vienna, Austria). Absor- bance readings for the standard curve and experimental samples were detected at 450 nm with 570 nM correction on the SpectraMax M3 from Molecular Devices. Con- centrations of detected proteins were extrapolated from standard curve and displayed as mean +/- the standard error of the mean in Fig. 2. Statistical analysis Data are presented as the mean ± standard error of mean (SEM). Statistics were performed using one-way analysis of variance (ANOVA) after confirmation that all data were normally distributedusing D’Agostino & Fig. 2 Composition of immune lineage cells differ in EVALI compared to otherwise healthy e-cigarette controls. Flow cytometric analysis was completed using a macrophage specific flow panel that included a viability dye, and antibodies specific for CD45, CD11c, CD11b, CD14, CD16, CD206, CD64 and CD163. (A)-(C) Total immune cells were detected as CD45 + cells in three groups healthy participants (A; healthy), e-cigarette controls (B; e-Cig), and EVALI subjects (C; EVALI). (D) The counts of CD45 + cells were normalized per mL of returned lavage fluid and shown as count/mL. (G-I) gating schematic and counts of (E) dendritic cells and (F) lymphocytes per mL of returned BAL fluid. (M-O) quantitation for (M) CD16lo and (N) CD16high inflammatory mono- cytes and (O) neutrophils and (J-L) the representative flow gating for each group. (J-L) shows the representative gating for alveolar macrophages, popu- lations are contained in the dark red box. The quantitation of (P) total alveolar macrophages (aMacs), (Q) CD163- M1 macrophages and (R) CD163 + M2 macrophages. The composition of total BAL of each of the above cell populations are displayed in S-U for each group. Statistical significance was deter- mined using an ordinary One-Way ANOVA with Dunnett’s post-test to determine between groups differences; indicates p < 0.05, * indicates p < 0.01, *** indicates p < 0.001. Sample size is n = 7 for healthy controls, n = 13 for e-Cig controls and n = 10 for EVALI subjects Warren et al. Respiratory Research (2023) 24:162 Table 1 Study Participant Characteristics Number Age (median) Reported Sex Female (%) Male (%) Race Caucasian (%) 7 28 2 (29) 5 (71) 3 (43) American Indian or Alaskan Native (%) 0 (0) 0 (0) Native Hawaiian or Other Pacific Islander (%) Other (%) Undisclosed/Refused (%) 0 (0) 4 (57) Ethnicity Non-Hispanic (%) Hispanic (%) Undisclosed/Refused (%) Years Vaped (median) Product Vaped Nicotine (%) THC/CBD (%) Dual-Use (%) 3 (43) 0 (0) 4 (57) n/a n/a n/a n/a EVALI 10 22 5 (50) 5 (50) E- Cig 13 23 7 (54) 6 (46) 8 (80) 9 (69) 0 (0) 1 (10) 1 (8) 0 (0) 1 (8) 0 (0) 1 (10) 2 (15) 7 (70) 10 (77) 1 (8) 2 (20) 1 (10) 2 (15) 3 2.5 6 (60) 10 (100) 6 (60)* 11 (85) 4 (31) 2 (15) Number Puffs Per Day <10 Puffs Per Day (%) >10 Puffs Per Day (%) 1 subject denied THC use but urine testing was positive for THC metabolites; this subject was re-classified as a dual-user 2 (20) 8 (80) n/a n/a 5 5 * Puffs/day not collected in 3 subjects Pearson testing. Tukey’s multiple comparison, post-tests were employed to compare differences across the three groups. All statistical analysis were completed on Graph- Pad Prism (version 9). In all analyses, P values less than 0.05 were considered statistically significant. Results Participant characteristics are summarized in Table 1. In general, subjects were young (< 35 years of age), Cauca- sian, and non-Hispanic. We enrolled 10 EVALI subjects, 13 e-Cig controls, and 7 healthy controls. Nine (90%) EVALI subjects were hospitalized for acute hypoxemic respiratory failure (requiring supplemental oxygen) and received antibiotics covering organisms that contribute to community acquired bacterial pneumonia; one (10%) subject was hospitalized but did not require supplemen- tal oxygen. The median time from hospital admission to Page 5 of 10 bronchoscopy was one day. Seven (70%) of EVALI sub- jects had a “confirmed” diagnosis by CDC criteria. The three remaining subjects met “probable” EVALI criteria: two tested positive for rhinovirus/enterovirus on nasal viral PCR, and the other had an incomplete rheumato- logic workup. The treating clinicians and research team classified these subjects as “probable” given that the clini- cal history was more consistent with EVALI than com- peting diagnoses; the two subjects with positive nasal swabs had negative viral panels on BAL and no medical history predisposing to respiratory failure from these viruses. Only one subject required intensive care unit (ICU) admission for their hypoxemia; the remaining subjects in the EVALI group who required supplemental oxygen support by simple nasal cannula were admitted to the medicine wards. Increased numbers of neutrophils are present in BAL of EVALI subjects BAL cytospins were examined as a crude means of under- standing general immune mediated lung inflammation. Numbers of eosinophils (Eos), lymphocytes (Lymphs), neutrophils (PMNs), macrophages (Macs) and air- way epithelial cells were determined in EVALI subjects, healthy controls, e-Cig controls (Fig. 1A). Lymphocytes and neutrophils were increased in EVALI in comparison to both healthy (p < 0.05 lymphocytes and p < 0.0001 for PMN) and e-Cig controls (p < 0.01 for lymphocytes and p < 0.0001 for PMN). Healthy and e-Cig controls main- tained higher numbers of macrophages in their BAL as compared to EVALI (p < 0.0001). Although two EVALI subjects had eosinophils detected in BAL, this difference did not achieve statistical significance (p = 0.0752). Only EVALI subjects had increased amounts of myeloperoxi- dase (MPO) compared to healthy (p < 0.0001) and e-Cig controls (p < 0.0001) (Fig. 1B). Finally, although readily visualized and quantified, there were no statistical differ- ences in the numbers of alveolar epithelial cells between groups (EVALI to healthy comparison p = 0.113). (p < 0.001) Cellular composition of BAL differs according to EVALI and e-Cig group status The concentration of CD45 + cells per mL were increased in EVALI compared to e-Cig (p < 0.01) and healthy (Fig. 2A-D). CD11c+CD11b− controls (DC) were detected with SSClo dendritic cells FSC+CD11b−CD11c−CD206−CD14− cells, that are likely lymphoid origin cell populations (Fig. 2E, F, G-I). DC and lymphoid origin cells were only statistically differ- ent in the EVALI subjects compared to healthy controls (p < 0.05); there were no differences between e-Cig sub- jects and EVALI subjects for either of these cell popula- tions. Total monocytes were detected, and discriminated as CD16 lo (Fig. 2M) and CD16 hi (Fig. 2N) monocytes, Warren et al. Respiratory Research (2023) 24:162 Page 6 of 10 to distinguish classical monocytes from inflammatory monocytes, respectively. EVALI subjects had a higher CD16 hi population of monocytes when comparing to healthy (p < 0.05) and e-Cig controls (p < 0.05), and healthy and e-Cig controls had a higher concentration of CD16 lo classical monocytes as compared to EVALI sub- jects (p < 0.01). As with the cytospin data, we were able to confirm an increased accumulation of neutrophils in the BAL fluid of EVALI subjects compared to both healthy controls (p < 0.05) and e-Cig controls (p < 0.05) (Fig. 2O). Finally, alveolar macrophages were examined (Fig. 2P) and discriminated along an M1-M2 spectrum based on CD163 and HLA-DR staining: M1-skewed macrophages were determined as HLA-DR hi CD163− (Fig. 2Q); while M2-skewed macrophages were also HLA-DRhi they were also CD163+ (Fig. 2R). Total macrophages were increased in EVALI subjects compared to healthy (p < 0.05). EVALI subjects had more CD163− macrophages in comparison to healthy (p < 0.01) and e-Cig controls (p < 0.01). Con- trastingly, e-Cig controls had more CD163+ macrophages as compared to EVALI subjects (p < 0.05), and healthy controls had comparable numbers of CD163+ macro- phages to the e-Cig controls. Finally, the composition of BAL cell populations were summarized in Fig. 2S and U. Markers of inflammation and airway leak are increased in BAL from EVALI subjects in comparison to e-Cig and healthy controls Prototypical markers of inflammation, including TNFα, IL-6, and the chemokines, IL-8 and CCL2, were increased in EVALI patients compared to healthy (p < 0.01) and e-Cig controls (p < 0.01)(Fig. 3A-D). Additional markers of lung pathogenesis, sICAM, RAGE, total protein, serum amyloid A and C-reactive protein, were all detected at higher levels in BAL fluid from EVALI patients compared to healthy (p < 0.01) and e-Cig (p < 0.01)(Fig. 3E-I). Of note, as with all the results thus far, healthy controls and e-Cig controls were similar and had virtually no markers of inflammation in their BAL fluid. Discussion Understanding the biologic underpinnings of EVALI is required to aid in the prevention, diagnosis, and treat- ment of this novel syndrome. We aimed to differentiate AM phenotypes in routine e-cigarette use from EVALI. Macrophage differentiation along a spectrum towards an M1 phenotype is a finding seen in various forms of acute lung injury and acute respiratory distress syndrome (ALI/ ARDS) in both murine and human studies [16, 23, 24]. In the case of ALI, macrophages are activated towards the M1 phenotype, with release of pro-inflammatory cytokines, such as tumor necrosis factor-alpha (TNF-α), interleukin-1 beta (IL-1β), and interleukin-6 (IL-6), and the production of reactive oxygen and nitrogen species (ROS and RNS, respectively). While this response is criti- cal for the initiation of the immune response to ALI, it is also nonspecific and can be triggered by a variety of insults, including viral infections, bacterial infections, and chemical exposure. Here, we demonstrate that AM from individuals with EVALI display features of an inflammatory, or M1-type, phenotype. In contrast, AM from e-Cig controls without EVALI express M2-skewed features, similar to AM from healthy controls. Neutro- philic inflammation, airway leak, and elevated IL-8 and CCL2 fit well with the M1 macrophage phenotype in the EVALI subjects. Neutrophilic infiltration in the air- ways contributes to the pathogenesis of COPD and other inflammatory lung diseases [25, 26]. Electronic-cigarette controls and healthy controls demonstrate comparable levels of CD163+ (M2-skewed) macrophages with no inflammatory cytokine profile detected. The CD163 marker was elevated on macrophages of e-Cig controls compared to EVALI subjects, however, indicating that macrophages from e-Cig controls maintain a relatively anti-inflammatory state. Together the data suggest that interventions to maintain AM in a normal, more anti- inflammatory state (associated with CD163 expression) might prevent or improve outcomes associated with EVALI. This line of reasoning regarding CD163 on AM of course requires more exploration before becoming clini- cally relevant. The specific components of the vaping mixture used in e-cigarette devices likely contribute to the development of EVALI [27, 28]. All EVALI subjects used THC prod- ucts, while only 30% (4/13) of the e-Cig group reported THC use. In the 2019 EVALI outbreak, > 85% of patients had presumed THC exposure and 94% had VEA detected in BAL fluid [3]. Mice exposed to aerosolized VEA devel- oped diffuse lung injury with a neutrophilic and mono- cytic inflammatory infiltrate as well as an increase in lung water and BAL protein levels [7]. The relative con- tributions of different vaping constituents for human EVALI is not yet clear but is the subject of active ongoing investigation. Chronic e-cigarette use may predispose users to devel- oping clinical disease by increasing susceptibility to respi- ratory infection [29]. In a study where mice were exposed to e-cigarette aerosols with and without nicotine, AM demonstrated intracytoplasmic inclusions and M2 mark- ers as compared to air-exposed animals [30]. However, mice exposed to e-cigarette aerosols had an excessive inflammatory response to the influenza A viral infection which resulted in increased levels of IFN-γ and TNFα. While histologic changes also demonstrated increased inflammation and edema, those animals exposed to e-cigarette vapors had decreased survival suggesting this hyperactive immune response was mounted due to vap- ing exposure that led to detrimental immune pathology Warren et al. Respiratory Research (2023) 24:162 Page 7 of 10 Fig. 3 Markers of inflammation and airway leak are increased in EVALI subjects compared to e-cigarette and healthy controls. As above, BAL was pre- cleared of cellular content and analyzed for markers of inflammation; (A) TNFα, (B) IL-6, (C) IL-8, (D) CCL2, (E) soluble ICAM and (F) RAGE by ELISA. Markers of airway leak were also detected; (G) C reactive protein (CRP), (H) serum amyloid A (SAA), and total protein (BCA; µg/mL). Statistical significance was de- termined using an ordinary One-Way ANOVA with Dunnett’s post-test to determine between groups differences; indicates p < 0.05, * indicates p < 0.01, * indicates p < 0.001, ** indicates p < 0.0001. Sample size is n = 7 for healthy controls, n = 13 for e-Cig controls and n = 10 for EVALI subjects in the delicate lung [30]; the cascade suggests e-ciga- rette aerosols prime the subject for an excessive inflam- matory response to viral infection. This was similarly reported in an in vitro model using e-cigarette extracts on human small airway epithelial cells [31]. In contrast, in our human study we did not see increased inflamma- tory proteins (e.g. TNFα, MPO) in e-Cig controls. Only two of the EVALI subject tested positive for rhinoviral Warren et al. Respiratory Research (2023) 24:162 Page 8 of 10 infection. Other studies have reported normally non- noxious pathogens in EVALI subjects [8]. Whether e-Cig use primes the lung for excess inflammatory response to these normally benign pathogens leading to EVALI in some individuals requires additional investigation. Our results add to prior studies examining the lungs’ immune responses to e-cigarette use. We did not evalu- ate pulmonary physiologic responses to e-cigarette use, as has been done in prior studies, which demonstrated alterations in heart rate, blood pressure, arterial stiff- ness, and oxygen tension [32–34] in subjects vaping nicotine-based solutions, but we did examine the inflam- matory profile and immune response like prior inves- tigators and arrived at different findings. As opposed to previous studies [35–37], we did not find increased BAL neutrophil or lymphocyte cellularity in our regular e-Cig users, though they were found in higher degrees in sub- jects with EVALI. Similarly, our EVALI subjects demon- strated increased BAL inflammatory markers like TNFα and IL-8, as opposed to the e-Cig group which largely resembled healthy controls. This finding differs from prior studies of the lungs’ response to e-liquids with nicotine, which have found abnormal inflammatory fea- tures [34, 35, 38] and evidence of direct airway damage [39], A limitation of our observational study is that we did not standardize exposures – such as the flavors per- mitted, nicotine/THC strength, frequency of use, or the time interval between the time of last e-cigarette use and the time of research bronchoscopy – in our e-Cig group; prior studies that demonstrated immune or inflamma- tory changes specifically accounted for these. It is thus feasible the negative results witnessed in the e-Cig group could be a result this variability, but further research is needed to confirm or refute this possibility. Generalizing findings from animal studies to human pathogenesis of disease is challenging. As our findings demonstrate, the alveolar cellular composition in e-Cig controls resembles healthy controls and most e-Cig con- trols do not go on to develop EVALI. It may be that the pathogenesis of EVALI is that of a multi-hit model where chronic exposure to e-cigarette vapors in the setting of differences in host characteristics, inadequately under- stood differences in e-cigarette constituents (such as exposure to VEA), and/or exposure to common patho- gens, result in clinical disease. There are several limitations to our study: first, there are no accepted standards on the quantity and quality of e-cigarette constitutes, devices, or post-market modifica- tions. We did not analyze the composition of e-cigarette vapors or e-liquid composition or assess the generation (1st – 4th ) of e-cigarette devices; recent reports have suggested that the later generation e-cigarette devices mitigate the progression to EVALI [40]. Post-marketing modifications to achieve higher temperatures or higher levels of THC or nicotine were not accounted for and we did not routinely confirm THC/nicotine metabo- lites present in our subjects (THC and nicotine use were self-reported). Only two of the ten EVALI subjects were tested for THC through routine clinical testing (not con- ducted as part of this study); both tested positive. Second, we did not examine the distinct morphologic features of AM obtained from EVALI subjects versus e-Cig controls to determine whether lipid-packed vacuoles were pres- ent in EVALI subjects and not in e-Cig controls, or vice versa. Initial reports of EVALI identified the presence of lipid-laden macrophages (LLM) in EVALI patients. However, LLM are present in BAL fluid of smokers of traditional cigarettes and are also seen in various other conditions such as lipoid pneumonia, pulmonary alveo- lar proteinosis, gastroesophageal reflux, and aspiration syndromes and thus, are not specific to EVALI [29, 41– 43]. Finally, unknown confounders may exist that could be important, including brand of vaping liquid, depth of inhalation, number of puffs/session, number of sessions/ day, and time from last vaping session to symptom onset/ hospital presentation; we received limited data from our subjects regarding this information. An important strength of this study is the compari- son of individuals with EVALI to e-Cig users without respiratory illness. While we examined several inflam- matory proteins, we were unable to identify a specific biomarker that underpins the transition from health to EVALI. There may be additional biomarkers of expo- sure and e-cigarette effect not included in the analysis. Future studies should evaluate sequencing differences and differences in lipid homeostasis in the macrophages and airway cells. Other markers of lung integrity (i.e., surfactants), and functional analysis of AM (phagocyto- sis) are important to establish in future studies, as well. This report is the first study that specifically investigates a comprehensive lung immune cell profile in subjects with EVALI, and compare EVALI subjects to both healthy and chronic e-Cig controls. In summary, we have demonstrated that EVALI induces a robust inflammatory response that is mediated at least in part by an inflammatory AM phenotype and that the AM in chronic e-Cig users more closely resembles healthy control subjects. Given these findings, we specu- late that EVALI likely develops as part of a sequence of insults that culminate in exuberant inflammation and respiratory compromise. Acknowledgements The Clinical Research Team at the University of Utah, in particular Lisa Weaver and Lindsey Waddoups, for their commitment and support of pulmonary clinical research. Authors’ contributions KJW generated the results in the figures, performed statistical analysis and wrote the paper. EMB and SJC designed the study, performed Warren et al. Respiratory Research (2023) 24:162 bronchoscopies, interpreted data, and contributed to the introduction, methods, and discussion. MH and EM generated data, interpreted data and approved the final version of the manuscript. DH, SM, JC, and MJL performed bronchoscopies and approved the final version of the manuscript. RP designed the study, evaluated data, and approved the final version of the manuscript. Funding National Institutes of Health grant U01HL123018-06S1 (awarded to SB, supporting EB, SJC, DH, DBP and MLL); Department of Veteran Affairs IK2 BX4004219 and 5I01 BX004637 (awarded to KJW) and 5I01BX001777-05 (awarded to RP); National Institutes of Health grant 5T32HL105321 (awarded to RP, supporting JRC); the University of Utah flow cytometry core is supported by the Office of the Director of the National Institute of Health under award number S10OD026959 and the NCI award number 5P30CA042014-24. Data availability All data generated and analysed during this study are included in this published article. Declarations Ethics approval and consent to participate The studies presented in this manuscript involved human subjects. All study protocols were reviewed and approved by the University of Utah IRB committee. Informed consent was acquired from each participant prior to inclusion in the study. Consent for publication Not applicable. Competing interests The authors declare that there are no competing interests. Received: 17 November 2022 / Accepted: 19 May 2023 References 1. Siegel DA, Jatlaoui TC, Koumans EH, Kiernan EA, Layer M, Cates JE, Kimball A, Weissman DN, Petersen EE, Reagan-Steiner S, Godfred-Cato S, Moulia D, Moritz E, Lehnert JD, Mitchko J, London J, Zaki SR, King BA, Jones CM, Patel A, Delman DM, Koppaka R. Lung Injury Response Clinical Working G, Lung Injury Response Epidemiology/Surveillance G. Update: Interim Guidance for Health Care Providers evaluating and caring for patients with suspected E-cigarette, or Vaping, Product Use Associated Lung Injury - United States, October 2019. MMWR Morb Mortal Wkly Rep. 2019;68:919–27. Perrine CG, Pickens CM, Boehmer TK, King BA, Jones CM, DeSisto CL, Duca LM, Lekiachvili A, Kenemer B, Shamout M, Landen MG, Lynfield R, Ghinai I, Heinzerling A, Lewis N, Pray IW, Tanz LJ, Patel A, Briss PA. Lung Injury Response Epidemiology/Surveillance G. characteristics of a Multistate Outbreak of Lung Injury Associated with E-cigarette Use, or Vaping - United States, 2019. MMWR Morb Mortal Wkly Rep. 2019;68:860–4. Blount BC, Karwowski MP, Shields PG, Morel-Espinosa M, Valentin-Blasini L, Gardner M, Braselton M, Brosius CR, Caron KT, Chambers D, Corstvet J, Cowan E, De Jesus VR, Espinosa P, Fernandez C, Holder C, Kuklenyik Z, Kusovschi JD, Newman C, Reis GB, Rees J, Reese C, Silva L, Seyler T, Song MA, Sosnoff C, Spitzer CR, Tevis D, Wang L, Watson C, Wewers MD, Xia B, Heitkemper DT, Ghinai I, Layden J, Briss P, King BA, Delaney LJ, Jones CM, Baldwin GT, Patel A, Meaney-Delman D, Rose D, Krishnasamy V, Barr JR, Thomas J, Pirkle JL. Lung Injury Response Laboratory Working G. vitamin E acetate in Bronchoalveolar- Lavage Fluid Associated with EVALI. N Engl J Med. 2020;382:697–705. Layden JE, Ghinai I, Pray I, Kimball A, Layer M, Tenforde MW, Navon L, Hoots B, Salvatore PP, Elderbrook M, Haupt T, Kanne J, Patel MT, Saathoff-Huber L, King BA, Schier JG, Mikosz CA, Meiman J. Pulmonary illness related to E-Cigarette Use in Illinois and Wisconsin - Final Report. N Engl J Med. 2020;382:903–16. Prevention CfDCa. Outbreak of Lung Injury Associated with the Use of E-Ciga- rette, or Vaping, Products. 2021 August 3, 2021 [cited 2022 September 12, 2. 3. 4. 5. Page 9 of 10 6. 2022]. Available from: https://www.cdc.gov/tobacco/basic_information/e- cigarettes/severe-lung-disease.html. Jasper AE, Sapey E, Thickett DR, Scott A. Understanding potential mecha- nisms of harm: the drivers of electronic cigarette-induced changes in alveolar macrophages, neutrophils, and lung epithelial cells. Am J Physiol Lung Cell Mol Physiol. 2021;321:L336–48. 7. Matsumoto S, Fang X, Traber MG, Jones KD, Langelier C, Hayakawa Serpa 8. P, Calfee CS, Matthay MA, Gotts JE. Dose-dependent pulmonary toxicity of aerosolized vitamin E acetate. Am J Respir Cell Mol Biol. 2020;63:748–57. Aberegg SK, Cirulis MM, Maddock SD, Freeman A, Keenan LM, Pirozzi CS, Raman SM, Schroeder J, Mann H, Callahan SJ. Clinical, bronchoscopic, and imaging findings of e-Cigarette, or Vaping, Product Use-Associated Lung Injury among Patients treated at an Academic Medical Center. JAMA Netw Open. 2020;3:e2019176. 9. Navon L, Jones CM, Ghinai I, King BA, Briss PA, Hacker KA, Layden JE. Risk factors for E-Cigarette, or Vaping, Product Use-Associated Lung Injury (EVALI) among adults who use E-Cigarette, or Vaping, Products - Illinois, July-October 2019. MMWR Morb Mortal Wkly Rep. 2019;68:1034–9. 10. Ghinai I, Navon L, Gunn JKL, Duca LM, Brister S, Love S, Brink R, Fajardo G, Johnson J, Saathoff-Huber L, King BA, Jones CM, Krishnasamy VP, Layden JE. Characteristics of persons who Report using only nicotine-containing products among interviewed patients with E-cigarette, or Vaping, Product Use-Associated Lung Injury - Illinois, August-December 2019. MMWR Morb Mortal Wkly Rep. 2020;69:84–9. 11. Taylor J, Wiens T, Peterson J, Saravia S, Lunda M, Hanson K, Wogen M, D’Heilly P, Margetta J, Bye M, Cole C, Mumm E, Schwerzler L, Makhtal R, Danila R, Lynfield R, Holzbauer S. Lung Injury Response Task F. characteristics of E-cigarette, or Vaping, Products used by patients with Associated Lung Injury and Products seized by Law Enforcement - Minnesota, 2018 and 2019. MMWR Morb Mortal Wkly Rep. 2019;68:1096–100. 12. Armatas C, Heinzerling A, Wilken JA. Notes from the Field: E-cigarette, or Vap- ing, Product Use-Associated Lung Injury cases during the COVID-19 response - California, 2020. MMWR Morb Mortal Wkly Rep. 2020;69:801–2. 13. Callahan SJ, Harris D, Collingridge DS, Guidry DW, Dean NC, Lanspa MJ, Bla- gev DP. Diagnosing EVALI in the time of COVID-19. Chest. 2020;158:2034–7. 14. Abbas AK, Lichtman AH, Pillai S. Basic immunology: functions and disorders of the immune system. St. Louis, Missouri: Elsevier; 2016. 15. Byrne AJ, Mathie SA, Gregory LG, Lloyd CM. Pulmonary macrophages: key players in the innate defence of the airways. Thorax. 2015;70:1189–96. 16. Chen X, Tang J, Shuai W, Meng J, Feng J, Han Z. Macrophage polarization and its role in the pathogenesis of acute lung injury/acute respiratory distress syndrome. Inflamm Res. 2020;69:883–95. 17. Morales-Nebreda L, Misharin AV, Perlman H, Budinger GR. The heterogene- ity of lung macrophages in the susceptibility to disease. Eur Respir Rev. 2015;24:505–9. Johnston LK, Rims CR, Gill SE, McGuire JK, Manicone AM. Pulmonary macro- phage subpopulations in the induction and resolution of acute lung injury. Am J Respir Cell Mol Biol. 2012;47:417–26. 18. 19. Schier JG, Meiman JG, Layden J, Mikosz CA, VanFrank B, King BA, Salvatore PP, Weissman DN, Thomas J, Melstrom PC, Baldwin GT, Parker EM, Courtney- Long EA, Krishnasamy VP, Pickens CM, Evans ME, Tsay SV, Powell KM, Kiernan EA, Marynak KL, Adjemian J, Holton K, Armour BS, England LJ, Briss PA, Houry D, Hacker KA, Reagan-Steiner S, Zaki S, Meaney-Delman D, Group CDCLIR. Severe Pulmonary Disease Associated with electronic-cigarette-product Use - Interim Guidance. MMWR Morb Mortal Wkly Rep. 2019;68:787–90. 20. Wells JM, Arenberg DA, Barjaktarevic I, Bhatt SP, Bowler RP, Christenson SA, Couper DJ, Dransfield MT, Han MK, Hoffman EA, Kaner RJ, Kim V, Kleerup E, Martinez FJ, Moore WC, O’Beirne SL, Paine R 3rd, Putcha N, Raman SM, Barr RG, Rennard SI, Woodruff PG, Curtis JL. Safety and Tolerability of Compre- hensive Research Bronchoscopy in Chronic Obstructive Pulmonary Disease. Results from the SPIROMICS bronchoscopy substudy. Ann Am Thorac Soc. 2019;16:439–46. 21. Collins AM, Rylance J, Wootton DG, Wright AD, Wright AK, Fullerton DG, Gordon SB. Bronchoalveolar lavage (BAL) for research; obtaining adequate sample yield. J Vis Exp 2014. 22. Misharin AV, Morales-Nebreda L, Reyfman PA, Cuda CM, Walter JM, McQuat- tie-Pimentel AC, Chen CI, Anekalla KR, Joshi N, Williams KJN, Abdala-Valencia H, Yacoub TJ, Chi M, Chiu S, Gonzalez-Gonzalez FJ, Gates K, Lam AP, Nicholson TT, Homan PJ, Soberanes S, Dominguez S, Morgan VK, Saber R, Shaffer A, Hinchcliff M, Marshall SA, Bharat A, Berdnikovs S, Bhorade SM, Bartom ET, Morimoto RI, Balch WE, Sznajder JI, Chandel NS, Mutlu GM, Jain M, Gottardi CJ, Singer BD, Ridge KM, Bagheri N, Shilatifard A, Budinger GRS, Perlman H. Warren et al. Respiratory Research (2023) 24:162 Page 10 of 10 Monocyte-derived alveolar macrophages drive lung fibrosis and persist in the lung over the life span. J Exp Med. 2017;214:2387–404. 23. Huang X, Xiu H, Zhang S, Zhang G. The role of Macrophages in the pathogen- nicotine disrupts airway barrier function and induces systemic inflammation and multiorgan fibrosis in mice. Am J Physiol Regul Integr Comp Physiol. 2018;314:R834–47. esis of ALI/ARDS. Mediators Inflamm. 2018;2018:1264913. 24. Morrell ED, Bhatraju PK, Mikacenic CR, Radella F, Manicone AM, Stapleton RD, Wurfel MM, Gharib SA. Alveolar macrophage Transcriptional Programs are Associated with Outcomes in Acute Respiratory Distress Syndrome. Am J Respir Crit Care Med. 2019;200:732–41. 25. Hikichi M, Mizumura K, Maruoka S, Gon Y. Pathogenesis of chronic obstruc- tive pulmonary disease (COPD) induced by cigarette smoke. J Thorac Dis. 2019;11:2129–S2140. 35. Been T, Traboulsi H, Paoli S, Alakhtar B, Mann KK, Eidelman DH, Baglole CJ. Differential impact of JUUL flavors on pulmonary immune modula- tion and oxidative stress responses in male and female mice. Arch Toxicol. 2022;96:1783–98. 36. Been T, Alakhtar B, Traboulsi H, Tsering T, Bartolomucci A, Heimbach N, Paoli S, Burnier J, Mann KK, Eidelman DH, Baglole CJ. Chronic low-level JUUL aerosol exposure causes pulmonary immunologic, transcriptomic, and proteomic changes. FASEB J. 2023;37:e22732. 26. Hoenderdos K, Condliffe A. The neutrophil in chronic obstructive pulmonary 37. Glynos C, Bibli SI, Katsaounou P, Pavlidou A, Magkou C, Karavana V, Topouzis S, disease. Am J Respir Cell Mol Biol. 2013;48:531–9. 27. Bhat TA, Kalathil SG, Bogner PN, Blount BC, Goniewicz ML, Thanavala YM. An animal model of inhaled vitamin E acetate and EVALI-like Lung Injury. N Engl J Med. 2020;382:1175–7. 28. Ranpara A, Stefaniak AB, Williams K, Fernandez E, LeBouf RF. Modeled respira- tory tract deposition of aerosolized oil diluents used in Delta(9)-THC-Based electronic cigarette Liquid Products. Front Public Health. 2021;9:744166. 29. Ghosh A, Ahmad S, Coakley RD, Sassano MF, Alexis NE, Tarran R. Lipid-laden macrophages are not unique to patients with E-Cigarette or Vaping Product Use-associated Lung Injury. Am J Respir Crit Care Med. 2021;203:1030–3. 30. Madison MC, Landers CT, Gu BH, Chang CY, Tung HY, You R, Hong MJ, Baghaei N, Song LZ, Porter P, Putluri N, Salas R, Gilbert BE, Levental I, Campen MJ, Corry DB, Kheradmand F. Electronic cigarettes disrupt lung lipid homeostasis and innate immunity independent of nicotine. J Clin Invest. 2019;129:4290–304. Kalomenidis I, Zakynthinos S, Papapetropoulos A. Comparison of the effects of e-cigarette vapor with cigarette smoke on lung function and inflammation in mice. Am J Physiol Lung Cell Mol Physiol. 2018;315:L662–72. 38. Masso-Silva JA, Moshensky A, Shin J, Olay J, Nilaad S, Advani I, Bojanowski CM, Crotty S, Li WT, Ongkeko WM, Singla S, Crotty Alexander LE. Chronic E-Ciga- rette Aerosol Inhalation alters the Immune State of the lungs and increases ACE2 expression, raising concern for altered response and susceptibility to SARS-CoV-2. Front Physiol. 2021;12:649604. 39. Staudt MR, Salit J, Kaner RJ, Hollmann C, Crystal RG. Altered lung biology of healthy never smokers following acute inhalation of E-cigarettes. Respir Res. 2018;19:78. 40. Hickman E, Payton A, Duffney P, Wells H, Ceppe AS, Brocke S, Bailey A, Rebuli ME, Robinette C, Ring B, Rager JE, Alexis NE, Jaspers I. Biomarkers of Airway Immune Homeostasis Differ significantly with generation of E-Cigarettes. Am J Respir Crit Care Med 2022. 31. Schaunaman N, Crue T, Cervantes D, Schweitzer K, Robbins H, Day BJ, Numata 41. Lahiri T. The utility of the lipid-laden macrophage index for the evaluation of M, Petrache I, Chu HW. Electronic cigarette vapor exposure exaggerates the pro-inflammatory response during influenza a viral infection in human distal airway epithelium. Arch Toxicol. 2022;96:2319–28. 32. Chaumont M, van de Borne P, Bernard A, Van Muylem A, Deprez G, Ullmo J, Starczewska E, Briki R, de Hemptinne Q, Zaher W, Debbas N. Fourth generation e-cigarette vaping induces transient lung inflammation and gas exchange disturbances: results from two randomized clinical trials. Am J Physiol Lung Cell Mol Physiol. 2019;316:L705–19. 33. Antoniewicz L, Brynedal A, Hedman L, Lundbäck M, Bosson JA. Acute Effects of electronic cigarette inhalation on the vasculature and the Conducting Airways. Cardiovasc Toxicol. 2019;19:441–50. 34. Crotty Alexander LE, Drummond CA, Hepokoski M, Mathew D, Moshensky A, Willeford A, Das S, Singh P, Yong Z, Lee JH, Vega K, Du A, Shin J, Javier C, Tian J, Brown JH, Breen EC. Chronic inhalation of e-cigarette vapor containing aspiration in children. Cancer Cytopathol. 2014;122:161–2. 42. Gibeon D, Zhu J, Sogbesan A, Banya W, Rossios C, Saito J, Rocha JP, Hull JH, Menzies-Gow AN, Bhavsar PK, Chung KF. Lipid-laden bronchoalveolar macro- phages in asthma and chronic cough. Respir Med. 2014;108:71–7. 43. Knauer-Fischer S, Ratjen F. Lipid-laden macrophages in bronchoalveo- lar lavage fluid as a marker for pulmonary aspiration. Pediatr Pulmonol. 1999;27:419–22. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Warren et al. Respiratory Research (2023) 24:162
10.1186_s12917-019-1925-6	Lelešius et al. BMC Veterinary Research (2019) 15:178 https://doi.org/10.1186/s12917-019-1925-6 R E S E A R C H A R T I C L E Open Access In vitro antiviral activity of fifteen plant extracts against avian infectious bronchitis virus Raimundas Lelešius1,2* Loreta Kubilienė5, Audrius Maruška3 and Algirdas Šalomskas2 , Agneta Karpovaitė2, Rūta Mickienė3, Tomas Drevinskas3, Nicola Tiso3, Ona Ragažinskienė4, Abstract Background: Avian infectious bronchitis (IB) is a disease that can result in huge economic losses in the poultry industry. The high level of mutations of the IB virus (IBV) leads to the emergence of new serotypes and genotypes, and limits the efficacy of routine prevention. Medicinal plants, or substances derived from them, are being tested as options in the prevention of infectious diseases such as IB in many countries. The objective of this study was to investigate extracts of 15 selected medicinal plants for anti-IBV activity. Results: Extracts of S. montana, O. vulgare, M. piperita, M. officinalis, T. vulgaris, H. officinalis, S. officinalis and D. canadense showed anti-IBV activity prior to and during infection, while S. montana showed activity prior to and after infection. M. piperita, O. vulgare and T. vulgaris extracts had > 60 SI. In further studies no virus plaques (plaque reduction rate 100%) or cytopathogenic effect (decrease of TCID50 from 2.0 to 5.0 log10) were detected after IBV treatment with extracts of M. piperita, D. canadense and T. vulgaris at concentrations of extracts ≥0.25 cytotoxic concentration (CC50) (P < 0.05). Both PFU number and TCID50 increased after the use of M. piperita, D. canadense, T. vulgaris and M. officinalis extracts, the concentrations of which were 0.125 CC50 and 0.25 CC50 (P < 0.05). Real-time PCR detected IBV RNA after treatment with all plant extracts using concentrations of 1:2 CC50, 1:4 CC50 and 1:8 CC50. Delta cycle threshold (Ct) values decreased significantly comparing Ct values of 1:2 CC50 and 1:8 CC50 dilutions (P < 0.05). Conclusions: Many extracts of plants acted against IBV prior to and during infection, but the most effective were those of M. piperita, T. vulgaris and D. canadense . Keywords: Avian infectious bronchitis, Plant extracts, Antiviral activity Background IB is a highly contagious respiratory and occasionally uro- genital disease in chickens [1]. IBV affects the upper re- spiratory tract and reduces egg production [2]. It is a coronavirus that belongs to the Coronaviridae family. IBV is an enveloped virus with a single-stranded positive-sense linear RNA molecule (approximately 27.6 kb in size) [3]. IB has a wide geographical distribution and is diagnosed worldwide [1]. IB outbreaks continuously and results in economic losses in the poultry industry. So far vaccination * Correspondence: raimundas.lelesius@lsmuni.lt 1Institute of Microbiology and Virology, Veterinary Faculty, Lithuanian University of Health Sciences, Kaunas, Lithuania 2Department of Veterinary Pathobiology, Veterinary Faculty, Lithuanian University of Health Sciences, Kaunas, Lithuania Full list of author information is available at the end of the article using inactivated or live vaccines [4] is regarded as the main method of prevention, but it is not having the de- sired effect [5–7]. The high level of mutations of IBV [8] leads to the emergence of new serotypes and genotypes, and limits the efficacy of routine prevention. Biological products derived from plants are used in medicine for different pharmacological reasons, including the treatment of infectious and non-infectious diseases [9, 10]. This class of antimicrobial plants is acknowledged and well investigated, and classes of active compounds have already been identified [11, 12]. The investigation of antiviral substances derived from plants is insufficient in comparison with the investigation of antimicrobial proper- ties. Fortunately, several experiments have shown that plants have positive antiviral activity in vitro and in vivo © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Lelešius et al. BMC Veterinary Research (2019) 15:178 Page 2 of 10 [13]. However, the same plants can have different antiviral activity against RNA or DNA viruses, either enveloped or non-enveloped, and even against different types or strains of a virus [14, 15]. A number of scientific publications have encouraged the use of polyphenolic compounds in the treatment and prophylaxis of chronic diseases [16]. The mixture of the geometric isomers and enantiomers of rosmarinic acid is accumulating in the families Lamiaceae Lindl, Astera- ceae Bercht. & J.Presl [17, 18]. Most large quantities of rosmarinic acid have been determined in the genus of plants such as Salvia L., Perilla L. Melissa L. and Echin- acea Moench. Rosmarinic acid has antioxidative, anti- inflammatory, antimutagenic, antibacterial and antiviral effects against the herpes simplex virus [19]. The Desmodium canadense herb contains flavonoids such as apigenin, apigenin-7-O-glucoside, luteolin, rutin, 2- vicenin, vitexin, isovitexin, vitexin rhamnoside, orientin, homoorientin, quercetin, hyperoside, astragalin and kaem- pherol [20]. In addition, it also contains saponins and phen- olic acids (chlorogenic acid, vanillic, 4- hydroxycinnamic, ferulic and caffeic). The Desmodium herb exhibits antioxi- dant, antibacterial, anti-inflammatory, hepatoprotective, di- uretic and analgesic activity [20]. C-glycosides of flavonoids are known to exhibit antioxidant, hepatoprotective, anti- inflammatory and antiviral effects [21]. The plants in this study were chosen for their medical, antibacterial and anti- viral properties. Ethanol extracts of medicinal plants be- longing to the families Lamiaceae (winter savory, perilla, blue giant hyssop, oregano, peppermint, lemon balm, thyme, hyssop, catnip and sage), Asteraceae (chamomile and purple coneflower), Geraniaceae (rock crane’s-bill), Apiaceae (garden angelica), and Fabaceae (showy tick tre- foil) were prepared. The majority of plants used for the preparation of extracts in this study belong to one of the famous medicinal aromatic plant families Lamiaceae. The medicinal plants from this family have long been used in traditional medicine worldwide. Many investigations of plant extracts have been per- formed with different coronaviruses. The main targets were proteins involved in coronaviral replication, prote- ases and ion channel conductance [22]. Only a few in- vestigations have been performed to test the anti-IBV activity of plant extracts. Several studies have found that the plant preparations inhibited IBV replication in vivo and vitro. Sambucus nigra, Houttuynia cordata, Alium sativum and Astragalus mongholicus inhibited IBV repli- cation [23–26]. The ethanol extract of Sambucus nigra inhibited IBV replication and reduced virus titres prior to infection [24], as did Houttuynia cordata essential oil mixed with an aqueous solution of sodium chloride so- lution [25]. It is suggested that the effect of extracts of, Alium sativum, Houttuynia cordata and Sambucus nigra can be associated with direct inactivation of envelope structures of a virus, which are necessary for adsorption to or entry into host cells, or might dissolute the IBV en- velope. Compounds that have a virucidal effect work like a disinfectant and do not require replication to inactivate the virus [15]. The mechanism of action of Astragalus polysaccharides has not been explained. Medicinal plants or substances derived from them are being tested as a tool for preventing infectious diseases such as IB in many countries, but the anti-IBV viral properties of the selected plants have not so far been tested. The objective of this study was to investigate ex- tracts of 15 selected medicinal plants for anti-IBV activity. Results Cytotoxicity of plant extracts All the extracts were more cytotoxic (P < 0.05) than the ethanol control (7.7 μl). A. foeniculum showed the high- est cytotoxic concentration (0.062 μg) and P. frutescens (0.77 μg) showed the lowest one. Antiviral effect against IBV According to the results of the antiviral effect assay, eight extracts were selected for determination of the virucidal effect. The selected extracts of S. montana, O. vulgare, M. piperita, M. officinalis, T. vulgaris, H. officinalis, S. offici- nalis and D. canadense showed anti-IBV activity in two of the four methods. All eight extracts showed an antiviral effect prior to infection (method 1). Furthermore, seven of these showed antiviral activity during infection (method 2), while only the extract of S. montana showed anti-IBV activity after infection (method 4). P. frutescens, N. cat- aria, E. purpurea, Ch. nobile and A. foeniculum showed an antiviral effect only in the first method, while G. macro- rrhizum and A. archangelica did not show an antiviral ef- fect in any method (Table 1). The above-mentioned eight plant extracts demonstrating anti-IBV activity were selected for further investigation. The 50% effective concentration (EC50) was determined in cells grown for 4 days (prior to infection). The EC50 values of ex- tracts of S. montana, O. vulgare, M. piperita, M. officinalis, T. vulgaris, H. officinalis, S. officinalis and D. canadense were between 0.003 and 0.076 μg, however S. officinalis ap- peared effective at the lowest concentration (0.003 μg) (Table 2). SI of M. piperita, O. vulgare, and T. vulgaris ex- tracts were 67.5, 65.0 and 63.1 respectively. The anti-IBV activity prior to infection in Vero cells The inhibitory effect of extracts against IBV prior to infec- tion in Vero cells was measured by virus plaque and TCID50 assays. The selected plant extracts had a virucidal effect and inhibited viral replication compared to the virus control (P < 0.05). No virus plaques or CPE were detected Lelešius et al. BMC Veterinary Research (2019) 15:178 Page 3 of 10 Table 1 Antiviral effect of plant extracts No. Latin name (family) Part Antiviral effect Virus pre-treatment with extract prior to infection during infection after infection 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. Satureja montana Chamaemelum nobile Perilla frutescens Agastache foeniculum Origanum vulgare Mentha piperita Geranium macrorrhizum Melissa officinalis Angelica archangelica Thymus vulgaris Hyssopus officinalis Nepeta cataria Echinacea purpurea Salvia officinalis Desmodium canadense herb herb herb herb herb herb herb herb leaves roots herb herb herb herb herb herb + + + + + + – + – – + + + + + + – – – – + + – + – – + + – – + + + – – – – – – – – – – – – – – – Cell pre- treatment prior to infection – – – – – – – – – – – – – – – – after IBV treatment with extracts, whose concentrations were equivalent to 1 and 0.5 CC50 (Table 3). No virus plaques or CPE were detected after IBV treat- ment with M. piperita, D. canadense or T. vulgaris ex- tracts at concentrations equivalent to between 1 and 0.25 CC50. The other five plant extracts exhibited a lower effect on inhibition of virus replication at the same concentra- tions. However, all extracts reduced the PFU number (PFU reduction rate was from 54.5 to 100.0%) and TCID50 (from 2.0 to 5.0 log10) significantly (P < 0.05). Table 2 Anti-IBV activity of some plant extracts in Vero cell cultures No. Plant extract 1. 5. 6. 8. 11. 12. 15. 16. Satureja montana Origanum vulgare Mentha piperita Melissa officinalis Thymus vulgaris Hyssopus officinalis Salvia officinalis Desmodium canadense CC50 (μg)a 0.75 EC50 (μg)b 0.044 0.52 0.27 0.59 0.63 0.64 0.11 0.29 0.008 0.004 0.015 0.010 0.076 0.003 0.017 SI 17.0 65.0 67.5 39.3 63.1 8.4 36.7 17.1 aThe assay for determination of CC50 was performed in octuplicate for each extract bThe experiments for determination of EC50 were repeated independently twice, and a mean is presented Virus plaques and CPE were detected after IBV treat- ment with all extract concentrations equivalent to 0.125 CC50. All extracts decreased the PFU number (PFU re- duction rate from 50.2 to 76.6%) and TCID50 (from 1.55 to 3.17 log10) significantly (P < 0.05). Both PFU number and TCID50 increased significantly after the use of M. piperita, D. canadense, T. vulgaris and M. officinalis, with extract concentrations of 0.125 CC50 compared to 0.25 CC50 (P < 0.05). Real-time RT-PCR assay Real-time RT-PCR detected IBV RNA after treatment with all plant extracts using concentrations equivalent to 1:2 CC50, 1:4 CC50 and 1:8 CC50. Delta Ct values decreased significantly comparing the Ct values of 1:2 CC50 and 1:8 CC50 dilutions (P < 0.05, Table 4). The quantity of IBV RNA decreased significantly only after every dilution of M. officinalis and H. officinalis extracts (Table 4). Multidimensional data analysis The performed tests provided highly multidimensional data. Hierarchical clusterisation was performed using seven attributes obtained from the results of all the ex- tracts. The seven attributes were: (1) CC50, (2) EC50, (3) SI, (4) CC50 dilution protecting 100% cells, (5) inhibition of cytopathic effect (CIA100) method 1, (6) CC50 dilution protecting 100% cells method 2 and (7) CC50 dilution protecting 100% cells method 3. Nevertheless, complete Lelešius et al. BMC Veterinary Research (2019) 15:178 Page 4 of 10 Table 3 Virucidal effect and virus yield reduction No. Plant extract Concentrations of plant extracts equivalent to CC50 1. S. montana 1 5. O. vulgare 6. M. piperita 0.5 0.25 0.125 1 0.5 0.25 0.125 1 0.5 0.25 0.125 8. M. officinalis 1 11. T. vulgaris 0.5 0.25 0.125 1 0.5 0.25 0.125 12. H. officinalis 1 0.5 0.25 0.125 15. S. officinalis 1 16. D. canadense 0.5 0.25 0.125 1 0.5 0.25 0.125 IBV (control) – PFU number (mean ± SD), log10 PFU reduction rate, % TCID50 (mean ± SD), log10 0 0 0.97 ± 0.19 1.03 ± 0.11 0 0 1.08 ± 0.00 1.40 ± 0.32 0 0 0 1.48 ± 0.05 x 0 0 0.15 ± 0.21 1.72 ± 0.01 x 0 0 0 1.70 ± 0.01 x 0 0 1.33 ± 0.01 2.00 ± 0.03 x 0 0 2.00 ± 0.00 2.19 ± 0.03 x 0 0 0 1.70 ± 0.01 x 4.40 ± 0.09 100 100 78.0 76.6 100 100 75.4 68.2 100 100 100 66.4 100 100 96.6 60.9 100 100 100 61.3 100 100 69.8 54.5 100 100 54.5 50.2 100 100 100 61.3 – 0 0 0.75 ± 0.64 a 1.83 ± 0.31 c x 0 0 1.00 ± 0.94 a 2.48 ± 0.23 d x 0 0 0a 2.63 ± 0.13 d x 0 0 0.56 ± 0.44 a 3.38 ± 0.21 e x 0 0 0 3.25 ± 0.22 ex 0 0 3.00 ± 0.29 b 3.35 ± 0.29 e 0 0 3.25 ± 0.25 b 3.45 ± 0.21 e 0 0 0a 2.75 ± 0.16 d 5.00 ± 0.19 xmeans that the PHU number and TCID50 difference within the group (plant extract concentrations 0.25 and 0.125 CC50) was statistically significant (P < 0.05); if groups do not share a common letter a and b it means that the TCID50 (plant extract concentration 0.25 CC50) difference between the groups was statistically significant (P < 0.05); if groups do not share a common letter c, d and e it means that the TCID50 (plant extract concentration 0.125 CC50) difference between the groups was statistically significant (P < 0.05) inhibition of the cytopathic effect (CIA100) method 4 was not used for calculations due to the fact that this ex- periment indicated inactivity for all tested extracts. As can be seen in Fig. 1a, three sub-clusters can be distin- guished: (1) anti-IBV inactive (A. foeniculum, C. nobile, G. macrorrhizum, A. archangelica roots and aerial part), (2) anti-IBV active (D. canadense, M. piperita, M. offici- nalis, O. vulgare, S. officinalis, T. vulgaris) and (3) hybrid sub-cluster (E. purpurea, S. montana, N. cataria, P. fru- tescens, H. officinalis). H. officinalis and S. montana were clustered belonging to the hybrid sub-cluster containing anti-IBV inactive and active plant extracts. This obser- vation can be explained by the fact that H. officinalis and S. montana exhibited the lowest cytotoxicity and the lowest anti-IBV effect among the anti-IBV active plant extracts. Lelešius et al. BMC Veterinary Research (2019) 15:178 Page 5 of 10 Table 4 Delta Ct values - comparison with the virus control by real-time RT-PCR assay No. Plant extract Delta Ct values x 6. 8. 5. 1. 1:4 CC50 Mentha piperita Melissa officinalis Satureja montana Origanum vulgare 1:2 CC50 1:8 CC50 14.06 ± 2.501 3.82 ± 0.952 3.32 ± 0.652 13.28 ± 2.731 7.40 ± 1.012 6.36 ± 1.012 18.32 ± 2.771 8.24 ± 1.062 7.70 ± 1.822 2.71 ± 0.943 12.49 ± 2.331 8.41 ± 2.222 12.65 ± 1.001 11.70 ± 1.171 2.66 ± 0.772 2.03 ± 1.103 10.02 ± 2.501 5.33 ± 2.102 1.95 ± 1.192 6.65 ± 1.191 8.25 ± 1.861 15. 16. Desmodium canadense 16.41 ± 2.441 10.58 ± 2.822 6.21 ± 0.642 xCt of IBV was 18.50 ± 1.66; 1, 2 and 3 means that the difference within the group was statistically significant (P < 0.05) 12. Hyssopus officinalis Thymus vulgaris Salvia officinalis 11. The comparison of different anti-IBV active plant ex- tracts was performed using a multidimensional scaling technique. The obtained results of anti-IBV extracts indi- cated highly multidimensional data (12 attributes), which cannot be visualised without additional mathematical means. The multidimensional scaling technique allows the projection of all dimensions to two dimensions, which are easily represented in the plot. In this study, 12 dimensions – (1) CC50, (2) EC50, (3) SI, (4) PFU number at dilution of 0.25, (5) PFU number at dilution of 0.125, (6) PFU reduc- tion rate at dilution of 0.25, (7) PFU reduction rate at dilu- tion of 0.125, (8) TCID50 at dilution of 0.25, (9) TCID50 at dilution of 0.125, (10) Delta Ct values at 1:2 CC50, (11) Delta Ct values at 1:4 CC50 and (12) Delta Ct values at 1:8 CC50 – were projected to two dimensionless projections (Coordinate 1 and Coordinate 2). Other attributes were not used for multidimensional scaling since they indicated similar numeric values. In the plot (Fig. 1b) four extreme points were identified: (i) S. montana (lowest cytotoxicity), (ii) H. officinalis (lowest anti-IBV activity) (iii) S. officinalis (highest cytotoxicity and anti-IBV activity) and (iv) M. piperita (highest SI). In the plot (Fig. 1b) there is a cluster of moderately active plant extracts: O. vulgare, M. officina- lis, T. vulgaris and D. canadense. It should be noted that M. piperita extract is more similar to the moderately ac- tive cluster than other extreme points (S. montana, H. offi- cinalis, S. officinalis). Interestingly, D. canadense is more similar to a moderately active cluster with dominating plants of family Lamiaceae, whereas D. canadense belongs to the family Fabaceae. Discussion A wide range of traditional medicinal plants and herbs have been reported to show antiviral activities against vari- ous viruses. In this study, the anti-IBV activities were ana- lysed of ethanol extracts from 16 medicinal plants that belong to 15 different species. In screening, the IBV Beaudette strain was used because it is adapted to the Vero cell culture system. A Vero cell line was used as it is one of the most common and well-established mamma- lian cell lines involved in assessing the effects of chemicals, toxins and other substances at the molecular level [27]. The cell line is also known to be susceptible to many vi- ruses, including IBV, and displays CPE upon infection. First, the cytotoxicity of plant extracts was evaluated before the possible mechanisms of virus inhibition were determined. Eight out of 16 extracts – D. canadense, M. piperita, M. officinalis, O. vulgaris, T. vulgaris, H. offici- nalis, S. officinalis, and S. montana – were chosen based on the results of the antiviral effect assay. The first seven extracts mentioned above possessed significant anti-IBV activities prior to/during infection, and only S. montana had an antiviral effect prior to and post-infection. All ex- tracts, except D. canadense, belong to the Lamiaceae family, members of which have been reported in many studies as showing antiviral activity [28–30]. So far, none A B Fig. 1 Visual representation of the experimental data. a Cluster dendrogram of all plant extracts. b Multidimensional Scaling plot of anti-IBV active plant extracts. Plant extracts: 1 – D. canadense, 2 – M. piperita, 3 – T. vulgaris, 4 – M. officinalis, 5 – O. vulgare, 6 – S. officinalis, 7 – H. officinalis, 8 – S. montana Lelešius et al. BMC Veterinary Research (2019) 15:178 Page 6 of 10 of these plants has been studied for antiviral activity against IBV. However, several preparations of botanicals inhibit viral replication in vitro and in vivo. Moreover, the results show that ethanol extract of Sambucus nigra inhibits viral replication and reduces virus titres prior to infection [24], as does Houttuynia cordata essential oil mixed with an aqueous solution of sodium chloride so- lution [25]. These results suggest that the effect of ex- tracts can be associated with the direct inactivation of the envelope structures of a virus, which are necessary for adsorption to or entry into host cells, or might dis- solute the IBV envelope. Thus, resistance to virucidal compounds due to mutations generated in the viral genome during replication is not likely [31]. Subsequently, the anti-IBV activities were examined of those eight plant extracts that showed the strongest anti- viral activity prior to infection, using plaque reduction assay to determine the virucidal effect. Furthermore, M. piperita, D. canadense and T. vulgaris extracts exhibited the strongest viral replication inhibition, and completely stopped IBV production at 1 to 0.25 log10 CC50 concen- tration. Comparable results were also obtained by TCID50 assay. The extract of M. officinalis showed a slightly lower virucidal effect and at 0.25 log10 the CC50 concentration PFU number (log10) was 0.15 ± 0.21. It has been reported that the aqueous extract of M. piper- ita shows anti-HIV-1 activity in MT-4 cells and inhibi- tory activity against HIV-reverse transcriptase [32]. Moreover, hydroalcoholic and aqueous extracts of M. piperita and T. vulgaris have been shown to exert sig- nificant anti-HSV-1 and anti-HSV-2 activities in vitro. M. piperita essential oil has been found to be signifi- cantly effective against both HSV -1 [28] and HSV-2 when the viruses are treated with peppermint oil prior to adsorption, but not after penetration into the host cell [31]. An increase in virion density prior to its attach- ment to the host cells is the most likely mechanism of action of the antiviral activity of the aqueous extract of M. piperita [33]. The results of real-time PCR showed that this method can be successfully used to evaluate and com- pare the antiviral effect of plant extracts. However, it is important to note that simple or RT-PCR detects viral RNA of both an inactivated virus and an intact one. Therefore, it is not possible to assess the per- centages of the nucleic acid of live and killed viruses in the sample. It is likely that the influence of the higher amounts of inactivated IBV on Delta Ct values could be significant, in some cases. It is more likely when the higher effective concentrations of plant ex- tracts were used. RT-PCR showed that the IBV RNA quantity was statistically higher in all cases in a com- parison after IBV treatment with extracts concentra- tions of 1:2 and 1:8 CC50. The RT-PCR confirmed that extracts of M. piperita and D. canadense had the highest effect on the synthesis of viral RNA. These results were unable to reveal the most anti-IBV effective plant extract. Therefore, multidimensional data analysis was carried out using hierarchical clusterisation. It identified an anti-IBV active cluster with M. piperita, T. vulgaris, D. canadense, M. officinalis, S. officinalis and O. vulgare. Multidimensional scaling technique identi- fied four extreme points: (i) S. montana, which provided the lowest cytotoxicity, (ii) H. officinalis, which provided the lowest anti-IBV activity, (iii) S. officinalis, which pro- vided the highest cytotoxicity and anti-IBV activity, and (iv) M. piperita, which provided the highest SI. In this study, the majority of plant extracts were found to have some antiviral activity and could inhibit IBV prior to and during infection. The ethanolic extracts had a mixture of compounds/fractions and it is possible that the antiviral activity of plants is decided by one or sev- eral compounds or combinations thereof [34]. Future studies are needed to detect the compounds or fractions responsible for anti-IBV activity and to investigate the mechanism of their action. Conclusions Many extracts of plants acted virucidally against IBV prior to and during infection, but those of M. piperita, T. vulgaris and D. canadense were the most effective. Materials and methods Plants and extracts All the plants were grown in the botanical garden of Vytautas Magnus University (Lithuania). Sixteen ethanol extracts were prepared from medicinal plants. The con- tent of biologically active compounds is dependent on the edaphoclimatic conditions of the plant’s cultivation, the vegetation phase, the phenotype and the method of preparation of the raw material. Previous studies com- paring the dried raw and fresh material of medicinal plants and different sample preparation methods have shown that drying and other conditions affect the quali- tative and quantitative composition of the raw material [35, 36]. In many cases, drying is used as a standardised preparation of raw material of medicinal plants since it reduces water content and the risk of microbiological spoilage of raw material. In the present study, drying was used to collect all the samples of different plants during the intensive blooming vegetation phase for the simultaneous determination of biological activities, in consideration of the fact that different medicinal plants differ in dynamics of the accumulation of biologically ac- tive compounds, vegetation, and therefore in the har- vesting of raw material [37]. The potentially antiviral plants were selected for extraction depending on the ac- cumulated compounds in herbs, leaves and roots. Lelešius et al. BMC Veterinary Research (2019) 15:178 Page 7 of 10 Preparation of plant extracts The solvent ethanol was diluted with sterile bidistilled water to 40% (vol.) concentration. Dried plant material from each plant (500 μg) was extracted with 10 ml solv- ent. The extraction was performed in an orbital shaker for 24 h at room temperature (20 °C). Each extract was filtrated using a paper filter and then polyvinyl difluoride membrane filter with 0.22-μm pore size. The concentra- tion of the extracts was 50 mg/ml, with reference to the starting material. All the prepared plant extracts were stored in a refrigerator at 4 °C. Cell line Vero cells (ATCC CCL-81) were provided by Dr. I. Jace- vičienė from the Department of Virus Research at the National Food and Veterinary Risk Assessment Institute in Lithuania. The cells were cultivated in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% foetal bovine serum (FBS) at 37 °C in 5% CO2 incu- bator. Nystatin (100 units/ml) and gentamycin (50 μg/ ml) were used to prevent microbial contamination. Virus The Vero-adapted Beaudette IBV strain was used. The virus was provided by Dr. M. H. Verheije of Utrecht University in The Netherlands. The virus stocks were prepared and stored at − 80 °C in aliquots. Cytotoxicity assay The cytotoxic concentration (CC50) was determined for each extract on Vero cells using MTT assay [38]. First, cells were seeded at a concentration of 1 × 104 cells/well in a 96-well plate and grown at 37 °C for 24 h. Each extract was tested in octuplicate once. After 72 h MTT reagent (10 μl, 5 mg/ml, Sigma-Aldrich) was added and incubated for 4 h at 37 °C. Then 100 μl dimethyl sulphoxide (DMSO) (Carl Roth, Germany) was added to each well and the plates were placed on the shaker for 5 min. The absorb- ance of each well was measured at 620 nm in a microplate reader (Multiskan™ FC Microplate Photometer) and the percentage of cell survival was calculated. Finally, dose- response curves were plotted to enable the calculation of CC50 that causes lysis and death in 50% of cells. Screening extracts for antiviral activity For determination of antiviral properties, one-day-old Vero cells (seeded 1 × 104 cells/well) in a 96-well plate were used. The virus was used at a multiplicity of infec- tion (MOI) of 0.05. Each extract was serially diluted two- fold to 1:128 in DMEM and assessed for the ability to inhibit IBV replication using four mechanisms. Every sample of extract was tested twice in quadruplicate. In addition, controls of cells, the virus and extracts were included. An inverted microscope (Leica, Germany) was used to observe cells after all procedures. In the first method, the virus was treated with the di- luted extract for 1 h in a separate 96-well plate and then poured onto the cells. The mixtures were discarded after incubation for 1 h at 37 C in 5% CO2 and then the cells were washed twice with PBS. After washing, DMEM containing 2% of FBS was added. Observation by mi- croscopy for inhibition of cytopathic effect (CPE) was performed after incubation for 72 h at 37 °C in 5% CO2. In the second method, the mixtures of virus and the di- luted extract were poured onto the cells immediately. The mixtures were discarded after incubation for 1 h at 37 °C in 5% CO2 and then the cells were washed twice with PBS. After washing, DMEM containing 2% of FBS was added. Observation by microscopy for inhibition of CPE was per- formed after incubation for 72 h at 37 °C in 5% CO2. In the third method, the cells were inoculated with the virus and then treated with extract. First the cells were inoculated with the virus and incubated for 1 h at 37 °C in 5% CO2. Then the unadsorbed virus was discarded and the cells were washed twice with PBS. After wash- ing, the cells were treated with the diluted extracts for 1 h at 37 °C in 5% CO2. After washing the cells twice with PBS, DMEM containing 2% of FBS was added. Observa- tion by microscopy for inhibition of CPE was performed after incubation for 72 h at 37 °C in 5% CO2. In the fourth method, the cells were treated with ex- tract prior to inoculation. First the cells were treated with the diluted extracts for 1 h at 37 °C in 5% CO2. Then the cells were washed twice with PBS and inocu- lated with the virus. After incubation for 1 h at 37 °C in 5% CO2 the cells were washed twice with PBS, and DMEM containing 2% of FBS was added. Observation by microscopy for inhibition of CPE was performed after incubation for 72 h at 37 °C in 5% CO2. The most promising plant extracts were selected for determination of EC50 and selectivity index (SI) based on the results of the antiviral effect assay. Determination of EC50 and SI Eight out of 16 plant extracts were chosen for the deter- mination of EC50 and SI using the first method. Extracts were titrated from 1 to 1:128 CC50 and used for virus treatment. After 72 h the MTT assay was performed as outlined above. The 50% effective concentrations (EC50) were calculated from the plot of percentages of cell via- bility against extract concentrations. Plaque reduction assay Concentrations of extracts from 1 CC50 to 0.125 CC50 and 25,000 plaque-forming units (PFU) of IBV were mixed and incubated at room temperature for 1 h. Lelešius et al. BMC Veterinary Research (2019) 15:178 Page 8 of 10 Plant extracts were diluted with DMEM to prepare four concentrations equivalent to 1 CC50, 0.5 CC50, 0.25 CC50, and 0.125 CC50, as calculated by cytotoxicity assay. These dilutions were then mixed with 25,000 plaque-forming units (PFU) of IBV and incubated at room temperature for 1 h. After incubation, a confluent monolayer of Vero cells in 6-well plates was inoculated with 1 ml of virus (MOI 0.05) and plant extract mixtures for 1 h at 37 °C in 5% CO2 and then discarded. The agar- ose 0.4% in maintenance medium was added to cells, and the plates were stored at room temperature for 15 min and incubated at 37 °C and 5% CO2. After 72 h the plates were microscopically examined for detection of CPE and then 0.2 ml MTT (5 mg/ml) was used for stain- ing. Plaques were counted after incubation at 37 °C in 5% CO2 for 4 h. The number of plaques was expressed as log10 and the reduction rate was calculated as follows: PFU number of virus control‐PFU number after the treatment with plant extract PFU number of virus control (cid:1) 100% Virus yield reduction A virus yield reduction was evaluated by means of virus titration and real-time reverse transcriptase polymerase chain reaction (RT-PCR). The cells were inoculated as outlined above in the plaque reduction assay section. Plant extracts were diluted with DMEM to prepare four concentrations equivalent to 1 CC50, 0.5 CC50, 0.25 CC50, and 0.125 CC50 as calculated by cytotoxicity assay. These dilutions were then mixed with 25,000 plaque- forming units (PFU) of IBV and incubated at room temperature for 1 h. After incubation a confluent mono- layer of Vero cells in 6-well plates was inoculated with 1 ml of virus (MOI 0.05) and plant extract mixtures for 1 h at 37 °C in 5% CO2, and then discarded. After inocu- lation DMEM containing 2% of FBS was added and the cells were incubated for 24 h. CPE of the virus was eval- uated using light microscopy. After that, the plates were frozen and thawed three times and the aliquots of the virus were prepared by centrifugation for 15 min at 2000 RPM. The prepared mixtures were used for quantifica- tion of both treated and untreated virus and viral nucleic acids by means of TCID50 assay and quantitative real- time RT-PCR respectively. TCID50 assay Determination of TCID50 of the control virus and the treated one was performed in 96-well plates. CPE was evaluated after 7 days. Virus titres were calculated using the Kärber method (Kärber, 1931). Real-time RT-PCR assay Ribonucleic acid (RNA) used in the real-time RT-PCR was extracted by means of TRIzol Reagent (Thermo Fisher Scientific, USA) according to the manufacturer’s instructions. Real-time RT-PCR was performed as de- scribed by Meir [39]. Briefly, a conserved region of 336 base pairs located at nucleotide position 741–1077 of the H120 strain N gene sequence (GenBank accession no. AM260960) was used to design primers and probe for the real-time RT-PCR assay. A downstream primer IBV-f (5-ATGCTCAACCTTGTCCCTAGCA-3) located at nucleotide position 811–832, an upstream primer IBV-r (5-TCAA-ACTGCGGATCATCACGT-3) located at nucleotide position 921–941, and a TaqMan® probe (FAM-TTGGAAGTAGAGTGACGCC- IBV-TM CAAACTTCA-BHQ1) located at nucleotide position 848–875 were used to amplify a 130-bp fragment. Both the primers and the probe were synthesised by Applied Biosystems, UK. The 25 μl real-time RT-PCR reaction contained 12.5 μl 2 × RT-PCR buffer mix (AgPath™ One- Step RT-PCR kit, Applied Biosystems), 1 μl 25 × RT-PCR enzyme mix (Applied Biosystems), primers to a final concentration of 400 nM, probe to a final concentration of 120 nM, 2 μl RNA template, and nuclease-free water. The reaction was carried out in StepOne™ Plus real-time PCR system (Applied Biosystems) at 45 °C for 10 min, 95 °C for 10 min, and 40 cycles of 95 °C for 15 s and 60 ° C for 45 s. Amplification plots were recorded and ana- lysed, and the threshold cycle (Ct) was determined with the Mastercycler RealPlex2 (Eppendorf ). The real-time RT-PCR was repeated four times, and then delta Ct values were calculated by subtracting the Ct values of virus control from Ct values of virus sam- ples treated with plant extracts. Means and standard de- viations of delta Ct values were then calculated to evaluate the effect of plant extracts on viral replication. Statistical and data analysis The differences between the methods and extracts were evaluated by Fisher’s criteria and the Student’s t-test. The data were regarded as significant when P < 0.05. Hierarchical clusterisation and multidimensional scaling (MDS) were performed using the software R-Studio. For hierarchical clusterisation, the Euclidean method was used. For MDS, the Euclidean method was used project- ing all dimensions to two dimensions. Abbreviations CC50: 50% cytotoxic concentration; CIA100: Complete inhibition of cytopathic effect; CPE: Cytopathic effect; Ct: Threshold cycle; DMEM: Dulbecco’s modified Eagle’s medium; DMSO: Dimethyl sulphoxide; EC50: 50% effective concentrations; FBS: Foetal bovine serum; IB: Avian infectious bronchitis; IBV: Avian infectious bronchitis virus; MDS: Multidimensional scaling; MOI: Multiplicity of infection; PBS: Phosphate-buffered saline; PCR: Polymerase chain reaction; PFU: Plaque-forming unit; RNA: Ribonucleic acid; RPM: Revolutions per minute; RT-PCR: Reverse transcription polymerase chain reaction; SI: Selectivity index; TCID50: 50% tissue culture infectious dose Acknowledgments ‘Not Applicable’. Lelešius et al. BMC Veterinary Research (2019) 15:178 Page 9 of 10 Authors’ contributions AM and AS proposed the experiment. OR, RM and LK chose medicinal plants. OR prepared the plants for extraction, TD and NT prepared the extracts of plants, RL and AK tested the extracts of plants. RL, AK, AS, and TD analysed the information. RL, AK, AS, AM, and TD reviewed the manuscript. All the authors read and approved the final manuscript. Funding The research was granted by Research Council of Lithuania, project No. MIP- 065/2015. Neither the design of the study, nor the collection, analysis, and in- terpretation of data and writing the manuscript was influenced by Research Council of Lithuania. Availability of data and materials The datasets supporting the results of this document are contained within the article. Any additional data may be requested to the corresponding author. Ethics approval and consent to participate ‘Not Applicable’. Competing interests The authors declare that they have no competing interests. Author details 1Institute of Microbiology and Virology, Veterinary Faculty, Lithuanian University of Health Sciences, Kaunas, Lithuania. 2Department of Veterinary Pathobiology, Veterinary Faculty, Lithuanian University of Health Sciences, Kaunas, Lithuania. 3Instrumental Analysis Open Access Centre, Vytautas Magnus University, Kaunas, Lithuania. 4Sector of Medicinal Plants, Kaunas Botanical Garden of Vytautas Magnus University, Kaunas, Lithuania. 5Faculty of Pharmacy, Lithuanian University of Health Sciences, Kaunas, Lithuania. Received: 15 March 2018 Accepted: 20 May 2019 References 1. 2. 3. 4. Cook JKA, Jackwood M, Jones RC. The long view: 40 years of infectious bronchitis research. Avian Pathol. 2012;41:239–50. Cavanagh D. Coronavirus avian infectious bronchitis virus. Vet Res. 2007; 38:281–97. Cavanagh D. Coronaviruses in poultry and other birds. Avian Pathol. 2005; 34:439–48. Lee HJ, Youn HN, Kwon JS, Lee YJ, Kim JH, Lee JB, Park SY, Choi IS, Song CS. Characterization of a novel live attenuated infectious bronchitis virus vaccine candidate derived from a Korean nephropathogenic strain. Vaccine. 2010;28:2887–94. 5. Mahmood ZH, Sleman RR, Uthman AU. Isolation and molecular characterization of Sul/01/09 avian infectious bronchitis virus, indicates the emergence of a new genotype in the Middle East. Vet Microbiol. 2011;150:21–7. 7. 8. Isolation and identification of four 6. Worthington KJ, Currie RJW, Jones RC. A reverse transcriptase – polymerase chain reaction survey of infectious bronchitis virus genotypes in Western Europe from 2002 to 2006. Avian Pathol. 2008; 37:247–57. Xu C, Zhao J, Hu X, Zhang G. infectious bronchitis virus strains in China and analyses of their S1 glycoprotein gene. Vet Microbiol. 2007;122:61–71. Liu HJ, Lee LH, Shih WL, Lin MY, Liao MH. Detection of infectious bronchitis virus by multiplex polymerase chain reaction and sequence analysis. J Virol Methods. 2003;109:31–7. Chandra S, Rawat DS. Medicinal plants of the family Caryophyllaceae: a review of ethno-medicinal uses and pharmacological properties. Integr Med Res. 2015;4:123–31. Shayganni E, Bahmani M, Asgary S, Rafieian-Kopaei M. Inflammaging and cardiovascular disease: management by medicinal plants. Phytomedicine. 2016;23:1119–26. Kama-Kama F, Midiwo J, Nganga J, Maina N, Schiek E, Omosa LK, Osanjo G, Naessens J. Selected ethno-medicinal plants from Kenya with in vitro activity against major African livestock pathogens belonging to the “Mycoplasma mycoides cluster”. J Ethnopharmacol. 2016;192:524–34. 11. 10. 9. 12. Gupta D, Dubey J, Kumar M. Phytochemical analysis and antimicrobial activity of some medicinal plants against selected common human pathogenic microorganisms. Asian Pacific J Trop Dis. 2016;6:15–20. 13. Gandhi GR, Barreto PG, Lima BD, Quintans JS, Araújo AA, Narain N, Quintans-Júnior LJ, Gurgel RQ. Medicinal plants and natural molecules with in vitro and in vivo activity against rotavirus: a systematic review. Phytomedicine. 2016;23:1830–42. 14. Visintini Jaime MF, Redko F, Muschietti LV, Campos RH, Martino VS, 15. 16. Cavallaro LV. In vitro antiviral activity of plant extracts from Asteraceae medicinal plants. Virol J. 2013;10:245. Schuhmacher A, Reichling J, Schnitzler P. Virucidal effect of peppermint oil on the enveloped viruses herpes simplex virus type 1 and type 2 in vitro. Phytomedicine. 2003;10:504–10. Kaškoniene V, Stankevičius M, Drevinskas T, Akuneca I, Kaškonas P, Bimbiraitė-Survilienė K, Maruška A, Ragažinskienė O, Kornyšova O, Briedis V, Ugenskiene R. Evaluation of phytochemical composition of fresh and dried raw material of introduced Chamerion angustifolium L. using chromatographic, spectrophotometric and chemometric techniques. Phytochemistry. 2015;115:184–93. 17. Alkan T, Nitta A, Mizoguguchi H, Itob A, Nabeshima T. A natural svacenger of peroxynitrites, rosmarinic acid, protects against impairment of memory incluced by Aβ25-35. Behav Brain Sci. 2007;180(2):139–45. 18. Dasmalchi K, Ollilainen V, Lackmanc P, Ollilainen V, Gennäs GB, Dorman HJD, Järvinen PP. Acetylcholinesterase inhibitory guided fractionation of Melissa officinalis L. Bioorgan Med Chem. 2009;17:867–71. 19. Razborsek MI, Voncina DB, Dolecek V, Voncina E. Determination of major 20. phenolic acids, phenolic Diterpenes and Triterpenes in rosemary (Rosmarinus officinalis L.) by gas chromatography and mass spectrometry. Acta Chim Slov. 2007;54:60–7. Inouye K, Takano H, Shiga A, Fujita Y, Makino H, Yanagisawa R, Kato Y, Yoshikawa T. Effects of volatile constituents of rosemary extract on lung inflammation induced by diesel exhaust particles Ken-ichiro. Basic Clin Pharmacol Toxicol. 2006;99:52–7. 21. Puodžiūnienė G, Janulis V, Ivanauskas L, Barsteigienė Z, Ribokaitė V, 22. Ragažinskienė O. Quantitative HPLC estimation of flavonoids in showy tick trefoil herbs. Pharm Chem J. 2009;43(5):263–6. Jackwood MW, Rosenbloom R, Petteruti M, Hilt DA, McCall AW, Williams SM. Avian coronavirus infectious bronchitis virus susceptibility to botanical oleoresins and essential oils in vitro and in vivo. Virus Res. 2010;148:86–94. 23. Mohajer Shojai T, Ghalyanchi Langeroudi A, Karimi V, Barin A, Sadri N. The effect of Allium sativum (garlic) extract on infectious bronchitis virus in specific pathogen free embryonic egg. Avicenna J Phytomed. 2016;6:458–67. 24. Chen C, Zuckerman DM, Brantley S, Sharpe M, Childress K, Hoiczyk E, Pendleton AR. Sambucus nigra extracts inhibit infectious bronchitis virus at an early point during replication. BMC Vet Res. 2014;10:24. 25. Yin J, Li G, Li J, Yang Q, Ren X. In vitro and in vivo effects of Houttuynia cordata on infectious bronchitis virus. Avian Pathol. 2011;40(5):491–8. 26. Zhang P, Liu X, Wang W, Liu H, Liu X, Li X, Wu X. Astragalus polysaccharides inhibit avian infectious bronchitis virus infection by regulating viral replication. Microb Pathog. 2017;114:124–8. 27. Ammerman NC, Beier-Sexton M, Azad AF. Growth and maintenance of Vero 28. cell lines. Curr Protoc Microbiol. 2008;Appendix 4:Appendix 4E. Kawahata T, Otake T, Mori H, Kojima Y, Oishi I, Oka S, Fukumori Y, Sano K. A novel substance purified from Perilla frutescens Britton inhibits an early stage of HIV-1 replication without blocking viral adsorption. Antivir Chem Chemother. 2002;13:283–8. 29. Mazzanti G, Battinelli L, Pompeo C. Inhibitory activity of Melissa officinalis L. extraction herpes simplex virus type 2 replication. Nat Prod Res. 2008;22: 1433–40. 30. Meneses R, Ocazionez RE, Martínez JR, Stashenko EE. Inhibitory effect of 31. essential oils obtained from plants grown in Colombia on yellow fever virus replication in vitro. Ann Clin Microbiol Antimicrob. 2009;8:8. Schnitzler P, Koch C, Reichling J. Susceptibility of drug-resistant clinical herpes simplex virus type 1 strains to essential oils of ginger, thyme, hyssop, and sandalwood. Antimicrob Agents Chemother. 2007;51(5):1859–62. 32. Yamasaki K, Nakano M, Kawahata T, Mori H, Otake T, Ueba N, Oishi I, Inami R, Yamane M, Nakamura M, Murata H, Nakanishi T. Anti-HIV-1 activity of herbs in Labiatae. Biol Pharm Bull. 1998;21(8):829–33. 33. Geuenich S, Goffinet C, Venzke S, Nolkemper S, Baumann I, Plinkert P, Reichling J, Keppler OT. Aqueous extracts from peppermint, sage and Lelešius et al. BMC Veterinary Research (2019) 15:178 Page 10 of 10 lemon balm leaves display potent anti-HIV-1 activity by increasing the virion density. Retrovirology. 2008;5:27. 34. Zhang XL, Guo YS, Wang CH, Li GQ, Xu JJ, Chung HY, Ye WC, Li YL, Wang GC. Phenolic compounds from Origanum vulgare and their antioxidant and antiviral activities. Food Chem. 2014;152:300–6. 35. Usha V, Suriyavathana M. Free radical scavenging activity of ethanolic 36. extract of Desmodium gangeticum. JACME. 2012;2:36–42. Ligor M, Stankevičius M, Wenda-Piesik A, Obelevičius K, Ragažinskiene O, Stanius Ž, Maruška A, Buszewski B. Comparative gas chromatographic-mass spectrometric evaluation of hop (Humulus lupulus L.) essential oils and extracts obtained using different sample preparation methods. Food Anal Methods. 2013;7:1433–42. Kaškonienė V, Kaškonas P, Jalinskaitė M, Maruška A. Chemical composition and chemometric analysis of variation in essential oils of Calendula officinalis L. during vegetation stages. Chromatographia. 2011;73(1):163–9. 38. Mosmann T. Rapid colorimetric assay for cellular growth and survival: 37. application to proliferation and cytotoxicity assays. J Immunol Methods. 1983;65(1–2):55–63. 39. Meir R, Maharat O, Farnushi Y, Simanov L. Development of a real-time TaqMan RT-PCR assay for the detection of infectious bronchitis virus in chickens, and comparison of RT-PCR and virus isolation. J Virol Methods. 2010;163(2):190–4. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s13058-023-01650-3	Lang et al. Breast Cancer Research (2023) 25:61 https://doi.org/10.1186/s13058-023-01650-3 Breast Cancer Research RESEARCH Open Access Head-to-head comparison of perfluorobutane contrast-enhanced US and multiparametric MRI for breast cancer: a prospective, multicenter study Manlin Lang1†, Ping Liang24†, Huiming Shen2, Hang Li3, Ning Yang4, Bo Chen5, Yixu Chen6, Hong Ding7, Weiping Yang8, Xiaohui Ji9, Ping Zhou10, ligang Cui11, Jiandong Wang12, Wentong Xu12, Xiuqin Ye13, Zhixing Liu14, Yu Yang15, Tianci Wei16, Hui Wang17, Yuanyuan Yan18, Changjun Wu19, Yiyun Wu20, Jingwen Shi21, Yaxi Wang22, Xiuxia Fang22, Ran li23 and Jie Yu24* Abstract Background Multiparametric magnetic resonance imaging (MP-MRI) has high sensitivity for diagnosing breast cancers but cannot always be used as a routine diagnostic tool. The present study aimed to evaluate whether the diagnostic performance of perfluorobutane (PFB) contrast-enhanced ultrasound (CEUS) is similar to that of MP-MRI in breast cancer and whether combining the two methods would enhance diagnostic efficiency. Patients and methods This was a head-to-head, prospective, multicenter study. Patients with breast lesions diag- nosed by US as Breast Imaging Reporting and Data System (BI-RADS) categories 3, 4, and 5 underwent both PFB- CEUS and MP-MRI scans. On-site operators and three reviewers categorized the BI-RADS of all lesions on two images. Logistic-bootstrap 1000-sample analysis and cross-validation were used to construct PFB-CEUS, MP-MRI, and hybrid (PFB-CEUS + MP-MRI) models to distinguish breast lesions. Results In total, 179 women with 186 breast lesions were evaluated from 17 centers in China. The area under the receiver operating characteristic curve (AUC) for the PFB-CEUS model to diagnose breast cancer (0.89; 95% confi- dence interval [CI] 0.74, 0.97) was similar to that of the MP-MRI model (0.89; 95% CI 0.73, 0.97) (P = 0.85). The AUC of the hybrid model (0.92, 95% CI 0.77, 0.98) did not show a statistical advantage over the PFB-CEUS and MP-MRI models (P = 0.29 and 0.40, respectively). However, 90.3% false-positive and 66.7% false-negative results of PFB-CEUS radiologists and 90.5% false-positive and 42.8% false-negative results of MP-MRI radiologists could be corrected by the hybrid model. Three dynamic nomograms of PFB-CEUS, MP-MRI and hybrid models to diagnose breast cancer are freely available online. Conclusions PFB-CEUS can be used in the differential diagnosis of breast cancer with comparable performance to MP-MRI and with less time consumption. Using PFB-CEUS and MP-MRI as joint diagnostics could further strengthen the diagnostic ability. †Manlin Lang and Ping Liang contributed equally to this work. Correspondence: Jie Yu jiemi301@163.com Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Lang et al. Breast Cancer Research (2023) 25:61 Page 2 of 15 Trial registration Clinicaltrials.gov; NCT04657328. Registered 26 September 2020. IRB number 2020-300 was approved in Chinese PLA General Hospital. Every patient signed a written informed consent form in each center. Keywords PFB-CEUS, MP-MRI, Hybrid, Diagnostic model, Breast lesions, Diagnostic performance Introduction There are more than twenty histological classifications for breast lesions. Diagnostic imaging for the identifi- cation of lesion features remains challenging due to the overlapping characteristics of some benign and malig- nant lesions [1, 2]. It is essential that imaging perfor- mance be optimized to meet the needs of patients with breast lesions. Blood perfusion is one of the vital indicators in dif- ferentiating breast lesions [3]. MP-MRI and CEUS are imaging methods that provide visualization of tumor blood perfusion [4]. Initially, studies suggested that mul- tiparametric MRI (MP-MRI) is more accurate than con- ventional ultrasound (US) or mammography [5, 6] and could improve DCE-MRI specificity [7]. However, pre- cautions must be taken in patients with renal dysfunc- tion with respect to the use of MRI contrast agents [8]. Cost, the timing of the MRI exam, claustrophobia, and patients with morbid obesity who cannot be accommo- dated within the MRI bore are clinically relevant and commonly considered limitations of MRI [9]. Contrast-enhanced ultrasound (CEUS) is a nonirradi- ating, accessible, and easy-to-implement imaging tech- nique that is a powerful supplementary problem-solving tool in the context of MRI [10]. However, CEUS is not always recommended for routine clinical diagnostic use by several breast lesion management guidelines, such as the American Cancer Society (ACS) [11], National Com- prehensive Cancer Network (NCCN) [6], or the Euro- pean Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB) [12]. The main reason cited is the lack of a specific pattern indicating malignancy because several retrospective studies have shown inconsistent results (Additional file 1: Table S1) [13–17]. Currently, sulfur hexafluoride (SHF) is the most widely available CEUS agent worldwide but has inconsist- ent diagnostic results. Image quality in the detection of breast cancer using SHF-CEUS is limited by insufficient signal generation from microbubbles released at high frequencies from linear transducers [18]. Upon recon- stitution with sterile water, stabilized microspheres of perfluorobutane (PFB), which are stabilized by a stable outer shell with hydrogenated egg phosphatidylserine, can resist acoustic pressure, leading to increased micro- bubble oscillations that minimize collapse and loss of signal [19–21]. Additionally, safety of PFB in the diagno- sis of breast lesions has been confirmed in a prospective open-label multicenter phase 3 study [14]. Therefore, PFB demonstrates theoretical potential to improve the qual- ity of breast CEUS examination with sufficient safety for patients. Therefore, we explored the application of feature analy- sis in a multicenter and larger patient cohort by compar- ing head-to-head PFB-CEUS and MP-MRI to investigate the difference in diagnostic ability between PFB-CEUS and MP-MRI for the identification of breast cancer and whether the combination of PFB-CEUS can improve the MP-MRI diagnostic capacity of breast cancer. Materials and methods Study design This study was a prospective, multicenter trial (Clinical- Trials.gov: NCT04657328) in which study participants were consecutively recruited from 17 centers in China from September 2020 to February 2021 (Additional file 1: Table S2). The inclusion criteria and exclusion cri- teria are shown in flowchart Fig. 1 and Additional file 1: Supplementary Materials and Methods S1. Twenty per- cent of the subjects were randomly selected to form a validation cohort, and the remaining 80% were used as a development cohort. Institutional review board and regulatory approval were granted by the Chinese PLA General, and all the patients provided written informed consent. All the investigators and authors had complete access to all the study results, and the authors had full control of the data and statistical results included in this report. Data collection Clinical characteristics and imaging features of PFB- CEUS and MP-MRI were prospectively collected in a Research Electronic Data Capture (REDCap) database. Imaging protocol A total of 25 and 27 radiologists with 5–10 years of expe- rience performed the PFB-CEUS examination and MP- MRI examination, respectively. PFB-CEUS and MP-MRI images from the 17 centers were stored in DICOM data format. Lang et al. Breast Cancer Research (2023) 25:61 Page 3 of 15 Fig. 1 Diagram of a patient selection and b model construction. PFB-CEUS perfluorobutane contrast-enhanced ultrasound, MP-MRI multiparametric magnetic resonance imaging PFB‑CEUS The PFB-CEUS examinations were performed using 15 devices (Additional file 1: Table S3). With the linear probe, pulse inversion harmonic imaging and a mechani- cal index of 0.18–0.24 were used for PFB-CEUS. A bolus injection of 0.015 ml/kg perfluorobutane-filled micro- bubble contrast agent (Sonazoid; GE Healthcare, Oslo, Norway) was administered via a ≧ 22-gauge catheter line placed in the antecubital vein. A 5-mL flush of 0.9% sodium chloride solution was administered after injec- tion of the contrast agent. The dispersion was prepared just prior to use and administered within 2 h of prepara- tion. The imaging timer was started simultaneous to the completion of the contrast agent injection with continu- ous assessment of the lesions for 1 min, followed by inter- mittent scanning for 10 s at the following time points: 1 min and 30 s, 2 min, 3 min, 4 min, and 5 min. Both the mass of interest and the breast involved were evaluated, and patients with multiple lesions were included based on the interval time between two injections being 20 min. MP‑MRI The MP-MRI examinations were performed using eight devices (Additional file 1: Table S4) and performed on 1.5 or 3.0 Tesla systems using a dedicated bilateral breast coil. All protocols followed ACR BI-RADS Mag- netic Resonance Imaging and EUSOMA recommenda- tions [22, 23], which included a T2-weighted sequence and a T1-weighted series acquired before and after the injection of a gadolinium-based contrast agent. In all centers, regions of interest (ROI) (typically 3 × 3 × 1 voxels) in the lesion were used to measure the time-sig- nal intensity curves, and ADC maps used for the evalu- ation were generated by inline monoexponential fitting of the highest and lowest b-value data by the scanner Lang et al. Breast Cancer Research (2023) 25:61 Page 4 of 15 software. A typical MRI exam occupies the MRI system for up to 40 min. Imaging analysis A panel of six radiologists (with each radiologist hav- ing ≥ 15 years of experience in breast lesion diagnosis) reviewed 50 cases of PFB-CEUS and MP-MRI images to identify and define the imaging characteristics. The MP- MRI imaging characteristics were identified by combin- ing the panel of six radiologist results with ACR MRI BI-RADS guideline items. Subsequently, for each imaging modality, two radiologists, each with > 5 years of expe- rience, were trained with a ≥ 0.75 kappa value in all the images. The readers were blinded to current and previous breast imaging and histological findings. Discrepancies were resolved by consensus and consultation with one of the radiologists with 20 years of experience. Quantitative PFB-CEUS parameters of the lesions were acquired using quantitative analysis software, i.e., Novo- Ultrasound Kit (Precision Health Institute, GE Health- care China). ROI 1 included the entire tumor boundary on PFB-CEUS imaging while avoiding the surround- ing parenchyma. ROI 2 encircled the normal-appearing parenchyma, including the same image acquisition plane as far as possible from the breast tumor [24] (Additional file 1: Figure S1). Reference standard For all of the lesions, histopathology was used as the reference standard. Tissue samples of the lesion were obtained by US-guided biopsy or surgical resection. Biopsy was performed using a 14/16-gauge core needle with real-time US/PFB-CEUS guidance. Specimens were reviewed by two senior pathologists from each academic practice. Statistical analysis To detect a difference of 0.1 between a diagnostic test using PFB-CEUS, with an area under the ROC curve (AUC) of 0.85, and using MRI, with an AUC of 0.95, a sample size of 81 malignant and 55 benign lesions was needed to achieve 90% power at a significance level of 0.05. Sample size calculations were performed using PASS version 11 (NCSS, LLC, Kaysville, Utah, USA). Univariate and multivariable logistic regression analy- ses were performed to select the risk factors for breast cancer diagnosis and construct three different models: the PFB-CEUS model, MP-MRI model and hybrid model. Each model contained clinical features (age, menopausal status and nulliparity) and radiological features. intranodular features of PFB-CEUS included the diameter of lesions, fall time (FT), rise time Radiological (RT), time-to-peak (TTP), mean transit time (mTT), arrival time (AT), derivation of washing time (earlier, later, or synchronous), degree of enhancement (hyper- enhancement, isoenhancement, or hypoenhancement), uptake pattern (centripetal, centrifugal, diffuse, or no enhancement), presence or absence of entire washout time > 5 min, heterogeneous pattern, rim-like enhance- ment, claw-shaped pattern, perfusion defects, size enlargement, noncircumscribed margin, and irregu- lar shape. Perinodular features of PFB-CEUS contained nourishing vessels and breast density. The diameter of lesions was defined as the maximum tumor diameter measured on conventional grayscale US. Radiological intranodularn features of MP-MRI included the diameter of lesions, pattern (noncircumscribed margin, irregu- lar shape, homogeneity, etc.), type of lesion (focus/foci only, mass, or nonmass enhancement), uptake pattern (centripetal, centrifugal, diffuse, or other), kinetics—ini- tial phase (slow, medium, or rapid), kinetics—delayed phase (persistent, plateau, or washout), signal intensity of T1 and T2, (high, equal, or low), DWI (high or low), and ADC value according to the Breast Imaging Report- ing and Data System MRI lexicon Perinodular features of MP-MRI contained breast density and background parenchymal enhancement (minimal, mild, moderate, or marked). The radiological features of the hybrid model contained all the radiological features of PFB-CEUS and MP-MRI. Model selection was performed under stepwise cri- teria when necessary. Bootstraps of 1000 resamples and fivefold and tenfold cross-validation were performed to evaluate the performance of the models (the bootstrap process is shown in Additional file 1: Supplementary Materials and Methods S2) [25]. Nomograms used are freely available online. Each of the websites was https:// ceus. shiny apps. io/ PFB- CEUS for the PFB-CEUS model, the MP-MRI https:// ceus. shiny apps. io/ MP- MRI/ for model, and https:// ceus. shiny apps. io/ Hybrid/ for the hybrid model. Discrimination was quantified by using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. Calibration was assessed with the Hosmer‒Lemeshow test and calibra- tion plot. Descriptive analysis summarized the patient characteristics. The operating point selection method is shown in Additional file 1: Supplementary Materials and Methods S3. All of the lesions were collected and ana- lyzed as follows: BI-RADS 4A+ means that in this mode, the lesion is categorized as 4A as a cutoff to consider a malignant lesion, and BI-RADS 3 was categorized as a benign lesion. The plus mark (BI-RASD 4B+ or 4C+) indicates higher categories as the malignant cutoff value. 4A+ Lang et al. Breast Cancer Research (2023) 25:61 Page 5 of 15 means that the malignancy is considered category 4A to 5; 4B+ means that the malignancy is considered category 4B to 5; and 4C+ means that the malignancy is consid- ered category 4C to 5. According to the definition in previous studies [26], the CEUS BI-RADS score was determined (Additional file 1: Supplementary Materials and Methods S4). MRI BI-RADS results were obtained according to the ACR BI-RADS Atlas [22]. The discriminatory performance of our newly established models was also compared to that of Luo et al. [27] 5-point scoring system (Luo model), Chen et al. [28] benign and malignant 6-pattern model (Chen model), and Yukio et al. [14] benign and malignant 2-pattern model (Yukio model). All analyses were per- formed using R and Stata (version 15). P values less than 0.05 were considered statistically significant. Table 1 Baseline characteristics of the study patients Overall lesions (n = 186) Development cohort (n = 151) Validation cohort (n = 35) 49 ± 11 (19, 76) 49 ± 11 (19, 76) 49 ± 10 (27, 68) 90 (48.4) 96 (51.6) 23.9 ± 3.5 (13.7, 37.0) 76 (50.3) 75 (49.7) 23.9 ± 3.3 (16.4, 37.0) 14 (40.0) 21 (60.0) 23.5 ± 4.2 (13.7, 34.1) Characteristic Age (years)b Age (years)b < 49 ≧ 49 Body mass index (kg/m2)b Body mass index (kg/m2) < 18.5 18.5 to < 25 ≥ 25.0 Dense breaste Yes No Parity status Nulliparous Parous Menopausal statusc Premenopausal Postmenopausal 6 (3.4) 110 (61.5) 63 (35.2) 137 (76.5) 42 (23.5) 16 (8.9) 163 (91.1) 107 (59.8) 72 (40.2) First-degree relatives with breast cancerd Presence Absence Diagnosis method Surgery Core biopsy Lesion size (cm)a Lesion size (cm)a < 1.5 ≧ 1.5 Histologic type Benign Malignant 169 (94.4) 10 (5.6) 73 (39.2) 113 (60.8) 1.8 (1.2, 2.4) 65 (34.9) 121 (65.1) 71 (38.1) 115 (61.8) 3 (2.1) 91 (62.3) 52 (35.6) 107 (73.3) 39 (26.7) 14 (9.6) 132 (90.4) 89 (61.0) 57 (39.0) 138 (94.5) 8 (5.5) 62 (41.1) 89 (58.9) 1.8 (1.2, 2.5) 52 (34.4) 99 (65.6) 58 (38.4) 93 (61.6) 3 (9.1) 19 (57.6) 11 (33.3) 30 (90.9) 3 (9.1) 2 (6.1) 31 (93.9) 18 (54.5) 15 (45.5) 31 (93.9) 2 (6.1) 11 (31.4) 24 (68.6) 1.8 (1.3, 2.4) 13 (37.1) 22 (62.9) 13 (37.1) 22 (62.9) Unless otherwise indicated, data are numbers of patients, and data in parentheses are percentages a Data are medians, and data in parentheses are the interquartile range b Data are means ± standard deviations, and data in parentheses are the range c Menopausal status: Women aged 60 years or older, reporting a history of hysterectomy, or reporting no periods within the past 12 months without the use of hormonal contraceptives were categorized as postmenopausal. Women reporting regular periods (12–18 times in the last 12 months) without the use of hormonal contraceptives were categorized as premenopausal d First-degree relatives (mother, sister, and daughter) with breast cancer e Breast density assessed with Magnetic Resonance Imaging according to the ACR BI-RADS Magnetic Resonance Imaging categories a, b, c, d. In this study, c and d categories of breast density are defined as dense breast. On the other hand, the, a and b categories of breast density are defined as nondense breast Lang et al. Breast Cancer Research (2023) 25:61 Page 6 of 15 Results Participant and imaging characteristics In total, 186 lesions from 179 female participants (mean age, 48 ± 11 years) with 115 malignant lesions and 71 benign lesions were enrolled after excluding 177 lesions (Fig. 1) from 17 tertiary centers and 13 provinces across China, allowing for the inclusion of a geographi- cally diverse patient population. The mean age (SD) of the overall cohort was 49 ± 11 years, and all 179 of the patients were women (Table 1). The details of the overall pathologic distribution and of each cohort are presented in Additional file 1: Table S5. The frequencies of the PFB-CEUS and MP-MRI characteristics in the two data cohorts are described in Additional file 1: Table S6. Model development and performance The univariate and multivariable logistic regression analyses of PFB-CEUS, MP-MRI, and the hybrid model are shown in Additional file 1: Tables S7, S8 and S9, respectively. In the external validation model, the hybrid model showed an AUC similar to that of both MP-MRI and PFB-CEUS with an AUC of 0.92 ([95% CI 0.77, 0.98]). The PFB-CEUS model was shown to have an AUC of 0.89 ([95% CI 0.74,0.97]), which was comparable to the PFB- CEUS BI-RADS (AUC 0.80 [95% CI 0.63,0.92), P = 0.43; better than the Luo model (AUC 0.74 [95% CI 0.57,0.87], P = 0.01); higher than the Chen model (AUC 0.69 [95% CI 0.51, 0.83], P = 0.02); and superior to the Yukio model (AUC 0.70 [95% CI 0.53, 0.85], P = 0.01). In addition, the MP-MRI model was shown to have an AUC of 0.89 ([95% CI 0.73, 0.97]), which was competitive with ACR BI- RADS models (AUC 0.73 [95% CI 0.55, 0.87], P = 0.15). (Additional file 1: Table S10, Fig. 2). The PFB-CEUS, MP-MRI and hybrid models were well calibrated, and all showed statistical significance (P > 0.05) in the Hosmer‒Lemeshow test. The calibra- tion plots are shown in Fig. 3. The decision curves are shown in Additional file 1: Figure S2. In addition, the respective dynamic nomograms of the PFB-CEUS, MP-MRI, and hybrid models are shown in the follow- ing links: https:// ceus. shiny apps. io/ PFB- CEUS/, https:// ceus. shiny apps. io/ MP- MRI/, https:// ceus. shiny apps. io/ Hybrid/. Figure 4 and Additional file 1: Figure S3 show a matched example in which the PFB-CEUS and MP-MRI nomograms showed a high malignant probability of Fig. 2 Receiver operating characteristic (ROC) curves of the models in the differentiation of breast lesions. a–c Results in the development cohort. d–f Results in the external validation cohort. PFB-CEUS perfluorobutane contrast-enhanced ultrasound, MP-MRI multiparametric magnetic resonance imaging Lang et al. Breast Cancer Research (2023) 25:61 Page 7 of 15 Fig. 3 . Plot shows calibration of the PFB-CEUS, MP-MRI and hybrid models. a PFB-CEUS model result in the development cohort; b MP-MRI model result in the development cohort; c hybrid model result in the development cohort; d PFB-CEUS model result in the external validation cohort. e MP-MRI model result in the external validation cohort. f hybrid model result in the external validation cohort. PFB-CEUS perfluorobutane contrast-enhanced ultrasound, MP-MRI multiparametric magnetic resonance imaging breast cancer. Figure 5 and Additional file 1: Figure S4 show a matched example in which PFB-CEUS showed a high malignant probability of cancer, but MP-MRI nom- ograms showed a low malignant probability of cancer in which the lesion was diagnosed as invasive carcinoma by pathology. Figure 6 shows a matched example in which the PFB-CEUS and MP-MRI nomograms showed a high malignant probability of cancer, but the hybrid nomograms showed a low malignant probability of can- cer in which the lesion was diagnosed as ductal hyper- plasia by pathology. Comparison among the three models PFB‑CEUS versus MP‑MRI model The PFB-CEUS model showed a comparable discrimi- nation ability for diagnosing breast cancer, with the MP-MRI model showing the same not only in the devel- opment cohort (AUC 0.90 [95% CI 0.84, 0.94]) versus (AUC 0.90 [95% CI: 0.85, 0.95], P = 0.80) but also in the validation cohort (AUC 0.89 [95% CI:0.74, 0.97]) ver- sus (AUC 0.89 [95% CI: 0.73, 0.97], P = 0.85) (Additional file 1: Table S10). PFB‑CEUS versus hybrid model The hybrid model showed a higher capacity to diagnose breast cancer compared with the PFB-CEUS model in the development cohort (AUC 0.95 [95% CI: 0.90, 0.98]) versus ((AUC 0.90, [95% CI 0.84,0.94]) P = 0.01) and a similar capacity with PFB-CEUS in the validation cohort ((AUC 0.92 [95% CI: 0.77, 0.98]) versus (AUC 0.89 [95% CI: 0.74, 0.97]), P = 0.29) (Additional file 1: Table S10). MP‑MRI versus hybrid model The hybrid model demonstrated a competitive capacity for diagnosing breast cancer compared with the MP-MRI model ((AUC 0.95 [95% CI: 0.90, 0.98]) vs. (AUC 0.90 [95% CI: 0.85, 0.95]), respectively; P = 0.078), not only in the development cohort but also in the validation cohort ((AUC, 0.92 [95% CI: 0.77, 0.98]) vs. (AUC 0.89 [95% CI: 0.73,0.97]), respectively; P = 0.401) (Additional file 1: Table S10). Model performance in subpopulations Of all the subgroups dichotomized by age and men- strual status, the PFB-CEUS model showed a similar Lang et al. Breast Cancer Research (2023) 25:61 Page 8 of 15 Fig. 4 PFB-CEUS and MP-MRI images for breast cancer diagnosis. This is a 67-year-old woman with an invasive carcinoma by pathological diagnosis. a At 23 s, the PFB-CEUS had arrived at the peak intensity. The lesion showed a size enlargement on PFB-CEUS compared with gray ultrasound. b TIC of PFB-CEUS shows a 14 s fall time of the lesion. c DWI shows a high signal lesion with an ADC value of 0.77 10–3 mm2/s. d DCE-MRI shows patients with mild BPE and a lesion with a noncircumscribed margin. e PFB-CEUS Nomogram shows that this case had a 98% MP at the https:// ceus. shiny apps. io/ PFB- CEUS/ link. f MP-MRI Nomogram shows that this case had a 95% MP at the https:// ceus. shiny apps. io/ MP- MRI/ link. MP-MRI multiparametric magnetic resonance imaging, PFB-CEUS perfluorobutane contrast-enhanced ultrasound, DWI diffusion weighted imaging, ADC apparent diffusion coefficient, DCE-MRI dynamic contrast-enhanced magnetic resonance imaging, BPE background parenchymal enhancement, TIC time–intensity curve, FT fall time, MP malignant probability Lang et al. Breast Cancer Research (2023) 25:61 Page 9 of 15 Fig. 5 A breast cancer was diagnosed by PFB-CEUS as a malignant lesion but was diagnosed as a benign lesion by MP-MRI. This is a 50-year-old woman with an invasive carcinoma by pathological diagnosis. a At the 8th second, PFB-CEUS had arrived at the peak intensity. The lesion on PFB-CEUS was the same size as that on gray ultrasound. b TIC of PFB-CEUS shows a 6.6 s fall time of the lesion. c DWI shows a high signal lesion with an ADC value of 1.6 * 10–3 mm2/s. d DCE-MRI shows the patients with a marked BPE and a lesion with a noncircumscribed margin. e PFB-CEUS Nomogram shows that this case had an 82% MP at the https:// ceus. shiny apps. io/ PFB- CEUS/ link. f MP-MRI Nomogram shows that this case had a 38% MP at the https:// ceus. shiny apps. io/ MP- MRI/ link. MP-MRI multiparametric magnetic resonance imaging, PFB-CEUS perfluorobutane contrast-enhanced ultrasound, DWI diffusion weighted imaging, ADC apparent diffusion coefficient, DCE-MRI dynamic contrast-enhanced magnetic resonance imaging, BPE background parenchymal enhancement, TIC time–intensity curve, FT fall time, MP malignant probability AUC to MP-MRI in the external validation cohort. No significant difference in AUC was noted between hybrid model comparisons of PFB-CEUS or MP-MRI in the external validation cohort (Additional file 1: Table S11). The diagnostic results for high-risk lesions (a lesion that would be appropriate for surgical consultation, including intraductal papilloma and phyllodes tumor) are presented in Additional file 1: Table S12. Model performance for diagnosing breast cancer compared with radiologists BI-RADS 4A, 4B, and 4C were used as the cut points of malignant lesions for comparison. PFB‑CEUS When compared with the on-site radiologists, the PFB- CEUS model achieved higher sensitivity in the BI-RADS 4C+ mode with lower specificity in the BI-RADS 4B+ mode. The hybrid model achieved higher sensitivity in the BI-RADS 4C+ mode with lower specificity in BI- RADS 4B+ and 4C+. When compared with the three senior reviewers, the PFB-CEUS model achieved higher sensitivity in the BI-RADS 4B+ and 4C+ modes, with lower specificity only in the BI-RADS 4C+ mode. The hybrid model achieved higher sensitivity in BI-RADS 4C+ mode and lower specificity in BI-RADS 4B+ mode (Table 2). Lang et al. Breast Cancer Research (2023) 25:61 Page 10 of 15 Fig. 6 A benign breast lesion was diagnosed as a cancer by PFB-CEUS and MP-MRI but was diagnosed as a benign lesion by the hybrid model. This is a 50-year-old woman with ductal hyperplasia by pathological diagnosis. a At the 12th second, PFB-CEUS had arrived at the peak intensity. The lesion on PFB-CEUS was the same size as that on gray ultrasound. b TIC of PFB-CEUS shows a 14 s fall time of the lesion. c DWI shows a high signal lesion with an ADC value of 1.2 * 10–3 mm2/s. d DCE-MRI shows the patients with a marked BPE and a lesion with a circumscribed margin. e PFB-CEUS Nomogram shows that this case had a 36% MP at the https:// ceus. shiny apps. io/ PFB- CEUS/ link. f MP-MRI Nomogram shows that this case had a 40% MP at the https:// ceus. shiny apps. io/ MP- MRI/ link. g Hybrid Nomogram shows that this case had a 14% MP at the https:// ceus. shiny apps. io/ hybrid/ link. MP-MRI multiparametric magnetic resonance imaging, PFB-CEUS perfluorobutane contrast-enhanced ultrasound, DWI diffusion weighted imaging, ADC apparent diffusion coefficient, DCE-MRI dynamic contrast-enhanced magnetic resonance imaging, BPE background parenchymal enhancement, TIC time–intensity curve, FT fall time, MP malignant probability MP‑MRI When compared with the on-site radiologists, the MP- MRI model achieved higher sensitivity in all three modes with lower specificity in the BI-RADS 4A+ and 4B+ modes, as well as the hybrid model. When compared with the three senior reviewers, the MP-MRI model achieved higher sensitivity in BI-RADS 4B+ mode and lower specificity in BI-RADS 4A+ mode. The hybrid model achieved comparable sensitivity in the BI-RADS 4C+ mode and higher specificity in the BI-RADS 4C+ mode (Table 2). Model performance with false‑positive and false‑negative correction rate PFB‑CEUS The false-positive correction rates of the PFB-CEUS and hybrid models were 80.6% and 90.3% for the on-site results and 82.2% and 88.9% for the reviewers’ results based on the BI-RADS 4A+ modes, respectively. The false-negative correction rates of the PFB-CEUS and hybrid models were 66.7% and 66.7% for the on- site results and 83.3% for the reviewers’ results in the BI-RADS 4A+ mode, respectively (Additional file 1: Table S13) (Fig. 7). MP‑MRI The false-positive correction rates of the MP-MRI and hybrid model were 77.7% and 86.1% for the on-site results and 83.0% and 90.5% for the reviewers’ results in the BI-RADS 4A+ mode, respectively. The false-nega- tive correction rates of the MP-MRI and hybrid models were 50.0% and 0.0% for the on-site results and 57.1% and 42.8% for the reviewers’ results in the BI-RADS 4A+ mode, respectively (Additional file 1: Table S13) (Fig. 7). Lang et al. Breast Cancer Research (2023) 25:61 Page 11 of 15 e u l a v P % p S e u l a v P % e S e u l a v P % p S e u l a v P % e S e u l a v P % p S e u l a v P % e S ) + c 4 . s v b 4 , a 4 , 3 ( S D A R ‑ I B ) + b 4 . s v a 4 , 3 ( S D A R ‑ I B ) + a 4 . s v 3 ( S D A R ‑ I B s e d o m S D A R - I B t n e r e ff d n i i l s t s i g o o d a r h t i i w l e d o m h c a e f o e c n a m r o i f r e p c i t s o n g a d r e c n a c t s a e r b f o n o s i r a p m o C 2 e l b a T f 4 3 1 0 . . ) 8 3 9 , . 4 1 8 ( e 7 1 0 0 . ) 5 1 1 / 4 0 1 ( 4 0 9 . f 2 5 0 0 . . ) 1 5 9 , . 5 3 8 ( f 7 2 6 0 . e 4 0 0 0 . f 1 1 7 0 . ) 1 7 / 6 3 ( 7 0 5 . . ) 8 2 6 , . 6 8 3 ( e – f – . ) 2 3 5 3 9 2 ( , . f 6 0 0 0 . ) 5 1 1 / 5 1 1 ( 0 0 0 1 . . ) 0 0 0 1 , . 8 6 9 ( . ) 8 9 9 , . 9 3 9 ( f 1 4 0 0 . e 2 0 0 0 . f – c 6 9 0 0 . d – c 2 3 3 0 . d 4 5 4 0 . e 2 3 3 0 . f 8 7 0 0 . e 5 1 8 0 . f 3 2 0 0 . ) 1 7 / 5 6 ( 5 1 9 . . ) 8 6 9 , . 5 2 8 ( ) 1 7 / 6 5 ( 9 8 7 . . ) 7 7 8 , . 6 7 6 ( ) 1 7 / 7 5 ( 3 0 8 . . ) 8 8 8 , . 1 9 6 ( ) 1 7 / 0 6 ( 5 4 8 . . ) 0 2 9 , . 0 4 7 ( ) 1 7 / 2 6 ( 3 7 8 . . ) 0 4 9 , . 3 7 7 ( ) 1 7 / 8 4 ( 6 7 6 . . ) 2 8 7 , . 5 5 5 ( ) 1 7 / 7 5 ( 3 0 8 . . ) 8 8 8 , . 1 9 6 ( ) 1 7 / 0 6 ( 5 4 8 . . ) 0 2 9 , . 0 4 7 ( c 9 2 0 0 . d 3 2 0 0 . c 4 0 0 0 . d 4 0 0 0 . ) 5 1 1 / 0 9 ( 3 8 7 . . ) 4 5 8 , . 6 9 6 ( ) 5 1 1 / 0 9 ( 3 8 7 . . ) 4 5 8 , . 6 9 6 ( ) 5 1 1 / 2 0 1 ( 7 8 8 . ) 5 1 1 / 4 0 1 ( 4 0 9 . . ) 8 3 9 , . 4 1 8 ( . ) 1 5 9 , . 5 3 8 ( ) 5 1 1 / 2 9 ( 0 0 8 . . ) 9 6 8 , . 5 1 7 ( ) 5 1 1 / 4 9 ( 7 1 8 . . ) 3 8 8 , . 5 3 7 ( c 1 0 0 0 < . d 1 0 0 0 < . c 1 0 0 0 . d 9 0 0 0 . e 1 4 0 0 . ) 5 1 1 / 2 0 1 ( 7 8 8 . e 1 0 0 0 < . ) 1 7 / 6 5 ( 9 8 7 . . ) 7 7 8 , . 6 7 6 ( ) 1 7 / 0 5 ( 4 0 7 . . ) 7 0 8 , . 4 8 5 ( ) 1 7 / 6 2 ( 6 6 3 . . ) 9 8 4 , . 5 5 2 ( ) 1 7 / 6 3 ( 7 0 5 . . ) 8 2 6 , . 6 8 3 ( ) 1 7 / 2 5 ( 2 3 7 . . ) 1 3 8 , . 4 1 6 ( ) 1 7 / 3 3 ( 5 6 4 . . ) 7 8 5 , . 5 4 3 ( ) 1 7 / 9 2 ( 8 0 4 . ) 5 1 1 / 6 0 1 ( 2 2 9 . . ) 4 6 9 , . 7 5 8 ( ) 5 1 1 / 3 0 1 ( 6 9 8 . . ) 5 4 9 , . 5 2 8 ( e 6 1 0 0 . ) 5 1 1 / 3 1 1 ( 3 8 9 . e 1 0 0 0 < . c 4 4 3 0 . d 1 0 0 0 . c – d – ) 5 1 1 / 8 0 1 ( 9 3 9 . ) 5 1 1 / 9 9 ( 1 6 8 . . ) 8 1 9 , . 4 8 7 ( . ) 5 7 9 , . 9 7 8 ( ) 5 1 1 / 2 1 1 ( 4 7 9 . . ) 5 9 9 , . 6 2 9 ( ) 5 1 1 / 5 1 1 ( 0 0 0 1 . . ) 0 0 0 1 8 6 9 ( , . c – d – ) 1 7 / 0 4 ( 3 6 5 . . ) 1 8 6 , . 0 4 4 ( ) 1 7 / 6 2 ( 6 6 3 . . ) 9 8 4 , . 5 5 2 ( A N c 1 0 0 0 < . d 9 2 2 0 . ) 1 7 / 8 1 ( 4 5 2 . . ) 1 7 3 , . 8 5 1 ( ) 1 7 / 5 3 ( 3 9 4 . . ) 4 1 6 , . 2 7 3 ( ) 1 7 / 8 1 ( 4 5 2 . . ) 1 7 3 , . 8 5 1 ( ) 1 7 / 8 ( 3 1 1 . . ) 0 1 2 , . 0 5 ( ) 1 7 / 8 1 ( 4 5 2 . . ) 1 7 3 , . 8 5 1 ( c – d – c – d – e – f – e – f – ) 5 1 1 / 2 1 1 ( 4 7 9 . . ) 5 9 9 , . 6 2 9 ( ) 5 1 1 / 9 0 1 ( 8 4 9 . . ) 1 8 9 , . 0 9 8 ( b s r e w e v e R i a e t i s - n O S U E C - B F P ) 5 1 1 / 5 1 1 ( 0 0 0 1 . l e d o m d i r b y H A N l e d o m S U E C . ) 0 0 0 1 , . 8 6 9 ( ) 5 1 1 / 3 1 1 ( 3 8 9 . . ) 8 9 9 , . 9 3 9 ( ) 5 1 1 / 8 0 1 ( 9 3 9 . . ) 5 7 9 , . 9 7 8 ( b s r e w e v e R i a e t i s - n O I R M - P M ) 5 1 1 / 5 1 1 ( 0 0 0 1 . l e d o m I R M . ) 0 0 0 1 , . 8 6 9 ( . ) 0 0 0 1 , . 8 6 9 ( ) 5 1 1 / 5 1 1 ( 0 0 0 1 . l e d o m d i r b y H i e c n e d fi n o c % 5 9 e h t d n a r o t a n m o n e d d n a r o t a r e m u n e h t e r a s e s e h t n e r a p n i a t a D . t n e s b a s i e d a r g 3 S D A R - I B s a s i s o n g a d n o i s e i l t s a e r b s ’ l e d o m S U E C - B F P e h t s n a e m A N . y t i c fi i c e p s p S , y t i v i t i s n e s e S . y t i v i t i s n e s ’ s r e w e v e r i / e t i s - n o e h t h c t a m o t y t i v i t i s n e s s ’ l e d o m e h t g n i t t e s y b d e n m r e t e d s a w y t i c fi i c e p s i s ’ l e d o m e h t ; i y t i c fi i c e p s ’ s r e w e v e r / e t i s - n o e h t h c t a m o t y t i c fi i c e p s s ’ l e d o m e h t g n i t t e s y b d e n m r e t e d s a w y t i v i t i s n e s i s ’ l e d o m e h T I i l R M - P M d e m r o f r e p o h w s t s i g o o d a r 7 2 d n a S U E C - B F P d e m r o f r e p o h w s t s i g o o d a r 5 2 a l i s l a v r e t n i s e g a m i I i i l R M - P M d e r o t s e h t d e w e v e r o h w s t s i g o o d a r 3 d n a S U E C - B F P d e r o t s e h t d e w e v e r o h w s t s i g o o d a r 3 b l i i i l e t i s - n o s t s i g o o d a r S U E C - B F P f o s t l u s e r c i t s o n g a d e h t h t i i w s n o s i r a p m o c r o f e u a v P c l e t i s - n o s t s i g o o d a r l i I R M - P M f o s t l u s e r c i t s o n g a d e h t h t i i w s n o s i r a p m o c r o f e u a v P e l i s r e w e v e r S U E C - B F P e h t f o s t l u s e r c i t s o n g a d e h t h t i i w s n o s i r a p m o c r o f e u a v P d l s r e w e v e r i I R M - P M f o s t l u s e r c i t s o n g a d h t i i w s n o s i r a p m o c r o f e u a v P l f Lang et al. Breast Cancer Research (2023) 25:61 Page 12 of 15 Fig. 7 FPCR and FNCR for on-site radiologists with models. a FPCR for on-site radiologists with models. b FNCR for on-site radiologists with models. On-site presented the radiologists who performed PFB-CEUS and MP-MRI. FPCR false-positive correction rate, FNIR false-negative correction rate Discussion MP-MRI has better sensitivity and specificity for the detection of breast cancer than dynamic contrast- enhanced MRI (DCE-MRI) [29]. However, this technol- ogy was not always available and was associated with prohibitively high costs for use as a routine diagnostic tool [30]. There is an increasing demand to develop a sur- rogate, easy-to-implement method to diagnose breast lesions. Our study is the first to compare the diagnostic performance between PFB-CEUS and MP-MRI, both advanced breast cancer diagnosis modalities, for the diagnosis of breast cancer. In this study, the AUC for the PFB-CEUS model was similar to that of the MP-MRI model, as well as in the subgroup analysis The hybrid model could improve the sensitivity and specificity of PFB-CEUS on-site radiolo- gists in all three modes, as well as MP-MRI. The FPCR and FNCR were excellent for PFB-CEUS, MP-MRI, and the hybrid model. The results indicate that PFB-CEUS could potentially improve the diagnostic ability of MP- MRI for breast cancer, and when patients are restricted by the contraindications of MP-MRI, PFB-CEUS could be performed as an alternative examination method for patients with breast lesions. Over the past decade, several comparative studies to investigate the ability of SHF-CEUS and DCE-MRI for breast cancer diagnosis have been published, and these have demonstrated inconsistent results. (Additional file 1: Lang et al. Breast Cancer Research (2023) 25:61 Page 13 of 15 Table S1). Half indicated that SHF-CEUS showed higher sensitivity and specificity than DCE-MRI, and the remain- der presented contrasting results. In all these studies, there was no rigorous imaging feature review or selec- tion and a lack of evaluation of interobserver variability between radiologists and inclusion analysis of quantitative SHF-CEUS indicators. Compared with the previous retro- spective cohort study between SHF-CEUS and DCE-MRI for breast cancer, it is worth mentioning that PFB-CEUS and MP-MRI images were prospectively collected accord- ing to the standard protocol in our study. Notably, our study has confirmed the theoretical advantages of PFB, which has a high resonance frequency and stable outer shell, leading to the depiction of clear enhancement signals such as tumor size and duration and improving the CEUS image quality in linear transducers. Enhancement size enlargement and FT became stable indicators through a bootstrap analysis that excluded the zero value, ensuring a more accurate estimation of PFB-CEUS. Compared with the previously constructed CEUS model, our study showed a higher AUC than all the others in both the development cohort and validation cohorts. In addition, MP-MRI was set as the control with PFB-CEUS in the present study, which can reflect the dif- fusion of water molecules in the intracellular and extra- cellular spaces and can lead to a higher specificity than DCE-MRI in the diagnosis of malignant lesions due to the high cell density and the small extracellular space in malignant lesions [29, 31, 32]. Encouragingly, PFB-CEUS did not sacrifice the diagnostic performance for breast cancer compared to MP-MRI with PFB-CEUS superior contrast resolution. When PFB was used as an agent to compare with MRI for breast cancer, only one result was reported, namely that PFB-CEUS reached a similar AUC compared with DCE-MRI in 127 patients [14]. Our study involved patients who were prospectively enrolled from 17 ter- tiary centers across China, providing for a head-to- head comparative study with a relatively large sample size that guaranteed the generalization results. Rigor- ous validation, including the 1000-sample bootstrap, five- and tenfold internal validation and independ- ent external validation, ensured robust and reproduc- ible results, even when using different device settings, and the comprehensive features were analyzed for the ability to distinguish breast cancer by using clinical parameters and qualitative and quantitative analysis of perinodular and intranodular features selected from PFB-CEUS and MP-MRI. Compared with the radiologists, the implementation of all three models was feasible in daily practice, requir- ing less time and effort to collect and analyze clinical and imaging characteristics (only 3–5 variables). Freely available online nomograms with relatively reliable results are structured for simple daily practice and are accessible to three different websites. Furthermore, a false-positive imaging diagnosis may lead to unnecessary biopsy invasion, patient anxiety, and health care costs. BI-RADS 4A is usually recommended for use in the clinic as the cutoff point for biopsy. In our study, the FPCR of PFB-CEUS, MP-MRI, and the hybrid model was excellent, particularly when BI-RADS 4A was used as the cutoff point to diagnose positive lesions, indi- cating that the models’ FPCR could supply a reference to appropriately avoid overdiagnosis for probable benign lesions. Our study had several limitations. First, the absolute accuracy and stability of the models may have been influ- enced by the relatively small sample size. The subgroup analysis in the different lesion size and breast density groups failed to achieve statistical results because of the small sample size in some groups, although this was cal- culated in the design of this prospective study. Second, because we enrolled only lesions with BI-RADS grades 3, 4, and 5 in the grayscale US, the prevalence of breast cancer in these participants may be higher than that found during the usual diagnostic examination for all surveillance-positive lesions. Therefore, the diagnostic performance of PFB-CEUS and MP-MRI is potentially biased. Last, the patients enrolled in this study were only Chinese without populations from other countries. In the future, a worldwide range trial could be conducted. In conclusion, the PFB-CEUS model was not only more efficient but also showed similar diagnostic ability as the MP-MRI model. Using PFB-CEUS and MP-MRI as joint diagnostics could further strengthen the abil- ity to better characterize breast cancer. This suggests that not only can PFB-CEUS be used as a surrogate method to diagnose breast lesions but also that when these models are used as an aid to radiologists in clini- cal practice, they can save time and effort and avoid overdiagnosis. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s13058- 023- 01650-3. Additional file 1. Supplementary tables and figures. Author contributions All authors read and approved the final manuscript. ML and JY wrote the main manuscript text. ML, PL, HS, HL, NY, BC, YC, HD, WY, XJ, PZ, LC, JW, WX, XY, ZL, YY, TW, HW, YY, CW, YW, JS, YW, XF, and RL collected the data. All authors reviewed the manuscript. . Funding .Not applicable. Lang et al. Breast Cancer Research (2023) 25:61 Page 14 of 15 Availability of data and materials Data generated or analyzed during the study are available from the corre- sponding author by request. Declarations Ethics approval and consent to participate The study was approved in the center of the principal investigator at Chinese PLA General Hospital in Beijing, China (IRB number: 2020-300). The Ethical Committee was “Chinese PLA General Hospital”. Approval was obtained on July 23rd, 2020. Every patient signed a written informed consent form in each center. The research was carried out in accordance with the Declaration of Helsinki. Competing interests The authors declare that they have no competing interests. Author details 1 Department of Interventional Ultrasound, Fifth Medical Center of Chinese PLA General Hospital & Chinese PLA Medical School, Beijing 100039, China. 2 Department of Ultrasound, Zhongda Hospital Southeast University, Nan- jing 210009, China. 3 Department of Breast Surgery, Affiliated Hospital of Putian University, Putian 351100, China. 4 Department of Ultrasound, Xingcheng People’s Hospital, Xingcheng 125100, China. 5 Department of Ultrasound Medi- cine, Lu’an People’s Hospital of Anhui Province, Liuan 237000, China. 6 Depart- ment of Ultrasound, The Fifth People’s Hospital of Chengdu, Chengdu 611130, China. 7 Department of Ultrasound, Huashan Hospital, Shanghai 200040, China. 8 Department of Ultrasound, Guangxi Medical University Cancer Hospital, Nan- ning 530021, China. 9 Department of Ultrasound, The Fourth Hospital of Hebei Medical University, Shijiazhuang 050011, China. 10 Department of Ultrasound, The Third Xiangya Hospital, Changsha 410000, China. 11 Department of Ultra- sound, Peking University Third Hospital, Beijing 100191, China. 12 General Surgery, Chinese PLA General Hospital, Beijing 100853, China. 13 Department of Ultrasound, First Affiliated Hospital of Southern University of Science and Technology, Second Clinical College of Jinan University, Shenzhen Medical Ultrasound Engineering Center, Shenzhen People’s Hospital, Shenz- hen 518020, China. 14 Department of Ultrasound Medicine, The First Affiliated Hospital of Nanchang University, Nanchang 330006, China. 15 Department of Ultrasound, Beijing Friendship Hospital, Beijing 100050, China. 16 Depart- ment of Ultrasound, The 2nd Affiliated Hospital of Harbin, Harbin 150001, China. 17 Department of Ultrasound, China-Japan Union Hospital of Jilin Uni- versity, Changchun 130033, China. 18 Department of Ultrasound, Zhengzhou Central Hospital, Zhengzhou 450000, China. 19 Department of Ultrasonogra- phy, The First Affiliated Hospital of Harbin Medical University, Harbin 150001, China. 20 Department of Ultrasound, Affiliated Hospital of Nanjing University of Chinese Medicine, Nanjing 210029, China. 21 Department of Ultrasound, Shengjing Hospital of China Medical University, Shenyang 110004, China. 22 Department of Ultrasound, The Affiliated Hospital of Inner Mongolia Medi- cal University, Hohhot 010050, China. 23 Department of Ultrasound, The First Affiliated Hospital of Xinxiang Medical University, Xinxiang 453100, China. 24 Department of Interventional Ultrasound, Fifth Medical Center of Chinese PLA General Hospital, Beijing 100039, China. Received: 21 October 2022 Accepted: 22 April 2023 References 1. Tosteson ANA, Fryback DG, Hammond CS, Hanna LG, Grove MR, Brown M, et al. Consequences of false-positive screening mammograms. JAMA Intern Med. 2014;174:954–61. 2. den Dekker BM, Bakker MF, de Lange SV, Veldhuis WB, van Diest PJ, Duvivier KM, et al. Reducing false-positive screening mri rate in women with extremely dense breasts using prediction models based on data from the DENSE trial. Radiology. 2021;301:283–92. Folkman J. What is the evidence that tumors are angiogenesis depend- ent? J Natl Cancer Inst. 1990;82:4–6. 3. 4. Sridharan A, Eisenbrey JR, Dave JK, Forsberg F. Quantitative nonlinear contrast-enhanced ultrasound of the breast. AJR Am J Roentgenol. 2016;207:274–81. 5. Riedl CC, Luft N, Bernhart C, Weber M, Bernathova M, Tea M-KM, et al. Triple-modality screening trial for familial breast cancer underlines the importance of magnetic resonance imaging and questions the role of mammography and ultrasound regardless of patient mutation status, age, and breast density. J Clin Oncol. 2015;33:1128–35. 6. Bevers TB, Helvie M, Bonaccio E, Calhoun KE, Daly MB, Farrar WB, et al. Breast cancer screening and diagnosis, version 3.2018, NCCN clinical practice guidelines in oncology. J Natl Compr Cancer Netw. 2018;16:1362–89. 7. Clauser P, Krug B, Bickel H, Dietzel M, Pinker K, Neuhaus V-F, et al. Diffu- sion-weighted imaging allows for downgrading MR BI-RADS 4 lesions in contrast-enhanced MRI of the breast to avoid unnecessary biopsy. Clin Cancer Res. 2021;27:1941–8. 8. Contrast Manual \| American College of Radiology [Internet]. [cited 2022 Sep 8]. https:// www. acr. org/ Clini cal- Resou rces/ Contr ast- Manual 9. Mann RM, Cho N, Moy L. Breast MRI: state of the art. Radiology. 2019;292:520–36. 10. Atri M, Jang H-J, Kim TK, Khalili K. Contrast-enhanced US of the liver and kidney: a problem-solving modality. Radiology. 2022;303:11–25. 11. Saslow D, Boetes C, Burke W, Harms S, Leach MO, Lehman CD, et al. American Cancer Society guidelines for breast screening with MRI as an adjunct to mammography. CA Cancer J Clin. 2007;57:75–89. 12. Sidhu PS, Cantisani V, Dietrich CF, Gilja OH, Saftoiu A, Bartels E, et al. The EFSUMB guidelines and recommendations for the clinical practice of contrast-enhanced ultrasound (CEUS) in non-hepatic applications: update 2017 (Long Version). Ultraschall Med. 2018;39:e2-44. 13. Du J, Wang L, Wan C-F, Hua J, Fang H, Chen J, et al. Differentiating benign from malignant solid breast lesions: combined utility of conventional ultrasound and contrast-enhanced ultrasound in comparison with mag- netic resonance imaging. Eur J Radiol. 2012;81:3890–9. 14. Miyamoto Y, Ito T, Takada E, Omoto K, Hirai T, Moriyasu F. Efficacy of sona- zoid (perflubutane) for contrast-enhanced ultrasound in the differentia- tion of focal breast lesions: phase 3 multicenter clinical trial. AJR Am J Roentgenol. 2014;202:W400-407. 15. Li C, Yao M, Shao S, Li X, Li G, Wu R. Diagnostic efficacy of contrast- enhanced ultrasound for breast lesions of different sizes: a comparative study with magnetic resonance imaging. Br J Radiol. 2020;93:20190932. 16. Liu W, Zong M, Gong H-Y, Ling L-J, Ye X-H, Wang S, et al. Comparison of diagnostic efficacy between contrast-enhanced ultrasound and DCE-MRI for mass- and non-mass-like enhancement types in breast lesions. Cancer Manag Res. 2020;12:13567–78. 17. Pan J, Tong W, Luo J, Liang J, Pan F, Zheng Y, et al. Does contrast- enhanced ultrasound (CEUS) play a better role in diagnosis of breast lesions with calcification? A comparison with MRI. Br J Radiol. 2020;93:20200195. 18. Sun C, Sboros V, Butler MB, Moran CM. In vitro acoustic characterization of three phospholipid ultrasound contrast agents from 12 to 43 MHz. Ultrasound Med Biol. 2014;40:541–50. 19. Boca Bene I, Dudea SM, Ciurea AI. Contrast-enhanced ultrasonography in the diagnosis and treatment modulation of breast cancer. J Pers Med. 2021;11:81. 20. Sontum PC. Physicochemical characteristics of Sonazoid, a new contrast agent for ultrasound imaging. Ultrasound Med Biol. 2008;34:824–33. 21. Piskunowicz M, Sridharan A, Poznick L, Silvestro E, Hwang M. Optimiza- tion of mechanical indices for clinical contrast-enhanced ultrasound. J Ultrasound Med. 2021;40:1963–70. 22. Morris E, Comstock C, Lee C. ACR BI-RADS magnetic resonance imaging. ACR BI-RADS atlas, breast imaging reporting and data system. Reston, VA: American College of Radiology; 2013. 23. Sardanelli F, Boetes C, Borisch B, Decker T, Federico M, Gilbert FJ, et al. Magnetic resonance imaging of the breast: recommendations from the EUSOMA working group. Eur J Cancer. 2010;46:1296–316. 24. Lee YJ, Kim SH, Kang BJ, Kim YJ. Contrast-enhanced ultrasound for early prediction of response of breast cancer to neoadjuvant chemotherapy. Ultraschall Med. 2019;40:194–204. 25. Ewout WS. Clinical prediction models. Springer, Cham. 26. Xiao X, Dong L, Jiang Q, Guan X, Wu H, Luo B. Incorporating contrast- enhanced ultrasound into the BI-RADS scoring system improves accuracy Lang et al. Breast Cancer Research (2023) 25:61 Page 15 of 15 in breast tumor diagnosis: a preliminary study in China. Ultrasound Med Biol. 2016;42(11):2630–8. https:// doi. org/ 10. 1016/j. ultra smedb io. 2016. 07. 005. (Epub 2016 Aug 17 PMID: 27544439). 27. Luo B, Xiao X, Ou B, Yang H, Wu H. Breast contrast-enhanced ultrasound: is a scoring system feasible? A preliminary study in China. PLoS ONE. 2014;9(8):e105517. 28. Luo J, Chen J-D, Chen Q, Yue L-X, Zhou G, Lan C, et al. Predictive model for contrast-enhanced ultrasound of the breast: is it feasible in malignant risk assessment of breast imaging reporting and data system 4 lesions? World J Radiol. 2016;8:600–9. 29. Zhu C-R, Chen K-Y, Li P, Xia Z-Y, Wang B. Accuracy of multiparametric MRI in distinguishing the breast malignant lesions from benign lesions: a meta-analysis. Acta Radiol. 2021;62:1290–7. 30. Pan I-W, Oeffinger KC, Shih Y-CT. Cost-sharing and out-of-pocket cost for women who received MRI for breast cancer screening. J Natl Cancer Inst. 2022;114:254–62. 31. O’Flynn EAM, DeSouza NM. Functional magnetic resonance: biomarkers of response in breast cancer. Breast Cancer Res. 2011;13:204. 32. Luo J, Hippe DS, Rahbar H, Parsian S, Rendi MH, Partridge SC. Diffusion tensor imaging for characterizing tumor microstructure and improving diagnostic performance on breast MRI: a prospective observational study. Breast Cancer Res. 2019;21:102. 33. Bland JM, Altman DG. Statistics notes: bootstrap resampling meth- ods. BMJ. 2015;350:h2622. https:// doi. org/ 10. 1136/ bmj. h2622. (PMID: 26037412). Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1242_dev.201186	© 2023. Published by The Company of Biologists Ltd \| Development (2023) 150, dev201186. doi:10.1242/dev.201186 RESEARCH ARTICLE Female reproductive dormancy in Drosophila is regulated by DH31-producing neurons projecting into the corpus allatum Yoshitomo Kurogi1, Eisuke Imura2,3,, Yosuke Mizuno1, Ryo Hoshino1, Marcela Nouzova4,5, Shigeru Matsuyama6, Akira Mizoguchi7, Shu Kondo8,9, Hiromu Tanimoto10, Fernando G. Noriega4,11 and Ryusuke Niwa3,‡ ABSTRACT including the fruit Female insects can enter reproductive diapause, a state of suspended egg development, to conserve energy under adverse environments. fly, Drosophila melanogaster, In many insects, reproductive diapause, also frequently called reproductive dormancy, is induced under low-temperature and short-day conditions by the downregulation of juvenile hormone (JH) biosynthesis in the corpus allatum (CA). In this study, we demonstrate that neuropeptide Diuretic hormone 31 (DH31) produced by brain neurons that project into the CA plays an essential role in regulating reproductive dormancy by suppressing JH biosynthesis in adult D. melanogaster. The CA expresses the gene encoding the DH31 receptor, which is required for DH31-triggered elevation of intracellular cAMP in the CA. Knocking down Dh31 in these CA-projecting neurons or DH31 receptor in the CA suppresses the decrease of JH titer, normally observed under dormancy-inducing conditions, leading to abnormal yolk accumulation in the ovaries. Our findings provide the first molecular genetic evidence demonstrating that CA-projecting peptidergic neurons play an essential role in regulating reproductive dormancy by suppressing JH biosynthesis. 1Degree Programs in Life and Earth Sciences, Graduate School of Science and Technology, University of Tsukuba, Tennodai 1-1-1, Tsukuba, Ibaraki 305-8572, Japan. 2Graduate School of Life and Environmental Sciences, University of Tsukuba, Tennodai 1-1-1, Tsukuba, Ibaraki 305-8572, Japan. 3Life Science Center for Survival Dynamics, Tsukuba Advanced Research Alliance (TARA), University of Tsukuba, Tennodai 1-1-1, Tsukuba, Ibaraki 305-8577, Japan. 4Department of Biological Sciences and BSI, Florida International University, 11200 SW 8th street, Miami, FL 33199, USA. 5Institute of Parasitology, Biology Center of the Academy of Sciences of the Czech Republic, 37005, Č eské Budě jovice, Czech Republic. 6Faculty of Life and Environmental Sciences, University of Tsukuba, Tennodai 1-1-1, Tsukuba, Ibaraki 305-8572, Japan. 7Division of Liberal Arts and Sciences, Aichi Gakuin University, 12 Araike, Iwasaki-cho, Nisshin, Aichi 470-0195, Japan. 8Department of Biological Science and Technology, Faculty of Advanced Engineering, Tokyo University of Science, Niijuku 6-3-1, Katsushika-ku, Tokyo 125-8585, Japan. 9Invertebrate Genetics Laboratory, National Institute of Genetics, Yata 111, Mishima, Shizuoka 411-8540, Japan. 10Graduate School of Life Sciences, Tohoku University, Katahira 2-1-1, Sendai, Miyagi 980-8577, Japan. 11Department of Parasitology, University of South Bohemia, Č eské Budě jovice 37005, Czech Republic. Present address: Department of Entomology, Institute for Integrative Genome Biology, University of California, Riverside, 900 University Ave, Riverside, CA 92521, USA. ‡Author for correspondence (ryusuke-niwa@tara.tsukuba.ac.jp) Y.K., 0000-0003-3286-5407; E.I., 0009-0001-1008-0847; Y.M., 0009-0001-9529- 8844; R.H., 0000-0002-0943-3131; M.N., 0000-0002-4562-1383; S.M., 0000-0003- 2673-5673; S.K., 0000-0002-4625-8379; H.T., 0000-0001-5880-6064; F.G.N., 0000-0001-8628-5374; R.N., 0000-0002-1716-455X This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed. Handling Editor: Irene Miguel-Aliaga Received 3 August 2022; Accepted 12 April 2023 KEY WORDS: Reproductive dormancy, Diapause, Juvenile hormone, Corpus allatum, Diuretic hormone 31, Drosophila INTRODUCTION Unfavorable seasonal changes for prolonged durations, such as extremely low temperatures and food shortages during winter, may be challenging for the survival of animals in temperate zones. During such unfavorable conditions, organisms often suspend or retard their normal development, growth and physiological functions, a process known as diapause (Hand et al., 2016). Diapause has been intensively studied in insects as its control could benefit many aspects of industry and agriculture (Denlinger, 2008, 2022; Denlinger et al., 2012). Previous studies have revealed that a fraction of insects enter diapause at the adult stage under diapause-inducing conditions, leading to multiple metabolic and behavioral changes, including slowed or halted reproductive maturation known as reproductive diapause (Denlinger, 2022; Hutfilz, 2022; Kurogi et al., 2021). As female insect adults allocate extensive energy resources toward oogenesis (Wheeler, 2003), reproductive diapause allows them to reduce their energy consumption and then resume reproduction under more favorable conditions. Reproductive diapause in insects is regulated by a complex interplay between multiple hormones and neurotransmitters (Denlinger, 2022; Denlinger et al., 2012; Kurogi et al., 2021). Among these, reduction in the titer of juvenile hormone (JH), an insect-specific sesquiterpenoid hormone (Qu et al., 2018; Riddiford, 2020), has been extensively studied, revealing its vital role in regulating reproductive diapause in female adult insects (Kurogi et al., 2021). As JHs are essential for promoting vitellogenesis under non-diapause-inducing conditions, a reduction in hemolymph JH leading to levels is required for suppressing vitellogenesis, reproductive diapause in females (Denlinger et al., 2012; Santos et al., 2019). In many insect species, the reduction of hemolymph JH levels correlates with the downregulation of JH biosynthesis in a specialized endocrine organ called the corpus allatum (CA) (Denlinger et al., 2012; Hand et al., 2016; Kurogi et al., 2021). Previous studies have demonstrated that surgical amputation of the nervous connection between the brain and the CA impairs the induction of reproductive diapause (Hodková, 1976; Kotaki and the presence of a Yagi, 1989). These observations suggest mechanism of signal transduction by which information about unfavorable environmental conditions is processed in the brain and transmitted to the CA to reduce JH biosynthesis. Previous studies have demonstrated that brain neurons projecting to the CA play crucial roles in the induction of reproductive diapause in multiple insect species (Denlinger, 2002, 2022; Denlinger et al., 2012). For example, some neurons with their cell bodies in the anterior midline of the brain ( pars intercerebralis; PI), project to the CA in the common ground-hopper (Tetrix undulata; 1 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 Poras, 1982), linden bug (Pyrrhocoris apterus; Hodková, 1976), blowfly (Protophormia terraenovae; Shiga and Numata, 2000; Shiga et al., 2000) and brown-winged green bug (Plautia stali; Matsumoto et al., 2013). A study reported that specific PI neurons in P. stali produce P. stali myoinhibitory peptide (Plast-MIP), a neuropeptide that inhibits JH biosynthesis in CA under diapause- inducing conditions (Hasebe and Shiga, 2021b; Hasegawa et al., 2020; Matsumoto et al., 2017). It has also been demonstrated that a group of neurons, the cell bodies of which are located on the lateral side of the brain ( pars lateralis; PL), project to the CA in P. terraenovae (Shiga and Numata, 2000), the bean bug (Riptortus pedestris; Shimokawa et al., 2008), the Colorado potato beetle (Leptinotarsa decemlineata; de Wilde and de Boer, 1969) and the grasshopper (Locusta migratoria; Poras et al., 1983). In all these cases, PL neurons appear to play an inhibitory role in ovarian development under diapause-inducing conditions. Moreover, cauterization experiments in L. decemlineata and L. migratoria have suggested that the PL plays an inhibitory role in JH biosynthesis under diapause-inducing conditions (Khan et al., 1986; Poras et al., 1983). In P. terraenovae, specific PL neurons projecting to the CA including cholecystokinin-8 and FMRF- produce neuropeptides, amide (Hamanaka et al., 2007, 2009). However, despite their crucial role in regulating reproductive diapause, the role of CA-projecting PL neurons in reproductive diapause has not been genetically confirmed. Moreover, it is unclear which neurotransmitters or neuropeptides in CA-projecting PL neurons are responsible for regulating reproductive diapause in insect species. for the regulation of The fruit fly, Drosophila melanogaster, also undergoes the suppression of oogenesis in low-temperature and short-day conditions (Saunders et al., 1989). In the most recent pieces of literature, the reproductive diapause-like status of D. melanogaster is frequently called ‘reproductive dormancy’ (see Kurogi et al., 2021 for detailed discussion). Therefore, in accordance with recent studies, we will use the term ‘reproductive dormancy’ to refer to suppressing oogenesis under diapause-inducing conditions. Studies on D. melanogaster reproductive dormancy with powerful genetic tools have contributed substantially to the discovery of several regulatory mechanisms of insect reproductive dormancy. Several studies have confirmed that JH biosynthesis and signaling are essential reproductive dormancy in D. melanogaster (Andreatta et al., 2018; Ojima et al., 2018; Saunders et al., 1990). In this study, we report the identification and characterization of CA-projecting PL neurons that produce the neuropeptide Diuretic hormone 31 (DH31). DH31 is known to play versatile roles in insects, particularly in D. melanogaster, such as diuretic action, establishment of daily temperature preference rhythms, locomotor activity, sleep regulation, midgut contraction immune responses and nutrient-dependent frequency, regulation of courtship (Benguettat et al., 2018; Furuya et al., 2000; Goda et al., 2016, 2018, 2019; Head et al., 2015; Kaneko et al., 2012; Kunst et al., 2014; Lin et al., 2022; Vanderveken and O’Donnell, 2014). In this study, we propose that DH31 signaling to the CA plays an important role in the inhibition of JH biosynthesis under dormancy-inducing conditions, leading to the induction of reproductive dormancy through the inhibition of JH-mediated maturation of eggs. intestinal RESULTS A subset of DH31-producing neurons projects into the D. melanogaster CA In our previous studies, we have identified several neurons including the CA of projecting to various endocrine organs, D. melanogaster (Imura et al., 2020; Mizuno et al., 2021; Shimada- Niwa and Niwa, 2014). To further identify CA-projecting neurons, we conducted immunostaining experiments using antibodies against different neuropeptides, and investigated whether any of these peptidergic-cell neuronal processes innervate the CA, as described previously for the hugin-positive neurons (Mizuno et al., 2021). Our results demonstrated that DH31-immunoreactive fibers and varicosities were present within the CA (Fig. 1A,B). A trace back of these CA-projecting DH31-producing neurons (Fig. 1A) led to three pairs of cell bodies located on the dorsal side of the central brain (Fig. 1A,C). Among these, two cell bodies in the brain hemisphere were clustered, whereas another cell body was located closer to the midline. We also found that the R21C09-GAL4 driver, in which the GAL4 transgene is expressed under the control of part of a Dh31 enhancer, labeled these CA-projecting neurons. The R21C09-GAL4-driven GFP signals were observed in three pairs of the central brain neurons and neuronal processes in the CA region, which were co-immunostained with an anti-DH31 antibody (Fig. S1A,B). A previous study identified one pair of CA-projecting lateral protocerebrum (CA-LP) 1 neurons and two pairs of CA-LP2 neurons, which innervate the CA in D. melanogaster larvae (Ádám et al., 2003; Siegmund and Korge, 2001). However, the neurotransmitters produced by these neurons remain unknown. As the positions of the cell bodies of the DH31-producing CA- projecting neurons were similar to those of CA-LP1 and CA-LP2 neurons, we examined whether the CA-LP1 and CA-LP2 neurons produced DH31. We found that Kurs21-GAL4-driven GFP, which reportedly marks CA-LP1 and CA-LP2 neurons (Fig. S1C; the Siegmund and Korge, 2001), neurites to the CA in the wandering third-instar larvae (Fig. S1D) as well as in adults (Fig. S1E,F). Moreover, the cell bodies and CA- projecting neurites of CA-LP1 and CA-LP2 neurons were immunostained with the anti-DH31 antibody in both larval and adult stages (Fig. S1C-F). Therefore, the DH31-producing-CA- projecting neurons in the D. melanogaster adult stage are identical to the CA-LP1 and CA-LP2 neurons described in its larval stage. Hereafter, we designate these three pairs of CA-projecting DH31- producing neurons, corresponding to both CA-LP1 and CA-LP2 neurons, as ‘CA-LP neurons’. labeled the projection of R21C09-GAL4 and Kurs21-GAL4 were active in many neurons in the central brain, in addition to the CA-LP neurons (Fig. S1G,H). We thus searched for a GAL4 driver that would be active in a restricted group of neuronal cells including the CA-LP neurons. We browsed large collections of adult fly images from the FlyLight database of the Janelia Research Campus, Howard Hugh Medical Institute (https://flweb.janelia.org/cgi-bin/flew.cgi). The FlyLight database provides large anatomical image datasets and a well- characterized collection of GAL4 lines, allowing us to visualize individual neurons in the D. melanogaster central nervous system (Jenett et al., 2012). We manually checked most of the images of GMR GAL4 lines deposited in FlyLight and found that R94H10- GAL4 labeled CA-LP neurons and a few additional neurons (Fig. 1D,E; Fig. S1I,J). In the central brain, R94H10-GAL4- positive neurons other than CA-LP neurons were DH31-negative (Fig. S1I). In the ventral nerve cord (VNC), in which some neurons are known to express Dh31 (Mandel et al., 2018), almost all R94H10-GAL4-positive neurons were DH31-negative, while a few cells were DH31-immunoreactives (Fig. S1J). However, as described later, we had evidence indicating that R94H10-GAL4- and DH31-double positive cells are not involved in reproductive in dormancy. Furthermore, although Dh31 expressed is 2 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 Fig. 1. Anatomical characterization of CA-LP neurons. (A) Immunostaining with anti-DH31 antibody in the adult central brain (outlined by dashed lines) and the corpus allatum (CA) of wild-type (w1118) adult females. The upper and lower parts of the photograph correspond to the dorsal and ventral sides of the central brain, respectively. Axonal processes from the cell bodies to the CA (large arrowhead) are indicated by the small arrowheads. Small arrows indicate the cell bodies of CA-LP neurons. (B) Immunostaining signal in the CA of wild-type (w1118) adult females 4 days after eclosion. The anti-DH31 antibody (green) was employed along with the anti- JHAMT antibody, which was used to visualize the CA (magenta). DH31-positive puncta are observed in the CA region. Note that DH31-immunoreacive varicosities were observed inside the CA. Dashed line indicates the outline of the CA. (C) A schematic representing the anatomy of brain, CA-LP neurons and CA. (D) Transgenic visualization of CA-LP neurons by GFP driven by R94H10-GAL4, which specifically labels CA-LP neurons and other small subsets of neurons. Samples were immunostained with anti- GFP (green) and anti-JHAMT (blue) antibodies. The CA is marked with a large arrowhead. Dashed line indicates the outline of the brain. Small arrows indicate the cell bodies of CA-LP neurons. (E) Immunostaining signal with anti-GFP (green) and anti-DH31 (magenta) antibodies in the brain region, including the soma of CA-LP neurons (arrows), and in the CA region in an adult female R94H10-GAL4 UAS-GFP UAS-mCD8::GFP. enteroendocrine cells (Veenstra et al., 2008), R94H10-GAL4 was not active in these gut cells (Fig. S1K). Therefore, in this study, we used the GAL4 driver to manipulate gene expression in CA-LP neurons. Thereafter, we examined the distribution of axonal termini and dendrites in the CA-LP neurons. Synaptotagmin::GFP (Syt::GFP) and DenMark transgenes, the translated products of which were localized at the axonal termini and dendrites respectively (Nicolaï et al., 2010; Zhang et al., 2002), were driven by R94H10-GAL4. We found that the Syt::GFP and DenMark signals were primarily observed in the CA region and in the dorsal side of the central brain respectively (Fig. 2A), suggesting that the DH31-immunoreactive fibers in the CA regions are axons. Furthermore, GFP reconstitution across synaptic partners (GRASP) analysis (Feinberg et al., 2008), in which two complementary fragments of GFP were expressed in CA-LP neurons and the CA, respectively, revealed that GRASP Fig. 2. Axonal termini of CA-LP neurons physically contact the CA. (A) (Left) Visualization of axons and dendrites of CA-LP neurons stained by synaptotagmin-GFP (SytGFP, green) and DenMark (magenta) driven by R94H10-GAL4. The CA (large arrowhead) is visualized by immunostaining with anti-JHAMT (blue). Dashed line indicates the outline of the brain. (Right) Magnified view of the left panel focusing on the CA region. SytGFP, but not DenMark, signals were observed in the CA region, indicating that axons of CA-LP neurons innervate the CA. Small arrows indicate the cell bodies of CA-LP neurons. Dashed line indicates the outline of the CA. (B) The GRASP signal was employed to visualize the close physical contact between CA-LP neurons and the CA. (C,D) GFP signals of GRASP-negative control adult females. (C) GRASP signals in the CA region of a virgin female expressing GAL4 in CA-LP neurons but not LexA in the CA (+/+; R94H10-GAL4/UAS-CD4::spGFP1-10 LexAOP>CD4::spGFP10). (D) GRASP signals in the CA region of a virgin female expressing LexA in the CA but not GAL4 in CA- LP neurons (+/JHAMT-LexA; +/UAS-CD4::spGFP1-10 LexAop- CD4::spGFP10). Dashed line indicates the outline of the CA. T N E M P O L E V E D 3 RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 signals were detected inside the CA (Fig. 2B-D). These results suggest that CA-LP neurons project to the CA and are in physical contact with the CA. A subset of circadian clock neurons innervates the CA-LP neurons To anatomically characterize CA-LP neurons, we browsed the publicly available connectome database of adult D. melanogaster brains to search for neurons that connect upstream of the CA-LP neurons (see Materials and Methods). In the available connectome data of neurons labeled with R21C09-GAL4, the two pairs of clustered CA-LP neurons have been annotated (Fig. S2A-D). The anatomical positions of the cell bodies and axons of these neurons suggested that they correspond to the CA-LP2 neurons (Fig. S1C,D; Siegmund and Korge, 2001). In contrast, another pair of CA-LP neurons that was similar to CA-LP1 neurons was not clearly identified in the available database. Therefore, in our connectome analysis, we focused on CA-LP2 neurons that exhibited 1780 connections with other neurons. Among these neurons, we found that several clock neurons, including the fifth sLN-v, DN1p, LNd and s-LNv, were connected to the CA-LP2 neurons (Fig. S2E). These results suggest that the circadian clock may modulate the function of the CA-LP neurons. s-LNvs are involved in regulating reproductive dormancy in D. melanogaster (Nagy et al., 2019). Moreover, the s-LNvs innervate neurons located in the PL, which closely corresponds to the LP region in the brain in which the CA projecting neurons are located in several insects, including the P. terraenovae (Hamanaka et al., 2005; Meuti and Denlinger, 2013; Shiga and Numata, 2009). Therefore, we examined the relationship between s-LNvs and CA-LP neurons in D. melanogaster. The morphology of s-LNvs can (PDF) be visualized using an anti-pigment dispersing factor antibody (Helfrich-Förster, 1995). We found that the axonal termini of s-LNvs were in close proximity to the neuronal processes of CA-LP neurons (Fig. S2F,G). We initially expected that the PDF receptor (Pdfr) would be present in CA-LP neurons; however, Pdfr knock-in T2A GAL4 was absent in these neurons (Fig. S2H). In contrast, the short neuropeptide F (sNPF), which is known to be present in s-LNvs (Johard et al., 2009), might be involved, as the s-NPF receptor (s-NPF-R) knock-in T2A GAL4 was positive in these neurons (Fig. S2I). These results suggest that CA-LP neurons, to an extent, have chemical connections with s-LNvs, at least through sNPF. The CA-LP neurons are required for inducing reproductive dormancy Considering that CA-LP neurons are located in the lateral protocerebrum and in proximity to the axon termini of s-LNvs, the morphological and anatomical features of D. melanogaster CA- LP neurons are substantially similar to those of P. terraenovae CA- projecting neurons (Meuti and Denlinger, 2013; Shiga et al., 2000). A previous study reported that CA-projecting neurons negatively regulate the induction of reproductive dormancy in P. terraenovae, as surgical amputation of the axons of CA-projecting neurons caused abnormal egg production, even under dormancy-inducing conditions (Matsuo et al., 1997). Therefore, we examined whether CA-LP neurons were also involved in reproductive dormancy in D. melanogaster. We first examined whether genetic mutants with Dh31 loss of function displayed any phenotypes of reproductive dormancy in virgin females. We confirmed that DH31 immunoreactivity in the CA region was diminished in the Dh31 loss-of-function flies (Fig. 3A), suggesting that DH31 is not active in the CA of mutant insects. Furthermore, under dormancy-inducing conditions (at 11±0.5°C under 10 light/14 h dark cycle), Dh31 loss of function in females led to significant enlargement of the ovaries compared with those in control females that exhibited the typical dormancy-induced reduction of ovarian development (Fig. 3B). Finally, Dh31 loss-of-function mutations in females led to mature egg production compared with the control females (Fig. 3C). We next examined whether the DH31 peptide, produced in CA-LP neurons, was a crucial regulator of reproductive dormancy. For this purpose, we knocked down Dh31 specifically in CA-LP neurons using a transgenic RNAi technique. An R94H10-GAL4- driven Dh31 inverted repeat (IR) construct eliminated almost all the DH31 protein in the CA (Fig. 3D). Under dormancy-inducing conditions, RNAi-treated females displayed greater ovarian development and higher production of mature eggs than the control females (Fig. 3E,F). We also investigated whether loss of Dh31 function affected mature egg formation under non-dormancy-inducing conditions. However, in this case, we did not observe any difference in mature egg production between control and Dh31 genetic mutant females (Fig. S3A) or between control and R94H10-GAL4-driven Dh31 RNAi animals (Fig. S3B). These results suggest that DH31 is dormancy-inducing oogenesis necessary conditions, but not under non-dormancy-inducing conditions. during inhibit to As we mentioned earlier, a few R94H10-GAL4-positive neurons in the VNC were DH31-positive (Fig. S1J; Fig. S3C). To exclude the possibility that these DH31-positive VNC neurons are involved in regulating reproductive dormancy, we used teashirt (tsh)-GAL80 to silence GAL4 activity in the VNC (Simpson, 2016) in combination with R94H10-GAL4 to knockdown Dh31 using RNAi. We confirmed that the GAL4 activity in R94H10-GAL4; tsh-GAL80 flies was efficiently silenced in the R94H10-GAL4- and DH31-double positive VNC neurons (Fig. S3D). Regarding the DH31 immunoreactivity in the CA and the number of mature eggs, R94H10-GAL4; tsh-GAL80-driven Dh31-IR flies exhibited almost identical phenotypes to R94H10-GAL4-driven Dh31-IR flies without tsh-GAL80 (Fig. 3G-I). These results suggest that DH31- positive VNC neurons are not involved in regulating reproductive dormancy. We implemented an additional genetic approach, in which a Dh31 transgene was driven specifically in CA-LP neurons in the genetic background of the Dh31 loss-of-function mutation. An R94H10-GAL4-driven Dh31 overexpression restored DH31 protein levels in the CA of the mutant females (Fig. 3J). Although loss of DH31 function in females failed to induce reproductive dormancy, reproductive Dh31 overexpression rescued the phenotypes of dormancy (Fig. 3K,L). Lastly, we examined the sufficiency of CA-LP neurons to suppress oogenesis. For this purpose, we overexpressed TrpA1, a temperature-sensitive cation channel gene (Hamada et al., 2008), in CA-LP neurons under non-dormancy-inducing conditions. Both the control and TrpA1-overexpressing virgin females at permissive temperature (21°C) had the normal number of mature eggs. On the other hand, at the TrpA1- overexpressing virgin females had less mature eggs compared with controls (Fig. S3E), suggesting that CA-LP neuronal activity is sufficient these results support the idea that DH31 produced by CA-LP neurons plays an essential role in the suppression of oogenesis under dormancy- inducing conditions. to suppress oogenesis. Taken together, the restrictive temperature (29°C), 4 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 Fig. 3. The loss of DH31 function in CA-LP neurons fails to induce reproductive dormancy. (A) Immunostaining signals of anti-DH31 antibody (green) in the CA region of control and Dh31 genetic mutant females. CA was visualized with an anti-JHAMT antibody (magenta). Note that DH31 puncta were diminished in the CA region of Dh31 trans-heterozygous mutant females. (B,C) Mature egg formation in control and Dh31 genetic mutant females under dormancy-inducing conditions. Representative images of the ovaries (B) and quantification of mature eggs per ovary (C). (D-I) Phenotypes of R94H10-GAL4 driven Dh31 RNAi adult females in the absence (D-F) and presence (G-I) of tsh-GAL80 transgene. (D,G) Immunostaining signals of the anti-DH31 antibody (green) in the CA region of control and Dh31 RNAi adult females. CA was visualized using an anti-JHAMT antibody (magenta). DH31 puncta were diminished in the CA region of Dh31 RNAi animals. (E,F,H,I) Mature egg formation in control and Dh31 RNAi females under dormancy-inducing conditions. Representative images of the ovaries (E,H) and quantification of mature eggs per ovary (F,I). (J) Immunostaining signals of the anti-DH31 antibody (green) in the CA region of Dh31 trans heterozygous mutant females in the presence or absence of CA-LP neuron-specific expression of Dh31 cDNA under dormancy- inducing conditions. DH31 puncta were recovered in the CA region of the transgenic rescue animals. (K,L) Mature egg formation in Dh31 trans-heterozygous mutant females with or without CA-LP neuron-specific expression under dormancy-inducing conditions. Representative images of the ovaries (K) and quantification of mature eggs per ovary (L). (M,N) Anti-DH31 immunoreactivity in the soma of CA-LP1 and CA-LP2 neurons between non-dormancy- and dormancy-inducing conditions. w1118 was used for the analysis. Representative images of anti-DH31 immunostaining signals in the soma of CA-LP1 and CA- LP2 neurons (M) and quantitative comparison of anti-DH31 immunoreactivity in the soma of CA-LP1 and CA-LP2 neurons (N). Small arrows indicate the cell bodies of CA-LP neurons. *P<0.001 (Wilcoxon rank sum test with Bonferroni’s correction for C,F,I and L; unpaired two-tailed Student’s t-test for N.). n.s., not significant. Box plots show the median values (middle bars) along with the first to third interquartile ranges (boxes). The whiskers represent 1.5 times the interquartile ranges. Samples were derived from virgin females 12 days after eclosion under dormancy-and non-dormancy-inducing conditions. Dashed line indicates the outline of the CA. T N E M P O L E V E D 5 RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 Comparison of Dh31 expression, DH31 protein levels and CA- LP neuronal activity between dormancy- and non-dormancy- inducing conditions To understand whether and how CA-LP neurons are regulated in response to dormancy-inducing conditions, we compared Dh31 expression and DH31 protein levels between dormancy- and non- dormancy-inducing conditions. Levels of Dh31 expression and DH31 protein in CA-LP neurons were confirmed by the Dh31 enhancer (R21C09)-GAL4 and by immunostaining with an anti- DH31 antibody, respectively. We found that Dh31 expression levels were upregulated in CA-LP1 neurons but remained unchanged in CA-LP2 neurons under dormancy-inducing conditions (Fig. S4A). In contrast, DH31 protein levels were notably reduced in both CA- LP1 and CA-LP2 neurons under dormancy-inducing conditions (Fig. 3M,N). These results might reflect enhancement of Dh31 transcription, at least in CA-LP1 neurons, and DH31 release from CA-LP neurons in response to dormancy-inducing conditions. To further investigate whether dormancy-inducing conditions affect the function of CA-LP neurons, we conducted a transcriptional reporter of intracellular Ca2+ (TRIC) assay, which is used to assess neuronal activities (Gao et al., 2015). However, we found no obvious elevation of TRIC signals, corresponding to intracellular Ca2+ levels, in CA-LP neurons under dormancy-inducing conditions (Fig. S4B,C). Therefore, although we have demonstrated that CA- LP neurons are regulated in response to dormancy-inducing conditions, the exact mechanism of this regulation is still unclear. The DH31 receptor in the CA is required for the induction of reproductive dormancy As CA-LP neurons project to and are in close contact with the CA, we examined whether a specific receptor for DH31 (DH31-R) (Johnson et al., 2005) was present in the CA. For this purpose, we generated a knock-in strain by inserting the T2A-GAL4 cassette into the Dh31-R locus (49). This transgenic line was used to confirm Dh31-R expression in the CA (Fig. 4A), and the expression levels of Dh31-R in the CA were not different between dormancy- and non-dormancy-inducing conditions (Fig. S5A). We then conducted phenotypic analyses of the genetic mutants that presented a loss of DH31-R function. Under dormancy-inducing conditions, Dh31-R mutant females produced more mature eggs compared with controls (Fig. 4B,C). We also took an approach using transgenic RNAi to knock down Dh31-R in the CA. Forced expression of two independent Dh31-R IR constructs resulted in ∼60% and 80% reduction of Dh31-R mRNA levels, respectively (Fig. S5B). We observed reproductive phenotypes in females in which Dh31-R was knocked down in the CA with Aug21-GAL4, known to be active in the CA (Ádám et al., 2003; Siegmund and Korge, 2001) (Fig. 4D,E). Aug21-GAL4-driven RNAi phenotypes were observed using two independent Dh31-R RNAi lines (Fig. 4F,G). In addition, JH acid O-methyltransferase (JHAMT)-GAL4-driven Dh31-R RNAi also exhibited reproductive phenotype under dormancy-inducing conditions (Fig. S5C). These results confirm that DH31-R in the CA is required for the induction of reproductive dormancy. Recent studies have revealed the essential role of the insulin- producing neurons (IPCs) in regulating reproductive dormancy (Kubrak et al., 2014; Kurogi et al., 2021; Nagy et al., 2019; Ojima et al., 2018). We therefore wondered whether DH31-R was also present and played an essential role in the IPCs. However, Dh31-R- T2A-GAL4-driven GFP signal was not observed in the IPCs marked by the presence of Drosophila insulin-like peptide 2 (DILP2; ILP2) (Fig. S5D). In addition, when we knocked down Dh31-R using Dilp2-GAL4-driven transgenic RNAi, mature egg production was not enhanced in the RNAi animals compared with control under dormancy-inducing conditions (Fig. S5E,F). These results suggest that DH31-R acts in the CA rather than in the IPCs to regulate reproductive dormancy. We also examined whether mature egg formation was impaired by loss of Dh31-R function under non-dormancy-inducing conditions. However, under non-dormancy-inducing conditions, either Dh31-R genetic mutant females or CA-specific Dh31-R RNAi females produced a number of mature eggs comparable with control animals (Fig. S5G-I). Therefore, these results suggest that DH31-R is required for suppressing oogenesis under dormancy-, but not in non-dormancy-inducing conditions. DH31 induces intracellular cAMP elevation in the CA through the DH31 receptor We employed live imaging techniques to further understand the role of DH31-R in transmitting DH31 signals to CA cells. Considering that DH31-R is coupled to cAMP as a secondary messenger (Johnson et al., 2005), we monitored the increase in intracellular cAMP levels using a CA ex vivo culture system that assays the activity of the cAMP sensor probe, Pink Flamindo (Harada et al., 2017). Fly tissues containing the CA expressing the Pink Flamindo transgene were dissected and treated with collagenase because the CA is surrounded by a thick layer of collagens (Fig. S6A). The collagenase-treated tissue was placed in a glass-bottomed dish filled with culture medium (Fig. 5A; Fig. S6B); subsequently, the Pink Flamindo signals were recorded in the presence and absence of extracellular chemical stimuli. We first confirmed that the Pink Flamindo signals reflected intracellular cAMP levels, as the administration of the adenylyl cyclase activator NKH477 elevated Pink Flamindo signals in the CA (Fig. S6C,D). Using this ex vivo system, we found that the administration of the synthetic DH31 peptide resulted in the prompt elevation of Pink Flamindo signals in the CA of both non-dormant flies (Fig. 5B,C) and dormant flies (Fig. 5D,E). Importantly, the DH31-stimulated elevation of Pink Flamindo signals was not observed in flies with the Dh31-R knocked down in the CA (Fig. 5F,G). These results indicate that DH31 signals, via DH31-R, lead to an increase in intracellular cAMP levels in CA cells. We also took another approach using a heterologously-expressed P2X2 purinoreceptor, which can activate neurons by bath application of ATP (Yao et al., 2012), to investigate whether the synaptic transmission of DH31 to the CA increases intracellular cAMP in the CA. In this case, we prepared adult females expressing P2X2 driven by R18G01-LexA, which labeled CA-LP neurons (Fig. S6E), along with JHAMT-GAL4-driven Pink Flamindo. After the brain-CA complexes were dissected out with their intact neuronal connections, we applied ATP to the medium, which should stimulate CA-LP neurons and Pink Flamindo signals in the CA. We observed a substantial, transient elevation of Pink Flamindo signals in the CA after ATP administration compared with control vehicle administration (Fig. 5H,I), although the difference between the groups did not reach statistical significance (Fig. 5I). We should note that the preparation of the brain-CA complex was technically challenging, at least in our hands. We also found that, in the brain- complexes after Pink Flamindo imaging, DH31 protein levels in CA-LP neurons were reduced in the ATP-treated samples compared with the vehicle-treated ones (Fig. 5J,K). This observation is consistent with our data showing that DH31 protein levels in CA-LP neurons were reduced in dormancy-inducing conditions (Fig. 3M,N). Therefore, our data suggest, to some extent, the involvement of the synaptic DH31 transmission in the elevation of intracellular cAMP in the CA. 6 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 Fig. 4. The loss of DH31 receptor function in the CA leads to failure in the induction of reproductive dormancy. (A) Expression of Dh31-R visualized using Dh31-R-T2A- GAL4-driven GFP (green). CA was revealed using an anti- JHAMT antibody (magenta). Dashed line indicates the outline of the CA. (B,C) Mature egg formation in control and Dh31-R mutant females under dormancy-inducing conditions. Representative images of the ovaries (B) and quantification of mature eggs per ovary (C). (D,E) Mature egg formation in the control and CA-specific Dh31-R RNAi females under dormancy-inducing conditions. Representative images of the ovaries (D) and quantification of mature eggs per ovary (E). Note that these data were obtained from the Vienna Drosophila Resource Center KK RNAi strain. (F,G) Mature egg formation in the control and CA-specific Dh31-R RNAi females under dormancy-inducing conditions. These data were obtained from the transgenic RNAi strain from the National Institute of Genetics. Representative images of the ovaries (F) and quantification of mature eggs per ovary (G). P<0.001 (Wilcoxon rank-sum test with Bonferroni correction for C,E,G). n.s., not significant. Box plots show the median values (middle bars) along with the first to third interquartile ranges (boxes). The whiskers represent 1.5 times the interquartile ranges. Samples were derived from virgin females 12 d after their transfer to dormancy-inducing conditions. Based on the results above, we hypothesized that the increased intracellular cAMP in the CA should lead to a suppression of JH biosynthesis, inducing reproductive dormancy. To test the hypothesis, we used a dominant negative form of protein kinase A (PKADN), which is the major effector protein downstream of cAMP. Consistent with the fact that a forced expression of PKADN in the CA leads to enhancing JH signaling (Andreatta et al., 2018), JHAMT-GAL4-driven PKADN expression resulted in producing more mature eggs under dormancy-inducing conditions (Fig. S7A,B). These results, along with those of a previous study (Andreatta et al., 2018), suggest that the increased level of intracellular cAMP in the CA negatively regulates oogenesis via the suppression of JH biosynthesis. DH31 signaling in the CA plays a role in decreasing hemolymph JH titers The induction of D. melanogaster reproductive dormancy is associated with the suppression of JH biosynthesis in the CA (Andreatta et al., 2018; Ojima et al., 2018; Saunders, 1990; Tatar and Yin, 2001; Tatar et al., 2001); therefore, we examined whether DH31 signaling in the CA influenced JH titers in the hemolymph. We used liquid chromatography coupled with electrospray tandem mass spectrometry (LC-MS/MS) (Ramirez et al., 2020) to measure hemolymph titers of JH III, a major form of JH in D. melanogaster adult females (Bownes and Rembold, 1987; Reiff et al., 2015; Sliter et al., 1987; Sugime et al., 2017). Females with Dh31 or Dh31-R loss of function exhibited a higher hemolymph JH III titer than the control females (Fig. 6A,B). In addition, under dormancy-inducing conditions, administration of the JH analog, methoprene, induced egg maturation, which phenocopied females with Dh31 loss of function (Fig. S7C,D). JH stimulates yolk protein synthesis in the fat body to promote mature egg production (Bownes, 1989; Flatt et al., 2005; Postlethwait and Weiser, 1973). Consistently, the mRNA levels of yolk protein 1, yolk protein 2 and yolk protein 3 were upregulated in females with loss of Dh31-R function (Fig. 6C; Fig. S7E,F). Altogether, these results suggest that DH31 signaling in the CA plays a crucial role in suppressing the hemolymph JH titer under dormancy-inducing conditions, leading to reproductive dormancy through the inhibition of JH-mediated oocyte maturation (Fig. 6D). DISCUSSION CA-projecting PL neurons are essential regulators of reproductive dormancy in insects (de Wilde and de Boer, 1969; Poras et al., 1983; Shiga and Numata, 2000; Shimokawa et al., 2008). In addition, CA-projecting PL neurons are known to play a crucial inhibitory role in JH biosynthesis (Khan et al., 1986; Poras et al., 1983). The current study provides the first genetic evidence of the role of CA-projecting PL neurons in JH-mediated reproductive dormancy. Our results suggest that the CA-LP neuron-CA axis modulates JH biosynthesis in response to dormancy-inducing conditions, revealing a regulatory neuroendocrine mechanism underlying reproductive dormancy. Our study confirmed that the DH31-producing CA-projecting neurons are identical to CA-LP neurons, which have been anatomically characterized in D. melanogaster larvae (Siegmund and Korge, 2001). Previous studies have indicated that CA-LP neurons have a negative effect on JH biosynthesis during pupal development, permitting a T N E M P O L E V E D 7 RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 Fig. 5. Synthetic DH31 peptide elevates intracellular cAMP level in the CA through the DH31 receptor. (A) A schematic showing ex vivo cAMP imaging. The dissected tissues, containing the brain, gut and CA, were incubated in Schneider’s Drosophila medium in the presence or absence of synthetic DH31 peptide, ATP or NKH477. (B-E) Pink Flamindo imaging in the CA, dissected out from non-dormant (B,C) and dormant (D,E) adult females, with or without DH31 peptide administration. (B,D) Changes in the relative fluorescence intensity of JHAMT-GAL4-driven Pink Flamindo in the CA with or without stimulation using 0.3 μM DH31 peptide (each n=6). F/F0 values are normalized by the signal intensity of JHAMT-GAL4-driven mCD8::GFP. (C,E) Quantification of normalized Pink Flamindo signals in the CA at 120 s after the stimulation. (F,G) Pink Flamindo imaging in the CA, dissected out from control and Dh31-R RNAi non-dormant adult females, with or without DH31 peptide administration. (F) Changes in the relative fluorescence intensity of JHAMT- GAL4-driven Pink Flamindo in the CA of control and Dh31-R RNAi females with stimulation using 0.3 μM DH31 peptide (each n=6). F/F0 values are normalized by the signal intensity of JHAMT-GAL4-driven mCD8::GFP. (G) Quantification of normalized Pink Flamindo signals in the CA at 120 s after the stimulation. (H,I) Pink Flamindo imaging in the CA, dissected out from dormant flies expressing P2X2 in CA-LP neurons, with or without ATP administration. (H) Changes in the relative fluorescence intensity of JHAMT-GAL4-driven Pink Flamindo in the CA of R18G01-LexA LexAOP-P2X2 females with stimulation using 2.5 mM ATP (each n=10). (I) Quantification of normalized Pink Flamindo signals in the CA at 56 s after the stimulation. (J,K) Anti-DH31 signal in the CA-LP neurons in the vehicle- and ATP-treated brain-CA complexes collected after Pink Flamindo imaging shown in H and I. (J) Representative images of anti-DH31 immunostaining signals in the soma of CA-LP1 and CA-LP2 neurons. (K) Quantitative comparison of anti-DH31 immunoreactivity in the soma of CA-LP1 and CA-LP2 neurons. Data are mean±s.e.m. P<0.05, *P<0.001 (unpaired two-tailed Student’s t-test for C,E,G,I and K). n.s., not significant. normal male genital rotation (Ádám et al., 2003), consistent with the inhibitory role of CA-LP neurons in JH biosynthesis during the adult stage. Nevertheless, we found that there was no phenotype of male genitalia rotation in genetic mutant males exhibiting either loss of Dh31 or Dh31-R functions (Fig. S8A,B). These results indicate that, at least in the larval and/or pupal stages, CA-LP neurons might produce an additional neuropeptide or neurotransmitter, different from DH31, which also has a negative impact on JH biosynthesis in the CA. Therefore, a complete list of the neuropeptides/neurotransmitters produced in CA-LP neurons is required. Consistent with our observations, a recent study reported that DH31 and DH31-R were not involved in D. melanogaster oogenesis under non-dormancy-inducing conditions (Ma et al., 2020). In addition, our data indicated that Dh31 transcription, at least in CA- LP1 neurons, is elevated, and DH31 protein level in the soma of CA- LP neurons is reduced in response to dormancy-inducing conditions. Based on these results, we hypothesize that dormancy-inducing conditions stimulate CA-LP neurons to produce Dh31 and release its products to the CA, which may underlie the induction of reproductive dormancy. However, it should be noted that this hypothesis requires T N E M P O L E V E D 8 RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 for Our time, studies, the first identified DH31 as a neurotransmitter derived from neurons that directly innervate the CA. In contrast, recent studies using D. melanogaster have revealed a number of circulating factors that are crucial for regulating reproductive dormancy (Kurogi et al., 2021). Most of these studies have repeatedly emphasized the importance of insulin-like peptides (ILPs) produced in IPCs located in PI. ILPs directly stimulate CA to enhance JH biosynthesis through the PI3-kinase-mTOR pathway under non-dormancy-inducing conditions (Ojima et al., 2018). Conversely, under dormancy-inducing conditions, the production and secretion of ILPs from IPCs are inhibited, resulting in the thus leading to reproductive suppression of JH biosynthesis, dormancy (Kurogi et al., 2021). The modulation of IPC activity is mediated by multiple neuropeptides and neurotransmitters, including PDF and sNPF produced by the s-LNv cells, as well as monoaminergic neurotransmitters, such as serotonin and dopamine (Andreatta et al., 2018; Nagy et al., 2019). Dopamine also suppresses JH biosynthesis via a dopamine receptor subtype in the CA (Andreatta et al., 2018). On the other hand, the release of allatostatin C (AstC) from DN3 neurons is essential for cold- activation of induced reproductive dormancy through the cholinergic AstC receptor type 2 in clock neurons (Meiselman et al., 2022). CA-LP neurons might cooperate with these humoral factors for the proper regulation of reproductive dormancy in vivo; however, it is not known how all these signals are integrated to regulate reproductive dormancy. A previous study has reported that cAMP-dependent protein kinase (PKA) negatively regulates JH biosynthesis (Andreatta et al., 2018). Considering this and our data from Pink Flamindo imaging studies, it is plausible that DH31 regulates JH biosynthesis by modulating intracellular cAMP levels; however, additional studies are required to elucidate the molecular basis of cAMP and PKA modulation of JH biosynthesis. Based on our immunostaining data, DH31 signaling did not seem to affect the protein levels of JHAMT, a crucial enzyme in the JH biosynthesis pathway (Niwa et al., 2008; Shinoda and Itoyama, 2003). regulating adult male physiological processes JH plays key roles in regulating numerous D. melanogaster adult female physiological processes, including vitellogenesis in non- dormancy-inducing conditions, ovulation, egg shape maintenance, gut remodeling, innate immunity, sleep regulation and aging (Flatt et al., 2008; Luo et al., 2021; Postlethwait and Weiser, 1973; Reiff et al., 2015; Schwenke and Lazzaro, 2017; Wu et al., 2018; Yamamoto et al., 2013). In addition, JH signaling is also essential in D. for melanogaster, including protein synthesis in accessory glands, sleep regulation, courtship motivation and courtship-associated memory retention (Lee et al., 2017; Wijesekera et al., 2016; Wu et al., 2018; Yamamoto et al., 1988; Zhang et al., 2021). In fact, we found that CA-LP neurons are also present in adult males (Fig. S8C) and therefore it is possible that CA-LP neurons might also regulate male adult JH biosynthesis. Further studies could reveal additional roles of CA-LP neurons in modulating JH-dependent events in adult male and female flies. Considering that DH31 is a well-conserved molecule in invertebrates (Cai et al., 2018), future studies could also investigate the role of CA-LP neurons in modulating JH- mediated reproductive dormancy in other insects. In particular, the morphology of D. melanogaster CA-LP neurons closely resembled that of the CA-projecting PL neurons in P. terraenovae (Shiga and Numata, 2000), suggesting that CA-LP neurons might be involved in the regulation of reproductive dormancy in other dipteran species. 9 T N E M P O L E V E D Fig. 6. Hemolymph JH III titers are increased after the loss of DH31 signaling. (A) Hemolymph JH III titers in control and Dh31 genetic mutant females under dormancy-inducing conditions. JH III amounts are normalized by body weight. (B) Hemolymph JH III titers in control and CA-specific Dh31- R RNAi females under dormancy inducing conditions. JH III amounts are normalized by body weight. (C) Quantification of yolk protein 1 mRNA in control and Dh31-R genetic mutant females under dormancy-inducing conditions by RT-qPCR. (D) A proposed model of DH31-dependent regulation of reproductive dormancy in D. melanogaster. Data are mean ±s.e.m. *P<0.001 (Tukey–Kramer’s HSD test for G and H). n.s., not significant. Samples were derived from virgin females 6 days after their transfer to dormancy-inducing conditions. further demonstration. In addition, another issue is that we did not observe differences in TRIC signals in CA-LP neurons between dormancy- and non-dormancy-inducing conditions. In future studies, it would be interesting to use additional neurobiological technologies besides TRIC assay to examine whether and how CA-LP neuronal activities are regulated in response to dormancy-inducing conditions. Previous research has reported that the circadian clock system is involved in the regulation of reproductive dormancy in many insect species (Denlinger, 2022; Hasebe and Shiga, 2021a; Meuti and Denlinger, 2013; Saunders, 2020), including D. melanogaster (Abrieux et al., 2020; Nagy et al., 2019; Saunders, 1990; Saunders et al., 1989). Therefore, we propose that, to an extent, circadian clock neurons, such as fifth s-LNv, s-LNv, LNd and DN1p are connected with CA-LP neurons and modulate their activity under dormancy- inducing conditions. In addition, sNPF secreted from s-LNv may stimulate the sNPF-R to modulate the activity of CA-LP neurons. Remarkably, a recent study showed that sNPF secreted from s-LNv cells positively regulates IPCs, leading to the suppression of reproductive dormancy in D. melanogaster (Nagy et al., 2019). Therefore, an alternative hypothesis is that sNPF from s-LNv simultaneously regulates CA-LP neurons and IPCs to control If sNPF plays a suppressive role in reproductive dormancy. reproductive dormancy, should negatively regulate CA-LP it neurons. Notably, an inhibitory effect of sNPF on central neurons has been reported in D. melanogaster (Vecsey et al., 2014). RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 Our study genetic provides evidence that conclusive D. melanogaster female reproductive dormancy is regulated by DH31-producing neurons projecting to the CA. Further functional analyses of the clock neurons–CA-LP neurons–CA axis in other insects could shed light on the conserved molecular mechanisms underlying the environmental adaptation of insects, including agricultural pests and infectious disease vectors. For example, a previous study has shown that DH31-positive neuronal processes innervate the CA in the cabbage root fly (Delia radicum), a serious pest for many Brassicaceae crops (Zoephel et al., 2012). These studies would provide important information that could help establish the foundation for the development of new approaches for the control of harmful insects. Lastly, as DH31 belongs to an evolutionarily conserved peptide family that includes mammalian calcitonin gene-related peptides (Coast et al., 2001), it would be intriguing to examine whether these related peptides are also involved in regulating dormancy or hibernation in other animal species. MATERIALS AND METHODS Drosophila strains and maintenance D. melanogaster flies were raised until eclosion on a standard yeast- cornmeal-glucose fly medium [0.275 g agar, 5.0 g glucose, 4.5 g cornmeal, 2.0 g yeast extract, 150 μl propionic acid and 175 μl 10% butyl p-hydroxybenzoate (in 70% ethanol) in 50 ml water] at 25°C under a 12 light/12 h dark cycle. To induce reproductive dormancy, adult virgin females were collected within 6 h of eclosion and reared on standard medium at 11±0.5°C under short-day length (SD: 10 h light/14 h dark cycle) conditions for 12 days, as described in previous studies (Andreatta et al., 2018; Ojima et al., 2018). To analyze the effects of JH analogs on reproductive dormancy, virgin female flies were collected within 6 h of eclosion and reared on standard medium supplemented with 1.5 mM methoprene (Sigma-Aldrich, PESTANAL 33375, racemic mixture; 1.5 M stock was prepared in ethanol) or 0.1% ethanol (control). The following D. melanogaster genetic mutant strains were used: Dh31#51 (from Fumika N. Hamada, Cincinnati Children’s Hospital Medical Center, USA), Dh31attp [Bloomington Drosophila Stock Center (BDSC), #84490] and Dh31-Rattp (BDSC, #84491) (from Yi Rao, Peking University School of Life Sciences, China). The Dh31#51 allele has a 730-bp deletion containing the second and third exons of the Dh31 gene (Head et al., 2015). Dh31attp is another allele in which a 1500-bp genomic region containing all exons of the Dh31 gene was replaced with a knock-in cassette (Deng et al., 2019). In addition, the following transgenic D. melanogaster strains were used: Actin5C-GAL4 (BDSC, #3954), Aug21-GAL4 (Siegmund and Korge, 2001) (BDSC, #30137), Df(2R)BSC273 (BDSC, #23169), Dilp2-GAL4 (BDSC, #37516), Dh31-R-T2A-GAL4 (Kondo et al., 2020), JHAMT-GAL4 (from Brigitte Dauwalder, University of Houston, USA; Wijesekera et al., 2016), JHAMT-LexA (from Naoki Yamanaka, U.C. Riverside, USA), Kurs21- GAL4 (from Stéphane Noselli, Université Côte D’Azur, France; Siegmund and Korge, 2001), LexAop-CD4::spGFP11,UAS-CD4::spGFP1-10 (BDSC, #58755), LexAop-mCD8::GFP (BDSC, #32203), LexAop-P2X2 (BDSC, #76030), Pdfr-T2A-GAL4 (Kondo et al., 2020), R18G01-LexA (BDSC, #52531), R21C09-GAL4 (BDSC, #48936), R94H10-GAL4 (BDSC, #47268), sNPF-R-T2A-GAL4 (BDSC, #84691), tsh-GAL80 (Simpson, 2016), UAS-DenMark,UAS-Syt1::GFP (BDSC, #33064), UAS-Dh31 (FlyORF, #F003632), UAS-Dh31-IR (BDSC, #41957), UAS-Dh31-R- IRNIG (NIG-FLY, #17043R-1), UAS-Dh31-R-IRKK (Vienna Drosophila Resource Center, #101995), UAS-Dicer2 (BDSC, #24651), UAS-mCD:: GFP (BDSC, #32219), UAS-mCD8::RFP, LexAop2-mCD8::GFP;nSyb- MKII::nlsLexADBDo;UAS-p65AD::CaM (for TRIC assay; BDSC, #61679), UAS-Pink Flamindo (this study), UAS-PKADN (BDSC, #35550), UAS-TrpA1;UAS-TrpA1 (the combined strain of BDSC #26263 and #26264, carrying TrpA1 transgenes on both the second and third chromosome), Viking-GFP (Drosophila Genomics Resource Center, #110692), and UAS-GFP,mCD8::GFP (from Kei Ito, University of Cologne, Germany; Ito et al., 1998). Heterozygous controls were obtained by crossing w1118 with strains of genetic mutants, GAL4 drivers or UAS effectors. Generation of mouse anti-DH31 antibody Peptides corresponding to the C-terminal 16 amino acid sequence (NH2- AKHLMGLAAANFAGGP-NH2) of Bombyx mori DH31 (GenBank accession number BAG49567.1), with an N-terminal addition of cysteine, were synthesized and conjugated with maleimide-activated bovine serum albumin (BSA) (Imject Maleimide-Activated BSA, Thermo Fisher Scientific, 77115). The BSA-conjugated DH31 partial peptides were dialyzed with phosphate buffer saline (PBS). Subsequently, 50 μl of 1 mg/ml dialyzed conjugates was mixed with 25 μl of ABISCO-100 adjuvant (Isconova) and 25 μl of PBS. The mixture was subcutaneously injected twice into the mice and whole blood was collected 12 days after the last immunization. Blood serum was heat-inactivated at 56°C for 30 min, followed by the addition of an equal volume of saturated ammonium sulfate solution to precipitate the proteins. The precipitate was dissolved and dialyzed twice in PBS. We would like to emphasize that there was only one amino acid difference in the C-terminal 16 amino acid sequence of DH31 between B. mori and D. melanogaster (NH2-AKHRMGLAAANFAGGP- NH2; GenBank accession number NP_523514.1); therefore, cross-reactivity was expected. The anti-Dh31 antibody signal in the CA disappeared in the the signal complete deletion mutant of Dh31 (Fig. 3A); conversely, reappeared in the Dh31-overexpressing animals on the Dh31 genetic mutant background (Fig. 3J), confirming the specificity of the antibody. Immunohistochemistry The tissues were dissected in PBS and fixed in 4% paraformaldehyde in PBS for 30-60 min at 25-27°C. The fixed samples were rinsed thrice in PBS, washed for 15 min with PBS containing 0.3% Triton X-100 (PBT), and treated with a blocking solution (2% BSA in PBT; Sigma-Aldrich, #A9647) for 1 h at 25-27°C or overnight at 4°C. The samples were incubated with a primary antibody in blocking solution overnight at 4°C. The primary antibodies used were as follows: chicken anti-GFP antibody (Abcam, #ab13970, 1:2000), rabbit anti-RFP antibody (Medical & Biological Laboratories, PM005, 1:2000), guinea pig anti-JHAMT antibody (Mizuno et al., 2021; 1:1000-2000); rabbit anti-JHAMT antibody (Niwa et al., 2008, 1:1000), mouse anti-DH31 antibody (this study; 1:200), guinea pig anti- DH31 (from Michael Nitabach, Yale University, USA; Kunst et al., 2014; 1:500), guinea pig anti-DILP2 (from Takashi Nishimura, Gunma University, Japan; Okamoto and Nishimura, 2015; 1:200) and rabbit antibody against PDF of the cricket Gryllus bimaculatus (from Outa Uryu and Kenji Tomioka, Okayama University, Japan; Abdelsalam et al., 2008; 1:2000). Notably, previous studies have confirmed that anti-cricket PDF antibodies cross-react with the D. melanogaster PDF protein (Miyasako et al., 2007; Umezaki et al., 2012). The samples were rinsed thrice with PBS and then washed for 15 min with PBT, followed by incubation with fluorophore (Alexa Fluor 488, 546, or 633)-conjugated secondary antibodies (Thermo Fisher Scientific; 1:200) in blocking solution for 2 h at room temperature or overnight at 4°C. The secondary antibodies (Thermo Fisher Scientific; 1:200) used were as follows: goat anti-chicken secondary antibody Alexa Fluor 488(A32931; RRID: AB_2762843), goat anti-mouse secondary antibody Alexa Fluor 488 (A32723; RRID: AB_2633275), goat anti-rabbit secondary antibody Alexa Fluor 488 (A32731; RRID: AB_2633280), goat anti-mouse secondary antibody Alexa Fluor 555 (A32727; RRID: AB_2633276), goat anti-rabbit secondary antibody Alexa Fluor 555 (A21435; RRID: AB_2535856) and goat anti-guinea pig secondary antibody Alexa Fluor 633 (A21105; RRID:AB_2535757). After the samples were rinsed thrice with PBS and then washed thrice for 15 min with PBT, they were mounted on glass slides using FluorSave reagent (Merck Millipore, #345789). Quantification of immunostaining signals was conducted using the ImageJ software version 1.53q (Schneider et al., 2012). Connectome analysis The connectome analysis was performed using NeuronBridge (Clements et al., 2020; Jenett et al., 2012; Meissner et al., 2023; Otsuna et al., 2018 10 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 preprint) (https://neuronbridge.janelia.org/) and neuPrint+ (Plaza et al., 2022; Scheffer et al., 2020) (https://neuprint.janelia.org/). NeuronBridge is a database in which neurons labeled by GAL4 drivers are mapped after identification using light and electron microscopic connectome analysis. The NeuronBridge body IDs corresponding to two of the three pairs of CA- LP neurons (Fig. S2A) were #295063181 and #5813067334. Neurons connected with #295063181 and #5813067334 were identified in the connectome database on neuPrint+, and the names of the connected neurons and data of the connection numbers were obtained. The percentages of neurons indicate the proportions of each clock neuron in relation to the whole connection to #295063181 and #5813067334 (Fig. S2A). Counting mature egg numbers in ovaries The ovaries of virgin females were dissected in PBS. The numbers of mature eggs (stage-14 oocytes) (King, 1970) in the ovaries were counted under a stereomicroscope (Leica MZ10F). Forced activation of CA-LP neurons by TrpA1 overexpression Flies carrying R94H10-GAL4 and two copies of UAS-TrpA1 were reared at 21°C from embryos to newly eclosed adults. Upon eclosion, flies were randomly assigned to two groups: one maintained at 21°C ( permissive temperature) where TrpA1 was not activated, and the other raised at 29°C (restrictive temperature) to activate TrpA1. After 4 days, mature egg numbers were counted. Generation of a UAS-Pink Flamindo transgenic line The pcDNA3.1 plasmid containing the Pink Flamindo coding sequence (Pink Flamindo-pcDNA3.1; Harada et al., 2017) was obtained from Addgene ( plasmid #102356). The following primers were used for PCR amplification, to obtain the Pink Flamindo coding sequences with EcoRI and XbaI sites at the 5′ and 3′ termini, respectively: Pink Flamindo F (5′-ACTGAATTCATGCTGGTGAGCAAGGGC-3′) and Pink Flamindo R (5′-CCTGCTCGACATGTTCATTAGATCTCAG-3′). PCR products were digested with EcoRI and XbaI, purified, and ligated with the EcoRI-XbaI- digested pWALIUM10-moe plasmid (Perkins et al., 2015). Transformants were generated using the phiC31 integrase system in the P{CaryP}attP2 strain (Groth et al., 2004) by WellGenetics. w+ transformants of pWALIUM10-moe were established using standard protocols. cAMP imaging in the CA after the administration of DH31 peptide In the implementation of live imaging in the CA, experimental conditions were optimized with reference to previous studies (Lee et al., 2017; Meiselman et al., 2017). cAMP transients in the CA were imaged in flies expressing UAS-Pink Flamindo and UAS-mCD8::GFP driven by JHAMT-GAL4. For preparing non- dormant flies, newly eclosed virgin females (0-6 h after eclosion) were cultured at 25°C under a 12 h light/12 h dark cycle for 1 day. For preparing dormant flies, newly eclosed virgin females were cultured at 11°C under a SD cycle for 12 days. Adult brain-CA-gut complexes were dissected in Schneider’s Drosophila Medium (SDM; Thermo Fisher Scientific, #21720024) without supplementation with fetal bovine serum. The brain-CA-gut complexes were then treated with 100 μl of collagenase solution (0.05 mg/ml collagenase in SDM; Sigma-Aldrich, #C0130) with gentle rotation for 9 min at 25-27°C and vortexed gently for 1 min, ensuring that the brain and gut were not physically separated. The samples were then washed twice with 1 ml of SDM and once with 500 μl of SDM. The dissected tissues were held in a glass-bottom dish (35×10 mm, IWAKI, #3910-035) with an insect pin (ϕ0.10 mm, Ento Sphinx Insect Pins) and silicone grease (Beckman), and 20 μl of SDM was added to cover the tissue. Live imaging was performed at 25-27°C using a laser scanning confocal microscope (LSM700, Carl Zeiss) with a 20× objective lens. mCD8::GFP and Pink Flamindo were excited with 488 nm and 555 nm lasers, respectively. Time-lapse images were acquired every 8 s for 472 s. Then, 112 s after starting live-imaging, 80 μl of SDM with or without 375 nM synthetic DH31 peptide (NH2- TVDFGLARGYSGTQEAKHRMGLAAANFAGGP-CONH2, synthesized by Eurofins Genomics; final concentration: 300 nM) or 125 μM NKH477 (Sigma-Aldrich, #N3290; final concentration: 100 μM) was applied on the tissue. For image processing, the CA was selected in a region of interest (ROI) over multiple time frames. The mean fluorescence intensities were measured along the time axis using the ImageJ software version 1.53q (Schneider et al., 2012). Data were analyzed using Microsoft Excel. cAMP imaging in the CA with bath-applied ATP-dependent neuronal activation through P2X2 Newly eclosed virgin females, which carry transgenes of JHMAT-GAL4, UAS-Pink Flamindo, R18G01-LexA and LexAop-P2X2, were cultured at 11°C under a SD cycle for 12 days. All experimental procedures were almost the same as those used for cAMP imaging in the CA with the administration of the DH31 peptide; except for the following two points. First, treated with collagenase. Second, the DH31 peptide was not used: 112 s after starting live-imaging, 80 μl of SDM with or without 3.125 mM ATP (Promega #E6011; final concentration: 2.5 mM) was applied on the tissue. After live imaging, the vehicle- and ATP-treated brain-CA-gut complexes were collected and used for immunostaining experiments with anti-DH31 antibody. Of note, the immunoreactive signals from anti-DH31 antibody were diminished in DH31 loss-of-function mutant animals (Fig. 3A). the brain-CA-gut complexes were not Measurement of JH III titers Newly emerged virgin females were collected within 6 h of eclosion and reared on standard medium at 11±0.5°C under SD conditions for 6 days. Forty adult female flies of each genotype were punctured using a tungsten needle and placed in a plastic tube with a hole at the bottom. The tube was then connected to a silanized glass vial (GL Science, #5183-4507) and centrifuged at 9100 g for 5 min. Pre-cooled PBS (150 µl) was added to the glass vial where the hemolymph was collected, and 6.25 pg/µl (in acetonitrile) of JH III-D3 (Toronto Research Chemicals, #E589402) was added as an internal control. Subsequently, 600 µl of hexane was added, and the samples were stirred for 1 min. The samples were then centrifuged at 2000 g for 5 min at 4°C, and 500 µl of the organic phase was transferred to a fresh silanized vial. The samples were dried under a gentle nitrogen flow and stored at −20°C until further analysis. JH titers from these hemolymph extracts were determined using LC-MS/MS as previously described (Ramirez et al., 2020). Reverse transcription-quantitative PCR (RT-qPCR) Total RNA was extracted from whole bodies of 6-day-old adult virgin female flies (under a SD cycle) after ∼4-6 h of light period. RNA was reverse-transcribed using ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo). Synthesized cDNA samples were used as templates for quantitative PCR using THUNDERBIRD SYBR qPCR Mix (Toyobo) on a Thermal Cycler Dice Real Time System (Takara Bio). The amount of target RNA was normalized to the endogenous control ribosomal protein 49 gene (rp49) and the relative fold change was calculated. The expression levels of yolk protein 1, yolk protein 2 and yolk protein 3 were compared using the ΔΔCt method. The following primers were used for this analysis: rp49 F (5′-CGGATCGATATGCTAAGCTGT-3′), rp49 R (5′-GCGCTTGTTC- GATCCGTA-3′), Dh31-R F (5′-TACATCCTTACGCCCTTTCGTCCT- 3′), Dh31-R R (5′-GGCAACGCACAGACCTTGAAATGA-3′) (Goda et al., 2018), yolk protein1 F (5′-CAGGCTCAGTACACCCACAC-3′), yolk protein1 R (5′-CTCAACGTTGTGGTGGATCTG-3′), yolk protein2 F (5′-ACCCTTAAGAAGCTGCAGGAG-3′), yolk protein2 R (5′-ATGGTT- GAACTGGGACAGATG-3′), yolk protein3 F (5′-CTCAAGAGCAGC- GACTACGAC-3′) and yolk protein3 R (5′-TAGCGTTTGAAGTTGGT- CAGG-3′). Statistical analysis All experiments were performed independently at least twice. The sample sizes were chosen based on the number of independent experiments required for statistical significance and technical feasibility. The experiments were not randomized, and the investigators were aware of which samples had received which treatment. All raw quantitative data are provided in Tables S1 and S2. All statistical analyses were performed using the ‘R’ software version 4.0.3. Details of the statistical analyses are described in figure legends. 11 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 Acknowledgements We thank Brigitte Dauwalder, Fumika N. Hamada, Kei Ito, Stéphane Noselli, Yi Rao, Julie Simpson, Asako Tsubouchi, Naoki Yamanaka, Bloomington Drosophila Stock Center, Vienna Drosophila Resource Center, National Institute of Genetics (NIG- FLY) and KYOTO Drosophila Stock Center in the Kyoto Institute of Technology for fly stocks; Takashi Nishimura, Michael Nitabach, Yoshiaki Tanaka, Kenji Tomioka and Outa Uryu for antibodies; Drosophila Genomics Resource Center and Addgene for plasmids; WellGenetics for generating transgenic flies; and Editage (www.editage. com) for English language editing. We are also grateful to Kazuki Harada and Takashi Tsuboi for their advice on the Pink Flamindo experiment; Chen Zhang and Young-Joon Kim for their advice on JH titer measurement; Tomotsune Ameku and Yuto Yoshinari for their advice on the fly works; and Masaharu Hasebe and Sakiko Shiga for critical reading of the manuscript. Competing interests The authors declare no competing or financial interests. Author contributions Conceptualization: Y.K., E.I., R.N.; Methodology: M.N., S.M., S.K., H.T., F.G.N.; Validation: Y.K., M.N., F.G.N., R.N.; Formal analysis: Y.K., M.N., F.G.N., R.N.; Investigation: Y.K., E.I., Y.M., R.H., M.N., F.G.N., R.N.; Resources: M.N., A.M., S.K., H.T., F.G.N.; Data curation: Y.K., M.N., F.G.N., R.N.; Writing - original draft: Y.K., F.G.N., R.N.; Writing - review & editing: Y.K., E.I., Y.M., R.H., M.N., S.M., A.M., S.K., H.T., F.G.N., R.N.; Visualization: Y.K., R.N.; Supervision: R.N.; Project administration: R.N.; Funding acquisition: Y.K., E.I., Y.M., M.N., H.T., F.G.N., R.N. Funding This work was supported by KAKENHI of the Japan Society for the Promotion of Science (21J20365 to Y.K., 17J00218 to E.I., 26250001 and 17H01378 to H.T.), a Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research on Innovative Areas (21H00226 to R.N.), and Japan Science and Technology Agency grant SPRING JPMJSP2124. This research was also funded by the Grantová Agentura Č eské Republiky ( project 22-21244S to M.N.) and the National Institute of Allergy and Infectious Diseases (R21AI167849 to F.G.N. and R21AI153689 to M.N.). Y.K., E.I., and R.H. received a fellowship from the Japan Society for the Promotion of Science. Open Access funding provided by University of Tsukuba: Tsukuba Daigaku. Deposited in PMC for immediate release. Data availability All relevant data can be found within the article and its supplementary information. Peer review history The peer review history is available online at https://journals.biologists.com/dev/ lookup/doi/10.1242/dev.201186.reviewer-comments.pdf. References Abdelsalam, S., Uemura, H., Umezaki, Y., Saifullah, A. S. M., Shimohigashi, M. and Tomioka, K. (2008). Characterization of PDF-immunoreactive neurons in the optic lobe and cerebral lobe of the cricket, Gryllus bimaculatus. J. Insect Physiol. 54, 1205-1212. doi:10.1016/j.jinsphys.2008.05.001 Abrieux, A., Xue, Y., Cai, Y., Lewald, K. M., Nguyen, H. N., Zhang, Y. and Chiu, J. C. (2020). EYES ABSENT and TIMELESS integrate photoperiodic and temperature cues to regulate seasonal physiology in Drosophila. Proc. Natl. Acad. Sci. USA 117, 15293-15304. doi:10.1073/pnas.2004262117 Ádám, G., Perrimon, N. and Noselli, S. (2003). The retinoic-like juvenile hormone controls the looping of left-right asymmetric organs in Drosophila. Development 130, 2397-2406. doi:10.1242/dev.00460 Andreatta, G., Kyriacou, C. P., Flatt, T. and Costa, R. (2018). Aminergic signaling controls ovarian dormancy in Drosophila. Sci. Rep. 8, 2030. doi:10.1038/s41598- 018-20407-z Benguettat, O., Jneid, R., Soltys, J., Loudhaief, R., Brun-Barale, A., Osman, D. and Gallet, A. (2018). The DH31/CGRP enteroendocrine peptide triggers intestinal contractions favoring the elimination of opportunistic bacteria. PLoS Pathog. 14, e1007279. doi:10.1371/journal.ppat.1007279 Bownes, M. (1989). The roles of juvenile hormone, ecdysone and the ovary in the control of Drosophila vitellogenesis. J. Insect Physiol. 35, 409-413. doi:10.1016/ 0022-1910(89)90115-7 Bownes, M. and Rembold, H. (1987). The titre of juvenile hormone during the pupal and adult stages of the life cycle of Drosophila melanogaster. Eur. J. Biochem. 164, 709-712. doi:10.1111/j.1432-1033.1987.tb11184.x Cai, W., Kim, C.-H., Go, H.-J., Egertová, M., Zampronio, C. G., Jones, A. M., Park, N. G. and Elphick, M. R. (2018). Biochemical, anatomical, and pharmacological characterization of Calcitonin-type neuropeptides in Starfish: Discovery of an ancient role as muscle relaxants. Front. Neurosci. 12, 382. doi:10. 3389/fnins.2018.00382 Clements, J., Goina, C., Kazimiers, A., Otsuna, H., Svirskas, R. and Rokicki, K. (2020). NeuronBridge codebase. https://janelia.figshare.com/articles/software/ NeuronBridge_Codebase/12159378 Coast, G. M., Webster, S. G., Schegg, K. M., Tobe, S. S. and Schooley, D. A. (2001). The Drosophila melanogaster homologue of an insect calcitonin-like diuretic peptide stimulates V-ATPase activity in fruit fly Malpighian tubules. J. Exp. Biol. 204, 1795-1804. doi:10.1242/jeb.204.10.1795 De Wilde, J. and de Boer, J. A. (1969). Humoral and nervous pathways in Insect photoperiodic induction of diapause in Leptinotarsa decemlineata. J. Physiol. 15, 661-675. doi:10.1016/0022-1910(69)90263-7 Deng, B., Li, Q., Liu, X., Cao, Y., Li, B., Qian, Y., Xu, R., Mao, R., Zhou, E., Zhang, W. et al. (2019). Chemoconnectomics: mapping chemical transmission in Drosophila. Neuron 101, 876-893.e4. doi:10.1016/j.neuron.2019.01.045 Denlinger, D. L. (2002). Regulation of Diapause. Annu. Rev. Entomol. 47, 93-122. doi:10.1146/annurev.ento.47.091201.145137 Denlinger, D. L. (2008). Why study diapause? Entomol. Res. 38, 1-9. doi:10.1111/ j.1748-5967.2008.00139.x Denlinger, D. L. (2022). Insect Diapause. Cambridge: Cambridge University Press. Denlinger, D. L., Yocum, G. D. and Rinehart, J. P. (2012). Hormonal Control of Diapause. In Insect Endocrinology (ed. L. I. Gilbert), pp. 430-463. Elsevier. Feinberg, E. H., VanHoven, M. K., Bendesky, A., Wang, G., Fetter, R. D., Shen, K. and Bargmann, C. I. (2008). GFP Reconstitution Across Synaptic Partners (GRASP) defines cell contacts and synapses in living nervous systems. Neuron 57, 353-363. doi:10.1016/j.neuron.2007.11.030 Flatt, T., Tu, M.-P. and Tatar, M. (2005). Hormonal pleiotropy and the juvenile hormone regulation of Drosophila development and life history. BioEssays 27, 999-1010. doi:10.1002/bies.20290 Flatt, T., Heyland, A., Rus, F., Porpiglia, E., Sherlock, C., Yamamoto, R., Garbuzov, A., Palli, S. R., Tatar, M. and Silverman, N. (2008). Hormonal regulation of the humoral innate immune response in Drosophila melanogaster. J. Exp. Biol. 211, 2712-2724. doi:10.1242/jeb.014878 Furuya, K., Milchak, R. J., Schegg, K. M., Zhang, J., Tobe, S. S., Coast, G. M. and Schooley, D. A. (2000). Cockroach diuretic hormones: characterization of a calcitonin-like peptide in insects. Proc. Natl. Acad. Sci. USA 97, 6469-6474. doi:10.1073/pnas.97.12.6469 Gao, X. J., Riabinina, O., Li, J., Potter, C. J., Clandinin, T. R. and Luo, L. (2015). A transcriptional reporter of intracellular Ca2+ in Drosophila. Nat. Neurosci. 18, 917-925. doi:10.1038/nn.4016 Goda, T., Tang, X., Umezaki, Y., Chu, M. L., Kunst, M., Nitabach, M. N. and Hamada, F. N. (2016). Drosophila DH31 neuropeptide and PDF receptor regulate night-onset temperature preference. J. Neurosci. 36, 11739-11754. doi:10.1523/ JNEUROSCI.0964-16.2016 rhythms of preferential Goda, T., Doi, M., Umezaki, Y., Murai, I., Shimatani, H., Chu, M. L., Nguyen, V. H., Okamura, H. and Hamada, F. N. (2018). Calcitonin receptors are ancient modulators for temperature in insects and body temperature in mammals. Genes Dev. 32, 140-155. doi:10.1101/gad.307884.117 Goda, T., Umezaki, Y., Alwattari, F., Seo, H. W. and Hamada, F. N. (2019). Neuropeptides PDF and DH31 hierarchically regulate free-running rhythmicity in Drosophila circadian locomotor activity. Sci. Rep. 9, 838. doi:10.1038/s41598- 018-37107-3 Groth, A. C., Fish, M., Nusse, R. and Calos, M. P. (2004). Construction of transgenic Drosophila by using the site-specific integrase from phage phiC31. Genetics 166, 1775. doi:10.1534/GENETICS.166.4.1775 Hamada, F. N., Rosenzweig, M., Kang, K., Pulver, S. R., Ghezzi, A., Jegla, T. J. and Garrity, P. A. (2008). An internal thermal sensor controlling temperature preference in Drosophila. Nature 454, 217-220. doi:10.1038/nature07001 Hamanaka, Y., Yasuyama, K., Numata, H. and Shiga, S. (2005). Synaptic connections between pigment-dispersing factor-immunoreactive neurons and neurons in the pars lateralis of the blow fly Protophormia terraenovae. J. Comp. Neurol. 491, 390-399. doi:10.1002/cne.20712 Hamanaka, Y., Tanaka, S., Numata, H. and Shiga, S. (2007). Peptide immunocytochemistry of neurons projecting to the retrocerebral complex in the blow fly, Protophormia terraenovae. Cell Tissue. Res. 329, 581-593. doi:10.1007/ s00441-007-0433-3 Hamanaka, Y., Tanaka, S., Numata, H. and Shiga, S. (2009). Morphological characterization of neurons projecting to the ring gland in the larval blow fly, Protophormia terraenovae. Zoolog. Sci. 26, 227-237. doi:10.2108/zsj.26.227 Hand, S. C., Denlinger, D. L., Podrabsky, J. E. and Roy, R. (2016). Mechanisms of animal diapause: recent developments from nematodes, crustaceans, insects, and fish. Am. J. Physiol. Regul. Integr. Comp. Physiol. 310, R1193-R1211. doi:10. 1152/ajpregu.00250.2015 Harada, K., Ito, M., Wang, X., Tanaka, M., Wongso, D., Konno, A., Hirai, H., Hirase, H., Tsuboi, T. and Kitaguchi, T. (2017). Red fluorescent protein-based cAMP indicator applicable to optogenetics and in vivo imaging. Sci. Rep. 7, 7351. doi:10.1038/s41598-017-07820-6 Hasebe, M. and Shiga, S. (2021a). Oviposition-promoting pars intercerebralis neurons show period-dependent photoperiodic changes in their firing activity in 12 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 the bean bug. Proc. Natl. Acad. Sci. USA 118, e2018823118. doi:10.1073/pnas. 2018823118 Hasebe, M. and Shiga, S. intercerebralis neurons, green bug, Plautia stali. Zoolog. Sci. 38, 317-325. doi:10.2108/zs210005 (2021b). Photoperiodic response in the pars in the brown-winged including Plast-MIP neurons, Hasegawa, T., Hasebe, M. and Shiga, S. (2020). Immunohistochemical and direct mass spectral analyses of Plautia stali myoinhibitory peptides in the cephalic ganglia of the brown-winged green bug Plautia stali. Zoolog. Sci. 37, 42-49. doi:10.2108/zs190092 Head, L. M., Tang, X., Hayley, S. E., Goda, T., Umezaki, Y., Chang, E. C., Leslie, J. R., Fujiwara, M., Garrity, P. A. and Hamada, F. N. (2015). The light on temperature preference in Drosophila. Curr. Biol. 25, influence of 1063-1068. doi:10.1016/j.cub.2015.02.038 Helfrich-Fö rster, C. (1995). The period clock gene is expressed in central nervous system neurons which also produce a neuropeptide that reveals the projections of circadian pacemaker cells within the brain of Drosophila melanogaster. Proc. Natl. Acad. Sci. USA 92, 612-616. doi:10.1073/pnas.92.2.612 Hodková, M. (1976). Nervous inhibition of corpora allata by photoperiod in Pyrrhocoris apterus. Nature 263, 521-523. doi:10.1038/263521a0 Hutfilz, C. (2022). Endocrine regulation of lifespan in insect diapause. Front. Physiol. 13, 825057. doi:10.3389/fphys.2022.825057 Imura, E., Shimada-Niwa, Y., Nishimura, T., Hü ckesfeld, S., Schlegel, P., Ohhara, Y., Kondo, S., Tanimoto, H., Cardona, A., Pankratz, M. J. et al. (2020). The Corazonin-PTTH Neuronal Axis Controls Systemic Body Growth by Regulating Basal Ecdysteroid Biosynthesis in Drosophila melanogaster. Curr. Biol. 30, 2156-2165.e5. doi:10.1016/j.cub.2020.03.050 Ito, K., Suzuki, K., Estes, P., Ramaswami, M., Yamamoto, D. and Strausfeld, N. J. (1998). The organization of extrinsic neurons and their implications in the functional roles of the mushroom bodies in Drosophila melanogaster Meigen. Learn. Mem. 5, 52-77. doi:10.1101/lm.5.1.52 Jenett, A., Rubin, G. M., Ngo, T.-T. B., Shepherd, D., Murphy, C., Dionne, H., Pfeiffer, B. D., Cavallaro, A., Hall, D., Jeter, J. et al. (2012). A GAL4-driver line resource for Drosophila neurobiology. Cell Rep. 2, 991-1001. doi:10.1016/j. celrep.2012.09.011 Johard, H. A. D., Yoishii, T., Dircksen, H., Cusumano, P., Rouyer, F., Helfrich- Fö rster, C. and Nä ssel, D. R. (2009). Peptidergic clock neurons in Drosophila: Ion transport peptide and short neuropeptide F in subsets of dorsal and ventral lateral neurons. J. Comp. Neurol. 516, 59-73. doi:10.1002/cne.22099 Johnson, E. C., Shafer, O. T., Trigg, J. S., Park, J., Schooley, D. A., Dow, J. A. and Taghert, P. H. (2005). A novel diuretic hormone receptor in Drosophila: evidence for conservation of CGRP signaling. J. Exp. Biol. 208, 1239-1246. doi:10.1242/jeb.01529 Kaneko, H., Head, L. M., Ling, J., Tang, X., Liu, Y., Hardin, P. E., Emery, P. and Hamada, F. N. (2012). Circadian rhythm of temperature preference and its neural control in Drosophila. Curr. Biol. 22, 1851-1857. doi:10.1016/j.cub.2012.08.006 Khan, M. A., Romberg-Privee, H. M. and Koopmanschap, A. B. (1986). Location of allatostatic centers in the pars lateralis regions of the brain of the Colorado potato beetle. Experimentia 42, 836-838. doi:10.1007/BF01941545 King, R. C. (1970). The meiotic behavior of the Drosophila oocyte. Int. Rev. Cytol. 28, 125-168. doi:10.1016/S0074-7696(08)62542-5 Kondo, S., Takahashi, T., Yamagata, N., Imanishi, Y., Katow, H., Hiramatsu, S., Lynn, K., Abe, A., Kumaraswamy, A. and Tanimoto, H. (2020). Neurochemical the Drosophila brain visualized by endogenously tagged organization of receptors. Cell Rep. 30, 284-297.e5. doi:10.1016/j.celrep. neurotransmitter 2019.12.018 Kotaki, T. and Yagi, S. (1989). Hormonal control of adult diapause in the brown- winged green bug, Plautia stali SCOTT (Heteroptera: Pentatomidae). Appl. Entomol. Zool. 24, 42-51. doi:10.1303/AEZ.24.42 Kubrak, O. I., Kuč erová, L., Theopold, U. and Nä ssel, D. R. (2014). The sleeping beauty: How reproductive diapause affects hormone signaling, metabolism, immune response and somatic maintenance in Drosophila melanogaster. PLoS ONE 9, e113051. doi:10.1371/journal.pone.0113051 Kunst, M., Hughes, M. E., Raccuglia, D., Felix, M., Li, M., Barnett, G., Duah, J. and Nitabach, M. N. (2014). Calcitonin gene-related peptide neurons mediate sleep-specific circadian output in Drosophila. Curr. Biol. 24, 2652-2664. doi:10. 1016/j.cub.2014.09.077 Kurogi, Y., Mizuno, Y., Imura, E. and Niwa, R. (2021). Neuroendocrine regulation of reproductive dormancy in the fruit fly Drosophila melanogaster: a review of juvenile hormone-dependent regulation. Front. Ecol. Evol. 9, 715029. doi:10. 3389/fevo.2021.715029 Lee, S. S., Ding, Y., Karapetians, N., Rivera-Perez, C., Noriega, F. G. and Adams, M. E. (2017). Hormonal signaling cascade during an early-adult critical period required for courtship memory retention in Drosophila. Curr. Biol. 27, 2798-2809.e3. doi:10.1016/j.cub.2017.08.017 Lin, H.-H., Kuang, M. C., Hossain, I., Xuan, Y., Beebe, L., Shepherd, A. K., Rolandi, M. and Wang, J. W. (2022). A nutrient-specific gut hormone arbitrates between courtship and feeding. Nature 602, 632-638. doi:10.1038/s41586-022- 04408-7 Luo, W., Liu, S., Zhang, W., Yang, L., Huang, J., Zhou, S., Feng, Q., Palli, S. R., Wang, J., Roth, S. et al. (2021). Juvenile hormone signaling promotes ovulation and maintains egg shape by inducing expression of extracellular matrix genes. Proc. Natl. Acad. Sci. USA 118, e2104461118. doi:10.1073/pnas.2104461118 Ma, T., Matsuoka, S. and Drummond-Barbosa, D. (2020). RNAi-based screens uncover a potential new role for the orphan neuropeptide receptor Moody in Drosophila female germline stem cell maintenance. PLoS ONE 15, e0243756. doi:10.1371/JOURNAL.PONE.0243756 Mandel, S. J., Shoaf, M. L., Braco, J. T., Silver, W. L. and Johnson, E. C. (2018). Behavioral aversion to AITC requires both painless and dTRPA1 in Drosophila. Front. Neural Circuits 12, 45. doi:10.3389/FNCIR.2018.00045 Matsumoto, K., Numata, H. and Shiga, S. (2013). Role of the brain in photoperiodic regulation of juvenile hormone biosynthesis in the brown-winged green bug Plautia stali. J. Insect Physiol. 59, 387-393. doi:10.1016/j.jinsphys.2013.01.007 Matsumoto, K., Suetsugu, Y., Tanaka, Y., Kotaki, T., Goto, S. G., Shinoda, T. and Shiga, S. (2017). Identification of allatostatins in the brown-winged green bug Plautia stali. J. Insect Physiol. 96, 21-28. doi:10.1016/j.jinsphys.2016.10.005 Matsuo, J., Nakayama, S. and Numata, H. (1997). Role of the corpus Allatum in the control of adult diapause in the blow fly, Protophormia terraenovae. J. Insect Physiol. 43, 211-216. doi:10.1016/S0022-1910(96)00103-5 Meiselman, M., Lee, S. S., Tran, R.-T., Dai, H., Ding, Y., Rivera-Perez, C., Wijesekera, T. P., Dauwalder, B., Noriega, F. G. and Adams, M. E. (2017). in Drosophila Endocrine network essential melanogaster. Proc. Natl. Acad. Sci. USA 114, E3849-E3858. doi:10.1073/ pnas.1620760114 reproductive success for Meiselman, M. R., Alpert, M. H., Cui, X., Shea, J., Gregg, I., Gallio, M. and (2022). Recovery from cold-induced reproductive dormancy is temperature-dependent AstC signaling. Curr. Biol. 32, by Yapici, N. regulated 1362-1375.e8. doi:10.1016/j.cub.2022.01.061 Meissner, G. W., Nern, A., Dorman, Z., DePasquale, G. M., Forster, K., Gibney, T., Hausenfluck, J. H., He, Y., Iyer, N. A., Jeter, J. et al. (2023). A searchable image resource of Drosophila GAL4-driver expression patterns with single neuron resolution. eLife 12, e80660. doi:10.7554/ELIFE.80660 Meuti, M. E. and Denlinger, D. L. (2013). Evolutionary links between circadian clocks and photoperiodic diapause in insects. Integr. Comp. Biol. 53, 131-143. doi:10.1093/icb/ict023 Miyasako, Y., Umezaki, Y. and Tomioka, K. (2007). Separate sets of cerebral clock neurons are responsible for light and temperature entrainment of Drosophila rhythms. J. Biol. Rhythms 22, 115-126. doi:10.1177/ circadian locomotor 0748730407299344 Mizuno, Y., Imura, E., Kurogi, Y., Shimada-Niwa, Y., Kondo, S., Tanimoto, H., Hü ckesfeld, S., Pankratz, M. J. and Niwa, R. (2021). A population of neurons that produce hugin and express the diuretic hormone 44 receptor gene projects to the corpora allata in Drosophila melanogaster. Dev. Growth Differ. 63, 249-261. doi:10.1111/dgd.12733 Nagy, D., Cusumano, P., Andreatta, G., Anduaga, A. M., Hermann-Luibl, C., Reinhard, N., Gesto, J., Wegener, C., Mazzotta, G., Rosato, E. et al. (2019). Peptidergic signaling from clock neurons regulates reproductive dormancy in Drosophila melanogaster. PLoS Genet. 15, e1008158. doi:10.1371/journal.pgen. 1008158 Nicolaï, L. J. J., Ramaekers, A., Raemaekers, T., Drozdzecki, A., Mauss, A. S., Yan, J., Landgraf, M., Annaert, W. and Hassan, B. A. (2010). Genetically encoded dendritic marker sheds light on neuronal connectivity in Drosophila. Proc. Natl. Acad. Sci. USA 107, 20553-20558. doi:10.1073/pnas.1010198107 Niwa, R., Niimi, T., Honda, N., Yoshiyama, M., Itoyama, K., Kataoka, H. and Shinoda, T. (2008). Juvenile hormone acid O-methyltransferase in Drosophila melanogaster. Insect Biochem. Mol. Biol. 38, 714-720. doi:10.1016/j.ibmb.2008. 04.003 Ojima, N., Hara, Y., Ito, H. and Yamamoto, D. (2018). Genetic dissection of stress- induced reproductive arrest in Drosophila melanogaster females. PLoS Genet. 14, e1007434. doi:10.1371/journal.pgen.1007434 Okamoto, N. and Nishimura, T. (2015). Signaling from glia and cholinergic neurons controls nutrient-dependent production of an insulin-like peptide for Drosophila body growth. Dev. Cell 35, 295-310. doi:10.1016/j.devcel.2015.10.003 Otsuna, H., Ito, M. and Kawase, T. (2018). Color depth MIP mask search: a new tool to expedite Split-GAL4 creation. bioRxiv. doi:10.1101/318006 Perkins, L. A., Holderbaum, L., Tao, R., Hu, Y., Sopko, R., McCall, K., Yang- Zhou, D., Flockhart, I., Binari, R., Shim, H.-S. et al. (2015). The transgenic RNAi project at Harvard medical school: resources and validation. Genetics 201, 843-852. doi:10.1534/genetics.115.180208 Plaza, S. M., Clements, J., Dolafi, T., Umayam, K., Neubarth, N. N., Scheffer, L. K. and Berg S. (2022). neuPrint: an open access tool for EM connectomics. Front. Neuroinform. 16, 896292. doi:10.3389/fninf.2022.896292 Poras, M. (1982). Le Contrô le endocrinien de la diapause imaginale des femelles de Tetrix undulata (Sowerby, 1806) (Orthoptere, Tetrigidae). Gen. Comp. Endocrinol. 46, 200-210. doi:10.1016/0016-6480(82)90202-7 Poras, M., Baehr, J. C. and Cassier, P. (1983). Control of corpus allatum activity during the imaginai diapause in females of Locusta migratoria L. Int. J. Invertebr. Reprod. 6, 111-122. doi:10.1080/01651269.1983.10510030 Postlethwait, J. H. and Weiser, K. (1973). Vitellogenesis induced by juvenile hormone in the female sterile mutant apterous-four in Drosophila melanogaster. Nat. New Biol. 244, 284-285. doi:10.1038/newbio244284a0 13 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201186. doi:10.1242/dev.201186 Qu, Z., Bendena, W. G., Tobe, S. S. and Hui, J. H. L. (2018). Juvenile hormone and sesquiterpenoids in arthropods: Biosynthesis, signaling, and role of MicroRNA. J. Steroid Biochem. Mol. Biol. 184, 69-76. doi:10.1016/j.jsbmb.2018.01.013 Ramirez, C. E., Nouzova, M., Michalkova, V., Fernandez-Lima, F. and Noriega, F. G. (2020). Common structural features facilitate the simultaneous identification and quantification of the five most common juvenile hormones by liquid chromatography-tandem mass spectrometry. Insect Biochem. Mol. Biol. 116, 103287. doi:10.1016/j.ibmb.2019.103287 Reiff, T., Jacobson, J., Cognigni, P., Antonello, Z., Ballesta, E., Tan, K. J., Yew, J. Y., Dominguez, M. and Miguel-Aliaga, I. (2015). Endocrine remodelling of the adult intestine sustains reproduction in Drosophila. eLife 4, e06930. doi:10. 7554/eLife.06930 Riddiford, L. M. (2020). Rhodnius, golden oil, and met: a history of juvenile hormone research. Front. Cell Dev. Biol. 8, 679. doi:10.3389/fcell.2020.00679 Santos, C. G., Humann, F. C. and Hartfelder, K. (2019). Juvenile hormone signaling in insect oogenesis. Curr. Opin. Insect Sci. 31, 43-48. doi:10.1016/j.cois. 2018.07.010 Saunders, D. S. (1990). The circadian basis of ovarian diapause regulation in Drosophila melanogaster: Is the period Gene causally involved in photoperiodic Biol. Rhythms time measurement? doi:10.1177/ 074873049000500404 315-331. 5, J. Saunders, D. S. (2020). Dormancy, diapause, and the role of the circadian system in insect photoperiodism. Annu. Rev. Entomol. 65, 373-389. doi:10.1146/annurev- ento-011019-025116 Saunders, D. S., Henrich, V. C. and Gilbert, L. I. (1989). Induction of diapause in Drosophila melanogaster: photoperiodic regulation and the impact of arrhythmic clock mutations on time measurement. Proc. Natl. Acad. Sci. USA 86, 3748-3752. doi:10.1073/pnas.86.10.3748 Saunders, D. S., Richard, D. S., Applebaum, S. W., Ma, M. and Gilbert, L. I. (1990). Photoperiodic diapause in Drosophila melanogaster involves a block to the juvenile hormone regulation of ovarian maturation. Gen. Comp. Endocrinol. 79, 174-184. doi:10.1016/0016-6480(90)90102-R Scheffer, L. K., Xu, C. S., Januszewski, M., Lu, Z., Takemura, S.-Y., Hayworth, K. J., Huang, G. B., Shinomiya, K., Maitin-Shepard, J., Berg, S. et al. (2020). A connectome and analysis of the adult Drosophila central brain. eLife 9, e57443. doi:10.7554/eLife.57443 Schneider, C. A., Rasband, W. S. and Eliceiri, K. W. (2012). NIH Image to ImageJ: 25 years of image analysis. Nat. Methods 9, 671-675. doi:10.1038/nmeth.2089 Schwenke, R. A. and Lazzaro, B. P. (2017). Juvenile hormone suppresses resistance to infection in mated female Drosophila melanogaster. Curr. Biol. 27, 596-601. doi:10.1016/j.cub.2017.01.004 Shiga, S. and Numata, H. (2000). The role of neurosecretory neurons in the pars intercerebralis and pars lateralis in reproductive diapause of the blowfly, Protophormia terraenovae. Naturwissenschaften 87, 125-128. doi:10.1007/ s001140050689 Shiga, S. and Numata, H. (2009). Roles of PER immunoreactive neurons in circadian rhythms and photoperiodism in the blow fly, Protophormia terraenovae. J. Exp. Biol. 212, 867-877. doi:10.1242/jeb.027003 Shiga, S., Toyoda, retrocerebral complex of Tissue Res. 299, 427-439. doi:10.1007/s004410050041 I. and Numata, H. (2000). Neurons projecting to the the adult blow fly, Protophormia terraenovae. Cell Shimada-Niwa, Y. and Niwa, R. (2014). Serotonergic neurons respond to nutrients and regulate the timing of steroid hormone biosynthesis in Drosophila. Nat. Commun. 5, 5778. doi:10.1038/ncomms6778 Shimokawa, K., Numata, H. and Shiga, S. (2008). Neurons important for the photoperiodic control of diapause in the bean bug, Riptortus pedestris. J. Comp. Physiol. A Neuroethol. Sens. Neural. Behav. Physiol. 194, 751-762. doi:10.1007/ s00359-008-0346-y Shinoda, T. and Itoyama, K. (2003). Juvenile hormone acid methyltransferase: A key regulatory enzyme for insect metamorphosis. Proc. Natl. Acad. Sci. USA 100, 11986-11991. doi:10.1073/pnas.2134232100 Siegmund, T. and Korge, G. (2001). Innervation of the ring gland of Drosophila doi:10.1002/1096- melanogaster. J. Comp. Neurol. 9861(20010319)431:4<481::AID-CNE1084>3.0.CO;2-7 481-491. 431, Simpson, J. H. (2016). Rationally subdividing the fly nervous system with versatile expression reagents. J. Neurogenet. 30, 185-194. doi:10.1080/01677063.2016. 1248761 Sliter, T. J., Sedlak, B. J., Baker, F. C. and Schooley, D. A. (1987). Juvenile hormone in Drosophila melanogaster: identification and titer determination during development. Insect Biochem. 17, 161-165. doi:10.1016/0020-1790(87)90156-9 Sugime, Y., Watanabe, D., Yasuno, Y., Shinada, T., Miura, T. and Tanaka, N. K. juvenile hormone titers in female Drosophila (2017). Upregulation of melanogaster through mating experiences and host food occupied by eggs and larvae. Zoolog. Sci. 34, 52-57. doi:10.2108/zs160150 Tatar, M. and Yin, C.-M. (2001). Slow aging during insect reproductive diapause: why butterflies, grasshoppers and flies are like worms. Exp. Gerontol. 36, 723-738. doi:10.1016/S0531-5565(00)00238-2 Tatar, M., Chien, S. A. and Priest, N. K. (2001). Negligible Senescence during Reproductive Dormancy in Drosophila melanogaster. Am. Nat. 158, 248-258. doi:10.1086/321320 Umezaki, Y., Yoshii, T., Kawaguchi, T., Helfrich-Fö rster, C. and Tomioka, K. (2012). Pigment-dispersing factor is involved in age-dependent rhythm changes in Drosophila melanogaster. J. Biol. Rhythms 27, 423-432. doi:10.1177/ 0748730412462206 Vanderveken, M. and O’Donnell, M. J. (2014). Effects of diuretic hormone 31, Drosokinin, and allatostatin A on transepitherial K+ transport and contraction frequency in the midgut and hindgut of larval Drosophila melanogaster. Arch. Insect Biochem. Physiol. 85, 76-93. doi:10.1002/arch.21144 Vecsey, C. G., Pı́rez, N. and Griffith, L. C. (2014). The Drosophila neuropeptides PDF and sNPF have opposing electrophysiological and molecular effects on central neurons. J. Neurophysiol. 111, 1033-1045. doi:10.1152/jn.00712.2013 Veenstra, J. A., Agricola, H.-J. and Sellami, A. (2008). Regulatory peptides in fruit fly midgut. Cell Tissue. Res. 334, 499-516. doi:10.1007/s00441-008-0708-3 Wheeler, D. (2003). The role of nourishment in oogenesis. Annu. Rev. Entomol. 41, 407-431. doi:10.1146/annurev.en.41.010196.002203 Wijesekera, T. P., Saurabh, S. and Dauwalder, B. (2016). Juvenile hormone is required in adult males for Drosophila courtship. PLoS ONE 11, e0151912. doi:10. 1371/journal.pone.0151912 Wu, B., Ma, L., Zhang, E., Du, J., Liu, S., Price, J., Li, S. and Zhao, Z. (2018). Sexual dimorphism of sleep regulated by juvenile hormone signaling in Drosophila. PLoS Genet. 14, e1007318. doi:10.1371/journal.pgen.1007318 Yamamoto, K., Chadarevian, A. and Pellegrini, M. (1988). Juvenile hormone action mediated in male accessory glands of Drosophila by calcium and kinase C. Science 239, 916-919. doi:10.1126/science.3124270 Yamamoto, R., Bai, H., Dolezal, A. G., Amdam, G. and Tatar, M. (2013). Juvenile hormone regulation of Drosophila aging. BMC Biol. 11, 85. doi:10.1186/1741- 7007-11-85 Yao, Z., Macara, A. M., Lelito, K. R., Minosyan, T. Y. and Shafer, O. T. (2012). functional neuronal connectivity in the Drosophila brain. Analysis of J. Neurophysiol. 108, 684-696. doi:10.1152/jn.00110.2012 Zhang, Y. Q., Rodesch, C. K. and Broadie, K. (2002). Living synaptic vesicle marker: Synaptotagmin-GFP. Genesis 34, 142-145. doi:10.1002/gene.10144 Zhang, S. X., Glantz, E. H., Miner, L. E., Rogulja, D. and Crickmore, M. A. (2021). Hormonal control of motivational circuitry orchestrates the transition to sexuality in Drosophila. Sci. Adv. 7, eabg6926. doi:10.1126/sciadv.abg6926 Zoephel, J., Reiher, W., Rexer, K.-H., Kahnt, J. and Wegener, C. (2012). Peptidomics of the agriculturally damaging larval stage of the cabbage root fly Delia radicum (Diptera: Anthomyiidae). PLoS ONE 7, e41543. doi:10.1371/ journal.pone.0041543 T N E M P O L E V E D 14
10.1186_s40249-020-0628-3	Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 https://doi.org/10.1186/s40249-020-0628-3 R E S E A R C H A R T I C L E Open Access Prevalence and incidence of nodding syndrome and other forms of epilepsy in onchocerciasis-endemic areas in northern Uganda after the implementation of onchocerciasis control measures Nolbert Gumisiriza1, Frank Mubiru2, Joseph Nelson Siewe Fodjo3, Martin Mbonye Kayitale4, An Hotterbeekx3, Richard Idro5, Issa Makumbi6, Tom Lakwo6, Bernard Opar6, Joice Kaducu6, Joseph Francis Wamala7 and Robert Colebunders3* Abstract Background: Around 2007, a nodding syndrome (NS) epidemic appeared in onchocerciasis-endemic districts of northern Uganda, where ivermectin mass distribution had never been implemented. This study evaluated the effect of community-directed treatment with ivermectin (CDTI) and ground larviciding of rivers initiated after 2009 and 2012 respectively, on the epidemiology of NS and other forms of epilepsy (OFE) in some districts of northern Uganda. Methods: In 2012, a population-based community survey of NS/epilepsy was carried out by the Ugandan Ministry of Health in Kitgum and Pader districts. In August 2017, we conducted a new survey in selected villages of these districts and compared our findings with the 2012 data. In addition, two villages in Moyo district (where CDTI was ongoing since 1993) served as comparative onchocerciasis-endemic sites in which larviciding had never been implemented. The comparison between 2012 and 2017 prevalence and cumulative incidence were done using the Fisher’s and Pearson’s Chi-square tests at 95% level of significance. Results: A total of 2138 individuals in 390 households were interviewed. In the selected villages of Kitgum and Pader, there was no significant decrease in prevalence of NS and OFE between 2012 and 2017. However, the cumulative incidence of all forms of epilepsy decreased from 1165 to 130 per 100 000 persons per year (P = 0.002); that of NS decreased from 490 to 43 per 100 000 persons per year (P = 0.037); and for OFE from 675 to 87 per 100 000 persons per year (P = 0.024). The median age of affected persons (NS and OFE) shifted from 13.5 (IQR: 11.0–15.0) years in 2012 to 18.0 (IQR: 15.0–20.3) years in 2017; P < 0.001. The age-standardized prevalence of OFE in Moyo in 2017 was 4.6%, similar to 4.5% in Kitgum and Pader. Conclusions: Our findings support the growing evidence of a relationship between infection by Onchocerca volvulus and some types of childhood epilepsy, and suggest that a combination of bi-annual mass distribution of ivermectin and ground larviciding of rivers is an effective strategy to prevent NS and OFE in onchocerciasis-hyperendemic areas. Keywords: Nodding syndrome, Epilepsy, Onchocerciasis, Prevalence, Incidence, Ivermectin, Larviciding, Uganda * Correspondence: robert.colebunders@uantwerpen.be 3Global Health Institute, University of Antwerp, Antwerp, Belgium Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 2 of 11 annual community-directed treatment with ivermectin (CDTI) started in 2009. Biannual CDTI was imple- mented in 2013. The average CDTI coverage in northern Uganda progressively improved from 33% in 2011 to 70% after 2015 [22]. In December 2012, the Government of Uganda launched vector control programs consisting of periodic ground larviciding of blackfly breeding areas in the three districts [22–24]. Aerial spraying was done as a one-off activity along the rivers Pager, Aswa and Agago which run through Kitgum, Pader and Lamwo districts resulting in a dramatic decrease of the blackfly population [24]. Since 2012, the number of new NS cases declined rap- idly, with no new cases reported altogether since 2013 [25]. However, it is not known whether the incidence of other forms of epilepsy (OFE) also decreased in the same region. Given that the number of new NS cases have been de- clining since 2012 [25], we hypothesized that following the implementation of onchocerciasis control measures in the districts of Kitgum and Pader, the incidence of OFE decreased alongside that of NS. In addition, we compared the 2017 prevalence of epilepsy in Kitgum and Pader with the prevalence of epilepsy in Moyo, a neighbouring onchocerciasis-endemic district where CDTI had been implemented since 1993 but without vector control activities. Methods Study sites The study was carried out in three northern Uganda dis- tricts of Kitgum, Pader, and Moyo (Fig. 1). Kitgum stretches northwards from the northern border of Pader district to the southern border of South Sudan. Moyo is located in the North-Western part of Uganda, with river Nile forming its eastern border, and South Sudan on its similar northern border. The three districts have Background Nodding syndrome (NS) is a neurological disorder that manifests with a unique epilepsy type characterized by repeated head nodding, often in association with pro- gressive neurocognitive impairment and physical decline [1–3]. It has been suggested that NS should be considered as one of the clinical presentations of onchocerciasis- associated epilepsy (OAE) [4, 5]. However, the pathophysi- ology of how Onchocerca volvulus may cause epilepsy re- mains unknown. Most cases of NS have been described in the onchocerciasis-endemic regions in northern Uganda [3, 6], western Uganda [7], South Sudan [8], and the Mahenge area in Tanzania [9]. More recent studies have reported nodding seizures among persons with epilepsy (PWE) in Cameroon [10] and the Democratic Republic of Congo [11]. A high prevalence of epilepsy has also been reported in many onchocerciasis meso- and hyper-endemic regions [12–15] particularly where transmission is poorly con- trolled [16–19]. A study in an onchocerciasis-endemic region in the Mbam valley in Cameroon showed that the risk of developing epilepsy increased with increasing O. volvulus microfilarial density [20]. onchocerciasis-endemic in the regions In Uganda, a NS epidemic occurred in the three onchocerciasis-endemic districts of Kitgum, Pader, and Lamwo in the northern part of the country [1] (Fig. 1). These three districts were among the most affected by the Lord’s Resistance Army (LRA) civil war which took place between 1986 and 2006, and consequently hosted the highest number of Internally Displaced Persons (IDP) lodged in camps [21]. Several of these camps were located in onchocerciasis-endemic areas in close proxim- ity to blackfly breeding sites. Due to the LRA insurgency, efforts to assess or control onchocerciasis in the area were compromised by insecurity (Table 1). By the year 2000, the mapping of onchocerciasis-affected areas had three districts and consistent been completed in all Fig. 1 Map of northern Uganda showing the 2017 study sites Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 3 of 11 Table 1 A timeline of onchocerciasis control activities, carried out in Kitgum, Pader and Moyo districts Year 1986–2006 1993 1994–2008 2008 2009 2012 2013 Kitgum and Pader districts Lord’s Resistance Army civil war interrupting onchocerciasis control efforts and programs Partial mapping of onchocerciasis Ivermectin only passively distributed Onchocerciasis mapping completed Start of consistent annual CDTI Moyo district Start of annual CDTI Start of vector control (river larviciding + aerial spraying) Start bi-annual CDTI Start of bi-annual CDTI CDTI Community-directed treatment with ivermectin environmental characteristics, marked by bushy vegetation and an extensive network of water bodies. The population is settled in clusters of extended family homesteads with subsistence farming as their main activity. Study procedures The 2012 survey in Kitgum, Pader and Lamwo districts In July 2012, the Uganda Ministry of Health with local and international partners (World Health Organization [WHO], Centers for Disease Control and Prevention in the United States [CDC], African Field Epidemiology Network [AFENET], Makerere School of Public Health [MakSPH], etc.) carried out a house-to-house survey in all the villages of Kitgum, Pader and Lamwo districts, to establish the prevalence and epidemiological distribution of NS and epilepsy. These districts were purposively se- lected because they were the most affected by the NS epidemic, based on the national surveillance health re- ports. Village leaders carried out a census of all house- holds in the village assisted by Village Health Teams (VHT). VHT are community volunteers who are identi- fied by their community members and are given basic training by the Uganda Ministry of Health on major health programs so that they can mobilize and sensitize communities to actively participate in and utilize the available health services, including mass drug adminis- tration for neglected tropical diseases. Prior to the 2012 survey, these VHT were trained to identify NS and epi- lepsy cases using simplified community-based definitions [1]. The VHT in conjunction with the village leaders vis- ited each household in their respective village and in- quired from the household head or a responsible adult if there were any household residents suspected of having NS or OFE including the deceased cases. They used a simplified paper-based tool to collect this information (see Additional file 1). The community case definition of NS was: any person with observed or reported episodes of head nodding (“luc luc”, local word for NS), and that of epilepsy as any person with epileptic seizures (“olili” or “lili” or “cimu”, local word for OFE). Differentiation of NS and OFE was taught to the VHTs through dramatization of commonly observed symptoms like; muscle jerks and contractions, mouth-frothing, loss of sphincter control and loss of consciousness. If a person presented with nodding seizures but also muscle jerks and contractions, mouth-frothing, loss of sphincter con- trol and loss of consciousness this person was consid- ered to have both “luc luc” and “lili”. As a follow-up of the 2012 survey described above, the CDC in conjunction with the Ugandan Ministry of Health (MOH) conducted a single-stage-cluster survey in March 2013. The aim of this survey was to systemat- ically assess and validate the prevalence of NS cases in northern Uganda [26]. Thirty parishes were selected by single-stage cluster sampling with probability propor- tional to size; 20–30 children (aged 5–18 years) with re- ported head nodding were selected per parish using simple random sampling without replacement. VHT called selected persons and their caregivers to a central meeting point at a specified date and time. Trained and supervised local clinicians subjected the children to a standardized tool that applied the 2012 consensus case definition of NS [1, 26]. The 2017 survey in Kitgum, Pader, and Moyo In August 2017, a house-to-house survey was carried out in selected villages of Kitgum, Pader and Moyo dis- tricts. Villages in Kitgum and Pader were randomly se- lected from the parishes with the highest prevalence of NS during the 2012 survey. The reason for selecting these high prevalence parishes was to allow us to dem- the onchocerciasis elimination onstrate the effect of measures with a relatively small population sample size. The parishes selected in Kitgum were: Lamit (Tumangu village) and Okidi (Kampala-Anyuka village). In Pader, Angole parish (Paikat-Akidi village) was selected. Given that the 2012 prevalence of NS in the parishes of Lamwo district was much lower (maximum of 2.9%) than most parishes in Kitgum and Pader, we did not include villages of Lamwo in the 2017 survey. In Moyo district, the villages of Pakarukwe and Pajakiri-North were ran- domly selected from among onchocerciasis-endemic parishes, as reported by the national health management information system (HMIS). Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 4 of 11 A two-stage epilepsy diagnosis method was used. Firstly, like in 2012, the trained VHT carried out a complete house-to-house screening of the entire village, during which demographic data was collected and sus- pected cases of NS and other forms of epilepsy identi- fied. A household was defined as family members who ate from a common pot. The VHT asked the head of the household or any responsible adult present, the same two questions that were asked during the 2012 survey i.e. whether a family member was affected by “luc luc” (NS) or “lili” (OFE). Additionally, the VHT administered a locally adapted and pre-tested 5-question screening data collection tool for epilepsy [27]. The five questions were: 1) Loss of consciousness with either urine on self and/or drooling; 2) Absence(s) or loss of contact with the surrounding of sudden onset and of brief duration; 3) Jerking or uncontrolled abnormal movement (convul- sion) of the limb(s) of sudden onset and lasting for a few minutes; 4) Sudden onset of brief, strange body sensa- tions, hallucinations or illusions, be they visual, auditory or olfactory; 5) Previously told that he/she has epilepsy [27]. In Moyo, the NS question was not asked by the VHT because a local word for NS did not exist. How- ever, the investigating clinicians questioned all suspected epilepsy cases about a history of nodding seizures; this was done in all study sites including Moyo. Any household member who answered positively to at least one screening question was suspected to have epi- lepsy. During the second stage, persons suspected to have epilepsy underwent a comprehensive clinical and neurological assessment by physicians or trained clinical officers to confirm or reject the diagnosis of epilepsy and to specify the seizure type. All consenting persons with probable NS or OFE were finger-pricked to test for the presence of O. volvulus-specific Ov16 IgG4 antibodies using rapid diagnostic tests (RDT) (Standard diagnostics Inc., Gyeonggi-do, Republic of Korea). Definitions A case of epilepsy was a person who had experienced at least two unprovoked seizures with a minimal time dif- ference of 24 h between the two events [28]. A case of NS was defined as a person with head nod- ding seizures (repetitive involuntary drops of the head towards the chest on two or more occasions) as con- firmed by a trained clinician, that occurs in a previously healthy child [1]. A case of OFE was a case of epilepsy, without a history of nodding seizures. Data management All data collection tools used during the 2012 and 2017 sur- vey are available as Additional file 1. During the 2017 sur- vey, VHT collected household survey data on paper which was then electronically captured using the software Epi- Info™ (CDC, Atlanta, USA) and then transferred to Micro- soft Excel 2016 (Microsoft, Seattle, USA) for further clean- ing. Physicians collected data on persons suspected to have epilepsy on electronic tablets via questionnaires developed on the Open Data Kit platform (https://opendatakit.org/). The collected data was further processed in Microsoft Excel 2016 and exported to Stata version 13 (StataCorp LLC, College Station,Texas, USA) for statistical analysis. Statistical analysis 2012 prevalence of NS and OFE in Kitgum and Pader We calculated the 2012 epilepsy prevalence using the popu- lation estimates of the selected villages during that period [29, 30]. Age-specific prevalence was obtained using the age structure obtained for each village during the 2017 survey. 2017 Prevalence of NS and OFE in Kitgum, Pader, and Moyo For the 2017 survey, crude estimates for NS and OFE prevalence were calculated by dividing the number of confirmed cases by the total population of the house- holds visited during the survey. For better comparison, differences in the age distribution of the study popula- tions were overcome by applying direct standardization on the crude age-specific prevalence rates. We used the population age structure (age groups: 0–9, 10–19, 20–29, 30–34, 35–39, 40–44, ≥ 40 years) reported by Uganda Housing and Population Census as a reference population [31]. To calculate the age-standardized prevalence of NS and OFE per village, we used the 2017 Uganda national bureau of statistics (UBOS) projected mid-year population estimates for each study village [31]. We calculated the sensitivity and specificity of a posi- tive answer to the questions “luc luc” and “lili” for a con- firmed diagnosis of respectively NS and OFE. Given that no confirmation of “luc luc” and “lili” diagnosis was done in 2012, and as the same VHT were involved in 2012 and the 2017 surveys, the sensitivity/specificity of the “luc luc” and “lili” diagnoses obtained during the 2017 validation process were applied to correct the 2012 find- ings. The comparison between 2012 and 2017 preva- lence was done using Fisher’s and Pearson’s Chi-square tests at 95% level of significance. Cumulative incidence of NS and OFE The cumulative incidence (or incidence proportion) of NS/OFE in Kitgum and Pader was estimated using data from the 2 yr preceding the 2012 and 2017 surveys. For the period before biannual CDTI and river larviciding, the total number of NS with onset between 2011 and 2012 as revealed by the 2012 survey was halved to obtain the average number of incident cases per year. This was then divided by the total population of the study sites as Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 5 of 11 of 2012. Similarly in 2017, we summed the number of cases with onset between 2016 and 2017, divided by two to obtain the average number of incident cases per year, and used the 2017 survey population as the denomin- ator. We assumed a stable village population for the 2 yr used to calculate incidence. The cumulative incidence was expressed per 100 000 persons per year. Given that cumulative incidence measures proportions, compari- sons were done using the Chi-squared test. Results The 2012 household survey in Kitgum and Pader district The number of new OFE cases every year had been rela- tively stable until the late 1990’s when there was a sharp rise that peaked in 2007, followed by a decline in new cases after 2008 (Fig. 2). The earliest NS cases according to the 2012 data appeared in 1989; after 1998, the num- ber of new NS cases began to rise in a similar time pat- tern as OFE cases. The 2017 household survey in Kitgum, Pader, and Moyo In the three districts of Kitgum, Pader, and Moyo, a total of 390 households were visited from which 2138 persons were screened for epilepsy (Table 2). More than 95% of the households in every selected vil- lage participated. A total of 1598/2138 (74.7%) partici- pants were aged below 30 years; the median age was 16 years (Interquartile Range [IQR]: 14–49). Death of a family member with epilepsy was reported by 21 (5.6%) households (24 deaths in 21 households). Overall, 1746 (81.7%) participants reported having taken ivermectin during the last mass distribution (April 2017). Of the household members screened, 163 were suspected to have epilepsy and the diagnosis was confirmed in 158/ 163 (96.9%) of them (Table 2). Of the five individuals not confirmed to have epilepsy, four were diagnosed as febrile seizures and one as mentally disabled. The median age of PWE was 19 years (IQR: 16–23), and the median age at onset of seizures was 9 years (IQR: 6–14). The most frequent seizure type, other than NS, was generalized tonic-clonic seizures. Among the PWE who had consented to Ov16 testing, the percent- age of Ov16 seropositivity was 27.7% in Kitgum, 50.0% in Pader, and 36.4% in Moyo. Nine (32.1%) of the 28 persons with NS were Ov16 seropositive compared to 28/79 (35.4%) persons with OFE. Based on 1147/1177 (97.5%) household members and 81/112 (72.3%) confirmed cases of NS/epilepsy in Kit- gum and Pader who had complete information, we assessed the performance of the “luc luc” and “lili” ques- tions for the diagnosis of confirmed NS/epilepsy as fol- lows: sensitivity is 98.8% (80/81) and specificity is 97.0% (1034/1066). each screening question in the discrimination between NS and OFE are summarized in Table 3. (The detailed characteristics of Prevalence of NS and OFE in 2017 in Kitgum, Pader, and Moyo The highest crude prevalence of NS was observed in Pader (5.1% in Paikati-Akidi village). In Kitgum district, Fig. 2 Number of new NS and OFE cases per year according to the 2012 community census in Kitgum and Pader districts Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 6 of 11 Table 2 Household and individual characteristics of the 2017 epilepsy survey (Kitgum, Pader and Moyo districts) Kitgum Pader Kitgum & Pader Moyo Household characteristics Number of households: n (%) Median household size: n (IQR) Agriculture as main activity: n (%) Family history of death from epilepsy: n (%) Age (years) at death of PWEa: median (IQR) Study population Number of participants: n (%) Age: median (IQR) Male gender: n (%) Ivermectin use in 2017: n (%) Age distribution (years): n (%) 0–9 10–19 20–29 30–39 ≥ 40 Confirmed persons with epilepsy Number of PWEa: Male gender: n (%) Female gender: n (%) Age in years: median (IQR) Age at seizure onset (years): median (IQR)b Year of onset of seizures: median (IQR)b Positive Ov16 test: n (%)c 142 (36.4) 6 (4–8) 134 (94.4) 15 (10.6) 16 (13–19) 861 (40.2) 16 (8–29) 419 (50.5) 716 (83.5) 258 (29.8) 260 (30.2) 131 (15.2) 77 (8.9) 135 (15.7) 83 49 (59.0) 55 (14.1) 6 (4–8) 54 (98.2) 2 (3.36) 33 (22–44) 316 (14.8) 17 (7–26) 163 (53.4) 249 (79.6) 98 (31.0) 89 (28.2) 61 (19.3) 20 (6.3) 48 (15.2) 29 17 (58.6) 197 (50.5) 6 (4–8) 188 (98.9) 17 (8.6) 16 (13–19) 1177 (55.1) 16 (7–28) 582 (51.3) 965 (82.5) 356 (30.3) 349 (29.7) 192 (16.3) 97 (8.2) 183 (15.6) 112 66 (58.9) 193 (49.5) 5 (3–7) 184 (95.3) 7 (3.6) 17 (14–28) 961 (45.0) 16 (8–30) 442 (48.2) 779 (81.2) 279 (29.0) 268 (27.9) 154 (16.0) 93 (9.7) 167 (17.9) 46 21 (45.7) 34 (41.0) 19 (16–22) 9 (7–12) 2007 (2005–2010) 12 (41.4) 16 (14–18) 8 (7–11) 2010 (2009–2012) 46 (41.1) 18 (16–20) 9 (7–12) 2008 (2006–2011) 25 (54.3) 23 (15–36) 10 (2–15) 2009 (2000–2012) 13 (27.7) 8 (50.0) 21 (33.3) 16 (36.4) IQR Interquartile range aPWE Persons with epilepsy, including those with nodding seizures b27 missing cOnly a limited number of persons were tested (test not available or participant declined to be tested) the crude prevalence of NS was 4.6 and 4.4% in Tumangu and Kampala-Anyuka villages respectively. On the other hand, OFE were more prevalent in the Kitgum district (7.8% in Tumangu and 4.2% in Kampala- Anyuka) compared to Pader (4.1% in Paikati-Akidi village). The 10–19 years age group was most affected by NS, while OFE were more frequent among the 20–29 years old. The prevalence of epilepsy in Moyo was 4.8%, and no nodding seizures were reported. The difference in the age-specific prevalence of OFE between Moyo and other study sites was most significant in the 10–19 years age group (P = 0.005) (Table 4). Using data collected during the 2017 survey, we ob- served that before 2007 there had been a gradual in- crease in new cases per year for both NS and OFE in Kitgum and Pader (Fig. 3). There was a peak in 2007 for Table 3 Performance of the screening questions “luc luc” and “lili” in diagnosing NS and OFE during the 2017 survey Yes to “luc luc”a or “Lili”b Yesc to “luc luc” only, or to both “luc Luc” and “Lili” Yes to “lili” only Confirmed nodding syndrome: n (%) Other forms of epilepsy: n (%) Non-epilepticd: n (%) Total 38 (46.9) 42 (51.9) 1 (1.2) 81 33 (63.5) 18 (34.6) 1 (1.9) 52 5 (17.2) 24 (83.8) 0 29 aLocal term for nodding syndrome bLocal term for other forms of epilepsy cIncludes a history of nodding syndrome, with or without other forms of seizures (NS and NS+) dSuspected case, but not confirmed as epilepsy NS Nodding Syndrome only, NS plus Nodding Syndrome and other seizures, OFE Other Forms of Epilepsy Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 7 of 11 Table 4 The crude and age-specific prevalence of nodding syndrome and other forms of epilepsy in the 2017 survey, in Kitgum, Pader and Moyo districts NS and NS plus cases only Crude prevalence of NS: % 95% confidence interval of prevalence Age-standardized prevalence rate of NS: % Age-specific prevalence of NS: n (%) 0–9 years 10–19 years 20–29 years 30–39 years ≥ 40 years OFE only Crude prevalence of OFE: % 95% confidence interval of prevalence Age-standardized prevalence of OFE: % Age-specific prevalence of OFE: n (%) 0–9 years 10–19 years 20–29 years 30–39 years ≥ 40 years All epilepsy cases (NS + OFE) Crude epilepsy prevalence: % 95% confidence interval of prevalence Kitgum n = 861 4.3 3.1–5.9 3.7 1 (0.4) 27 (10.4) 9 (6.9) 0 0 5.3 4.0–7.1 5.1 2 (0.8) 15 (5.8) 25 (19.1) 1 (1.3) 3 (2.2) 9.6 7.8–11.9 Age-standardized prevalence rate of epilepsy: % 8.8 Pader n = 316 5.1 3.0–8.3 4.4 1 (1.0) 14 (15.7) 1 (1.6) 0 0 4.1 2.3–7.1 3.7 2 (2.0) 8 (9.0) 2 (3.4) 1 (5.3) 0 9.2 6.3–13.1 7.9 Kitgum & Pader n = 1177 Moyo n = 961 P-value 4.5 3.4–5.9 3.8 2 (0.6) 41 (11.8) 10 (5.2) 0 0 5.0 3.9–6.5 4.5 4 (1.1) 23 (6.6) 27 (14.1) 2 (2.1) 3 (1.6) 9.5 7.9–11.4 8.3 0 0 0 0 0 0 0 0 4.8 3.6–6.4 4.6 9 (3.2) 5 (1.9) 17 (11.1) 4 (4.3) 11 (6.6) 4.8 3.6–6.4 4.6 NA NA NA NA NA NA NA 0.831 0.98 0.089a 0.005b 0.412b 0.442a 0.026a < 0.001 < 0.001 IQR Interquartile range, NS Nodding Syndrome only, NS plus Nodding Syndrome and other seizures, OFE Other Forms of Epilepsy, NA Not applicable, CI Confidential interval aFisher’s exact chi-square test bPearson chi-square test OFE and in both 2007 and 2009 for NS. After 2013, there was a steady decline in the number of new cases of NS and OFE; eventually, no new case of NS was recorded for the year 2017. Comparison of prevalence and incidence of NS and OAE in selected villages in Kitgum and Pader districts between 2012 and 2017 Over a five-year period (2012–2017), the prevalence of NS and OFE did not change significantly in Kitgum and Pader. However, the cumulative incidence of all forms of epilepsy decreased from 1165 to 130 per 100 000 persons per year (P = 0.002); that of NS decreased from 490 to 43 per 100 000 persons per year (P = 0.037); and for OFE from 675 to 87 per 100 000 persons per year (P = 0.024) (Table 5). The number of new NS cases per year dropped from four in 2012, to zero in 2017 (Fig. 3). Comparison of ages of persons with NS and OFE in Kitgum and Pader districts between 2012 and 2017 Compared to the 2012 data, there was an age shift of PWE to older age groups in 2017. The median age of all persons with epilepsy shifted from 13.5 years (IQR: 11.0–15.0) in 2012 to 18.0 years (IQR: 15.0–20.3) in 2017; P < 0.001 (Fig. 4). Discussion The 2012 epilepsy survey shows that in the absence of onchocerciasis elimination measures, a simultaneous sharp increase in NS and OFE cases could occur as was observed in Kitgum and Pader (Fig. 2). We discussed the potential reasons for this increase in a previous publica- tion [32]. Most likely, they are multifactorial and include the close proximity of IDP camps to the blackfly breed- ing sites, coupled with no access to ivermectin. Other factors that may have played a role could be the poor Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 8 of 11 Fig. 3 Number of new NS and OFE cases per year of onset in Kitgum, Pader and Moyo districts according to the 2017 survey nutritional status of the children during the war, making them susceptible to more severe O. volvulus infection as well as other health conditions. Another event that oc- curred prior to the NS epidemic was the theft of about 300 000 cows from the Acholi people in northern Uganda [33]. This is relevant for onchocerciasis because these Onchocerca ochengi-infected cows may have played a protective role against human O. volvulus infection in these villages in the past [34]. Our 2017 epilepsy survey shows the effects of optimal onchocerciasis elimination interventions (aerial spraying of breeding sites, ground larviciding of rivers, and bi- annual ivermectin distribution) on the incidence and prevalence of both NS and OFE. In 2017, the prevalence of epilepsy in Kitgum and Pader remained high. This was expected because 5 yr of optimal onchocerciasis control may not be enough to alter the prevalence of OAE, especially if PWE is treated with anti-epileptic Table 5 Comparison of crude prevalence and incidence of nodding syndrome and other forms of epilepsy in Kitgum and Pader districts during the 2012 and 2017 Selected villages All forms of epilepsy 2012 2017 Nodding syndrome Other forms of epilepsy P-value 2012 2017 P-value 2012 2017 P-value Prevalence in Kitgum district Kampala-anyuka: n (%) Tumangu: n (%) Prevalence in Pader district Paikati-akidi: n (%) 54 (11.5) 24 (10.1) 47 (8.5) 35 (12.3) 0.110 0.392 22 (4.6) 10 (4.2) 24 (4.4) 13 (4.6) 0.798 0.811 32 (6.9) 14 (5.8) 23 (4.2) 22 (7.8) 0.061 0.381 32 (11.9) 30 (9.5) 0.347 13 (4.8) 16 (5.1) 0.905 19 (7.0) 14 (4.4) 0.166 Kitgum and Pader districts combined Prevalence: n (%) 95% CI of prevalence Cumulative incidencea 110 (11.2) 10.6–14.8 112 (9.7) 8.1–11.6 0.251 45 (4.6) 3.4–6.1 53 (4.6) 3.5–6.0 1.00 65 (6.6) 5.2–8.4 59 (5.1) 3.9–6.6 0.134 1165 130 0.002 490 43 0.037 675 87 0.024 2012 population distribution of: Kampala-anyuka: n = 469, Tumangu: n = 241, Paikati-akidi: n = 269 2017 population distribution of: Kampala-anyuka: n = 552, Tumangu: n = 284, Paikati-akidi: n = 317 All comparisons done using Pearson chi-square test CI Confidential interval aCumulative incidence per 100 000 persons per year Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 9 of 11 load in the population. Contrary to previous reports from the Ugandan Ministry of Health [25], we encoun- tered 3 NS cases who reported an onset of the seizures after 2013. Potentially, a recall bias concerning the exact year of onset of NS could explain these late-onset cases. The age shift observed in persons with NS and OFE to older age groups is most likely explained by a decreased incidence of OAE in the 5–18 years old age group and an increased survival because of increased access to anti- epileptic treatment and better nutrition. Similar epi- demiological trends of NS and OFE suggests that both conditions may be sharing the same risk factors and etiological agent. The Ov16 seroprevalence obtained using the rapid test was similar among persons with NS and OFE, and much lower than the 66.7% Ov16 ELISA prevalence reported among NS cases in the same area in 2009 [6]. The low Ov16 prevalence in 2017 corroborates with the findings of Weil et al. which showed a reduced sensitivity of the Ov16 rapid tests especially after treat- ment with ivermectin [35]. The low level of Ov16 preva- lence among OFE cases is also explained by the fact that in 2017 less PWE presented with OAE. load during childhood [20]. While it Our study illustrates the importance of strengthening onchocerciasis elimination programs, particularly in areas with high epilepsy prevalence. Recent epilepsy sur- veys in Cameroon and Tanzania have shown that annual CDTI has only a limited effect on the incidence of OAE [19, 36]. A recent prospective study in Cameroon also showed that the risk to develop epilepsy depends on the microfilarial is known that the microfilarial load in an infected individ- ual tends to increase with age, it has also been reported that children develop NS at an earlier age than OFE, while onchocercal blindness is observed in general only after the age of 20 years. This suggests that the microfi- larial threshold to develop epilepsy is lower than for blindness; it also explains why annual CDTI may have decreased onchocercal blindness dramatically but does not seem to decrease the community microfilarial load enough to substantially impact the incidence of epilepsy. While our study supports the growing epidemiological evidence that infection with O. volvulus may be directly or indirectly associated with epilepsy, the pathophysiological mechanism of how this may happen still needs to be eluci- dated. A recently published post-mortem study of five persons from northern Uganda who died of NS suggested that NS is a tauopathy and a neurodegenerative disease [37]. However, these histopathological findings are most likely the consequence of NS rather than its cause. Indeed, we carried out another post-mortem study in the same area, among five persons with NS and four persons with another type of OAE. This study revealed similar neuro- inflammatory histopathological changes with mild to tau-immunoreactive neurofibrillary sparse deposits of Fig. 4 Age of persons with NS and OFE identified during the 2012 and 2017 surveys in Kitgum and Pader drugs such that they survive and therefore do not exit the patient pool for several years. The high prevalence of epilepsy in Moyo is not sur- prising because the villages included in this study were also located in an onchocerciasis-endemic region. Des- pite long-standing annual, then later bi-annual distribu- tion of ivermectin in Moyo, the number of new epilepsy cases remained relatively stable every year. The average CDTI coverage in Moyo in the past decade has been greater than 85% [24], and the 2017 Ov16 seropreva- lence among PWE in Moyo was 36.4%. Whether on- going onchocerciasis transmission could be the cause of the persistently high epilepsy incidence observed in this district needs to be investigated. The medical history of most PWE excluded a number of common causes of epi- lepsy such as perinatal anoxia, brain trauma, cerebral malaria, and meningitis. is possible however that It neurocysticercosis was responsible for part of this high epilepsy prevalence. In contrast to villages Moyo district, there was a steady decline of new epilepsy cases in Kitgum and Pader following the introduction of annual CDTI in 2009 and a further decrease after river larviciding was initiated in 2012. This suggests that in areas of high onchocerciasis endemicity, the combination of bi-annual ivermectin dis- tribution and ground larviciding of rivers may offer bet- ter and faster protection of the population against OAE because it results in a rapid decrease of the microfilarial Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 10 of 11 tangles in 4/5 persons with NS and 2/4 persons with an- other form of OAE, suggesting that NS and other forms of OAE are caused by a similar phenomenon [38]. The tau deposits most likely are induced by repetitive seizures [39] with seizure-associated hypoxia [40] and possibly repeated head injuries [41]. The decreasing incidence of both NS and OFE after strengthening onchocerciasis elimination measures suggests that both conditions may be O. volvu- lus-related. The strength of our study is that it is a population- based study comparing the same villages over time and comparing two different onchocerciasis foci exposed to different onchocerciasis control measures. However, our study also has several limitations. The 2012 and 2017 survey used slightly different methodologies, complicat- ing the comparison of the prevalence and incidence data. Moreover, the diagnosis of epilepsy was mainly made by clinical officers or medical doctors but was not con- firmed by a neurologist. In addition, no laboratory inves- tigations were performed besides the Ov16 testing in some participants. Other causes of epilepsy were only excluded by medical history. The incidence of epilepsy in our study was based on interviewing PWE and not prospectively by an epilepsy surveillance system. A set- back of this approach is the introduction of recall bias, especially regarding the year of epilepsy onset for PWE who had died before the surveys. Therefore our inci- dence data becomes less reliable as the year of onset is further in the past. Another limitation is the fact that a different team of clinical officers participated in the survey in Moyo; however, they had received the same training as those who investigated Kitgum and Pader districts and were supervised by the same medical doctor (NG). Because there was no local word for NS in Moyo, it is possible that some NS cases could have gone unrecognized. In 2012, in the high epilepsy prevalence villages in Kitgum and Pader, the prevalence of NS might have been overestimated. Indeed, because of national and international attention about NS, chil- dren with OFE could have been reported by family members to have NS in order to attract more support and interest of funders (Wamala J, personal commu- nication). Our 2017 survey results showed that 34.6% of the “luc luc” screened by the VHT were not con- firmed to be NS cases but rather as OFE. This figure is only slightly lower than the 45% reported in the CDC/MOH validation study performed in 2013 [26]. The high specificity and sensitivity of the questions “luc luc” or “lili” for screening PWE, and the fact that these questions were asked during both surveys, in- creases the reliability of our findings when consider- ing the overall prevalence and incidence of epilepsy (including NS) between 2012 and 2017. Conclusions Our study illustrates the importance of the public health problem, not only caused by NS but also by OFE in places where onchocerciasis elimination measures are not implemented or are sub-optimal. The lesson learned from the epilepsy epidemic in northern Uganda is that onchocerciasis elimination strategies need to be revised and retargeted, taking into account OAE. Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s40249-020-0628-3. Additional file 1. 2012 Village Health Team (VHT) nodding syndrome case report form. Abbreviations AFENET: African Field Epidemiology Network; CDC: Centers for Disease Control and Prevention; CDTI: Community directed treatment with ivermectin; IDP: Internally displaced persons; IQR: Interquartile pange; LRA: Lord’s Resistance Army; MakSPH: Makerere School of Public Health; MOH: Ministry of Health; NS: Nodding syndrome; NS +: Nodding syndrome plus.; OFE: Other forms of epilepsy; RDT: Rapid diagnostic test; VHT: Village health teams; WHO: World Health Organization Acknowledgments We are grateful to all the doctors, clinical officers and VHT who performed the study; and the population of the villages for their participation. Authors’ contributions MI, OB, RI, WJF, KJ, OB and LT participated in the 2012 survey. CR, GN, and MKM wrote the protocol. GN and CR coordinated the 2017 survey. GN and CR wrote the first draft of the paper. MF, SFJN, and HA performed the analysis and assisted in the writing of the paper. All authors critically reviewed and approved the final manuscript. Funding R Colebunders received funding from the European Research Council (ERC) grant number 671055, project title NSETHIO. Availability of data and materials All collected data is confidentially kept at both the Global Health Institute, University of Antwerp (Belgium) and the Infectious Disease Institute in Kampala (Uganda). The datasets are available from the corresponding author on a reasonable request. Ethics approval and consent to participate The 2012 survey was performed by the Ugandan Ministry of Health as a non-research public health response activity. Ethical approval for the 2017 survey was obtained from the ethical committees of Lacor hospital in Gulu and the University of Antwerp and Ugandan National Council for Science and Technology (study reference HS77ES). Informed consent was obtained from all study participants and assent from children > 12 years old. Consent for publication Written informed consent was obtained from all persons described in this paper. Competing interests The authors declare that they have no competing interests. Author details 1Busitema University, Mbale, Uganda. 2Infectious Disease Institute, Makerere University, Kampala, Uganda. 3Global Health Institute, University of Antwerp, Antwerp, Belgium. 4Department of Population Studies, School of Statistics, Makerere University, Kampala, Uganda. 5Department of Paediatrics and Child Health, College of Health Sciences, Makerere University, Kampala, Uganda. Gumisiriza et al. Infectious Diseases of Poverty (2020) 9:12 Page 11 of 11 6Ministry of Health, Kampala, Uganda. 7World Health Organization, Juba, South Sudan. Received: 28 August 2019 Accepted: 14 January 2020 References 1. World Health Organisation. International scientific meeting on nodding syndrome, Kampala. 2012. https://www.who.int/neglected_diseases/ diseases/Nodding_syndrom_Kampala_Report_2012.pdf?ua=1. Accessed 9 Dec 2019. Dowell SF, Sejvar JJ, Riek L, Vandemaele KAH, Lamunu M, Kuesel AC, et al. Nodding syndrome. Emerg Infect Dis. 2013;19:1374–84. Sejvar JJ, Kakooza AM, Foltz JL, Makumbi I, Atai-Omoruto AD, Malimbo M, et al. Clinical, neurological, and electrophysiological features of nodding syndrome in Kitgum, Uganda: an observational case series. Lancet Neurol. 2017;12:166–74. 3. 2. 4. Wamala JF, Malimbo M, Tepage F, Lukwago L, Okot CL, Cannon RO, et al. 5. 6. 7. 8. Nodding syndrome may be only the ears of the hippo. PLoS Negl Trop Dis. 2015;9:1–7. Colebunders R, Njamnshi AK, van Oijen M, Mukendi D, Kashama JM, Mandro M, et al. Onchocerciasis-associated epilepsy: from recent epidemiological and clinical findings to policy implications. Epilepsia Open. 2017;2:145–52. Foltz JL, Makumbi I, Sejvar JJ, Malimbo M, Ndyomugyenyi R, Atai-Omoruto AD, et al. An epidemiologic investigation of potential risk factors for nodding syndrome in Kitgum District, Uganda. PLoS One. 2013. https://doi.org/10.1371/journal.pone.0066419. Kaiser C, Rubaale T, Tukesiga E, Kipp W, Asaba G. Case report: nodding syndrome, Western Uganda, 1994. Am J Trop Med Hyg. 2015;93:198–202. Tumwine JK, Vandemaele K, Chungong S, Richer M, Anker M, Ayana Y, et al. Clinical and epidemiologic characteristics of nodding syndrome in Mundri county, southern Sudan. Afr Health Sci. 2012;12:242–8. 9. Winkler AS, Wallner B, Friedrich K, Pfausler B, Unterberger I, Matuja W, et al. 10. 11. 12. 13. A longitudinal study on nodding syndrome-a new African epilepsy disorder. Epilepsia. 2014;55:86–93. Siewe JFN, Ngarka L, Tatah G, Mengnjo MK, Nfor LN, Chokote ES, et al. Clinical presentations of onchocerciasis-associated epilepsy (OAE) in Cameroon. Epilepsy Behav. 2019;90:70–8. Siewe Fodjo JN, Mandro M, Mukendi D, Tepage F, Menon S, Nakato S, et al. Onchocerciasis-associated epilepsy in the Democratic Republic of Congo: clinical description and relationship with microfilarial density. PLoS Negl Trop Dis. 2019. https://doi.org/10.1371/journal.pntd.0007300. Kipp W, Kasoro S, Burnham G. Onchocerciasis and epilepsy in Uganda. Lancet. 1994;343:183–4. Kaiser C, Asaba G, Leichsenring M, Kabagambe G. High incidence of epilepsy related to onchocerciasis in West Uganda. Epilepsy Res. 1998;30: 247–51. 14. Boussinesq M, Pion SDS, Demanga-Ngangue A, Kamgno J. Relationship between onchocerciasis and epilepsy: a matched case-control study in the Mbam Valley, Republic of Cameroon. Trans R Soc Trop Med Hyg. 2002;96: 537–41. 15. Pion SDS, Kaiser C, Boutros-Toni F, Cournil A, Taylor MM, Meredith SEO, et al. Epilepsy in onchocerciasis endemic areas: systematic review and meta- analysis of population-based surveys. PLoS Negl Trop Dis. 2009. https://doi. org/10.1371/journal.pntd.0000461. Levick B, Laudisoit A, Tepage F, Ensoy-Musoro C, Mandro M, Bonareri Osoro C, et al. High prevalence of epilepsy in onchocerciasis endemic regions in the Democratic Republic of the Congo. PLoS Negl Trop Dis. 2017;11:1–14. Lenaerts E, Mandro M, Mukendi D, Suykerbuyk P, Dolo H, Rossi DW, et al. High prevalence of epilepsy in onchocerciasis endemic health areas in Democratic Republic of the Congo. Infect Dis Poverty. 2018;7:68. 16. 17. 18. Colebunders R, Carter JY, Claver P, Puok K, Bhattacharyya S, Menon S, et al. High prevalence of onchocerciasis-associated epilepsy in villages in Maridi County, republic of South Sudan: a community-based survey. Seizure. 2018; 63:93–101. 19. Mmbando BP, Suykerbuyk P, Mnacho M, Kakorozya A, Matuja W, Hendy A, et al. High prevalence of epilepsy in two rural onchocerciasis endemic villages in the Mahenge area, Tanzania, after 20 years of community directed treatment with ivermectin. Infect Dis Poverty. 2018;7:64. 20. Chesnais CB, Nana-Djeunga HC, Njamnshi AK, Lenou-Nanga CG, Boulle C, Bissek A-CZ-K, et al. The temporal relationship between onchocerciasis and epilepsy: a population-based cohort study. Lancet Infect Dis. 2018;18:1278–86. 21. Bureau for democracy, conflict, and humanitarian assistance. Uganda- Complex Emergency. 2006. http://www.cidi.org/wp-content/uploads/ uganda-complex-sr2008-02-2008-10-30.pdf. Accessed 24 Nov 2019. Katabarwa MN, Lakwo T, Habomugisha P, Unnasch TR, Garms R, Hudson- Davis L, et al. After 70 years of fighting an age-old scourge, onchocerciasis in Uganda, the end is in sight. Int Health. 2018;10:i79–88. 22. 23. Ministry of Health, Republic of Uganda. National onchocerciasis control program. 2018. http://health.go.ug/programs/national-onchocerciasis- control-program-0. Accessed 15 Nov 2019. 24. Oguttu DW. Interventions and progress towards onchocerciasis elimination in Uganda. In: DNDi Partners’ meeting. Kampala. 2018. https://www.dndi. org/wp-content/uploads/2018/05/Oguttu_CountryProgressUganda_ UgandaMeeting2018.pdf. Accessed 8 Dec 2019. 25. Ministry of Health, Republic of Uganda. Statement of nodding syndrome in 26. Uganda. 2018. http://health.go.ug/download/file/fid/1800. Accessed 8 Dec 2019. Iyengar P, Wamala J, Ratto J, Blanton C, Malimbo M, Lukwago L, et al. Prevalence of nodding syndrome-Uganda, 2012-2013. Morb Mortal Wkly Rep. 2014;63:603–6. 27. Diagana M, Preux PM, Tuillas M, Ould Hamady A, Druet-Cabanac M. 28. Depistage de l’epilepsie en zones tropicales: validation d’un questionnaire en Mauritanie. Bull Soc Pathol Exot. 2006;99:103–7. Fisher RS, Acevedo C, Arzimanoglou A, Bogacz A, Cross JH, Elger CE, et al. ILAE official report: a practical clinical definition of epilepsy. Epilepsia. 2014;55:475–82. 29. Uganda Bureau of Statistics. The national population and housing census 2014 – Area specific profiles, Kitgum district. 2017. https://www.ubos.org/ wp-content/uploads/publications/2014CensusProfiles/KITGUM.pdf. Accessed 12 Nov 2019. 30. Uganda Bureau of Statistics. The national population and housing census 2014 – area specific profiles, Pader District 2017. Available from: https://www.ubos.org/wp-content/uploads/publications/2014 CensusProfiles/PADER.pdf. Accessed 12 Nov 2019. 31. Uganda Bureau of Statistics. National population and housing census, 2014- main report. 2016. https://www.ubos.org/wp-content/uploads/publications/ 03_20182014_National_Census_Main_Report.pdf. Accessed 12 Nov 2019. 32. Colebunders R, Siewe FJN, Hotterbeekx A. Onchocerciasis-associated epilepsy, an additional reason for strengthening Onchocerciasis elimination programs. Trends Parasitol. 2018;34:208–16. 33. Mirzeler M, Young C. Pastoral politics in the northeast periphery in Uganda: AK-47 as change agent. J Mod Afr Stud. 2000;38:407–29. 34. Wahl G, Enyong P, Ngosso A, Schibel JM, Moyou R, Tubbesing H, et al. Onchocerca ochengi: epidemiological evidence of cross-protection against Onchocerca volvulus in man. Parasitology. 1998;116:349–62. 35. Weil GJ, Steel C, Liftis F, Li B, Mearns G, Lobos E, et al. A rapid-format 36. antibody card test for diagnosis of Onchocerciasis. J Infect Dis. 2000;182: 1796–9. Siewe Fodjo JN, Tatah G, Tabah EN, Ngarka L, Nfor LN, Chokote SE, et al. Epidemiology of onchocerciasis-associated epilepsy in the Mbam and Sanaga river valleys of Cameroon: impact of more than 13 years of ivermectin. Infect Dis Poverty. 2018;7:114. 37. Pollanen MS, Onzivua S, Robertson J, Mckeever PM, Olawa F, Kitara DL, et al. Nodding syndrome in Uganda is a tauopathy. Acta Neuropathol. 2018;136: 691–7. 38. Hotterbeekx A, Lammens M, Idro R, Akun PR, Lukande R, Akena G, et al. Neuroinflammation and not Tauopathy is a predominant pathological signature of nodding syndrome. J Neuropathol Exp Neurol. 2019;78:1049–58. Tai XY, Koepp M, Duncan JS, Fox N, Thompson P, Baxendale S, et al. Hyperphosphorylated tau in patients with refractory epilepsy correlates with cognitive decline: a study of temporal lobe resections. Brain. 2016;139:2441–55. 39. 40. Wu JW, Hussaini SA, Bastille IM, Rodriguez GA, Mrejeru A, Rilett K, et al. 41. Neuronal activity enhances tau propagation and tau pathology in vivo. Nat Neurosci. 2016;19:1085–92. Santangelo R, Cecchetti G, Bernasconi MP, Cardamone R, Barbieri A, Pinto P, et al. Cerebrospinal fluid amyloid-beta 42, total tau and phosphorylated tau are low in patients with normal pressure hydrocephalus: analogies and differences with Alzheimer’s disease. J Alzheimers Dis. 2017;60:183–200.
10.1265_ehpm.22-00245	Environmental Health and Preventive Medicine RESEARCH ARTICLE Environmental Health and Preventive Medicine (2023) 28:22 https://doi.org/10.1265/ehpm.22-00245 Cross-sectional associations between early mobile device usage and problematic behaviors among school-aged children in the Hokkaido Study on Environment and Children’s Health Chihiro Miyashita1, Keiko Yamazaki1, Naomi Tamura1, Atsuko Ikeda-Araki1,2, Satoshi Suyama3, Takashi Hikage4, Manabu Omiya5, Masahiro Mizuta6 and Reiko Kishi1＊＊Correspondence: rkishi@med.hokudai.ac.jp 1Center for Environmental and Health Sciences, Hokkaido University, Sapporo, Japan. 2Faculty of Health Sciences, Hokkaido University, Sapporo, Japan. 3Funded Research Division of Child and Adolescent Psychiatry, Hokkaido University Hospital, Sapporo, Japan. 4Graduate School, Faculty of Information Science and Technology, Hokkaido University, Sapporo, Japan. 5Information Initiative Center, Hokkaido University, Sapporo, Japan. 6Center for Training Professors in Statistics, The Institute of Statistical Mathematics, Tokyo, Japan. Abstract Background: Concerns have been raised about the adverse health impacts of mobile device usage. The objective of this cross-sectional study was to examine the association between a child’s age at the ﬁrst use of a mobile device and the duration of use as well as asso- ciated behavioral problems among school-aged children. Methods: This study focused on children aged 7–17 years participating in the Hokkaido Study on Environment and Children’s Health. Between October 2020 and October 2021, the participants (n = 3,021) completed a mobile device use-related questionnaire and the strengths and diﬃculties questionnaire (SDQ). According to the SDQ score (normal or borderline/high), the outcome variable was behavioral problems. The independent variable was child’s age at ﬁrst use of a mobile device and the duration of use. Covariates included the child’s age at the time of survey, sex, sleep problems, internet addiction, health-related quality of life, and history of developmental concerns assessed at health checkups. Logistic regression analysis was performed for all children; the analysis was stratiﬁed based on the elementary, junior high, and senior high school levels. Results: According to the SDQ, children who were younger at their ﬁrst use of a mobile device and used a mobile device for a longer duration represented more problematic behaviors. This association was more pronounced among elementary school children. Moreover, subscale SDQ analysis showed that hyperactivity, and peer and emotional problems among elementary school children, emotional problems among junior high school children, and conduct problems among senior high school children were related to early and long usage of mobile devices. Conclusions: Elementary school children are more sensitive to mobile device usage than older children, and early use of mobile devices may exacerbate emotional instability and oppositional behaviors in teenagers. Longitudinal follow-up studies are needed to clarify whether these problems disappear with age. Keywords: Hokkaido Study on Environment and Children’s Health, Mobile devices, Children, Behavioral problems 1 Introduction The usage of mobile devices, including smartphones and tablets, has rapidly increased across social classes. In 2018, the smartphone penetration rate was approximately 75% and 45% in developed and developing countries, re- spectively [24]. Globally, the age at which a child ﬁrst uses a mobile device is constantly tipping toward earlier ages. Some parents allow their children to use mobile devices at early ages for entertainment purposes. Concerns have been raised about the negative early health impacts of regular contact with mobile devices, especially in relation to neu- robehavioral developmental delays and imbalances in healthy activities in children [1, 26, 28]. The World Health Organization and some developed countries, including the United States of America, Canada, and Australia, have recommended that parents should avoid giving screen- based devices to infants younger than 18 months and re- strict screen time to <1 hour daily for preschool children aged 2–5 years [2, 3, 23, 30, 31]. While these recommendations are based on previous studies that mainly targeted traditional devices such as © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Environmental Health and Preventive Medicine (2023) 28:22 2 of 11 the television, there is limited scientiﬁc evidence regarding the associations between the early use of contemporary mobile devices and development in children. Unstructured play with the hands and body and practical social commu- nication are important for the development of the central nervous system in early infancy [2, 3]. For establishing healthy behaviors, daytime activities, nighttime sleep, and routine mealtimes are essential. Early mobile device usage can interfere with comprehensive neurodevelopment and engagement in healthy activities, both of which foster lan- guage and cognitive development and social skills in in- fancy. According to a Korean study, children aged 1–3 years who spent longer times with touch screens displayed increased emotional problems and depression and anxiety symptoms [15]. Another study reported that the regular use of screen-based media among infants at 4 months was associated with poor performance on self-regulation tests but not cognitive ﬂexibility or working memory tests at assessed at 14 months [17]. In Japan, the internet penetration rate among school- aged children was 93.2% in 2019. The penetration rate rapidly increased from 12.5% to 49.5%, for smartphones and from 15.3% to 41.0%, for tablets, among elementary school students, between 2014 and 2019. More than 50% of children have contact with the internet in their early life, with 4.7% of those aged 0 years and 50.2% of those aged 3 years using the internet [18]. A Japanese study suggested that regular and frequent use of mobile devices among ﬁrst grade elementary school children was associated with in- creased emotional and behavioral problems [9]. However, this study did not assess school children aged >7 years. The association between mobile device usage in early life and developmental eﬀects was not evaluated. Therefore, the current study aimed to assess the association between a child’s age at ﬁrst use of a mobile device and the duration of use and the behavioral problems among elementary, junior high, and senior high school students. 2 Methods 2.1 Subjects This study focused on children aged 7–17 years between 2003 and 2012; the children were followed up until Oc- tober 2020 in a prospective birth cohort study of the Hokkaido Study on Environment and Children’s Health [11–13]. We mailed the strengths and diﬃculties question- naire (SDQ) and a questionnaire regarding mobile device use and lifestyle to 5,221 parent–child pairs between Oc- tober 2020 and January 2021 based on a random selection. A total of 3,364 responses were received by October 2021 (response rate = 64.4%). From the questionnaire, we as- sessed the associations between exposure to mobile de- vices (the child’s age at ﬁrst use of a mobile device and duration of use), the outcomes (child behavioral problems as determined by the SDQ), and potential covariates, in- cluding the child’s sleep problems, internet addiction, health-related quality of life, and history of developmental concerns assessed at health checkups (Fig. 1). The child- ren’s age at survey were recorded based on the response date on the questionnaire. The child’s date of birth and sex were obtained from medical records. Additional informa- tion, including the parents’ age and educational and house- hold incomes, were obtained from the baseline question- naire during maternal pregnancy [11, 13]. Of the 3,364 questionnaires returned, we excluded two pre-school chil- dren and 341 participants with missing response data. A total of 3,021 children, including 1,433 elementary school, 1,121 junior high school, and 467 senior high school chil- dren, were ﬁnally included in the study (Table 1 and Fig. 1). 2.2 Ethics statement The institutional ethics board for epidemiological studies at Hokkaido University Graduate School of Medicine and Hokkaido University Center for Environmental and Health Sciences approved the study protocol (approval number 19 - 118). Informed consent was obtained from all study participants before enrollment. 2.3 Exposure assessment We used 4-type exposure factors to monitor child mobile device usage (Table 2); the ﬁrst exposure factor was the child’s age at the ﬁrst use of a mobile device, according to parents’ answer to the following question: “At what age did your child ﬁrst use a mobile device? For example, you showed movies to your child on mobile devices, or your child used a mobile device by themselves?” The second exposure factor was the duration of mobile device usage, meaning usable years, calculated based on the child’s age at the ﬁrst use of a mobile device and that at the time of Participants were born from 2003 to 2012, and have been followed up until October 2020 in the Hokkaido Study on Environment and Children’s Health (n = 13,899) Random selection of participants to receive mobile device usage and child health questionnaire (n = 5,221) Participants returned mobile device usage and child health questionnaire (n = 3,364) Excluded 2 participants who were pre-school children Excluded 341 participants who had missing data Study participants (n = 3,021) Fig. 1 Flow chart of study participants Environmental Health and Preventive Medicine (2023) 28:22 3 of 11 Table 1 Characteristics of participants (children and their parents). All children Categories Number (%) Mean « SD or Median (IQR) School type Elementary school Number (%) Mean « SD or Median (IQR) Junior high school Number (%) Mean « SD or Median (IQR) Senior high school Number (%) Mean « SD or Median (IQR) p 3021 12.4 « 2.4 1433 10.2 « 1.2 1121 13.7 « 0.9 467 15.8 « 0.5 <0.001 Child Age at survey Sex Siblings History of developmental concerns Boy Girl No Yes No Yes 1499 (49.6) 1522 (50.4) 401 (16.0) 2100 (84.0) 2726 (90.2) 295 (9.8) Personal mobile device use and restriction 564 (18.7) No Having personal mobile devices 2457 (81.3) Yes Restricted use of mobile devices on weekdays No Yes No Yes Restricted use of mobile devices on holidays Child health quality Health-related quality of life Sleep problems Internet addiction 1310 (43.4) 1709 (56.6) 1456 (48.2) 1565 (51.8) 3021 3021 3021 721 (50.3) 712 (49.7) 192 (17.2) 926 (82.8) 1271 (88.7) 162 (11.3) 434 (30.3) 999 (69.7) 389 (27.2) 1042 (72.8) 472 (32.9) 961 (67.1) 550 (49.1) 571 (50.9) 151 (15.4) 830 (84.6) 1028 (91.7) 93 (8.3) 129 (11.5) 992 (88.5) 540 (48.2) 581 (51.8) 589 (52.5) 532 (47.5) 228 (48.8) 239 (51.2) 58 (14.4) 344 (85.6) 427 (91.4) 40 (8.6) 1 (0.2) 466 (99.8) 381 (81.6) 86 (18.4) 395 (84.6) 72 (15.4) 0.766 0.342 0.025 <0.001 <0.001 <0.001 43.0 (38.0, 47.0) 1433 44.0 (40.0, 48.0) 1121 43.0 (37.0, 47.0) 467 41.0 (35.0, 46.0) <0.001 24.0 (22.0, 26.0) 21.0 (16.0, 26.0) 1433 1433 24.0 (22.0, 26.0) 19.0 (15.0, 24.0) 1121 1121 23.0 (22.0, 26.0) 21.0 (17.0, 26.0) 467 467 23.0 (21.0, 25.0) <0.001 23.0 (19.0, 27.0) <0.001 Mother Age at time of survey Educational level 3021 44.1 « 5.1 1433 42.1 « 4.9 1121 45.3 « 4.6 467 47.2 « 4.2 <0.001 ¯9 10–12 13–15 >16 80 (2.6) 1152 (38.1) 1376 (45.5) 413 (13.7) 37 (2.6) 553 (38.6) 630 (44.0) 213 (14.9) 32 (2.9) 421 (37.6) 522 (46.6) 146 (13.0) 11 (2.4) 178 (38.1) 224 (48.0) 54 (11.6) 0.478 Father Age at time of survey Educational level ¯9 10–12 13–15 >16 <3.0 3.0–4.9 5.0–7.9 ²8 Annual household income (million Japanese Yen) 2984 45.7 « 5.9 1419 43.7 « 5.7 1103 46.9 « 5.4 462 48.9 « 5.2 <0.001 167 (5.6) 1099 (36.8) 785 (26.3) 938 (31.4) 510 (19.0) 1230 (45.8) 722 (26.9) 226 (8.4) 82 (5.8) 508 (35.6) 385 (27.0) 451 (31.6) 252 (19.5) 594 (46.0) 347 (26.9) 98 (7.6) 56 (5.1) 409 (37.1) 292 (26.5) 346 (31.4) 194 (19.6) 446 (45.1) 263 (26.6) 87 (8.8) 29 (6.3) 182 (39.6) 108 (23.5) 141 (30.7) 64 (15.7) 190 (46.7) 112 (27.5) 41 (10.1) 0.637 0.481 SD, standard deviation. SDQ, Strength and Diﬃculties Questionnaire. IQR, interquartile range is the 75th and 25th percentile p value by one-way ANOVA, »2 test, and Kruskal-Wallis test, which were used to compare data among elementary, junior high, and senior high school children. survey. The third exposure factor was the child’s age at which they were given personal mobile devices, according to answers by parents to the following question: “At what age did your child ﬁrst receive a personal mobile device such as a cell phone or tablet?” The fourth exposure factor was the duration of personal mobile device usage, indicat- ing that holding years were calculated using the age at which a child had a personal mobile device and that at the time of survey. 2.4 Assessment outcome We used the SDQ of the common methods for assessing behavioral and mental health problems among children and adolescents aged 4–17 in questionnaires, which were completed by the children’s parents or teachers [7]. The SDQ consists of 25 items, each rated as being not true (0), somewhat true (1), or certainly true (2). The items are divided into ﬁve subscales covering conduct problems, hyperactivity, emotional symptoms, peer problems, and Environmental Health and Preventive Medicine (2023) 28:22 4 of 11 Table 2 Child’s age at ﬁrst use of a mobile device and the duration of use. Age at ﬁrst use of a mobile device Duration of mobile device usage Age at ﬁrst having a personal mobile device Duration for having a personal mobile device All children Median (IQR) 7.0 (5.0, 10.0) 5.0 (3.0, 7.0) 12.0 (8.0, 13.0) 2.0 (1.0, 3.0) School type Elementary school children Median (IQR) 6.0 (4.0, 8.0) 4.0 (2.0, 6.0) 8.0 (7.0, 10.0) 2.0 (1.0, 3.0) Junior high school children Median (IQR) 9.0 (6.0, 11.0) 5.0 (3.0, 8.0) 12.0 (11.0, 13.0) 2.0 (1.0, 3.0) Senior high school children Median (IQR) 10.0 (7.0, 12.0) 6.0 (3.0, 9.0) 14.0 (12.0, 15.0) 2.0 (1.0, 4.0) p <0.001 <0.001 <0.001 0.49 IQR, interquartile range. p value by Kruskal-Wallis test, which were used to compare median among elementary, junior high, and senior high school children. prosocial behavior [7]. Summing up the scores on the four subscales, i.e., excluding prosocial behavior, gives the SDQ total diﬃculties score (TDS), which can range from 0 to 40. The TDS from the Japanese version of the SDQ has a cut-oﬀ score of 12/13 for normal/borderline and 15/ 16 for borderline/high [16]. In this study, according to the parents’ answers, the children were divided into two groups—normal or borderline/high—based on the TDS and ﬁve subscales. The cut-oﬀ score for the ﬁve subscales of conduct problems, hyperactivity, emotional symptoms, peer problems, and prosocial behavior were 3/4, 5/6, 3/4, 3/4, and 6/5 for the normal and borderline/high groups, respectively [8]. 2.5 Covariates Based on previous studies, we used several covariates, in- cluding the children’s age at the time of survey, sex, and history of developmental concerns at health checkups, health-related quality of life, sleep problems, internet ad- diction, school type, and interaction between a child’s mo- bile device usage and school type, all of which had poten- tial confounding eﬀects on a child’s mobile device usage and child behavioral problems based on the associations noted in this study (Table 4 and Supplemental Table 2 and 3) [9, 16, 19]. In fact, we assessed the generic health- related quality of life for children using the KIDSCREEN- 10 questionnaire [21]; this questionnaire consists of 10 items, including the physical, psychological, and social di- mensions of wellbeing [25]. We also assessed the tendency toward internet dependence using a modiﬁed Internet Ad- diction Test, which comprised 11 items [29, 32]. These questionnaires were completed by the children. We as- sessed sleep problems in the children based on 19 items from the short version of the sleep questionnaire for chil- dren [22], which was answered by their parents. Moreover, we used covariates to assess the interactions between a child’s mobile device usage and school type. The children included in this cross-sectional study had a wide birth-year period of 10 years, and their mobile device penetration rate had rapidly changed in the meantime [18]. The children’s behavioral problems not only changed as they developed but were also related to schooling from elementary to high school. We conducted school-speciﬁc analyses because we believed that the school type aﬀected both the exposure and outcomes. Meanwhile, we did not use covariates of paren- tal household income, educational levels, and the presence of siblings as these factors were not associated with a child’s mobile device usage in this study (Table 4). This indicates that the association between mobile device usage and social class could be weakening. 2.6 Statistical analysis Simple associations between the parents’ and children’s characteristics and the child’s age at ﬁrst use of a mobile device and duration of use as well as the children’s behav- ioral problems were assessed using one-way ANOVA, the »2 test, and the Kruskal–Wallis test. A logistic regression analysis was performed for all children, stratiﬁed by ele- mentary, junior high, and senior high school; the outcome (TDS and sub-analyses of child behavioral problems by SDQ) was considered the dependent variable, whereas ex- posure (child’s age at ﬁrst use of a mobile device and the duration of use) was considered the independent variable. The covariates included child age at time of survey, sex, sleep problems, internet addiction, health-related quality of life, history of developmental concerns assessed at health checkups, and school type. The logistic regression analysis among children stratiﬁed by school type was adjusted for the same variables, except for school type. p < 0.05 was considered statistically signiﬁcant. All statistical analyses were performed using SPSS software for Windows (ver- sion 21.0J; IBM, Armonk, NY, USA). 3 Results A total of 3,021 children, including 1,433 elementary school, 1,121 junior high school, and 467 senior high school children, were involved in this study. The children’s and parents’ characteristics and the diﬀerences among school types are shown in Table 1. The children’s and parents’ ages at the time of survey, the child’s internet addi- ction score, and the rate of having personal mobile devices increased as the children progressed from elementary to senior high school. The rate of restricted mobile device usage on weekdays and holidays and the health-related quality of life scores decreased as the children progressed from elementary to senior high school. Moreover, the his- tory of developmental concerns assessed at health check ups and sleep problem scores diﬀered among elementary, junior high, and senior high school children (Table 1). Environmental Health and Preventive Medicine (2023) 28:22 5 of 11 Table 3 Number (%) of child behavioral problems based on the TDS and subscales according to SDQ. TDS Categories Normal Borderline/High Overall Number (%) 2586 (85.6) 435 (14.4) School type Elementary school Number (%) 1205 (84.1) 228 (15.9) Junior high school Number (%) 972 (86.7) 149 (13.3) Senior high school Number (%) 409 (87.6) 58 (12.4) p 0.072 Conduct problems Normal Borderline/High 2727 (90.3) 294 (9.7) 1260 (87.9) 173 (12.1) Hyperactivity/inattention Normal Borderline/High 2727 (90.3) 294 (9.7) 1257 (87.7) 176 (12.3) Emotional problems Normal Borderline/High 2598 (86.0) 423 (14.0) 1228 (85.7) 205 (14.3) Peer problems Normal Borderline/High 2594 (85.9) 427 (14.1) 1255 (87.6) 178 (12.4) 1032 (92.1) 89 (7.9) 1036 (92.4) 85 (7.6) 964 (86.0) 157 (14.0) 949 (84.7) 172 (15.3) 435 (93.1) 32 (6.9) 434 (92.9) 33 (7.1) 406 (86.9) 61 (13.1) 390 (83.5) 77 (16.5) <0.001 <0.001 0.798 0.031 Prosocial behavior 1011 (70.7) 418 (29.3) TDS, Total diﬃculties score. SDQ, Strength and Diﬃculties Questionnaire. P value by »2 test, which was used to compare the data among elementary, junior high, and senior high school children. Normal Borderline/High 2006 (66.6) 1005 (33.4) 715 (64.0) 403 (36.0) 280 (60.3) 184 (39.7) <0.001 The median age at ﬁrst use of a mobile device and the duration of use was 7.0 and 5.0 years overall, 6.0 and 4.0 among elementary school children, 9.0 and 5.0 among junior high school children, and 10.0 and 6.0 among senior high school children, respectively (Table 2). The distribu- tion of child age at ﬁrst use of a mobile device is shown in Supplemental Table 1. The number of borderline/high (cases) according to TDS was 435 (14.4%) overall, 228 (15.9%) among elementary school children, 149 (13.3%) among junior high school children, and 58 (12.4%) among senior high school children (Table 3). The ﬁve subscales of childhood behavioral problems according to SDQ are shown in Table 3. The associations between the age at which a child ﬁrst used a mobile device and basic partic- ipant information are shown in Table 4, both unstratiﬁed and stratiﬁed by school type. Among all children, we ob- served positive associations between the child’s age at ﬁrst use of a mobile device and the child’s and parent’s age at the time of survey. Negative associations were observed between the health-related quality of life score and sleep problems score. The age of the children at ﬁrst use of a mobile device diﬀered by their sex, history of develop- mental concerns assessed at health checkups, possession of personal mobile devices, and mobile device restriction on weekdays and holidays. When the children were strati- ﬁed by school type, their age at ﬁrst use of a mobile device was associated with the child’s and parent’s age at the time of survey, their sex, their history of developmental con- cerns assessed at health checkups, their health-related quality of life, their sleep problems, and their internet ad- diction, among at least one school type (Table 4). The associations between child behavioral problems (TDS) and the characteristics of participants among all children and among those stratiﬁed by school type are shown in Supplemental Tables 2 and 3. Among all chil- dren, the prevalence of child behavioral problems diﬀered by the child’s and mother’s age at the time of survey, their sex, presence of siblings, their history of developmental concerns assessed by medical checkup, mobile device re- strictions on weekdays and holidays, their health-related quality of life, their sleep problems, their internet addic- tion, and their maternal educational levels (Supplemental Tables 2 and 3). When stratiﬁed by school type, the prev- alence of the children’s behavioral problems diﬀered by the children’s age at the time of survey, their sex, presence of siblings, their history of developmental concerns as- sessed at health checkups, mobile device restrictions on weekdays and holidays, their health-related quality of life, their sleep problems, and their internet addiction, among at least one school type (Supplemental Tables 2 and 3). According to the logistic regression analysis among all children, compared to the normal group, the adjusted odds ratios of the borderline/high group signiﬁcantly decreased with increasing child age at ﬁrst use of a mobile device (95% CI = 0.85 [0.77, 0.93]); by contrast, it signiﬁcantly increased with increasing duration for use of a mobile device (95% CI = 1.20 [1.08, 1.33]) (Table 5 and Fig. 2). In the logistic regression analysis stratiﬁed by school type, compared to the normal group, the adjusted odds ratios of child behavioral problems for the borderline/high group signiﬁcantly decreased with increasing child’s age at ﬁrst use of a mobile device (95% CI = 0.87 [0.81, 0.93]) and signiﬁcantly increased with increasing duration for use of a mobile device (95% CI = 1.15 [1.08, 1.23]) among ele- mentary school children. However, the adjusted odds ra- tios for the child behavioral problems were not signiﬁ- cantly associated with the child’s age and duration of mobile device use among junior and senior high school Environmental Health and Preventive Medicine (2023) 28:22 6 of 11 Table 4 Associations between child’s age at ﬁrst use of a mobile device and characteristics of participants. Categories All children Median (IQR) r School type Elementary school children Median (IQR) r Junior high school children Median (IQR) r High school children r Median (IQR) Child Age at time of survey Sex Siblings History of developmental concerns Boy Girl No Yes No Yes 6.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 6.0 (5.0, 9.0) Personal mobile device use and restriction Having personal mobile devices No Yes 6.0 (4.0, 8.0) 7.0 (5.0, 10.0) Restricted use of mobile devices on weekdays Restricted use of mobile devices on holidays Child health quality Health-related quality of life Sleep problems Internet addiction Mother Age at time of survey Educational level Father Age at time of survey Educational level Annual household income (million Japanese Yen) No Yes No Yes 8.0 (5.0, 10.0) 7.0 (5.0, 9.0) 7.5 (5.0, 10.0) 7.0 (5.0, 9.0) ¯9 10–12 13–15 >16 ¯9 10–12 13–15 >16 <3.0 3.0–4.9 5.0–7.9 ²8 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 7.0 (5.0, 10.0) 0.466 0.272 0.131 ¹0.031 5.0 (3.0, 7.0) 6.0 (4.0, 8.0) 6.0 (4.0, 8.0) 6.0 (4.0, 8.0) 6.0 (4.0, 8.0) 5.5 (3.0, 7.0) 6.0 (4.0, 7.0) 6.0 (3.0, 8.0) 6.0 (4.0, 8.0) 6.0 (4.0, 8.0) 6.0 (4.0, 8.0) 6.0 (4.0, 8.0) 8.0 (5.0, 10.0) 10.0 (6.0, 12.0) 8.0 (6.0, 11.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0)* 7.0 (6.0, 10.0) 8.0 (6.0, 10.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0) 10.0 (6.0, 12.0) 10.0 (7.0, 12.0) 10.5 (6.0, 12.3) 10.0 (7.0, 12.0) 10.0 (7.0, 12.0) 10.0 (6.0, 12.0) 15.0 10.0 (7.0, 12.0) 10.0 (6.0, 12.0) 10.0 (7.0, 13.0) 10.0 (6.0, 12.0) 10.0 (7.3, 13.0) ¹0.097 ¹0.118** 0.002 0.013 ¹0.054* ¹0.118** ¹0.072* ¹0.048 ¹0.072* ¹0.080 ¹0.113* ¹0.024 0.251 0.157 0.061* 0.030 5.0 (4.0, 8.0) 6.0 (4.0, 7.0) 6.0 (4.0, 8.0) 6.0 (4.0, 8.0) 8.0 (5.0, 10.0) 8.0 (6.0, 11.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0) 10.0 (6.0, 12.0) 10.0 (6.0, 12.0) 10.0 (7.0, 12.0) 10.5 (7.8, 14.0) 0.251 0.144 0.112** 0.082 5.0 (3.8, 7.0) 6.0 (3.0, 7.0) 6.0 (4.0, 8.0) 6.0 (4.0, 8.0) 6.0 (4.0, 8.0) 6.0 (4.0, 7.0) 6.0 (4.0, 8.0) 6.0 (3.8, 8.0) 10.0 (6.3, 12.0) 9.0 (6.0, 11.0) 8.0 (6.0, 10.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0) 9.0 (6.0, 11.0) 10.0 (7.0, 12.0) 10.0 (6.0, 12.0) 10.0 (6.0, 12.0) 10.0 (7.0, 12.0) 10.0 (6.0, 12.0) 10.0 (6.0, 12.0) 10.0 (7.0, 13.0) 10.0 (6.5, 12.0) IQR, interquartile range is the 75th and 25th percentile. r, Spearman’s rank correlation coeﬃcient. P < 0.05, P < 0.01 by one-way ANOVA, »2 test, and Kruskal-Wallis test, which were used to compare data among all children, and among those stratiﬁed by school type children (Table 5 and Fig. 2). The eﬀects of the interaction between exposure (age at ﬁrst use of a mobile device and the duration) and the school type were statistically signiﬁ- cant (Table 5, and Supplemental Table 4 and 8). The adjusted odds ratios of the ﬁve subscales—conduct problems, hyperactivity, emotional symptoms, peer prob- lems, and prosocial behavior—according to the logistic regression analysis results for all children and those strati- ﬁed by the school type are shown in Table 6 and Fig. 2 (for the borderline/high and normal groups). Among all the children and among elementary school children only, the adjusted odds ratios for hyperactivity and peer problems signiﬁcantly decreased with increasing child’s age at ﬁrst use of a mobile device and decreasing child’s duration for the use of a mobile device. Among all the children and among those in elementary and junior high school, the adjusted odds ratios for emotional symptoms signiﬁcantly decreased with increasing age at ﬁrst use of a mobile de- vice and decreasing duration for use of a mobile device. Among all the children in senior high school, the adjusted Environmental Health and Preventive Medicine (2023) 28:22 7 of 11 Table 5 OR for child behavioral problems (TDS) according to child’s mobile device usage. Exposure All children OR (95% CI)a P for interaction Elementary school Junior high school Senior high school OR (95% CI)b OR (95% CI)b OR (95% CI)b Age at ﬁrst use of a mobile device 0.85 (0.77, 0.93)* 0.009 1.20 (1.08, 1.33)** 0.005 Duration for use of a mobile device Age at ﬁrst having personal mobile devices 0.776 0.94 (0.79, 1.11) 0.988 Duration for having personal mobile devices 1.10 (0.90, 1.34) TDS; total diﬃculties score. a (ALL); The logistic regression analysis models were adjusted for child age at time of survey, sex, history of developmental concerns, health-related quality of life, sleep problems, internet addiction, school type, and interaction between exposure and school type. b (Stratiﬁed by school type); The logistic regression analysis models were adjusted for child age at time of survey, sex, history of developmental concerns, health-related quality of life, sleep problems, and internet addiction. P < 0.05, P < 0.01 P for interaction; each exposure (child’s age at their ﬁrst use of a mobile device and duration of use) and school type. 0.87 (0.81, 0.93)* 1.15 (1.08, 1.23)** 0.89 (0.77, 1.02) 1.13 (0.98, 1.30) 1.02 (0.96, 1.09) 0.98 (0.92, 1.04) 0.95 (0.85, 1.05) 1.06 (0.95, 1.17) 0.99 (0.91, 1.08) 1.01 (0.93, 1.10) 0.90 (0.80, 1.02) 1.11 (0.98, 1.25) odds ratios for conduct problems signiﬁcantly decreased with increasing age at ﬁrst use of personal mobile devices and decreasing duration for having personal mobile de- vices (Table 6). In the supplemental logistic regression analysis stratiﬁed by the children’s sex, boys who were younger at their ﬁrst use of a mobile device and used such devices for longer durations were found to be hyperactive, have emotional instabilities, and experience peer problems in elementary school but displayed opposite behaviors in senior high school. Girls who were younger at their ﬁrst use of mobile devices and used such devices for longer durations were found to have emotional instabilities in junior high school but displayed opposite behaviors in senior high school (Supplemental Tables 4–7). 4 Discussion In this cross-sectional study, children who were younger at their ﬁrst use of a mobile device and used such devices for longer durations represented more problematic behaviors according to the SDQ. Moreover, when stratiﬁed by school type, the above associations remained statistically signiﬁ- cant for elementary school children, but not for junior high school and older children (Table 5). Our results suggest that elementary school children are more sensitive to mo- bile device usage than junior high and senior high school children because they are in the early stages of socializa- tion and their behavioral and mental development is rap- idly growing. Four exposure factors were determined in this study: the child’s age at ﬁrst use of a mobile device and the duration of use, and the child’s age at their ﬁrst owning of a personal mobile device and the duration of owning. Given the rapid spread of mobile devices in recent years, elementary school children have started using such devices earlier than high school children did. By contrast, elementary school children used mobile devices for shorter periods than high school children. However, elementary school children represented more problematic behaviors, suggesting that starting to use mobile devices at an early age causes negative eﬀects on developmental immature neural behaviors. Through a cross-sectional and longitudi- nal survey, the Danish National Birth Cohort has reported negative prenatal and postnatal eﬀects of cell phone use on emotional and behavioral diﬃculties in children aged 7 and 11 [4, 5, 27]; our study corroborated the results of this study, showing that early exposure to mobile devices can cause developmental impacts on school-aged children. Using SDQ subscales, our study assessed problematic behaviors among school children aged 7–17 years. Ele- mentary school children who were younger at their ﬁrst use of a mobile device and used these devices for longer durations had increased hyperactivity–inattention and dis- played peer and emotional problematic behaviors. More- over, junior high school children who were younger at their ﬁrst use of a mobile device and used such devices for longer durations represented more emotional problem- atic behaviors. One possible reason for these diﬀerences in the relationships with the school type is that the contents of mobile device use may diﬀer by the school type. Diﬀer- ent eﬀects with several content types were evaluated as screen time exposure, which has been noted as a risk factor for sensory development impacts, emotional and behavior- al problems, sleep disturbances, and internet addiction among school-aged children [6, 9]. A Swedish study tar- geting teenagers described that using mobile devices for social networking services is associated with increased communication skills as a positive eﬀect; however, it is also associated with increased anxiety as a negative eﬀect [10]. The results of our study corroborated those of the Swedish study, which stated that early exposure to mobile devices could exacerbate emotional instabilities in teen- agers. While the duration of personal mobile device usage did not diﬀer by the children’s school type (Table 2), se- nior high school children who were younger and had a personal mobile device for a longer period represented more conducted problematic behaviors. This suggests a correlation between having a mobile device and exhibiting oppositional and deﬁant behaviors among high-school teenagers. When stratiﬁed by sex, only boys exhibited an association between mobile device usage and hyperactivity and peer problems. Moreover, associations between mo- bile device usage and emotional problems were observed Environmental Health and Preventive Medicine (2023) 28:22 8 of 11 Fig. 2 Behavioral problems in children and child age at ﬁrst use of a mobile device OR: odds ratio. CI; conﬁdence interval. TDS; total each diﬃculties score. The OR (95% CI) for children of borderline/high group, who was compared to children of normal group based on total each TDS and the ﬁve subscales—conduct problems, hyperactivity, emotional symptoms, peer problems, and prosocial behavior were calculated by the logistic regression analysis, which was adjusted for child age at time of survey, sex, history of developmental concerns, health-related quality of life, sleep problems, internet addiction, school type, and interaction between exposure and school type among all children. The logistic regression analysis among children stratiﬁed by school type was adjusted for the same variables excluding school type and interaction between exposure and school type. P < 0.05, P < 0.01 Environmental Health and Preventive Medicine (2023) 28:22 9 of 11 Table 6 OR of subscale of SDQ according to child’s mobile device usage. Exposure All children OR (95% CI)a Conduct problems Age at ﬁrst use of a mobile device 0.97 (0.87, 1.08) Duration of mobile device use 1.02 (0.91, 1.14) Age at ﬁrst having personal mobile devices 0.95 (0.78, 1.15) Duration of having personal mobile devices 1.07 (0.85, 1.33) Elementary school P for interaction OR (95% CI)b Junior high school Senior high school OR (95% CI)b OR (95% CI)b 0.543 0.741 0.696 0.831 0.99 (0.93, 1.06) 1.01 (0.94, 1.08) 0.91 (0.78, 1.06) 1.11 (0.95, 1.30) 1.01 (0.94, 1.08) 0.99 (0.93, 1.07) 0.95 (0.85, 1.08) 1.05 (0.93, 1.19) 1.01 (0.91, 1.12) 0.99 (0.89, 1.10) 0.87 (0.76, 0.99) 1.15 (1.01, 1.32)* Hyperactivity/inattention Age at ﬁrst use of a mobile device 0.89 (0.79, 0.99)* 0.173 0.264 Duration of mobile device use 1.12 (0.99, 1.25) Age at ﬁrst having personal mobile devices 0.91 (0.74, 1.12) 0.829 0.875 Duration of having personal mobile devices 1.09 (0.86, 1.39) Age at ﬁrst use of a mobile device Duration of mobile device use Age at ﬁrst having personal mobile devices 1.01 (0.85, 1.19) Duration of having personal mobile devices 1.03 (0.85, 1.26) Emotional problems 0.91 (0.83, 1.00)* 0.541 1.15 (1.04, 1.26)** 0.129 0.652 0.996 Peer problems Age at ﬁrst use of a mobile device 0.93 (0.85, 1.02) Duration of mobile device use 1.09 (0.99, 1.20) Age at ﬁrst having personal mobile devices 0.90 (0.77, 1.06) Duration of having personal mobile devices 1.16 (0.96, 1.40) 0.490 0.350 0.191 0.130 0.93 (0.87, 1.00)* 1.08 (1.00, 1.15)* 0.94 (0.80, 1.11) 1.06 (0.90, 1.25) 0.96 (0.89, 1.04) 1.04 (0.96, 1.12) 0.91 (0.80, 1.04) 1.09 (0.96, 1.24) 1.01 (0.90, 1.13) 0.99 (0.88, 1.11) 0.95 (0.81, 1.11) 1.05 (0.90, 1.23) 0.92 (0.87, 0.98)* 1.08 (1.02, 1.15)* 0.97 (0.84, 1.11) 1.03 (0.90, 1.18) 0.92 (0.87, 0.97) 1.09 (1.03, 1.15) 0.98 (0.88, 1.08) 1.03 (0.92, 1.14) 0.99 (0.91, 1.08) 1.01 (0.93, 1.10) 0.97 (0.86, 1.10) 1.03 (0.91, 1.16) 0.93 (0.87, 0.99)* 1.07 (1.01, 1.15)* 0.93 (0.81, 1.07) 1.07 (0.93, 1.23) 0.97 (0.92, 1.02) 1.03 (0.98, 1.09) 0.96 (0.88, 1.06) 1.04 (0.94, 1.14) 1.00 (0.92, 1.07) 1.00 (0.93, 1.08) 1.13 (0.99, 1.28) 0.89 (0.78, 1.01) Prosocial behavior Age at ﬁrst use of a mobile device 0.98 (0.91, 1.05) Duration of mobile device use 1.01 (0.95, 1.09) Age at ﬁrst having personal mobile devices 1.05 (0.93, 1.19) Duration of having personal mobile devices 0.90 (0.77, 1.04) SDQ; The strengths and diﬃculties questionnaire. a (ALL); The logistic regression analysis models were adjusted for child age at time of survey, sex, history of developmental concerns, health-related quality of life, sleep problems, internet addiction, school type, and interaction between exposure and school type. b (Stratiﬁed by school type); The logistic regression analysis models were adjusted for child age at time of survey, sex, history of developmental concerns, health-related quality of life, sleep problems, and internet addiction. P for interaction; each exposure (child’s age at their ﬁrst use of a mobile device and duration of use) and school type. P < 0.05, *P < 0.01 0.98 (0.94, 1.03) 1.02 (0.97, 1.07) 1.05 (0.95, 1.17) 0.95 (0.85, 1.06) 1.01 (0.97, 1.05) 0.99 (0.95, 1.03) 1.04 (0.96, 1.12) 0.96 (0.89, 1.04) 1.00 (0.95, 1.06) 1.00 (0.95, 1.06) 0.99 (0.91, 1.07) 1.01 (0.93, 1.10) 0.513 0.690 0.606 0.235 in elementary school boys and junior high school girls (Supplemental Tables 4–7). The above results may be related to diﬀerences in developmental properties and tim- ing based on the children’s sex [16, 19]. The adverse eﬀects of having personal mobile devices are inconclusive in this cross-sectional analysis, and a longitudinal evalua- tion throughout adolescence is needed in future studies. Parental supervision of their children’s use of mobile devices is recommended. However, the results of this study remained unchanged even after adjusting the paren- tal usage restrictions for weekdays and holidays (Supple- mental Tables 8 and 9); taken together, our study demon- strates the importance of delaying the use of mobile de- vices rather than imposing parental restrictions. In our study, 15.3% of the children aged ¯3 years were found to have used mobile devices according to their parents, which is lower than the rate of 50.2% for children who used the internet according to a 2019 Japanese survey [18]. This diﬀerence may be attributed to this study targeting a wide birth-year period of 10 years and excluding the use of internet with a wired connection (TV or personal com- puter). In our study, 69.7% of the elementary school-aged children had personal mobile devices, higher than the rates of 49.5% and 41.0% for smartphones and tablets, respec- tively, in 2019 of Japanese’s survey [18]. This diﬀerence may be attributed to the fact that this study included mo- bile game devices and other devices as well. In addition, during the survey period in 2020, the penetration of mobile devices among the school-aged children in Hokkaido pre- fecture increased because the administration started to pro- vide mobile devices to a part of school children for online classes. 4.1 Strengths This study was conducted from 2020 to 2021 and allowed us to examine recent associations between mobile device usage and behavioral problems in children aged 7–17 years. This study used four exposure factors, including the age at their ﬁrst use of a mobile device and the duration of usage before or after owing a personal mobile device. Environmental Health and Preventive Medicine (2023) 28:22 10 of 11 This facilitated the evaluation of both early exposure to and having mobile devices. Previous studies have reported that the early use of mobile devices disrupts healthy activ- ities and increases the risk of sleep problems and internet addiction and lowers the health-related quality of life [2, 3, 31]. Children with developmental disorders often exhibit symptoms such as insomnia and anxiety [1]. These factors may be mutually related with mobile device usage and behavioral problems in children. Our results were obtained after adjusting for mutually related potential confounders, including sleep problems, internet addiction, health-related quality of life, and developmental concerns. 4.2 Limitations This cross-sectional study could not establish a clear causal direction. In fact, a previous study reported that children with developmental disorders are inclined toward heavy use of mobile devices owing to their weak self-regulation and the presence of restricted interests and repetitive behav- iors [14]. Also, parents of inattentive and hyperactive chil- dren are more likely to use mobile devices when calming down their children [20]. Furthermore, in this study, the parents retrospectively provided the age at which their chil- dren ﬁrst used a mobile device and the duration of use; hence, recall bias may be possible. The diﬀering results by school type may be associated with diﬀerences in the sample size, the rate of having personal mobile devices, and the varying content used among school types. 5 Conclusions Our ﬁndings suggest that elementary school children are more sensitive to mobile device usage than older children. Children who are younger at their ﬁrst use of a mobile device and use such devices for longer durations may be prone to emotional instabilities as teenagers. Children who are younger and have had a personal mobile device for longer may show oppositional behaviors as teenagers. However, longitudinal follow-up studies are needed to clarify whether these problems disappear with age. Supplementary information The online version contains supplementary material available at https://doi.org/ 10.1265/ehpm.22-00245. Additional ﬁle 1: Supplemental Table 1. Distribution of child’s age at ﬁrst use of a mobile device. Supplemental Table 2. Associations between child behavioral problems (TDS) and characteristics of participants (continuous variable). Supplemental Table 3. Associations between child behavioral problems (TDS) and characteristics of participants (categorical variable). Supplemental Table 4. OR for child behavioral problems (TDS) according to the child’s mobile device usage (boys). Supplemental Table 5. OR for subscales of SDQ according to children’s mobile device usage (boys). Sup- plemental Table 6. OR for child behavioral problems (TDS) according to child’s mobile device usage (girls). Supplemental Table 7. OR of subscale of SDQ according to child’s mobile device usage (girls). Supplemental Ta- ble 8. OR for child behavioral problems (TDS) according to child’s mobile device usage (Additional adjustment). Supplemental Table 9. OR of sub- scale of SDQ according to child’s mobile device usage (Additional adjust- ment). Declaration Ethics approval and consent to participate The institutional ethical board for human gene and genome studies at Hokkaido University Center for Environmental and Health Sciences (reference no. 139, August 30, 2022) and Hokkaido University Graduate School of Medicine (May 31, 2003) approved the study protocol. Written informed consent was obtained from all participants at the time of enrollment. Consent for publication Not applicable. Availability of data and material The datasets generated and/or analyzed during the current study are not publicly available because the study involves human participants with a nondisclosure provision of individual data stated in the written informed consent in order to prevent compromise of study participants’ privacy but are available from the corresponding author upon reasonable request. Competing interests The authors declare no conﬂict of interest. Funding This work was supported by the Grant-in-Aid for Health Science Research from the Japanese Ministry of Internal Affairs and Communications (JPMI10001). Authors’ contributions Reiko Kishi designed the study and developed the methodology. Chihiro Miyashita, Keiko Yamazaki, Naomi Tamura, and Atsuko Ikeda-Araki, collected the data and performed the analyses. Chihiro Miyashita, Keiko Yamazaki, and Naomi Tamura drafted the manuscript. Reiko Kishi, Satoshi Suyama, Takashi Hikage, Manabu Omiya, and Masahiro Mizuta provided critical revision of the manuscript. All authors take full responsibility for the content of this paper. All authors read and approved the ﬁnal manuscript. Acknowledgements We would like to express our appreciation to all of the study participants of the Hokkaido Study on Environment and Children’ Health. We also express our profound gratitude to all personnel in the hospitals and clinics that collaborated including Sapporo Toho Hospital, Keiai Hospital, Endo Kikyo with the study, Maternity Clinic, Shiroishi Hospital, Memuro Municipal Hospital, Aoba Ladies Clinic, Obihiro-Kyokai Hospital, Akiyama Memorial Hospital, Sapporo Medical University Hospital, Hokkaido University Hospital, Kitami Red Cross Hospital, Hoyukai Sapporo Hospital, Gorinbashi Hospital, Hashimoto Clinic, Asahikawa Medical College Hospital, Hakodate Central General Hospital, Ohji General Hospital, Nakashibetsu Municipal Hospital, Sapporo Tokushukai Hospital, Asahi- kawa Red Cross Hospital, Wakkanai City Hospital, Kushiro Rosai Hospital, Sapporo-Kosei General Hospital, Shibetsu City General Hospital, Nikko Memorial Hospital, Sapporo City General Hospital, Kohnan Hospital, Hakodate City Hospital, Hokkaido Monbetsu Hospital, Tenshi Hospital, Hakodate Goryoukaku Hospital, Nakamura Hospital, Kin-ikyo Sapporo Hospital, Kitami Lady’s Clinic, Engaru-Kosei General Hospital, Kushiro Red Cross Hospital, Nayoro City General Hospital, and Obihiro-Kosei General Hospital. Received: 11 October 2022, Accepted: 3 March 2023 Published online: 12 April 2023 References 1. Byun YH, Ha M, Kwon HJ, Hong YC, Leem JH, Sakong J, et al. Mobile phone use, blood lead levels, and attention deﬁcit hyperactivity symptoms in children: a longitudinal study. PLoS One. 2013;8:e59742. https://doi.org/ 10.1371/journal.pone.0059742. 2. Strasburger VC, Hogan MJ, Mulligan DA, Ameenuddin N, Christakis DA, Cross C, et al. Children, adolescents, and the media. Pediatrics. 2013;132: 958–61. https://doi.org/10.1542/peds.2013-2656. 3. Hill D, Ameenuddin N, Chassiakos R, Cross C, Hutchinson J, Levine A, et al. Media and young minds. Pediatrics. 2016;138:e20162591. https://doi. org/10.1542/peds.2016-2591. Environmental Health and Preventive Medicine (2023) 28:22 11 of 11 4. Divan HA, Kheifets L, Obel C, Olsen J. Cell phone use and behavioural problems in young children. J Epidemiol Community Health. 2012;66: 524–9. https://doi.org/10.1136/jech.2010.115402. 5. Divan HA, Kheifets L, Obel C, Olsen J. Prenatal and postnatal exposure to cell phone use and behavioral problems in children. J Epidemiol. 2008;19: S94–SS5. 6. Eirich R, McArthur BA, Anhorn C, McGuinness C, Christakis DA, Madigan S. Association of screen time with internalizing and externalizing behavior problems in children 12 years or younger: A systematic review and meta- analysis. JAMA Psychiatry. 2022;79:393–405. https://doi.org/10.1001/ jamapsychiatry.2022.0155. 7. Goodman R. The strengths and difﬁculties questionnaire: a research note. J Child Psychol Psychiatry. 1997;38:581–6. https://doi.org/10.1111/j.1469- 7610.1997.tb01545.x. 8. Goodman R. Psychometric properties of the strengths and difﬁculties questionnaire. J Am Acad Child Adolesc Psychiatry. 2001;40:1337–45. https://doi.org/10.1097/00004583-200111000-00015. 9. Hosokawa R, Katsura T. Association between mobile technology use and in early elementary school age. PLoS One. 2018;13: child adjustment e0199959. https://doi.org/10.1371/journal.pone.0199959. 10. Jenkins RH, Shen C, Dumontheil I, Thomas MSC, Elliott P, Röösli M, et al. Social networking site use in young adolescents: association with health- related quality of life and behavioural difﬁculties. Comput Human Behav. 2020;109. https://doi.org/10.1016/j.chb.2020.106320. 11. Kishi R, Araki A, Minatoya M, Hanaoka T, Miyashita C, Itoh S, et al. The Hokkaido Birth Cohort Study on Environment and Children’s Health: cohort proﬁle-updated 2017. Environ Health Prev Med. 2017;22:46. https://doi.org/ 10.1186/s12199-017-0654-3. 12. Kishi R, Ikeda-Araki A, Miyashita C, Itoh S, Kobayashi S, Ait Bamai Y, et al. Hokkaido birth cohort study on environment and children’s health: cohort proﬁle 2021. Environ Health Prev Med. 2021;26:59. https://doi.org/10.1186/ s12199-021-00980-y. 13. Kishi R, Kobayashi S, Ikeno T, Araki A, Miyashita C, Itoh S, et al. Ten years of progress in the Hokkaido birth cohort study on environment and children’s health: cohort proﬁle—updated 2013. Environ Health Prev Med. 2013;18:429–50. https://doi.org/10.1007/s12199-013-0357-3. 14. Lawrence AC, Narayan MS, Choe DE. Association of Young Children’s Use of Mobile devices with their self-regulation. JAMA Pediatr. 2020;174:793–5. https://doi.org/10.1001/jamapediatrics.2020.0129. 15. Lin HP, Chen KL, Chou W, Yuan KS, Yen SY, Chen YS, et al. Prolonged touch screen device usage is associated with emotional and behavioral problems, but not language delay, in toddlers. Infant Behav Dev. 2020;58. https://doi.org/10.1016/j.infbeh.2020.101424. 16. Matsuishi T, Nagano M, Araki Y, Tanaka Y, Iwasaki M, Yamashita Y, et al. Scale properties of the Japanese version of the strengths and difﬁculties questionnaire (SDQ): a study of infant and school children in community samples. Brain Dev. 2008;30:410–5. https://doi.org/10.1016/j.braindev. 2007.12.003. 17. McHarg G, Ribner AD, Devine RT, Hughes C, Blair C, Hughes C. Infant screen exposure links to toddlers’ inhibition, but not other EF constructs: A propensity score study. Infancy. 2020;25:205–22. https://doi.org/10.1111/ infa.12325. 18. Japanese Cabinet Ofﬁce; 2020. Survey on internet usage environment for young people in Japan. https://www8.cao.go.jp/youth/kankyou/internet_ torikumi/tyousa/r01/net-jittai/pdf/sokuhou.pdf. Accessed Jan 21, 2023 (in Japanese). 19. Moriwaki A, Kamio Y. Normative data and psychometric properties of the strengths and difﬁculties questionnaire among Japanese school-aged children. Child Adolesc Psychiatry Ment Health. 2014;8:1. https://doi.org/ 10.1186/1753-2000-8-1. 20. Muñoz-Silva A, Lago-Urbano R, Sanchez-Garcia M, Carmona-Márquez J. Child/adolescent’s ADHD and parenting stress: the mediating role of family impact and conduct problems. Front Psychol. 2017;8:2252. https://doi.org/ 10.3389/fpsyg.2017.02252. 21. Nezu S, Iwasaka H, Saeki K, Obayashi K, Ishizuka R, Goma H, et al. Reliability and validity of Japanese versions of KIDSCREEN-27 and KIDSCREEN-10 questionnaires. Environ Health Prev Med. 2016;21:154– 63. https://doi.org/10.1007/s12199-016-0510-x. 22. Okada M, Kitamura S, Iwadare Y, Tachimori H, Kamei Y, Higuchi S, et al. Reliability and validity of a brief sleep questionnaire for children in Japan. J Physiol Anthropol. 2017;36:35. https://doi.org/10.1186/s40101-017-0151-9. 23. Okely AD, Ghersi D, Hesketh KD, Santos R, Loughran SP, Cliff DP, et al. A collaborative approach to adopting/adapting guidelines—the Australian 24- hour Movement Guidelines for the early years (birth to 5 years): an integration of physical activity, sedentary behavior, and sleep. BMC Public Health. 2017;17:869. https://doi.org/10.1186/s12889-017-4867-6. 24. Poushter J, Bishop C, Chwe H. Social media use continues to rise in developing countries but plateaus across developed ones; 2018. https:// assets.pewresearch.org/wp-content/uploads/sites/2/2018/06/15135408/ Pew-Research-Center_Global-Tech-Social-Media-Use_2018.06.19.pdf. Pew Research Center. 25. Ravens-Sieberer U, Auquier P, Erhart M, Gosch A, Rajmil L, Bruil J, et al. The KIDSCREEN-27 quality of life measure for children and adolescents: psychometric results from a cross-cultural survey in 13 European countries. Qual Life Res. 2007;16:1347–56. https://doi.org/10.1007/s11136-007- 9240-2. 26. Roser K, Schoeni A, Röösli M. Mobile phone use, behavioural problems and concentration capacity in adolescents: A prospective study. Int J Hyg Environ Health. 2016;219:759–69. https://doi.org/10.1016/j.ijheh.2016.08. 007. 27. Sudan M, Olsen J, Arah OA, Obel C, Kheifets L. Prospective cohort analysis of cellphone use and emotional and behavioural difﬁculties in children. J Epidemiol Community Health. 2016;70:1207–13. https://doi.org/ 10.1136/jech-2016-207419. 28. Thomas S, Benke G, Dimitriadis C, Inyang I, Sim MR, Wolfe R, et al. Use of mobile phones and changes in cognitive function in adolescents. Occup Environ Med. 2010;67:861–6. https://doi.org/10.1136/oem.2009.054080. 29. Tobe H, Takeuchi K, Hori M. The relationship between the tendency toward internet dependence and mental health and the psychosocial problems of students. Jpn J Sch Health. 2010;52:125–34. 30. Tremblay MS, Chaput JP, Adamo KB, Aubert S, Barnes JD, Choquette L, et al. Canadian 24-hour movement guidelines for the early years (0–4 years): an integration of physical activity, sedentary behaviour, and sleep. BMC Public Health. 2017;17:874. https://doi.org/10.1186/s12889-017- 4859-6. 31. World Health Organization. Guidelines on physical activity, sedentary behaviour and sleep for children under 5 years of age. Geneva: World Health Organization; 2019. 32. Young KS. Caught in the net: how to recognize the sign of Internet addic- tion and a winning strategy for recovery. New York: John Wiley & Sons. (Odashima Y, Mainichi Publishing Co.; 1998, translated in Japanese); 1996.
10.1186_s12933-021-01314-6	Nesti et al. Cardiovasc Diabetol (2021) 20:124 https://doi.org/10.1186/s12933-021-01314-6 Cardiovascular Diabetology ORIGINAL INVESTIGATION Open Access Mechanisms of reduced peak oxygen consumption in subjects with uncomplicated type 2 diabetes Lorenzo Nesti1,2† Iacopo Fabiani3, Domenico Trico1,4, Stefano Masi2 and Andrea Natali1,2 , Nicola Riccardo Pugliese2†, Paolo Sciuto1, Nicolò De Biase2, Matteo Mazzola2, Abstract Background: Type 2 diabetes mellitus (T2D) increases the risk of incident heart failure (HF), whose earliest finger- print is effort intolerance (i.e. impaired peak oxygen consumption, or VO2peak). In the uncomplicated T2D population, however, the prevalence of effort intolerance and the underpinning mechanistic bases are uncertain. Leveraging the multiparametric characterization allowed by imaging-cardiopulmonary exercise testing (iCPET), the aim of this study is to quantify effort intolerance in T2D and to dissect the associated cardiopulmonary alterations. ± 5.4 mL/min/kg, p < 0.0001). Despite a comparable cardiac output, patients with effort intolerance 3.1 vs 12.7 Methods: Eighty-eight adults with well-controlled and uncomplicated T2D and no criteria for HF underwent a maximal iCPET with speckle tracking echocardiography, vascular and endothelial function assessment, as well as a comprehensive biohumoral characterization. Effort intolerance was defined by a VO2peak below 80% of maximal predicted oxygen uptake. Results: Forty-eight patients (55%) had effort intolerance reaching a lower VO2peak than T2D controls (16.5 min/kg, vs 21.7 0.002), lower VO2/work slope showed reduced peak peripheral oxygen extraction (11.3 0.009) 3.0, p 2.8 vs 15.2 (9.9 and global longitudinal strain (peak-rest ΔGLS 1.7 0.03) than subjects with VO2peak above 80%. Diastolic function, vascular resistance, endothelial function, biohumoral exams, right heart and pulmonary function indices did not differ between the two groups. Conclusions: Effort intolerance and reduced VO2peak is a severe and highly prevalent condition in uncomplicated, otherwise asymptomatic T2D. It results from a major defect in skeletal muscle oxygen extraction coupled with a sub- tle myocardial systolic dysfunction. 3.3 mL/dL, p 1.4, p < 0.0001), impaired left ventricle systolic reserve (peak S’ 13.5 1.2 vs 11.2 1.5 vs 2.5 3.2 mL/ 1.8, p = ± ± ± ± ± ± = = ± ± ± Keywords: Type 2 diabetes, Effort intolerance, Heart failure with preserved ejection fraction, Exercise physiology, Cardiopulmonary exercise test, Diabetic cardiomyopathy Correspondence: lorenzo.nesti@phd.unipi.it †Lorenzo Nesti and NicolaRiccardo Pugliese contributed equally to this work 1 Metabolism, Nutrition, and Atherosclerosis Laboratory, Department of Clinical and Experimental Medicine, University of Pisa, Via Savi 10, 56126 Pisa, Italy Full list of author information is available at the end of the article Background Defined as the inability to perform physical exercise at the maximal intensity expected (according to age, gen- der, body mass index (BMI), and habitual levels of physi- cal activity [1]), effort intolerance can be quantified and objectively measured by peak oxygen consumption (VO2peak) during a graded maximal exercise test. Effort © The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 2 of 13 intolerance is the hallmark of heart failure (HF), is part of the definition of the HF syndrome and is intimately linked to its pathophysiology [2]. Given that type 2 diabe- tes mellitus (T2D) portends an increased risk of develop- ing HF—especially HF with preserved ejection fraction (HFpEF)—that is neither entirely explained by tradi- tional cardiovascular risk factors nor by coronary heart disease, an early, primary cardiopulmonary impairment has been postulated, but never clearly demonstrated [3]. Effort intolerance was previously reported in otherwise asymptomatic subjects with T2D, with slower oxygen uptake kinetics and reduced peak values in comparison to normal subjects during cardiopulmonary exercise test (CPET) [4]. Notably, reduced VO2peak is a reliable pre- dictor of cardiovascular disease, all-cause mortality [5], development of HF [6], and reduced quality of life in T2D [7]. According to current guidelines, the early appear- ance of effort intolerance is a marker of subclinical HF (American Heart Association stage B), whose early rec- ognition justifies a more aggressive diagnostic-therapeu- tic workup. To date, however, the prevalence of effort intolerance in T2D is unclear, and so are the underlying mechanisms [4], leaving the prevention strategies for HF in T2D uncertain. With the present study, we aim at quantifying the prevalence of effort intolerance in an outpatient, uncom- plicated, and otherwise asymptomatic T2D population. Participants underwent exercise echocardiography dur- ing a maximal imaging-CPET (iCPET), which allows the dissection of the pathophysiological mechanisms under- lying a reduced VO2peak by simultaneous measurement of the major determinants of exercise physiology. Patients and methods Study population We prospectively enrolled 114 patients from the Dia- betes Outpatient Clinic at the Santa Chiara University Hospital of Pisa. Inclusion criteria were: men or women of 40 to 80 years of age with a clinical diagnosis of T2D according to the ADA criteria [8]; HbA1c values between 53 and 69 mmol/mol (7.0 to 8.0%); on stable hypogly- cemic and cardioactive therapy for at least 3 months; baseline echocardiographically-assessed left ventricle ejection fraction (LVEF) above 50%; without a dignosis of HF according to guidelines [9]. Exclusion criteria were: symptoms or diagnosis of HF, serum BNP above 100 pg/ mL, any established cardiovascular disease, presence of retinopathy (at ophtalmology), peripheral artery disease, peripheral or cardiac neuropathy, respiratory insuffi- ciency or diagnosis of chronic obstructive pulmonary disease (more than moderate airflow obstruction (forced expiratory volume in 1 s [FEV1] to forced vital capacity [FVC] ratio < 0.70 and FEV1 < 50% of predicted FEV1) and/or restrictive pattern (< 80% of predicted FVC)); pul- monary hypertension; any acute or chronic inflamma- tory disease; severe obesity defined as body mass index (BMI) > 40 kg/m2; uncontrolled blood pressure defined as BP > 160/100 mmHg; impaired kidney function defined as estimated glomerular filtration rate (eGFR) < 60 mL/ min/1.73 m2; uncontrolled arrhythmias (including atrial fibrillation); any more than mild valvular disesase; inabil- ity to cycle due orthopedic limitations; poor echocar- diographical acoustic windows; ongoing pregnancy or breastfeeding. The definitive inclusion/exclusion criteria were reas- sessed both after the baseline evaluation and after the iCPET. The Local Ethic Committee approved the study protocol. All patients gave written informed consent before enrolment. Patient characterization i. Clinical characterization A full clinical history was obtained. Baseline demo- graphic data, anthropometric variables (height, weight and body mass index, BMI), functional sta- tus, cardiovascular risk factors (e.g. family history of cardiovascular disease, alcohol and smoking habits), comorbidities (e.g. arterial hypertension, dyslipidemia), and medication were also recorded. A thorough physical exam was also performed, including resting vital parameters. All patients underwent a resting cardiac echocardiography (see later for “baseline, speckle tracking, and exercise echocardiography”) and electrocardiogram. ii. Biohumoral characterization Blood cell count, HbA1c, blood lipids, creatinine, elec- trolytes, uric acid, hepatic function, urinalysis, erythrocytes sedimentation rate, high-sensitivity C-reactive protein, BNP, urine albumin-to-creati- nine ratio (ACR) were recorded at baseline. eGFR was calculated through the CKD-EPI formula. iii. Vascular assessment Patients underwent peripheral vascular disease assess- ment through the cardio-ankle vascular index (CAVI) and ankle-brachial index (ABI) with the Vascular Screening System VaSera VS-1500 N® (Fukuda Denshi, Japan) to rule out peripheral vas- cular disease. Endothelial function was assessed by downstream hyperemic response to ischemia using an EndoPAT device (EndoPAT 2000, Itamar Medical Ltd., Caesarea, Israel), according to stand- ard procedures [10]. The reactive hyperemia index (RHI) was calculated as the ratio between post- and Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 3 of 13 pre-occlusion amplitudes of the pulse, normalized to the contralateral arm. iv. Screening of neuropathy All patients underwent a thorough neurological clini- cal examination by a trained physician, as well as screening of neurological complications by the Semmens-Weinstein monofilament test and Neu- rotester® to exclude the presence of neurological complications. The monofilament test was per- formed according to standard procedures [11] for the screening of peripheral sensory-motor neu- ropathy. The screening of the cardiac autonomic neuropathy (CAN) was performed with the Neu- rotester® (Meteda srl, San Benedetto del Tronto, Italy): variations in RR intervals in response to the Valsalva manoeuvre, lying-to-standing, and deep breathing, according to standard procedures [12]. The three tests were repeated three times each for each patient. Two or more abnormal tests, based on age-related normal values, identified the pres- ence of CAN. Cardiopulmonary exercise test (CPET) protocol A symptom-limited graded ramp bicycle exercise test was performed in the semi-supine position on a tilting, dedi- cated, microprocessor-controlled stress echocardiogra- phy cycle ergometer (Ergoline ergoselect 2000 GmbH, Germany). A 12-lead electrocardiogram and non-inva- sive arterial saturation and blood pressure (BP) were monitored continuously. Heart rate (HR) and brachial BP were measured at rest and every minute during exercise using a validated automatic device (Omron M6 Com- fort, Kyoto, Japan). The expected VO2peak, estimated on the bases of patient age, height, weight and clinical his- tory [1], was used to adjust the ramp increments (Watt) to reach the patient’s estimated VO2peak in 8 to 12 min. The protocol included two minutes of unloaded pedalling and four minutes of recovery after peak effort. Then, we excluded from the analysis patients who did not reach a RER > 1.0 during the exercise test. Breath-by-breath min- ute ventilation, carbon dioxide production (VCO2), and VO2 were measured using a dedicated cardiopulmonary diagnostic device (Blue Cherry, Geratherm Respiratory GmbH, Germany). We defined VO2peak as the highest median value of the two 30-s intervals of the last minute of exercise, as previously extensively validated [13–17]. An automatic procedure determined anaerobic threshold (AT) based on the V-slope, ventilatory equivalents and end-tidal partial pressure methods; AT was verified visu- ally and, if necessary, recalculated [1]. The chronotropic response was calculated as the change in HR from rest to peak exercise, divided by the difference between the age-predicted maximal HR and the resting HR (i.e. HR reserve). Chronotropic incompe- tence was defined as the failure to achieve ≥ 80% of the HR reserve during exercise [18]. In patients on β-blockers or calcium-channel blockers, chronotropic incompetence was defined as the failure to achieve 62% of HR reserve [18]. Baseline, speckle tracking, and exercise stress echocardiography protocol All patients underwent a comprehensive transthoracic echocardiography examination at rest (GE healthcare vivid e95, Milwaukee, WI, USA) according to the Inter- national Recommendations [19]. Data collected at each stage, that is at baseline, after 4 min, at the AT, and at peak effort, included: left ventricle (LV) and atrial (LA) volumes, stroke volume (SV), peak E-wave and A-wave velocities, tissue Doppler imaging (TDI)-derived S’ and e’ at the septal and lateral mitral annulus, tricuspid regur- gitation velocity and systolic pulmonary artery pres- sure (sPAP), tricuspid annular plane systolic excursion (TAPSE); LV volumes and LVEF were calculated from the apical two- and four-chamber views using the modified Simpson’s rule. SV was calculated by multiplying the LV outflow tract area at rest by the LV outflow tract veloc- ity–time integral measured by pulsed-wave Doppler dur- ing each activity level, as previously validated [20–23]. Cardiac output was calculated as the multiplication of SV and HR. The Δ(a-v)O2 was estimated indirectly with a validated and previously used approach by different groups using both our combined iCPET approach [24] and in a different setting with CPET and right heart cath- eterization [25]. Images were acquired concurrently with breath-by-breath gas exchange measurements at both baseline and peak of exercise. All measurements were reported as the average of three beats. We measured global longitudinal strain (GLS) from the apical long-axis view and two- and four-chamber views, ensuring a frame rate > 50 Hz (GE healthcare EchoPAC BT 12). We reported the average values from the three apical views at rest and low-load effort, within the first 4 min of exercise. We excluded poorly tracked segments and patients were not analysed if more than one segment per view was deemed unacceptable. STE-derived meas- urements were reported as the average of three beats. Statistical analysis Analyses were performed using JMP Pro software version 13.2.1 (SAS Institute, Cary, NC). Values are presented as mean ± SD, or as median and interquartile range (IQR), for variables with normal and non-normal distribution, respectively. Variables with a non-normal distribution at Kolmogorov–Smirnov test were logarithmically trans- formed for parametric analysis. Comparisons between Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 4 of 13 groups were made by the Student t-test for unpaired data for continuous variables and by the chi-square test for categorical variables. All tests were conducted at a two- sided α level of 0.05. In order to exclude drug-related chronotropic incom- petence, we performed a sensitivity analysis after having excluded the subgroup taking beta-blockers, which con- firmed the results of the analysis on the whole population (see Additional file 1: Table S4). Factors with ascertained or potential influence on Δ(a-v)O2 were selected for uni- variate linear regression analysis (age, sex, BMI, duration of diabetes, HbA1c, smoke, presence of hypertension, treatment with metformin, creatinine, ACR, PCR, hemo- globin, RHI, CAVI, ABI, MBP at peak, SVR at peak, CO at peak, LVEF at rest and at peak, ΔLVEF, GLS at rest and at 4 min, ΔGLS, S’ at rest and at peak, ΔS’, E/e’ peak, VE/ VCO2 slope). In order to reduce overfitting, clinical, bio- humoral, echocardiography, and CPET variables showing a p value < 0.100 at univariate analysis, were pooled into the multiple for multivariate analysis. Factors with ascer- tained or potential influence on VO2peak age, sex, BMI, duration of diabetes, HbA1c, creatinine, eGFR, ACR, PCR, hemoglobin, RHI, CAVI, HR at peak, chronotropic incompetence, MBP at peak, SVR at peak, CO at peak, LVEF at rest and at peak, ΔLVEF, GLS at rest and at 4 min, ΔGLS, S’ at rest and at peak, ΔS’, E/e’ peak, VE/ VCO2 slope) were selected for univariate linear regres- sions. Shapiro–Wilk test was used for normal data, and Breusch-Pagan/Cook-Weisberg test was used for het- eroskedasticity in the multiple regression models. A p value < 0.05 was considered statistically significant. Results Baseline characteristics of the study population According to the inclusion and exclusion criteria, 114 consecutive patients were recruited for the study from December 2017 to July 2020; after baseline evaluation, 26 were subsequently excluded because of definitive exclu- sion criteria (10 for suboptimal ultrasound images during the exercise, 10 for incapacity of performing a maximal CPET due to discomfort, 4 for ECG signs suggestive of ischemia, 2 for evidence of autonomic neuropathy); the analysis was performed on 88 T2D subjects who met the definitive inclusion/exclusion criteria. We defined exercise intolerance as the incapacity of reaching a VO2peak > 80% of predicted VO2max, which occurred in 48 subjects (55%). Baseline characteristics of the whole pop- ulation and according to the presence or absence of effort intolerance are reported in Table 1. The two groups had similar numerosity, sex prevalence, BMI, glycemic control, blood pressure values, prevalence of comorbidities, treatment for diabetes and cardio-active therapy. The group with preserved exercise capacity was slightly older, with a congruent marginally lower eGFR despite comparable values of serum creatinine. The group reaching a lower VO2peak showed lower HDL-cholesterol values, despite comparable values of other blood lipids and prevalence of lipid lowering treatment, diuretics and beta blockers. At baseline echocardiography (Table 2), all patients showed normal RV and LV dimensions, mass, diastolic, and 2D systolic function (LVEF), with no dif- ference between the two groups. Conversely, there was a difference in the TDI and speckle-tracking indices, so that the group achieving a lower VO2peak showed a 10% lower baseline S’ and GLS values at rest. Cardiopulmonary exercise test All patients performed a maximal exercise test, as defined by the maximal respiratory exchange ratio steadily greater than 1.05 at peak exercise according to guidelines [1]. The group with reduced exercise capacity achieved a 24% lower VO2peak and a 10% lower peak workload and peak heart rate, while the mean systolic and diastolic blood pressure values were comparable throughout the test. We observed similar results at the sensitivity analy- sis to exclude drug-related chronotropic insufficiency, observing similar results (see Additional file 1: Table S4); however, these data should be interpreted considering the decrease in sample size. No difference was identi- fied in ventilatory or gas exchange parameters, while the anaerobic threshold (AT) was reached earlier in the subjects with effort intolerance both in absolute terms, as well as when expressed as % of VO2peak. The group with reduced exercise capacity showed a higher preva- lence of chronotropic incompetence, a lower VO2/work- load slope, a reduced oxygen pulse at peak, as well as an impaired peripheral oxygen extraction [(a-v)ΔO2] at peak. These results are reported in Table 2 and in Fig. 1. Exercise echocardiography During exercise, cardiac output (CO), LVEF, S’, GLS, and E/e’ all increased linearly with the workload. A reduced systolic reserve was observed in the group with effort intolerance in the form of: lower prevalence of subjects with a normal contractility reserve (i.e. an increase in LVEF > 7.5%), reduced ΔS’ (peak vs baseline), as well as in reduced early (4 min of exercise) GLS and GLS change from baseline (ΔGLS). LV diastolic indices, stroke vol- ume, and cardiac output did not differ between the two groups throughout the iCPET test, as well as the right heart indices did not change throughout the test (TAPSE, sPAP, TAPSE/sPAP, TAPSE/CO) (Table 2 and Fig. 1). Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 5 of 13 Table 1 Baseline characteristics of the study population Effort Intolerance p value Demographic and clinical data Male gender (n, %) Age (years) BMI (kg/m2) Systolic BP (mmHg) Diastolic BP (mmHg) Alcohol (n, %) Smoke (n, %) Hypertension (n, %) Dyslipidemia (n, %) Duration of diabetes (years) Therapy ACEi/ARBs (n, %) Beta-blockers (n, %) Mineralocortic. Rec. Ant. (n, %) Hydrochlorothyazide (n, %) Furosemide (n, %) Statin (n, %) Metformin, n (%) GLP1R-A (n, %) DPP4i (n, %) SGLT2i (n, %) Insulin (n, %) Biohumoral data HbA1c (mmol/mol) Cholesterol (mg/dL) HDL-c (mg/dL) LDL-c (mg/dL) Ttriglycerides (mg/dL) Hemoglobin (g/dL) Creatinin (mg/dL) eGFR (mL/min/1.73mq) Uric acid (mg/dL) ACR (mg/g) CRP (mg/dL) BNP (pg/mL) Vascular and endothelial data RHI endoPAT CAVI mean All patients 88) (n = 71 (81%) ± 9.2 4.7 63.8 29.0 ± 135.9 83.1 ± 5 (6%) 15.5 ± 9.6 Yes (n = 48) 42 (88%) No (n = 40) 29 (73%) ± 9.4 5.4 62.0 29.8 ± 136.5 83.3 ± 3 (6%) 16.0 ± 9.7 ± 8.5 3.6 65.9 28.0 ± 135.1 82.9 ± 2 (5%) 15.1 ± 9.7 14 (16%) 67 (76%) 66 (75%) 8.0 9.3 ± 56 (63%) 23 (26%) 3 (3%) 13 (15%) 3 (3%) 63 (71%) 76 (85%) 2 (2%) 2 (2%) 0 (0%) 14 (16%) 10 (21%) 38 (79%) 35 (73%) 7.7 9.9 ± 30 (63%) 12 (25%) 2 (4%) 6 (13%) 3 (6%) 32 (67%) 41 (85%) 1 (2%) 2 (4%) 0 (0%) 10 (21%) 4 (10%) 30 (73%) 31 (76%) 8.4 8.6 ± 25 (63%) 11 (27%) 1 (2%) 7 (17%) 0 (0%) 30 (76%) 34 (85%) 1 (2%) 0 (0%) 0 (0%) 4 (10%) 56.8 ± 163.6 49.5 ± 100.0 10.4 18.4 ± 13.3 32.9 129.9 57.5 ± ± 1.3 14.3 0.89 0.22 86.5 15.8 5.39 1.39 ± ± ± ± 6.0 (0.0 – 14.3) 11.1 56.9 ± 157.6 46.9 ± 94.6 ± 137.3 14.2 39.2 ± 11.2 33.2 57.6 ± 1.4 0.89 0.20 89.5 17.1 5.49 1.61 ± ± ± ± 5.8 (1.9 – 27.6) 9.7 56.5 ± 172.3 53.2 ± 107.5 35.4 ± 14.4 31.1 119.7 56.6 ± ± 1.1 14.3 0.90 0.24 82.9 13.6 ± ± ± 5.22 1.10 ± 6.1 (0.0 – 9.9) 0.278 0.450 ± 16 (10 – 33) 0.360 0.590 ± 14 (10 – 37) 0.193 0.189 ± 16 (10 – 29) 0.62 0.28 9.32 1.59 ± ± 0.63 0.25 9.01 1.69 ± ± 0.60 0.32 9,67 1.40 ± ± ns 0.0446 ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns 0.0283 ns ns ns ns 0.0476 ns ns ns ns ns ns The study population is reported as a whole and divided in two groups based on the achievement of a peak oxygen uptake > 80% of the maximal theorical oxygen uptake. P values were calculated with a student t-test and were reported as “ns” if non significant Regression analysis According to the Fick’s equation, whole-body oxy- gen uptake is determined by CO and �(a − v)O2 ; in our study, peripheral oxygen extraction explains the impaired exercise capacity (Fig. 2). Therefore, we focused the regression analysis on Δ(a-v)O2. In the whole population, univariate determinants of Δ(a-v) O2 at peak were: sex, hemoglobin levels, peak CO, peak systemic vascular resistance, peak mean arterial pres- sure, GLS at 4 min and ΔGLS. In a multiple regression Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 6 of 13 Table 2 Results of cardiopulmonary exercise test and exercise echocardiography All patients 88) (n = Effort Intolerance p value “Adjusted” p value Yes (n = 48) No (n = 40) Cardiopulmonary exercise test Workload (W) Time of effort (min) HR rest (bpm) HR peak (bpm) HR peak (%max) HR reserve (bpm) Chronotr. incomp. (n, %) MBP rest (mmHg) MBP peak (mmHg) RER peak VO2/work slope VO2 rest (mL/min/kg) VO2 AT (mL/min/kg) VO2 AT (%peakVO2) VO2 peak (mL/min/kg) VO2 peak (%VO2max) VE/VCO2 slope VD/VT (%) O2 pulse peak (mL/bpm) O2 pulse peak (%max) AV O2 diff rest (mL/dL) AV O2 diff peak (mL/dL) Exercise echocardiography EDVi (mL/m2) LVMi (g/m2) LAVi (mL/m2) SV rest (mL) SV peak (mL) CO rest (L/min) CO peak (L/min) LVEF rest (%) LVEF peak (%) ΔEF Contractility reserve (n, %) GLS rest (%) GLS 4 min (%) ΔGLS S’ mean rest (cm/sec) S’ mean peak (cm/s) ΔS’ mean E/e’ rest (cm/s) E/e’ peak (cm/sec) SVR rest (dynes/cm) SVR peak (dynes/cm) TAPSE/sPAP peak TAPSE/CO peak 118 30 11.4 2.0 ± ± 79.9 ± 132.9 86.1 ± 13.6 18.6 ± 12.0 15.7 74.7 ± 46 (53%) 113 25 10.8 1.8 ± ± 80.5 ± 129.2 82.6 ± 12.9 19.0 ± 12.4 14.6 76.1 ± 34 (71%) 126 34 12.0 2.0 ± ± 79.3 ± 137.4 90.2 ± 14.6 17.2 ± 10.2 16.9 73.0 ± 12 (30%) 102.9 10.3 ± 146.5 16.3 ± 0.06 1.08 ± 10.5 1.5 ± 9.1 4.9 4.1 ± 16.2 84.5 11.9 18.8 5.1 79.0 17.0 27.5 3.8 15.6 4.1 11.7 2.8 ± ± ± ± ± ± ± 94.2 18.0 6.5 ± 11.9 ± 2.3 3.2 ± 103.3 10.9 146.9 18.2 ± ± 0.07 ± 1.2 1.2 1.09 9.9 ± 3.7 ± 13.7 ± 81.5 ± 16.45 66.7 2.7 13.7 3.19 ± 8.8 27.2 4.1 16.0 3.9 11.3 2.5 ± ± ± ± 84.3 14.0 6.1 ± 11.3 ± 2.4 3.1 ± 102.5 9.6 ± 146.1 14.0 ± 0.06 1.07 ± 11.2 1.4 ± 1.3 ± 5.2 7.7 4.6 ± 19.5 88.8 ± 21.69 93.6 5.39 ± 11.9 27.8 3.3 15.2 4.4 ± ± ± 3.0 12.3 ± 106.0 15.0 6.9 ± 12.7 ± 2.3 3.3 ± 14.8 22.8 51.0 11.1 86.3 16.6 ± ± ± 24.7 7.6 69.1 ± 102.1 ± 1.2 5.5 ± 13.7 3.7 ± ± 58.9 4.3 67.6 5.6 ± 3.7 8.6 ± 53 (60%) 16.3 2.6 18.4 3.0 ± ± 1.5 1.8 3.0 ± 2.1 2.1 ± 9.2 ± 14.3 5.1 ± ± 8.3 2.2 2.0 8.6 ± 1,583 367 ± 218 911 0.95 0.23 2.24 0.62 ± ± ± 13.3 26.6 51.7 10.5 86.6 14.4 ± ± ± 24.5 8.2 69.5 ± 103.3 ± 1.3 5.5 ± 13.4 3.7 ± ± 58.6 0.7 66.5 6.1 ± 4.1 8.0 ± 23 (26%) 15.7 2.5 17.5 2.9 ± ± 1.5 1.8 2.8 ± 1.9 1.7 ± 8.8 ± 13.5 4.7 ± ± 8.6 2.3 2.1 8.7 ± 1,578 383 ± 230 932 0.96 0.21 2.33 0.64 ± ± ± 51.0 13.0 87.4 20.8 25.5 7.5 ± ± ± 68.6 ± 100.7 13.0 22.1 5.4 ± 14.0 58.6 ± 1.2 3.8 1.0 ± ± 68.9 4.8 ± .31 9.5 ± 30 (34%) 17.0 2.6 19.5 2.8 ± ± 1.8 1.7 3.0 ± 2.3 2.5 ± 9.6 ± 15.2 5.6 ± ± 8.0 2.2 1.9 8.4 ± 1,588 353 ± 203 885 0.94 0.26 2.12 0.58 ± ± ± 0.0460 0.0050 ns 0.0361 0.0023 ns 0.0001 ns ns ns < 0.0001 0.0015 < 0.0001 0.0115 < 0.0001 < 0.0001 ns ns ns < 0.0001 ns 0.0399 ns ns ns ns ns ns ns ns ns ns 0.0154 0.0321 0.0044 0.0293 0.0364 0.0085 ns ns ns ns ns ns ns < 0.0001 0.0008 ns 0.0015 0.0015 ns 0.0005 ns ns ns < 0.0001 0.0005 < 0.0001 0.0047 < 0.0001 < 0.0001 ns ns < 0.0001 < 0.0001 0.0218 0.0002 ns ns ns ns ns ns ns ns ns 0.0412 0.0386 ns 0.0048 0.0093 ns 0.0006 0.0023 ns ns ns ns ns ns The whole population and the two groups divided by the presence or absence of effort intolerance, are reported. P values were calculated as Student’s t test for comparisons of the means of the two groups, and “adjusted” p values were calculated after adjustment for age, sex, and BMI Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 7 of 13 * 0 25 50 75 100 125 Workload [W] C ] L d / L m [ 2 O ) v - a ( ∆ F i ] n m / L [ E V 20 15 10 5 0 18 16 14 12 10 8 Normal tolerance Effort intolerance * B i ] n m / L m [ O C 0 25 50 75 100 125 20 15 10 5 0 0 25 50 75 100 125 Workload [W] Workload [W] A ] 1 - g k i 1 - n m L m [ 2 O V 25 20 15 10 5 0 D 175 ] m p b [ e t a R t r a e H 150 125 100 75 50 25 G 12 ] o i t a r [ ' / e E 10 8 6 4 0 25 50 75 100 125 Workload [W] E ] 1 - m c c e s e n y d [ R V S H 2500 2000 1500 1000 500 ] c e s / m c [ n a e m ' S 20 15 10 5 0 25 50 75 100 125 34 35 Workload [W] 36 37 VCO2 [L/min] 38 39 40 * * * I ] % [ S L G 25 20 15 10 0 25 50 75 100 125 0 25 50 75 100 125 Workload [W] Workload [W] 4 0 Time of exercise [min] Fig. 1 Graphic representation of nine key variables obtained during imaging-cardiopulmonary exercise test. Grey lines represent the group with effort intolerance, black lines represent the control group. Significant differences are highlighted by a star (*) model, only male sex (st-ß 0.35, p = 0.0002), hemo- globin levels (st-ß 0.22, p = 0.0133), and ΔGLS (st-ß 0.21, p = 0.0486) were independent predictors of peak Δ(a-v)O2 (see Table 3). To gain insight on the link between reduced peak oxy- gen utilization and systolic dysfunction, linear regression analyses were performed on the two corresponding sets of variables in the whole population (Fig. 3). Significant positive linear correlations were found between S’ at peak and VO2peak (Panel A), as well as between ΔS’ and VO2peak (panel B), between GLS at 4 min and VO2peak (panel C) and ΔGLS and VO2peak (panel D). Also, we observed a significant correlations between Δ(a-v)O2 and GLS at 4 min (panel E) and change in GLS (Panel F). Discussion Effort intolerance in type 2 diabetes We examined 88 older adults with well-controlled and iCPET. uncomplicated T2D undergoing a maximal The observed value of VO2peak in the whole population (18 mL/kg/min) falls far below the reference values for the general population of this age group (women 31 mL/ kg/min; men 39 mL/kg/min) [26], confirming previous reports indicating poor exercise capacity in T2D subjects [4]. An identical mean value of VO2peak (18.0 ± 6.6 mL/ min/kg) has been recently reported in a larger and even younger (by 10 years) cohort of 224 asymptomatic sub- jects with T2D, falling into the lower 10% of the age- matched male general population distribution and the lower 20% of female general population [27]. Exercise intolerance, defined by VO2peak below 80% of the pre- dicted value according to Wasserman equation, is widely used to define negative prognosis in subjects suffering from heart disease [1]; still, it was present in most (55%) of our study population despite the absence of either vas- cular and autonomic diabetic complications, criteria for a definite diagnosis of HF, or any detectable significant car- diac impairment at resting assessment. The prevalence Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 8 of 13 Table 3 Determinants of peripheral oxygen extraction Univariate Variable st-β Multivariate st-β p p ns 0.0138 0.35 0.0002 ns ns ns ns ns ns ns ns ns 0.01 0.07 0.01 0.01 0.04 0.02 0.02 − − − 0.01 0.02 1.02 0.50 − − 0.70 0.0103 0.22 0.0133 − 0.06 0.20 0.08 − 0.05 0.01 0.52 0.01 − − 0.01 0.10 0.12 0.34 0.19 0.06 0.06 0.16 0.12 0.06 − ns ns ns 0.0175 < 0.0001 < 0.0001 ns ns ns ns − 0.02 0.01 0.30 − ns ns ns 0.0060 < 0.0001 0.10 0.21 ns 0.0486 ns ns ns ns ns Age Sex BMI Duration of diabetes HbA1c Smoke Hypertension metformin Creatinine ACR (log) PCR (log) Hemoglobin RHI endoPAT CAVI ABI MBP peak SVR peak CO peak LVEF rest LVEF peak Δ LVEF GLS rest GLS 4 min Δ GLS S’ rest S’ peak Δ S’ E/e’ peak VE/VCO2 slope of exhaustion during exercise and in higher fatigue with respect to controls at any given workload, even when adjusted for the reduced VO2peak [29, 30]. Still, the rea- sons for the decreased exercise tolerance are far from being clear, possibly encompassing any combination of myocardiogenic, skeletal myogenic, vasculogenic, or neu- rogenic determinants [4]; we sought to determining the associated alterations in the different organs and systems. Mechanisms of effort intolerance in type 2 diabetes Our first finding is that effort intolerance is not due to a defect in mechanical efficiency (as was suggested for obese individuals [31, 32]), given that the slope of VO2 vs work-rate is steeper in subjects with preserved exercise tolerance (Fig. 1) and that we can exclude an impairment in ventilatory parameters. A reduced O2 supply could be related to central (lung and/or heart) or peripheral (hematologic, vascular, or mitochondrial) impairment Fig. 2 Linear regressions between systolic indices, peripheral oxygen extraction, and peak oxygen uptake. A significant positive linear correlation exists between S’ (A) at peak/change in S’ and VO2peak (B), as well as between GLS at 4 min and VO2peak (C) and change in GLS and VO2peak (D). A positive linear correlation was also observed between Δ(a-v)O2 and GLS at 4 min (E) and change in GLS (F) of this severe condition in uncomplicated T2D was not reported in previous studies, wherein the severity of effort intolerance has probably been underestimated. Indeed, the subjects in our cohort with effort intolerance show VO2peak values that are commonly found in patients with overt HF, a population wherein such a reduced VO2peak portends a rather poor prognosis [28]. The study population was homogeneous in demo- graphic parameters, glycemic control, duration of diabe- tes, cardio-active and glucose-lowering therapy, vascular and endothelial function parameters, as well as the prev- alence of comorbidities (Table 1). It is thus very difficult to predict effort intolerance based on the resting clinical phenotype alone. The difference in HDL cholesterol was small and seemingly driven by the slightly higher preva- lence of male and overweight subjects in the group with low VO2peak. This, together with the presence of a small difference in age, prompted the decision to verify the dif- ferences in iCPET data after adjusting for age, sex and BMI (Table 2). Since effort intolerance is the hallmark of HF irrespective of LVEF [2], and that cardiorespira- tory fitness is known to be a strong predictor of incident HFpEF in the T2D population [6], we sought to deter- mine the associated alterations and mechanisms under- pinning the reduced VO2peak in T2D patients to gain insight on the earliest defects at the bases of the their higher HF vulnerability. Previous findings reported early development of fatigue in T2D as a perceived limitation of force-generating capacity that requires higher intensity of effort that might eventually reduce the exercise dura- tion, and that can be highlighted by an early appearance Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 9 of 13 5 A 150 ] x a m 2 O V % [ k a e p 2 O V 125 100 75 50 25 C 150 ] x a m 2 O V % [ k a e p 2 O V 125 100 75 50 25 E 25 ] % i [ n m 4 t a S L G 20 15 10 r2 = 0.06 p = 0.02 10 15 S' meanpeak [cm/sec] 20 r2 = 0.10 p = 0.003 r2 = 0.06 p = 0.03 B 150 ] x a m 2 O V % [ k a e p 2 O V 125 100 75 50 25 25 0 2 r2 = 0.11 p = 0.004 D 150 ] x a m 2 O V % [ k a e p 2 O V 125 100 75 50 25 6 4 ΔS' mean [cm/sec] 8 10 12 10 15 20 GLS at 4 min [%] 25 0 1 2 3 4 ∆GLS [%] 5 6 7 r2 = 0.09 p = 0.006 r2 = 0.13 p = 0.001 F 8 6 4 2 0 ] % [ S L G ∆ 5 10 15 ∆(a-v)O2 [mL/dL] 20 5 10 15 ∆(a-v)O2 [mL/dL] 20 Fig. 3 Determinants of peak oxygen uptake according to Fick’s principle. VO2 is determined by cardiac output (CO) and peripheral oxygen extraction (Δ(a-v)O2). The p value for the difference between the two study groups is shown on top of the diagrams [4]. We excluded lung disease, as all the patients under- went spirometry before exercise, whilst hematologic diseases were excluded after analysis of blood exams before enrollment. Then, imaging-CPET provides the opportunity to dissect the different components of the Fick’s equation, thanks to the possibility of measuring stroke volume with the simultaneous echocardiographic assessment. According to the Fick’s principle, a reduced peripheral oxygen extraction explains the impaired car- diopulmonary function in our population (Fig. 3). Previ- ous studies have reported either a reduced or a normal peripheral extraction in T2D [33, 34]. Notably, both the study by Baldi et al. [33] and the more recent one by Kob- ayashi [35] performed in a similar population of asymp- tomatic T2D patients confirm our findings of a normal cardiac output with a Δ(a-v)O2 that was reduced by 20% compared to the control group and that correlated with the reduced VO2peak. Notably, in line with our findings, the Authors conclude that a reduced peripheral oxygen extraction might be regarded as a limitation to whole- body oxygen uptake. The older study showing normal Δ(a-v)O2 was conducted in a very small group of female adolescents [34], thus with poor clinical applicability. Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 10 of 13 The two groups of our study did not differ in biohu- moral values, endothelial function, indices of pulmonary function, diastolic function indices, right heart indices, mean arterial pressure, and systemic vascular resistances throughout the entire iCPET. The subjects with effort intolerance showed higher prevalence of chronotropic incompetence. The large difference, however, is driven by the fact the patients of our study group fall close to the 80% threshold (mean HR peak%: 86.1%) thus a small difference in peak HR generates a major recruitment of subject with a diagnosis of chronotropic incompetence. In quantitative terms, the difference between the groups was small; subjects with effort intolerance exploited their HR reserve only 8% less than the others (82.6 vs 90.2%) with a difference in peak HR of just 8 beats/min- ute (129 vs 137 bpm). If we also consider that no subject had evidence of CAN at conventional tests and, more importantly, that HR kinetics and peak CO were super- imposable in the two groups throughout the whole iCPET, it is unlikely that chronotropic incompetence is the cause of effort intolerance in our patient; it might rather be the consequence of their lower fitness. Whilst crude indices of systolic performance such as SV, CO and LVEF were not different between the two study groups, less load-dependent indices (S’ and GLS) showed a gradient that was evident both in resting con- ditions and in exercise-induced changes. It is widely known that a reduced baseline GLS is an early marker of LV subclinical systolic dysfunction, being present both in HFpEF patients irrespective of the diabetic state [24] and in T2D subjects without HF [36, 37]—where it also pre- dicts incident HF [38]. Our findings confirm most previ- ous reports (although not all [39]) describing reduced S’ velocity of the mitral annulus measured through tissue Doppler in patients with T2D during exercise [40–42], an observation that was also related to myocardial fibrosis as measured though cardiac magnetic resonance [43]. Given the large prevalence of diastolic dysfunction in T2D sub- jects [38] and the results of a recent report by Gulsin et al. [44], we were surprised not to see alteration in E/e’ in our population with effort intolerance, neither at rest nor during exercise. In the work of Gulsin and coll., how- ever, diastolic indices were not measured during exercise, and the association was essentially driven by a minority of subjects with a baseline E/e’ > of 12.5, and in the whole population the effect size was small with + 1 units of E/e’ justifying -0.3 ml/kg/min of VO2peak. Based on our data, diastolic dysfunction is not relevant for explaining effort intolerance in T2D. Determinants of peripheral oxygen extraction In multivariate analysis, the determinants of Δ(a-v)O2— the main factor explaining the reduced VO2peak in our population—were sex, hemoglobin levels, and ΔGLS (Table 3). This confirms previous observations on the sex-related differences in effort tolerance and possibly the different risk of HF development seen between T2D males and females [45]. It also supports the determinis- tic role of hemoglobin in ambient oxygen availability [1]. At rest, diabetic individuals show reduced ATP release from red blood cells in response to hemoglobin desatu- ration activated through endothelial purinergic recep- tors that trigger nitric oxide-dependent and independent arteriolar vasodilation, and that significantly impacts on muscle blood flow [46]; however, its relevance dur- ing exercise in unknown. Interestingly, SGLT-2 inhibi- tors, which protect T2D patients from HF incidence and decompensation through unknown mechanisms, signifi- cantly rise hemoglobin as a side effect [47]. Since SGLT-2 inhibitors have been recently demonstrated to amelio- rate aerobic fitness in T2D subjects both without HF [48] and with HFpEF [49], one can speculate that the hemo- concentration with increased hemoglobin obtained with their pharmacological effect might increase Δ(a-v)O2 and partly explain the increased whole-body oxygen uptake. Finally, the strength of the correlation between subclini- cal systolic dysfunction to peak Δ(a-v)O2 and VO2peak is a novel finding suggesting a strong link between skeletal and cardiac muscle pathology in T2D. This relationship was previously demonstrated in animal models of T2D [50], while in humans previous studies have reported either a reduced GLS or S’ at rest [51], reduced longitudi- nal systolic reserve [40], or reduced Δ(a-v)O2 in T2D sub- jects with effort intolerance [33], but none has reported a direct relationship between peripheral oxygen extraction and systolic indices, neither at rest nor during exercise. Clinical value Taken together, the results of the present study suggest a myogenic limitation of whole-body oxygen uptake in T2D limiting exercise tolerance, with a tight interplay between myocardial and skeletal muscles. Whether this is secondary to a reduced number of mitochondria, a mitochondrial functional impairment, altered myofi- brillar structure and/or composition, muscle microvas- culature, or to systemic regulators of muscle perfusion [4] goes beyond the purpose of this study. However, the lack of relationships with total peripheral resistances and endothelial function supports a primitive muscle cell impairment involving both skeletal and myocardial mus- cle. The lower anaerobic threshold, also when expressed in terms of % VO2peak, indicates a reduced aerobic capac- ity and strongly supports the hypothesis of a mitochon- drial defect either in number or in function, as previously observed in this population [52, 53]. The observation that exercise training can increase whole-body oxygen Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 11 of 13 consumption through an amelioration of skeletal muscle energetics further sustains this point [54, 55]. mechanisms of effort intolerance within the T2D popula- tion is unknown). The combination of reduced VO2peak, Δ(a-v)O2, and GLS of the subjects with T2D and effort intolerance observed in the present study represents a phenotype which is also shared by the patients with HFpEF wherein reduced peripheral oxygen extraction and/or GLS appear as the major determinants of effort intolerance [37, 56, 57]. Being each trait less pronounced in our population, we speculate that this condition might represent an inter- mediate phenotype and—if eventually confirmed to be so prevalent in T2D—might explain the excess prevalence of HFpEF among these patients [2, 56]. Of note, our popu- lation did not show diastolic dysfunction as a key deter- minant of effort intolerance, at least not as importantly as in HFpEF subjects [2], probably marking a further step forward towards overt HFpEF. Due to the “diabesity” pandemic and the high incidence (and costs) of HFpEF, new strategies for the early identification of the patients at risk of HFpEF are needed. In this context, the iCPET might reveal a useful screening tool [58]. Longitudi- nal trials evaluating the transition from T2D with effort intolerance to overt HFpEF would provide support to our hypothesis, as well as clinical intervention trials aiming at verifying whether ameliorating peripheral oxygen extrac- tion (e.g. exercise) is possible to prevent the development of HFpEF in T2D patients. Concluding remarks Effort intolerance is severe and highly prevalent in uncomplicated, otherwise asymptomatic T2D, and is mainly driven by a primitive muscular impairment involving both skeletal and myocardial muscle in the form of impaired peripheral oxygen extraction and a reduced systolic reserve, despite preserved LVEF and car- diac output. These alterations closely resemble the major clinical features of HFpEF and could represent an inter- mediate pathological condition. Strengths and limitations One strength of the present study is the sample size (greater than previous reports), the multi-parametric analysis performed (previous works were focused on specific organ dysfunctions, or did not use exercise echo- cardiography, or considered resting variables), the care- ful exclusion of micro- and macrovascular complications (previous works frequently included complicated patients and highlighted the contribution of specific complica- tions to reduced aerobic capacity), and the study of effort intolerance within the diabetic population (whilst previ- ous works were focused on differences between diabetic and non-diabetic subjects, and the exact prevalence and We recognize some limitations of the present work. This is a single-centre, cross-sectional study with a rela- tively small sample size. We only focused on asympto- matic, uncomplicated T2D; therefore, the results should not be applied to different cohorts. We acknowledge that the technical challenge of acquiring echocardiogra- phy images during exercise may affect SV and CO meas- urements, despite the technique has been extensively validated and used by different groups [24]. Also Δ(a-v) O2 was not directly measured, however our method has been extensively validated, used by several investigators [24, 25]; and, most importantly, our values are in line with observations reporting both non-invasive and inva- sive oxygen extraction data [57, 59]. Our imaging proto- col was performed in a semi-supine position for a better echocardiographic evaluation [13]; caution is advised to extend our results to other types of exercise (supine or upright). Abbreviations AT: Anaerobic threshold; BMI: Body mass index; CO: Cardiac output; CPET: Car- diopulmonary exercise test; SV: Stroke volume; E’: Early left ventricular filling phase velocity; HF: Heart failure; HFpEF: Heart failure with preserved ejection fraction; HR: Heart rate; METs: Metabolic equivalents; PaCO2: Arterial partial pressure of carbon dioxide; PETCO2: End-tidal carbon dioxide partial pressure; S’: Peak systolic annular velocity; T2D: Type 2 diabetes mellitus; VCO2: Carbon dioxide emission measured at the mouth; VD/VT: Physiological dead space/ tidal volume ratio; VE: Pulmonary minute ventilation; VE/VCO2: Ventilatory equivalent for carbon dioxide or “ventilatory efficiency”; VO2: Oxygen uptake measured at the mouth; VO2/HR: Oxygen pulse; VO2max: Maximal theorical oxygen uptake; VO2peak: Peak oxygen uptake; Δ(a-v)O2: Peripheral oxygen extraction. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12933- 021- 01314-6. Additional file 1: Table S4. Sensitivity analysis performed after having excluded all subjects taking beta-blockers. And repeating the statistical analysis as described for the whole population. Acknowledgements The authors are grateful to all the volunteers and the personnel of the Metab- olism, Nutrition, and Atherosclerosis Lab and the Cardiopulmonary Test Lab, Department of Clinical and Experimental Medicine, University of Pisa, Italy. Authors’ contributions LN performed the clinical and biohumoral characterization of patients, the screening of diabetic complications, kept patients’ records, created the database, performed the statistical analysis, wrote the manuscript, and ideated and produced the tables and the figures. NRP performed the imaging-cardiopulmonary exercise tests, provided a substantial contribu- tion to the interpretation of the data, and critically revised the manuscript. PS performed the screening of eligible patients, kept patients’ records, and performed the screening of diabetic complications. NDB and MM performed the cardiopulmonary exercise tests, kept patients’ records, and critically revised the manuscript. IF performed the cardiopulmonary exercise tests, provided Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 12 of 13 a substantial contribution to the interpretation of the data, and critically revised the manuscript. DT performed the screening of patients, the clinical and biohumoral characterization, provided a substantial contribution to the interpretation of the data, revised the figures, and critically revised the manu- script. SM provided a substantial contribution to the interpretation of the data and critically revised the manuscript. AN conceived and designed the paper, provided a substantial contribution to the analysis and interpretation of the data, and critically revised the manuscript. All authors revised the manuscript, read and approved the final manuscript. Funding No funding was required for writing this paper. Availability of data and materials Not applicable. Declarations Ethics approval and consent to participate Not applicable. Consent for publication All the Authors gave their consent to publication. Competing interests The authors have no conflict of interest to declare. Author details 1 Metabolism, Nutrition, and Atherosclerosis Laboratory, Department of Clinical and Experimental Medicine, University of Pisa, Via Savi 10, 56126 Pisa, Italy. 2 Cardiopulmonary Laboratory, Department of Clinical and Experimental Medi- cine, University of Pisa, Pisa, Italy. 3 Fondazione Toscana G. Monasterio, Pisa, Italy. 4 Department of Surgical, Medical and Molecular Pathology and Critical Care Medicine, University of Pisa, Pisa, Italy. Received: 22 March 2021 Accepted: 3 June 2021 References 1. American Thoracic S, American College of Chest P. ATS/ACCP state- ment on cardiopulmonary exercise testing. Am J Respir Crit Care Med. 2003;167(2):211–77. Tannenbaum S, Sayer GT. Advances in the pathophysiology and treat- ment of heart failure with preserved ejection fraction. Curr Opin Cardiol. 2015;30(3):250–8. Lehrke M, Marx N. Diabetes mellitus and heart failure. Am J Cardiol. 2017;120(1S):S37–47. 2. 3. 4. Nesti L, Pugliese NR, Sciuto P, Natali A. Type 2 diabetes and reduced exer- cise tolerance: a review of the literature through an integrated physiology approach. Cardiovasc Diabetol. 2020;19(1):134. 5. Church TS, LaMonte MJ, Barlow CE, Blair SN. Cardiorespiratory fitness and body mass index as predictors of cardiovascular disease mortality among men with diabetes. Arch Intern Med. 2005;165(18):2114–20. 6. Pandey A, Patel KV, Bahnson JL, Gaussoin SA, Martin CK, Balasu- bramanyam A, Johnson KC, McGuire DK, Bertoni AG, Kitzman D, et al. Association of intensive lifestyle intervention, fitness, and body mass index with risk of heart failure in overweight or obese adults with type 2 diabetes mellitus: an analysis from the look AHEAD trial. Circulation. 2020;141(16):1295–306. Trikkalinou A, Papazafiropoulou AK, Melidonis A. Type 2 diabetes and quality of life. World J Diabetes. 2017;8(4):120–9. 7. 8. American Diabetes A. 2. Classification and diagnosis of diabetes: stand- ards of medical care in diabetes-2020. Diabetes Care. 2020;43(Suppl 1):S14–31. 9. Ponikowski P, Voors AA, Anker SD, Bueno H, Cleland JGF, Coats AJS, Falk V, Gonzalez-Juanatey JR, Harjola VP, Jankowska EA, et al. 2016 ESC Guide- lines for the diagnosis and treatment of acute and chronic heart failure: the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) Developed with the special contribution of the Heart Failure Association (HFA) of the ESC. Eur Heart J. 2016;37(27):2129–200. 10. Trico D, Nesti L, Frascerra S, Baldi S, Mengozzi A, Natali A. A protein/lipid preload attenuates glucose-induced endothelial dysfunction in individu- als with abnormal glucose tolerance. Nutrients. 2020; 12(7). 11. Mueller MJ. Identifying patients with diabetes mellitus who are at risk for lower-extremity complications: use of Semmes-Weinstein monofila- ments. Phys Ther. 1996;76(1):68–71. 12. Dimitropoulos G, Tahrani AA, Stevens MJ. Cardiac autonomic neuropathy in patients with diabetes mellitus. World J Diabetes. 2014;5(1):17–39. 13. Guazzi M, Bandera F, Ozemek C, Systrom D, Arena R. cardiopulmonary exercise testing: what is its value? J Am Coll Cardiol. 2017;70(13):1618–36. 14. Kosmala W, Rojek A, Przewlocka-Kosmala M, Mysiak A, Karolko B, Marwick TH. Contributions of nondiastolic factors to exercise intoler- ance in heart failure with preserved ejection fraction. J Am Coll Cardiol. 2016;67(6):659–70. 15. Eisman AS, Shah RV, Dhakal BP, Pappagianopoulos PP, Wooster L, Bailey C, Cunningham TF, Hardin KM, Baggish AL, Ho JE, et al. Pulmonary capillary wedge pressure patterns during exercise predict exercise capacity and incident heart failure. Circ Heart Fail. 2018;11(5):e004750. 16. Topilsky Y, Rozenbaum Z, Khoury S, Pressman GS, Gura Y, Sherez J, Man A, Shimiaie J, Edwards S, Berookhim J, et al. Mechanisms of effort intoler- ance in patients with heart failure and borderline ejection fraction. Am J Cardiol. 2017;119(3):416–22. 17. Borlaug BA, Kane GC, Melenovsky V, Olson TP. Abnormal right ventricular- pulmonary artery coupling with exercise in heart failure with preserved ejection fraction. Eur Heart J. 2016;37(43):3293–302. 18. Brubaker PH, Kitzman DW. Chronotropic incompetence: causes, conse- quences, and management. Circulation. 2011;123(9):1010–20. 19. Lang RM, Badano LP, Mor-Avi V, Afilalo J, Armstrong A, Ernande L, Flachskampf FA, Foster E, Goldstein SA, Kuznetsova T, et al. Recommenda- tions for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging. Eur Heart J Cardiovasc Imaging. 2015;16(3):233–70. 20. Pugliese NR, Fabiani I, Mandoli GE, Guarini G, Galeotti GG, Miccoli M, Lombardo A, Simioniuc A, Bigalli G, Pedrinelli R, et al. Echo-derived peak cardiac power output-to-left ventricular mass with cardiopulmo- nary exercise testing predicts outcome in patients with heart failure and depressed systolic function. Eur Heart J Cardiovasc Imaging. 2019;20(6):700–8. 21. Pugliese NR, De Biase N, Gargani L, Mazzola M, Conte L, Fabiani I, Natali A, Dini FL, Frumento P, Rosada J, et al. Predicting the transition to and progression of heart failure with preserved ejection fraction: a weighted risk score using bio-humoural, cardiopulmonary, and echocardiographic stress testing. Eur J Prev Cardiol. 2020. https:// doi. org/ 10. 1093/ eurjpc/ zwaa1 29. 22. Pugliese NR, Mazzola M, Fabiani I, Gargani L, De Biase N, Pedrinelli R, Natali A, Dini FL. Haemodynamic and metabolic phenotyping of hypertensive patients with and without heart failure by combining cardiopulmonary and echocardiographic stress test. Eur J Heart Fail. 2020;22(3):458–68. 23. Pugliese NR, De Biase N, Conte L, Gargani L, Mazzola M, Fabiani I, Natali A, Dini FL, Frumento P, Rosada J, et al. Cardiac reserve and exercise capacity: insights from combined cardiopulmonary and exercise echocardiogra- phy stress testing. J Am Soc Echocardiogr. 2021;34(1):38–50. 24. Shimiaie J, Sherez J, Aviram G, Megidish R, Viskin S, Halkin A, Ingbir M, Nesher N, Biner S, Keren G, et al. Determinants of effort intolerance in patients with heart failure: combined echocardiography and cardiopul- monary stress protocol. JACC Heart Fail. 2015;3(10):803–14. 25. Wolsk E, Kaye D, Komtebedde J, Shah SJ, Borlaug BA, Burkhoff D, Kitz- man DW, Lam CSP, van Veldhuisen DJ, Ponikowski P, et al. Central and peripheral determinants of exercise capacity in heart failure patients with preserved ejection fraction. JACC Heart Fail. 2019;7(4):321–32. 26. Aspenes ST, Nilsen TI, Skaug EA, Bertheussen GF, Ellingsen O, Vatten L, Wisloff U. Peak oxygen uptake and cardiovascular risk factors in 4631 healthy women and men. Med Sci Sports Exerc. 2011;43(8):1465–73. 27. Rapp D, Scharhag J, Wagenpfeil S, Scholl J. Reference values for peak oxygen uptake: cross-sectional analysis of cycle ergometry-based car- diopulmonary exercise tests of 10 090 adult German volunteers from the Prevention First Registry. BMJ Open. 2018;8(3):e018697. Nesti et al. Cardiovasc Diabetol (2021) 20:124 Page 13 of 13 28. Corra U, Agostoni PG, Anker SD, Coats AJS, Crespo Leiro MG, de Boer RA, Harjola VP, Hill L, Lainscak M, Lund LH, et al. Role of cardiopulmonary exercise testing in clinical stratification in heart failure. A position paper from the Committee on Exercise Physiology and Training of the Heart Failure Association of the European Society of Cardiology. Eur J Heart Fail. 2018;20(1):3–15. of aerobic exercise capacity in adults with type 2 diabetes. Diabetes Care. 2020;43(9):2248–56. 45. Athithan L, Chowdhary A, Swarbrick D, Gulsin GS, Singh A, Jex N, Jain M, Khan JN, Graham-Brown MPM, Wormleighton JV, et al. Male sex adversely affects the phenotypic expression of diabetic heart disease. Ther Adv Endocrinol Metab. 2020;11:2042018820927179. 29. Huebschmann AG, Reis EN, Emsermann C, Dickinson LM, Reusch JE, 46. Thaning P, Bune LT, Hellsten Y, Pilegaard H, Saltin B, Rosenmeier JB. Bauer TA, Regensteiner JG. Women with type 2 diabetes perceive harder effort during exercise than nondiabetic women. Appl Physiol Nutr Metab. 2009;34(5):851–7. 30. Kim YS, Seifert T, Brassard P, Rasmussen P, Vaag A, Nielsen HB, Secher NH, van Lieshout JJ. Impaired cerebral blood flow and oxygenation during exercise in type 2 diabetic patients. Physiol Rep. 2015;3(6):e12430. 31. Lafortuna CL, Proietti M, Agosti F, Sartorio A. The energy cost of cycling in young obese women. Eur J Appl Physiol. 2006;97(1):16–25. 32. Mancusi C, de Simone G, Best LG, Wang W, Zhang Y, Roman MJ, Lee ET, Howard BV, Devereux RB. Myocardial mechano-energetic efficiency and insulin resistance in non-diabetic members of the Strong Heart Study cohort. Cardiovasc Diabetol. 2019;18(1):56. 33. Baldi JC, Aoina JL, Oxenham HC, Bagg W, Doughty RN. Reduced exercise arteriovenous O2 difference in Type 2 diabetes. J Appl Physiol (1985). 2003;94(3):1033–8. 34. Gusso S, Hofman P, Lalande S, Cutfield W, Robinson E, Baldi JC. Impaired stroke volume and aerobic capacity in female adolescents with type 1 and type 2 diabetes mellitus. Diabetologia. 2008;51(7):1317–20. 35. Kobayashi Y, Christle JW, Contrepois K, Nishi T, Moneghetti K, Cauwen- berghs N, Myers J, Kuznetsova T, Palaniappan L, Haddad F. Peripheral oxygen extraction and exercise limitation in asymptomatic patients with diabetes mellitus. Am J Cardiol. 2021. https:// doi. org/ 10. 1016/j. amjca rd. 2021. 03. 011. 36. Vukomanovic V, Suzic-Lazic J, Celic V, Cuspidi C, Petrovic T, Grassi G, Tadic M. The relationship between functional capacity and left ventricular strain in patients with uncomplicated type 2 diabetes. J Hypertens. 2019;37(9):1871–6. 37. Kraigher-Krainer E, Shah AM, Gupta DK, Santos A, Claggett B, Pieske B, Zile MR, Voors AA, Lefkowitz MP, Packer M, et al. Impaired systolic function by strain imaging in heart failure with preserved ejection fraction. J Am Coll Cardiol. 2014;63(5):447–56. 38. Wang Y, Yang H, Huynh Q, Nolan M, Negishi K, Marwick TH. Diagnosis of nonischemic stage B heart failure in type 2 diabetes mellitus: optimal parameters for prediction of heart failure. JACC Cardiovasc Imaging. 2018;11(10):1390–400. 39. Roberts TJ, Burns AT, MacIsaac RJ, MacIsaac AI, Prior DL, La Gerche A. Exercise capacity in diabetes mellitus is predicted by activity status and cardiac size rather than cardiac function: a case control study. Cardiovasc Diabetol. 2018;17(1):44. 40. Ha JW, Lee HC, Kang ES, Ahn CM, Kim JM, Ahn JA, Lee SW, Choi EY, Rim SJ, Oh JK, et al. Abnormal left ventricular longitudinal functional reserve in patients with diabetes mellitus: implication for detecting subclinical myocardial dysfunction using exercise tissue Doppler echocardiography. Heart. 2007;93(12):1571–6. 41. Jellis CL, Jenkins C, Leano R, Martin JH, Marwick TH. Reduced end-systolic pressure-volume ratio response to exercise: a marker of subclini- cal myocardial disease in type 2 diabetes. Circ Cardiovasc Imaging. 2010;3(4):443–9. 42. Zahiti BF, Gorani DR, Gashi FB, Gjoka SB, Zahiti LB, Haxhiu BS, Kamberi LS. Left ventricular diastolic dysfunction in asymptomatic type 2 diabetic patients: detection and evaluation by tissue Doppler imaging. Acta Inform Med. 2013;21(2):120–3. 43. Jellis C, Wright J, Kennedy D, Sacre J, Jenkins C, Haluska B, Martin J, Fen- wick J, Marwick TH. Association of imaging markers of myocardial fibrosis with metabolic and functional disturbances in early diabetic cardiomyo- pathy. Circ Cardiovasc Imaging. 2011;4(6):693–702. 44. Gulsin GS, Henson J, Brady EM, Sargeant JA, Wilmot EG, Athithan L, Htike ZZ, Marsh AM, Biglands JD, Kellman P, et al. Cardiovascular determinants Attenuated purinergic receptor function in patients with type 2 diabetes. Diabetes. 2010;59(1):182–9. 47. Zinman B, Wanner C, Lachin JM, Fitchett D, Bluhmki E, Hantel S, Mattheus M, Devins T, Johansen OE, Woerle HJ, et al. Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes. N Engl J Med. 2015;373(22):2117–28. 48. Kumar N, Garg A, Bhatt DL, Sabongui S, Gupta N, Chaudhry S, Arena R, Verma S. Empagliflozin improves cardiorespiratory fitness in type 2 diabetes: translational implications. Can J Physiol Pharmacol. 2018;96(11):1184–7. 49. Carbone S, Canada JM, Billingsley HE, Kadariya D, Dixon DL, Trankle CR, Buckley LF, Markley R, Vo C, Medina de Chazal H, et al. Effects of empa- gliflozin on cardiorespiratory fitness and significant interaction of loop diuretics. Diabetes Obes Metab. 2018;20(8):2014–8. 50. Chen Y, Zeng W, Chen W, Zhang Y, Zhu T, Sun J, Liang Z, Wang L, Yang Z, Wu B, et al. Evaluating the correlation of the impairment between skeletal muscle and heart using MRI in a spontaneous type 2 diabetes mellitus rhesus monkey model. Acta Diabetol. 2020;57(6):673–9. 51. Enomoto M, Ishizu T, Seo Y, Yamamoto M, Suzuki H, Shimano H, Kawakami Y, Aonuma K. Subendocardial systolic dysfunction in asympto- matic normotensive diabetic patients. Circ J. 2015;79(8):1749–55. 52. Koliaki C, Roden M. Alterations of mitochondrial function and insulin sensitivity in human obesity and diabetes mellitus. Annu Rev Nutr. 2016;36:337–67. 53. Schrauwen-Hinderling VB, Kooi ME, Hesselink MK, Jeneson JA, Backes WH, van Echteld CJ, van Engelshoven JM, Mensink M, Schrauwen P. Impaired in vivo mitochondrial function but similar intramyocellular lipid content in patients with type 2 diabetes mellitus and BMI-matched control sub- jects. Diabetologia. 2007;50(1):113–20. 54. Malin SK, Francois ME, Eichner NZM, Gilbertson NM, Heiston EM, Fabris C, Breton M. Impact of short-term exercise training intensity on beta-cell function in older obese adults with prediabetes. J Appl Physiol (1985). 2018;125(6):1979–86. 55. Eser P, Marcin T, Prescott E, Prins LF, Kolkman E, Bruins W, van der Velde AE, Pena-Gil C, Iliou MC, Ardissino D, et al. Clinical outcomes after cardiac rehabilitation in elderly patients with and without diabetes mellitus: the EU-CaRE multicenter cohort study. Cardiovasc Diabetol. 2020;19(1):37. 56. Montero D, Diaz-Canestro C. Determinants of exercise intolerance in heart failure with preserved ejection fraction: a systematic review and meta-analysis. Int J Cardiol. 2018;254:224–9. 57. Dhakal BP, Malhotra R, Murphy RM, Pappagianopoulos PP, Baggish AL, Weiner RB, Houstis NE, Eisman AS, Hough SS, Lewis GD. Mechanisms of exercise intolerance in heart failure with preserved ejection fraction: the role of abnormal peripheral oxygen extraction. Circ Heart Fail. 2015;8(2):286–94. 58. Ross R, Blair SN, Arena R, Church TS, Despres JP, Franklin BA, Haskell WL, Kaminsky LA, Levine BD, Lavie CJ, et al. Importance of assessing cardiores- piratory fitness in clinical practice: a case for fitness as a clinical vital sign: a scientific statement from the American Heart Association. Circulation. 2016;134(24):e653–99. 59. Haykowsky MJ, Brubaker PH, John JM, Stewart KP, Morgan TM, Kitzman DW. Determinants of exercise intolerance in elderly heart failure patients with preserved ejection fraction. J Am Coll Cardiol. 2011;58(3):265–74. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations.
10.1186_s43058-023-00433-3	Pluta et al. Implementation Science Communications (2023) 4:50 https://doi.org/10.1186/s43058-023-00433-3 RESEARCH Implementation Science Communications Open Access Data envelopment analysis to evaluate the efficiency of tobacco treatment programs in the NCI Moonshot Cancer Center Cessation Initiative Kathryn Pluta1, Sarah D. Hohl2,3, Heather D’Angelo4, Jamie S. Ostroff5, Donna Shelley6, Yasmin Asvat7, Li‑Shiun Chen8, K. Michael Cummings9, Neely Dahl10, Andrew T. Day11, Linda Fleisher12, Adam O. Goldstein13, Rashelle Hayes14, Brian Hitsman15, Deborah Hudson Buckles16, Andrea C. King17, Cho Y. Lam18, Katie Lenhoff19, Arnold H. Levinson20, Mara Minion2, Cary Presant21, Judith J. Prochaska22, Kimberly Shoenbill13, Vani Simmons23, Kathryn Taylor24, Hilary Tindle25, Elisa Tong26, Justin S. White27, Kara P. Wiseman10, Graham W. Warren9, Timothy B. Baker3, Betsy Rolland2,28, Michael C. Fiore2,3 and Ramzi G. Salloum1* Abstract Background The Cancer Center Cessation Initiative (C3I) is a National Cancer Institute (NCI) Cancer Moonshot Pro‑ gram that supports NCI‑designated cancer centers developing tobacco treatment programs for oncology patients who smoke. C3I‑funded centers implement evidence‑based programs that offer various smoking cessation treat‑ ment components (e.g., counseling, Quitline referrals, access to medications). While evaluation of implementation outcomes in C3I is guided by evaluation of reach and effectiveness (via RE‑AIM), little is known about technical efficiency—i.e., how inputs (e.g., program costs, staff time) influence implementation outcomes (e.g., reach, effective‑ ness). This study demonstrates the application of data envelopment analysis (DEA) as an implementation science tool to evaluate technical efficiency of C3I programs and advance prioritization of implementation resources. Methods DEA is a linear programming technique widely used in economics and engineering for assessing relative performance of production units. Using data from 16 C3I‑funded centers reported in 2020, we applied input‑oriented DEA to model technical efficiency (i.e., proportion of observed outcomes to benchmarked outcomes for given input levels). The primary models used the constant returns‑to‑scale specification and featured cost‑per‑participant, total full‑time equivalent (FTE) effort, and tobacco treatment specialist effort as model inputs and reach and effectiveness (quit rates) as outcomes. Results In the DEA model featuring cost‑per‑participant (input) and reach/effectiveness (outcomes), average con‑ stant returns‑to‑scale technical efficiency was 25.66 (SD = 24.56). When stratified by program characteristics, technical efficiency was higher among programs in cohort 1 (M = 29.15, SD = 28.65, n = 11) vs. cohort 2 (M = 17.99, SD = 10.16, n = 5), with point‑of‑care (M = 33.90, SD = 28.63, n = 9) vs. no point‑of‑care services (M = 15.59, SD = 14.31, n = 7), larger *Correspondence: Ramzi G. Salloum rsalloum@ufl.edu Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Pluta et al. Implementation Science Communications (2023) 4:50 Page 2 of 13 (M = 33.63, SD = 30.38, n = 8) vs. smaller center size (M = 17.70, SD = 15.00, n = 8), and higher (M = 29.65, SD = 30.99, n = 8) vs. lower smoking prevalence (M = 21.67, SD = 17.21, n = 8). Conclusion Most C3I programs assessed were technically inefficient relative to the most efficient center benchmark and may be improved by optimizing the use of inputs (e.g., cost‑per‑participant) relative to program outcomes (e.g., reach, effectiveness). This study demonstrates the appropriateness and feasibility of using DEA to evaluate the relative performance of evidence‑based programs. Keywords Data envelopment analysis, Efficiency, Program performance, Implementation costs, Smoking cessation, Implementation science, Tobacco treatment, Cancer Contributions to the literature • This study demonstrates the utility of data envelop- ment analysis (DEA) as a novel implementation science tool for evaluating program efficiency. • DEA allows for the identification of program fac- tors associated with higher levels of relative efficiency, which can be leveraged to increase efficiency across peer programs • Decision makers can use findings from DEA to improve efficiency of existing tobacco treatment pro- grams within oncology settings by identifying the types of programs that maximize reach and effectiveness while minimizing costs. • Maximizing efficiency of tobacco treatment programs can promote better program sustainability long term. Background Tobacco use is a preventable risk factor that can exacer- bate adverse health outcomes for individuals with cancer, increasing risk for all-cause mortality, cancer-specific mortality, cancer recurrence, and worsening response to cancer treatment [1, 2]. Failed first-line cancer therapy associated with continued tobacco use adds a major bur- den to the US healthcare system, amounting to approxi- mately $3.4 billion per year, or $10,678 annual cost per patient [3]. Thus, timely tobacco treatment in patients with cancer is imperative to mitigate the harmful effects of tobacco use on individuals’ health and rising health care costs. The importance of smoking cessation for individuals with cancer is widely recognized by cancer organiza- tions and the Surgeon General [1, 2]. However, patients with cancer face barriers to accessing tobacco treatment as part of their cancer care [1, 4, 5]. Screening patients for smoking using electronic health records and referring them to smoking cessation programs can effectively facil- itate their engagement in these programs [6]. However, tobacco cessation interventions are not routinely offered as a part of standard care in oncology. As such, patients with cancer who smoke are not consistently connected with necessary tobacco treatment resources [7]. In response to this gap, the Cancer Center Cessa- tion Initiative (C3I) was launched in 2017 as part of the National Cancer Institute (NCI) Cancer Moonshot pro- gram with the aim of integrating tobacco cessation treat- ments into routine cancer care. The goal of this multilevel implementation initiative has been to foster and sustain evidence-based tobacco cessation programs for patients with cancer and to encourage system-level change by identifying and referring patients who use tobacco to ces- sation treatments [8, 9]. C3I includes 52 NCI-designated cancer centers, which have implemented evidence-based tobacco treatments into their standard of care [10]. C3I provided funding to cancer centers across three cohorts: 42 centers received funding for two years over two cycles (i.e., cohort 1: 2017–2019; cohort 2: 2018–2020), and 10 centers received funding for 1 year (i.e., cohort 3: 2020–2021). Eleven cohort 1 and 2 centers were funded for an additional year as enhancement sites. Each funded cancer center offers evidence-based smoking cessa- tion treatments (e.g., in-person/telephone/video-based/ point-of-care counseling, cessation mediation, patient education material, Quitline referral), with variability in the provision of type and number of specific treatments across centers. C3I is guided by the reach, effectiveness, adoption, implementation, and maintenance (RE-AIM) framework, which facilitates multilevel assessment of the programs’ health-related impact (i.e., individual, organizational, community) [11–13]. Identifying factors that contrib- ute to the fidelity and effectiveness of C3I programs is imperative for their sustainability [14, 15]. Cancer centers participating in C3I reported key implementation out- comes biannually for continued program evaluation and to inform future resource allocation needs. Program reach and effectiveness are two key out- comes of interest that are reported by C3I-funded can- cer centers as standardized outcomes and have been used to evaluate the success of the tobacco treatment programs. Reach is defined as the proportion of patients who received at least one component of evidence-based, Pluta et al. Implementation Science Communications (2023) 4:50 Page 3 of 13 tobacco treatment. Effectiveness is defined as patient- reported 7-day point prevalence abstinence at 6-month follow-up. C3I centers’ reach and effectiveness rates varied as a function of center characteristics, including cancer center size, implementation strategies used, and types of treatments offered. For instance, Hohl et al. [16] found that cancer center size (i.e., number of patients served) was positively associated with effectiveness and negatively associated with reach. C3I centers that imple- mented tobacco treatment programs center-wide had similar effectiveness and higher reach than centers that engaged in partial implementation [16]. Further, centers that offered tobacco treatment through interactive voice response (i.e., automated calls) had the highest median reach and lowest effectiveness, whereas centers that implemented in-person face-to-face counseling had the highest effectiveness but the lowest reach. Additionally, offering six or fewer (vs. seven) types of tobacco treat- ments within a program was associated with higher reach and effectiveness. In addition to reach and effectiveness, cost is an impor- tant factor that can affect program sustainability [17]. An economic evaluation of 15 C3I sites found that monthly operating costs per site ranged from $5129 to $20,751 (median = $11,045), with most costs going towards per- sonnel [18]. Cost per participant ranged from $70 to $2093 (median = $454) and cost per quit was less than $3500 across centers. Overall, C3I centers achieved sat- isfactory quit rates at reasonable costs [14], and the pro- grams were expected to become more cost-effective as they continued to scale up. Identifying factors associ- ated with high reach and effectiveness while minimizing costs is imperative for maximizing program efficiency and sustainability. However, operating costs of C3I pro- grams in relation to their reach and effectiveness has not been examined. Therefore, the objective of this study was to compare program outcomes relative to expended resources and to identify best practices across can- cer centers, including which program components and implementation strategies were associated with optimal efficiency. To compare C3I program outcomes relative to expended resources, we applied a mathematical optimi- zation method called data envelopment analysis (DEA). DEA is widely used in economics and engineering for measuring the relative performance of production units [19, 20]. One advantage of DEA is that it does not require any parametric assumptions regarding data distributions, and data are not restricted to any functional form [20]. DEA assesses the ratio of outputs to inputs when evaluat- ing performance and produces a “best practice frontier” representing the best performing units. Performance of the remaining units is calculated as a relative score compared to the unit(s) located on the best practice fron- tier. Thus, DEA can be particularly useful in assessing which C3I programs are operating most efficiently. We assessed efficiency of C3I programs to demonstrate the application of this method in implementation science. In this study, each C3I program was compared against the best practice frontier which consists of the C3I program(s) with the most efficient performance (i.e., pro- portion of observed outcomes to benchmarked outcomes for given input levels). Multiple inputs and outcomes can be considered simultaneously while using DEA, which allows for evaluation of several factors related to perfor- mance, such as cost, reach, and effectiveness. Although these implementation outcomes are commonly assessed in implementation research, efficiency is rarely evalu- ated as it relates to the implementation of evidence-based practices. Thus, this study also seeks to demonstrate the utility of using DEA as a program evaluation tool within the field of implementation science. By assessing the reach and effectiveness relative to resources expended, decision makers can be better informed regarding which factors contribute to the most effective and sustainable program components in order to maximize impact of their programs. Methods Overview This is a descriptive study using DEA to examine reach and effectiveness of the C3I program relative to resources expended. DEA applications in health services give insights into which organizations are more efficient than others using program outcomes as outputs and resources expended as inputs [21]. This study examined program efficiency in cohorts 1 and 2 of C3I participating can- cer centers that had implemented tobacco treatment into oncology care. C3I sites implemented variations in implementation tobacco treatment components and strategies, requiring investments in different types and proportions of resources, including expenditures on staff- ing, medications, and electronic health record systems. The heterogeneity of components and implementation strategies, as well as the presence of multiple outcomes of interest, pose challenges for evaluating the relative per- formance of these programs. Cancer centers with exist- ing tobacco treatment programs have historically focused on different outcomes as their primary objectives (e.g., by emphasizing reach vs. effectiveness) [22]. The diversity in the way centers invest in resources and prioritize out- comes is reflective of the differences in implicit valuation that cancer centers assign to various program compo- nents and outcomes. DEA allows for multiple inputs and outcomes to be considered simultaneously without any parametric assumptions on data distributions. DEA is Pluta et al. Implementation Science Communications (2023) 4:50 Page 4 of 13 appropriate for comparing C3I programs due to its char- acterization of the implicit valuation placed on program components, which varies by site, and its ability to simul- taneously model efficiency for multiple outcomes, such as reach and effectiveness. Sixteen of 42 NCI-designated cancer centers from cohorts 1 and 2 that had complete data for input and outcome measures of interest (i.e., tobacco treatment specialist (TTS), full time equivalent (FTE) of overall staff, cost-per-patient, reach, and effec- tiveness) were included in this study. We stratified the analysis by cancer center characteristics because identify- ing factors associated with cancer centers that maintain high reach and effectiveness given budget constraints is important to foster sustainability of C3I programs. Strati- fying the analysis by cancer center characteristics clarifies which components are associated with greater efficiency and informs how efficiency can be improved at underper- forming centers. Data collection procedures treatment program evaluation data were Tobacco reported to the C3I Coordinating Center, based at the University of Wisconsin-Madison Carbone Cancer Center. The Coordinating Center assisted grantees with integrating evidence-based tobacco treatment services into cancer care [10] and created standardized met- rics to evaluate the tobacco treatment programs. All C3I cancer centers received an online questionnaire via Qualtrics (Provo, UT) every 6 months from the Coordi- nating Center, which assessed center characteristics (e.g., size, TTS FTE, treatments offered) and outcomes (e.g., reach, effectiveness). Specific methods regarding C3I measurement are detailed elsewhere [10, 23]. C3I cent- ers were given the option to report implementation costs and other resources expended (e.g., number of tobacco treatment specialists, program staff FTE) using an addi- tional biannual Qualtrics survey [14]. Cost data used in this study were reported during the January to June 2020 reporting cycle. To be included in this study, centers must have reported reach, effectiveness, cost, total program FTE, and tobacco treatment specialist FTE. This study was classified by the University of Wisconsin-Madison and University of Florida Institutional Review Boards as program evaluation and therefore exempt. period. Smoking prevalence was assessed by the pro- portion of cancer patients within the center who were documented in the electronic health record system as currently smoking cigarettes. Centers that offered point-of-care counseling for tobacco cessation (i.e., in- person or telehealth) included programs with a brief intervention delivered by a health care provider during routine oncology appointments to discuss evidence- based tobacco treatment options and offer tobacco ces- sation-related advice [10, 16, 23]. C3I programs were also categorized by cohort, whereby cohort 1 sites received funding from 2017 to 2019, and cohort 2 sites received funding between 2018 and 2020. The reporting period was the same for both cohorts, and we did not control for lead time among cohort 1 sites. Therefore, cohort 1 sites had more implementation experience than cohort 2 sites, on average, for each assessment. Tobacco treatment program components C3I sites reported the types of evidence-based treatments offered in their programs. These treatments included the following: in-person individual or group counseling, tele- phone-based counseling, point-of-care counseling, inter- active voice response system track and triage services (i.e., TelASK), Quitline referral, SmokefreeTXT text mes- saging service, online resources (e.g., smokefree.gov), and smoking cessation medications. Input measures Measures indicating presence of a TTS on site, total FTE, and cost were collected through the biannual cost surveys and used as input measures. C3I cancer centers reported FTE of tobacco treatment specialists employed in the program. Sites also reported FTE associated with all tobacco treatment program staff by personnel type, which was summed across personnel types to derive the total FTE measure. “Other personnel FTE” was cal- culated by subtracting TTS FTE from total FTE. We calculated cost-per-patient by dividing total monthly operating costs of each participating C3I center by the number of patients participating in a tobacco treatment program within the 6-month reporting period. Details of how total monthly operating costs were calculated can be found elsewhere [14]. Site characteristics Outcome measures Data reported included: size of the cancer center, smok- ing prevalence for patients at the center, presence of a point-of-care tobacco cessation intervention (i.e., in- person, telehealth) [24], and whether sites were part of the first or second C3I cohort. Cancer center size was assessed by number of unique adult cancer patients served by the center during the 6-month reporting Reach was assessed as the proportion of unique patients seen during the 6-month reporting period who used tobacco and received at least one type of evidence-based tobacco treatment (e.g., tobacco cessation medications, Quitline referral, point-of-care counseling [24]), relative to the total number of patients who smoked at each C3I center. Effectiveness was assessed as the proportion of Pluta et al. Implementation Science Communications (2023) 4:50 Page 5 of 13 patients currently using tobacco who engaged in tobacco treatment and reported abstinence from tobacco use for a minimum of seven days at six months follow-up. The number of total patients using tobacco was assessed using two items on the C3I 6-month survey: (1) for how many patients who received tobacco treatment in the July–December 2020 reporting period do you have fol- low-up effectiveness data? (2) For how many patients who received tobacco treatment in the July-December 2020 reporting period are 6-month effectiveness data miss- ing? While a small number of programs implemented biochemical verification as part of their assessments, this was not standard across all programs. Therefore, we only used self-reported abstinence for our assessments of pro- gram effectiveness. A complete response approach was used wherein each center determined their own denomi- nator for effectiveness based on their center’s reporting practices. Assessment of program performance We applied DEA to assess the relative performance of C3I centers [20]. We used the DEA optimization method, which has been applied to estimate the technical or cost efficiency of healthcare programs [25–27]. DEA deter- mines how efficiently a program converts inputs into outcomes compared with other programs and produces a best practice frontier comprising the most efficient programs. Efficiency scores We used DEA to estimate efficiency scores for each pro- gram as the ratio of the weighted sum of outcomes to the weighted sum of inputs, and graphically plotted the efficiency scores according to cost and reach/effective- ness. We applied the input-oriented DEA approach with constant returns-to-scale [19, 28]. Under the input ori- entation, the efficiency measure is based on the propor- tion to which the observed input levels can be produced for given outcome levels. Compared to efficiency scores, rankings are robust as they are not based on unstable solutions of linear programming models. We compared efficiency scores across subgroups of sites, by funding cycle, core components, and implementation strate- gies used. The most efficient program(s) are used as the benchmark for comparison with other programs. The efficiency of any program is relative to the efficiency of other programs in the sample, and the relative efficiency of any given program can change when compared to a different set of programs. Slacks amount of slack among inefficient C3I programs, rela- tive to the most efficient program(s), for each input and outcome (i.e., distance between inefficient programs and the most efficient program). We reported the percent- age of change needed to eliminate inefficiencies and to achieve performance consistent with the most efficient program(s) on the best practice frontier. Analyses Three DEA models assessed the relative efficiency of the sixteen C3I programs with complete data. Model 1 input: cost per participant; outcomes: reach, effectiveness. Model 2 inputs: TTS, other personnel; outcomes: reach. Model 3 inputs: TTS, other personnel; outcomes: effectiveness. Analyses were also stratified by C3I center characteris- tics. We conducted all analyses using the PIM-DEA V.3.2 software. Results Table 1 includes descriptive statistics summarizing site characteristics, inputs, and outcomes of the included C3I centers. Cancer centers served an average of 24,652 (standard deviation, SD = 21,596, median = 22,075, range = 507–89,485) patients during the 6-month report- ing period, and median smoking prevalence was 9.3% (range = 2.2–47.1%) across centers; 44% of cancer centers Table 1 Site characteristics, inputs and outputs of the C3I programs (n = 16) Site characteristics Median Mean SD Min Max Cancer center size (number of patients served) Smoking prevalence 22,075 24,652 21,596 507 89,485 9.3% 10.5% 2.2% 47.1% 6.5% N (%) Point of care intervention 7 (44%) Cohort 1 (vs. cohort 2) 11 (69%) Inputs Median Mean SD Min Max Tobacco treatment specialist 0.65 Total FTE Cost per patient Cost per quit Outputs Reach Reach percent Effectiveness 0.66 1.39 0.60 0.74 $572 $518 0.00 0.42 $70 2.00 2.42 $2093 1.31 $454 $2765 $2981 $2015 $330 $9628 Median Mean SD Min Max 108.0 25.0% 33.0 19.9% 254.13 257.03 46 935 24.4% 14.1% 2.5% 47.8% 38.3 44.6 7 197 20.4% 10.6% 2.6% 35.3% Slacks represent excess input utilization or shortages in outcomes within DEA [29]. We assessed the mean Effectiveness percent SD Standard deviation Pluta et al. Implementation Science Communications (2023) 4:50 Page 6 of 13 Fig. 1 Efficiency frontier for C3I programs: reach (%) and effectiveness (%) relative to cost‑per‑participant (n = 16). P1 is on the best practice frontier (i.e., the most efficient unit) had implemented point-of-care interventions and 69% were part of cohort 1 (vs. cohort 2). Mean number of tobacco treatment specialist FTE was 0.66 (SD = 0.60), and mean total FTE was 1.39 (SD = 0.74). Average cost- (SD = $518, median = $474), per-patient was $572 (SD = $2015, and average cost-per-quit was $2981 median = 2765). Overall, programs reached 24.4% of patients who smoked (SD = 14.1, range = 2.5–47.8%) and had a 20.4% effectiveness (SD = 10.6, range = 2.6–35.3%), on average. In the first DEA model (Fig. 1), we assessed reach and effectiveness (as outcomes) relative to cost-per-partic- ipant (as the input). Only one program was located on the best practice frontier (i.e., benchmark program(s) with the most efficient performance), while the majority of programs clustered near the origin (i.e., away from the best practice frontier). This distribution suggests gener- ally low effectiveness and reach relative to costs, in com- parison to the one program on the best practice frontier. Six programs had relatively higher effectiveness (vs. reach), and 10 programs had relatively higher reach (vs. effectiveness). Next, we used DEA to assess effectiveness and reach relative to total costs, stratified by program character- istics (Supplementary Materials). C3I centers in cohort 1 were less clustered around the origin than centers in cohort 2, and the best practice frontier was farther from the origin for cohort 1 (vs. cohort 2) (Supplemen- tary Fig. 1). Only one C3I center was located on the best practice frontier for each cohort. Similarly, C3I centers that did not implement point-of-care interventions were clustered closer to the origin, suggesting that point-of- care was associated with higher reach and effectiveness relative to costs (Supplementary Fig. 2). Only one C3I center was located on the best practice frontier for each model assessing point-of-care. Larger cancer centers (i.e., above median size) were more efficient and had gener- ally greater reach than smaller centers (Supplementary Fig. 3). One C3I center was located on the best practice frontier in each model assessing larger and smaller cent- ers. Centers with below-median smoking prevalence were clustered more closely to the origin, suggesting lower reach and effectiveness relative to total cost than centers with above-median smoking prevalence (Supple- mentary Fig. 4). One C3I center was located on the best practice frontier in the model for greater than median smoking prevalence, and two C3I centers were located on the best practice frontier in the model for lower than median smoking prevalence. Additionally, we used DEA to assess the reach and effectiveness of C3I centers relative to the personnel involved in tobacco treatment administration (i.e., TTS, other personnel). For the first set of models, the inputs were TTS and other personnel, and the output was reach. C3I centers clustered around the origin, suggesting that most had generally low other personnel and TTS rela- tive to reach. Only one C3I center was located on the best practice frontier. Nine C3I centers had higher FTE Pluta et al. Implementation Science Communications (2023) 4:50 Page 7 of 13 Fig. 2 Efficiency frontier for C3I programs: TTS and other personnel relative to reach (n = 16). P9 is on the best practice frontier (i.e., the most efficient unit) for other personnel (vs. TTS) relative to reach, whereas seven C3I centers had higher TTS FTE (vs. other person- nel) relative to reach (Fig. 2). Next, we used DEA to assess the effectiveness of C3I centers relative to personnel involved in tobacco treatment administration (i.e., other personnel, TTS). C3I centers were clustered near the origin, suggesting lower use of TTS and other personnel relative to effectiveness. Two C3I centers were located on the best practice frontier. Seven C3I centers had higher use of other personnel relative to effectiveness, whereas six centers had higher use of TTS relative to effectiveness (Fig. 3). Table 2 shows the average efficiency scores overall, by cohort, by whether the program had a point-of-care intervention, by cancer center size, and by smoking prevalence. Sites in cohort 1, those with point-of-care interventions, those within larger-than-median cancer centers, and those with higher-than-median smoking prevalence had higher efficiency scores, on average. The performance analysis identified the slacks, rep- resenting either excess input utilization or shortages of output production. Table 3 shows the average slack in programs deemed inefficient. These results represent the combined scores of slack for all inefficient programs, for each input and output. Table 3 also shows the percent- age of change in the number of inputs or outputs needed to eliminate the inefficiencies and achieve target levels. Based on our preliminary sample, cost per participant should be reduced by an average of 74.34%, TTS FTE should be reduced by an average 10.98%, and other per- sonnel FTE by 52.18% to maximize efficiency. Discussion This study demonstrated utility of DEA for implemen- tation research by assessing reach and effectiveness of tobacco treatment programs within NCI-designated can- cer centers relative to their operating costs. We identified factors associated with the most optimal programs that could be leveraged to increase efficiency of tobacco treat- ment programs across centers that function less opti- mally. Programs that were in cohort 1 (i.e., typically more advanced in implementation), programs that had imple- mented point-of-care interventions, and programs in cancer centers that were larger in size tended to be more efficient. This information is particularly useful for program evaluation because it directly compares how well C3I programs converted their available resources into meas- urable outcomes (i.e., reach, effectiveness). Variability in efficiency was high across C3I centers, which is unsur- prising given that some centers had existing infrastruc- ture for tobacco treatment programs, whereas others implemented these programs for the first time. Existing program infrastructure may have contributed to pro- gram efficiency, given that these sites would have already implemented some tobacco treatment program-related protocols into their workflows. There was also variabil- ity in how long tobacco treatment programs had been Pluta et al. Implementation Science Communications (2023) 4:50 Page 8 of 13 Fig. 3 Efficiency frontier for C3I programs: TTS and other personnel relative to effectiveness (n = 16). P2 and P5 are on the best practice frontier (i.e., the most efficient units) implemented across centers. Thus, DEA is an impor- tant tool for program evaluation because it can identify which programs effectively maximize their resources given budget constraints. Identifying how resources can be allocated to foster sustainability of C3I centers has implications for other tobacco cessation programs in oncology settings. For instance, DEA may be used to characterize treatment efficiencies in others areas of oncology practice, such as value-based care, enrollment in clinical trials, and improving palliative care and cancer survivorship. Although research using DEA to assess tobacco treat- ment programs for oncology patients is limited, DEA has been applied in other healthcare settings. For example, DEA has been applied in examining the efficiency of pri- mary healthcare centers, including inputs and outputs such as number of patients and staff, costs, procedures, prescriptions, and referrals [30]. Additionally, DEA has been used to assess public health concerns regarding healthcare systems and the optimal organization of pri- mary care service delivery, using inputs such as primary care governance, workforce development, and economic conditions, and outputs such as comprehensiveness, access, coordination and service delivery indicators of access continuity and comprehensiveness of care [30]. Moreover, application of DEA is not limited to assess- ing efficiency of programs and systems, and it has been used to support decision-making in clinical settings. For example, DEA was used for real-time benchmarking in radiotherapy treatment planning, where it was associated with improvement of most treatment plans [31]. Thus, DEA has a broad range of applications within healthcare, including within the oncology domain. DEA can be used as a stand-alone analysis, given its unique ability to assess the relative efficiency of produc- tion units. DEA has been widely applied in other fields such as economics [19, 20] as well as in clinical settings as described above. DEA can also be used in combina- tion with other methods, (e.g., qualitative interviews, longitudinal surveys), to glean a more holistic perspec- tive regarding how to improve program efficiency. For example, conducting qualitative interviews or surveys with personnel directly involved with the implementa- tion procedures could elucidate specific recommenda- tions regarding how to improve efficiency, beyond which inputs and outputs are affecting efficiency [16]. As such, DEA can be used either independently to assess program efficiency or complimentarily with other analyses. Overall, we found that many C3I sites had low effi- ciency relative to the best practice frontier. In the DEA model assessing reach and effectiveness relative to pro- gram costs, only one program was located on the best practice frontier, and this program appeared to achieve substantially higher efficiency compared to other pro- grams. We examined additional DEA models strati- fied by key organizational characteristics. C3I centers that were part of cohort 1 (vs. cohort 2), had deployed point-of-care tobacco cessation interventions (vs. no Pluta et al. Implementation Science Communications (2023) 4:50 Page 9 of 13 Table 2 Technical efficiency scores and returns to the scale of tobacco treatment programs in C3I CRS technical efficiency VRS technical efficiency Scale efficiency CRS, N (%) DRS, N (%) All programs (n = 16) Mean SD Min Cohort 1 (n = 11) Mean SD Min Cohort 2 (n = 5) Mean SD Min No point‑of‑care (n = 7) Mean SD Min Point‑of‑care (n = 9) Mean SD Min Center size: < 22,075 (n = 8) Mean SD Min Center size: ≥ 22,075 (n = 8) Mean SD 25.66 24.56 2.91 29.15 28.65 2.91 17.99 10.16 7.94 15.59 14.31 2.91 33.50 28.63 7.94 17.70 15.00 2.91 33.63 30.38 Min 7.94 Smoking prevalence: < 6.5% (n = 8) Mean 21.67 SD Min Smoking prevalence: ≥ 6.5% (n = 8) Mean SD Min 17.21 2.91 29.65 30.99 5.69 76.82 22.98 39.2 80.46 23.21 46.75 68.82 22.76 39.20 62.07 23.07 39.20 88.30 15.89 57.79 74.88 22.70 46.75 78.76 24.66 39.20 72.83 23.00 39.20 80.81 23.80 46.75 31.40 23.73 5.88 32.54 26.92 5.88 28.88 16.99 10.26 24.01 15.56 5.88 37.15 28.10 10.26 22.38 16.41 5.88 40.42 27.41 10.26 29.03 18.69 5.88 33.77 29.06 10.69 7 (44) 9 (56) 5 (45) 6 (55) 2 (40) 3 (60) 4 (57) 3 (43) 3 (33) 6 (67) 4 (50) 4 (50) 3 (38) 5 (62) 3 (38) 5 (62) 4 (50) 4 (50) CRS Constant returns to scale, DRS Decreasing returns to scale, SD Standard deviation, VRS Variable returns to scale point-of-care), were larger (vs. smaller) in size, and had higher (vs. lower) smoking prevalence tended to be more efficient (i.e., greater reach and effectiveness relative to cost). Thus, centers may prioritize imple- menting point-of-care interventions over other types of tobacco treatment interventions to maximize effi- ciency. Although point-of-care interventions may be expensive, it is noteworthy that their implementation was associated with greater program efficiency (i.e., ratio of reach and effectiveness relative to costs). Simi- larly, larger NCI-designated cancer centers generally achieved higher reach and effectiveness while miti- gating costs, suggesting that C3I programs are more sustainable when implemented in larger (vs. smaller) cancer centers, and that smaller cancer centers or com- munity oncology practices may require more resources to sustainably implement tobacco treatment programs. Pluta et al. Implementation Science Communications (2023) 4:50 Page 10 of 13 Table 3 Slacks evaluation for inefficient programs (n = 16) Mean (SD) Percentage of change 0.11 (0.30) 0.15 (0.39) 597.08 (660.76) 30.14 (20.07) 5.31 (3.90) – 10.98 – 52.18 – 74.34 267.67 38.22 Input slacks Tobacco treatment specialist Other personnel Cost per participant Output slacks Reach (percent) Effectiveness (percent) SD standard deviation Additionally, centers that had higher smoking preva- lence were generally higher in efficiency. It is possible that centers with lower smoking prevalence were less efficient because there were fewer eligible patients to enroll in the program. C3I centers with lower smok- ing prevalence had lower reach relative to effectiveness, which suggests that these programs may be underuti- lized and consequently operate less efficiently than pro- grams with higher enrollment rates. This information can be particularly useful for establishing tobacco ces- sation programs for individuals with cancer. Selecting locations that are more likely to maintain low costs rel- ative to reach and effectiveness may increase the like- lihood that cancer centers will sustain these programs long-term. However, this practice may come at the cost of marginalizing patients in settings with limited resources, which may require greater costs to imple- ment and sustain tobacco treatment programs. Moreover, whether C3I centers had existing infra- structure for tobacco cessation treatments prior to the initiative may have impacted relative program effi- ciency. For instance, centers that had independently focused on promoting tobacco cessation programs before joining C3I may be more efficient than centers that initiated tobacco cessation programs as part of C3I. Implementation readiness has been shown to be associ- ated with higher chances that a cancer center provides tobacco cessation treatments to its patients [32]. Thus, the stage of program implementation is another impor- tant factor that may contribute to a center’s efficiency and sustainability. Further, programs that employed less TTS FTE and other personnel FTE on average achieved greater effi- ciency. Previous research regarding the effects of a TTS for tobacco cessation are mixed. Recent studies found that C3I centers with lower TTS-to-patient ratios tended to have higher reach and lower effectiveness [16] and that counseling delivered by TTSs was asso- ciated with higher smoking cessation rates [33]. Future research should investigate whether reducing TTS and personnel is associated with greater efficiency among C3I centers in general, or whether centers with particu- lar characteristics may benefit from an increase, in TTS and other personnel (e.g., large centers and/or those with particularly high smoking prevalence and less program staff ). More research is needed to identify the most efficient TTS staff-to-patient ratio and explore the contexts in which TTS and other personnel are essen- tial for maximizing program efficiency. It is important to note that it can be challenging to find a balance between maximizing research and delivering an effective intervention, particularly in the oncology setting. Individuals who continue smoking after their cancer diagnosis can be especially difficult to treat even with a high intensity intervention. Therefore, assessing the external validity and cumulative impact of smok- ing cessation interventions in oncology settings is of utmost importance. We assessed reach and effective- ness as separate outcomes, however, both must be con- sidered to assess population impact. The cumulative impact of an intervention is a function of every step of dissemination and participation (e.g., proportion of staff that take part, patients that accept participation, patients that benefit from the intervention and continue benefit- ting 6 months later) [34]. Even interventions that have high effectiveness may yield low population impact after accounting for participation and retention issues at every level of dissemination. As such, consistent and transpar- ent reporting about participation and representative- ness at all levels of dissemination are vital for evaluating the cumulative impact of interventions. Future research should evaluate the cumulative impact of C3I programs on tobacco cessation outcomes across various contexts. Limitations This study is not without limitations. First, our conveni- ence sample of 16 NCI-designated C3I centers may not be representative of many cancer care programs. This sample consisted of cancer centers that received sup- plemental funding to improve or expand tobacco cessa- tion resources; thus, results may not be generalizable to other cancer centers and should be interpreted with cau- tion. The programs in this sample reported low propor- tions of individuals who were American Indian or Alaska Native (≤ 1%), Asian, Native, or Pacific Islander (≤ 1%), or Hispanic (3%); therefore, generalization of findings to these populations may be limited. However, smoking prevalence among C3I centers was similar to estimates of national rates of tobacco use among individuals who have had cancer [35–37]. It is possible that the results do not accurately repre- sent the experiences with efficiency across all tobacco Pluta et al. Implementation Science Communications (2023) 4:50 Page 11 of 13 treatment programs in C3I. Therefore, a larger sample of C3I centers is needed before factors affecting pro- gram efficiency can be reliably assessed and interpreted. Despite this limitation, we achieved the primary goal of this study, which was to demonstrate the benefits of using DEA as a tool for assessing program implementation and performance. Specifically, we showed that DEA can be used to inform program efficiency by assessing read- ily available practice parameters, such as program reach, effectiveness, and cost. Another limitation is the reporting of outcomes and program features was voluntary; therefore, data collected from C3I centers may be partially incomplete. Missing data, whether deliberate or coincidental, can skew find- ings [38]; therefore, more automated data reporting of tobacco treatment program measures would improve future data quality. There was also a lack of uniformity regarding which program personnel reported data to C3I, which may have resulted in between-reporter inconsist- encies. Each center also determined their own denomi- nator for effectiveness based on their center’s reporting practices, which exacerbates variability in reporting across centers. Finally, some of the data for this study were collected during the COVID-19 pandemic, when many non-emer- gent appointments were postponed or canceled. Other pandemic-related changes, such as limited staff due to illness, staff changes, and program changes with imple- mentation of telehealth, may have impacted program costs and efficiency. Reach and effectiveness may have been affected during this time due to pandemic-related restrictions and barriers. On the other hand, reach may have increased with the pandemic related transforma- tion to telehealth treatment models. Data were reported at the level of the C3I center; therefore, we did not have access to individual-level data. Consequently, we were unable to investigate more granular factors that may be associated with program efficiency, such as which spe- cific tobacco treatments patients were receiving, at what frequency, and whether efficiency was moderated by patient characteristics (e.g., age, cancer site, treatment) [35]. Given these restrictions, we were also unable to undertake a thorough analysis of potential confounding factors that may account for observed relationships, such as differences in the types of patients who received differ- ent treatments, and effects of other program features that were not measured or reported. Conclusion DEA is a useful tool for assessing the relative efficiency of organizations that implement evidence-based pro- grams in a way that is not possible with other ana- lytic methods. In the case of C3I, identifying factors associated with high reach and effectiveness, while maintaining low operating costs is important for the sustainability of tobacco treatment programs. Deci- sion makers can use findings from DEA to improve efficiency of existing tobacco treatment programs within oncology settings and evaluate how cancer cent- ers could most effectively support implementation of tobacco treatment programs. This study demonstrated that DEA provides valuable information that can foster more sustainable implementation of tobacco treatment programs in oncology settings. Abbreviations C3I NCI TTS FTE Cancer Center Cessation Initiative National Cancer Institute Tobacco treatment specialist Full time equivalent Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s43058‑ 023‑ 00433‑3. Additional file 1: Supplementary Figure 1. Efficiency frontier for C3I programs in Cohort 1 (left) and Cohort 2 (right): reach and effectiveness relative to cost‑per‑participant. Supplementary Figure 2. Efficiency frontier for C3I programs with (left) and without (right) a point‑of‑care intervention: reach and effectiveness relative to cost‑per‑participant. Supplementary Figure 3. Efficiency frontier for C3I programs at centers larger (left) and smaller (right) than median size: reach and effectiveness relative to cost‑per‑participant. Supplementary Figure 4. Efficiency frontier for C3I programs at centers with higher (left) and lower (right) than median smoking prevalence: reach and effectiveness relative to cost‑per‑participant. Acknowledgements We would like to thank the tobacco treatment program staff at participating cancer centers for their support in data collection. Authors’ contributions RGS and GWW conceptualized the study. SH, MM, HD, and BR contributed to the data collection and coordination. RGS analyzed the data. KP and RGS wrote the first draft. YA, L‑SC, KMC, ATD, AOG, BH, DHB, ACK, CYL, KL, AHL, JJP, KS, KT, HT, ET, and JSW contributed to the acquisition of data and manuscript revision. JSO, DS, ND, LF, RH, CP, VS, KPW, TBK, and MCF contributed to succes‑ sive drafts. All authors approved the final manuscript. Funding A contract from the 17GZSK0031) supported the implementation of tobacco treatment programs and data reporting at the NCI‑designated cancer centers and coordination efforts at the University of Wisconsin—Madison. Ramzi G. Salloum was supported by NCTATS award UL1TR001427. Jamie S. Ostroff was supported by NCI award P30CA008748‑52S1. Li‑Shiun Chen was supported by NCI awards P30CA091842‑19S5, P50CA24443 and the Siteman Investment Program. Cary A. Presant was supported by NCI grants P30CA033572 and P30CA03572‑37S5. Neely Dahl and Kara Wiseman were supported by NCI award P30CA044579. Kara Wiseman was supported by the UVA iTHRIV Schol‑ ars Program, NCATS UL1T R003015 and KL2TR003016. Katie Lenhoff was sup‑ ported by NCI award P30CA023108‑43. Kimberly Shoenbill was supported by NCI awards P30CA016086‑43S1 and P30CA016086‑44S5. Hilary A. Tindle was supported by NCI award 3P30CA068485‑24S3. Graham Warren was supported by NCI awards P30CA138313‑09S4 and A22‑0010–01. Judith J. Prochaska was supported by NCI awards P30CA124435‑11S1 and P30CA124435‑13S2. Timothy Baker was supported by NCI award P01 CA180945. Justin S. White was supported by award P30CA082103‑19S2. Pluta et al. Implementation Science Communications (2023) 4:50 Page 12 of 13 Availability of data and materials The datasets generated and analyzed during the study are not publicly avail‑ able due to the sensitive nature of some data. The cost dataset is available from the corresponding author upon reasonable request. All other data are available from the C3I coordinating center on reasonable request. Declarations Ethics approval and consent to participate Data reported in this manuscript were collected for the purpose of program evaluation and quality improvement. The need for informed consent of participants was waived because programs participating in the evaluation provided deidentified, aggregated data. The evaluation was categorized as program evaluation and deemed exempt by the Institutional Review Board of the University of Wisconsin‑Madison. Consent for publication We consent to publication. Competing interests Dr. Baker occupies the Glaxo Wellcome Chair in the Department of Medicine at the University of Wisconsin School of Medicine and Public Health. Dr. Prochaska has provided consultation to pharmaceutical and technology com‑ panies that make medications and other treatments for quitting smoking and also has served as an expert witness in lawsuits against tobacco companies. The other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Author details 1 Department of Health Outcomes and Biomedical Informatics, University of Florida College of Medicine, 2004 Mowry Rd, Gainesville, FL 32610, USA. 2 University of Wisconsin Carbone Cancer Center, 600 Highland Ave, Madison, WI 53705, USA. 3 School of Medicine and Public Health, University of Wiscon‑ sin, 750 Highland Ave, Madison, WI 53705, USA. 4 National Cancer Institute, 9609 Medical Center Dr, Rockville, MD 20850, USA. 5 Memorial Sloan Kettering Cancer Center, 1275 York Ave, New York, NY 10065, USA. 6 New York University School of Global Public Health, 708 Broadway, New York, NY 10003, USA. 7 Rush University Medical Center and Rush Cancer Center, 1725 W Harrison St, Suite 1010, Chicago, IL 60612, USA. 8 Washington University Siteman Cancer Center, 4921 Parkview Pl, St. Louis, MO 63110, USA. 9 Medical University of South Carolina, 171 Ashley Ave, Charleston, SC 29425, USA. 10 University of Virginia Comprehensive Cancer Center, 1240 Lee St, Charlottesville, VA 22903, USA. 11 University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dal‑ las, TX 75390, USA. 12 Fox Chase Cancer Center, 333 Cottman Ave, Philadelphia, PA 19111, USA. 13 University of North Carolina Lineberger Cancer Center, 450 West Dr, Chapel Hill, NC 27599, USA. 14 Virginia Commonwealth University Department of Psychiatry, 501 N 2Nd St, Suite 400B, Richmond, VA 23219, USA. 15 Northwestern University Feinberg School of Medicine and Lurie Compre‑ hensive Cancer Center of Northwestern University, 420 E Superior St, Chicago, IL 60611, USA. 16 Indiana University Simon Comprehensive Cancer Center, 535 Barnhill Dr, Indianapolis, IN, USA. 17 University of Chicago Medicine Compre‑ hensive Cancer Center, 5758 S Maryland Dr, Chicago, IL 60637, USA. 18 Hunts‑ man Cancer Institute, University of Utah, 1950 Circle of Hope Dr, Salt Lake City, UT 84112, USA. 19 One Medical Center Drive, Dartmouth‑Hitchcock Norris Cotton Cancer Center, Lebanon, NH 03756, USA. 20 University of Colorado Comprehensive Cancer Center, 1665 North Aurora Court, Aurora 200480045, USA. 21 City of Hope Comprehensive Cancer Center and Beckman Research Institute, 1500 E Duarte Rd, Duarte, CA 91010, USA. 22 Stanford Cancer Institute, Stanford University, 265 Campus Dr, Ste G2103, Stanford, CA 94305, USA. 23 H. Lee Moffitt Cancer Center, 3011 Holly Dr, Tampa, FL 33612, USA. 24 George‑ town University Lombardi Comprehensive Cancer Center, 3800 Reservoir Rd NW, Washington, DC 20007, USA. 25 Vanderbilt University Medical Center Vanderbilt‑Ingram Cancer Center, 2220 Pierce Ave, Nashville, TN 37232, USA. 26 University of California Davis Comprehensive Cancer Center, 2279 45Th St, Sacramento, CA 95817, USA. 27 Philip R. Lee Institute for Health Policy Studies, University of California San Francisco, 490 Illinois St, Floor 7, San Francisco, CA 94158, USA. 28 University of Wisconsin Institute for Clinical and Translational Research, 750 Highland Ave, Madison, WI 53705, USA. Received: 8 December 2022 Accepted: 2 May 2023 References 1. U.S. Department of Health and Human Services. Smoking Cessation: A Report of the Surgeon General. Atlanta; 2020. 2. Warren GW, Alberg AJ, Kraft AS, Cummings KM. The 2014 surgeon gen‑ eral’s report: “The health consequences of smoking‑50 years of progress”: a paradigm shift in cancer care. Cancer. 2014;120(13):1914–6. 3. Warren GW, Cartmell KB, Garrett‑Mayer E, Salloum RG, Cummings KM. Attributable failure of first‑line cancer treatment and incremental costs associated with smoking by patients with cancer. JAMA Netw Open. 2019;2(4):e191703. 4. Gritz ER, Toll BA, Warren GW. Tobacco use in the oncology setting: 5. advancing clinical practice and research. Cancer Epidemiol Biomark Prev. 2014;23(1):3–9. Land SR, Toll BA, Moinpour CM, Mitchell SA, Ostroff JS, Hatsukami DK, et al. Research priorities, measures, and recommendations for assess‑ ment of tobacco use in clinical cancer research. Clin Cancer Res. 2016;22(8):1907–13. 6. Baker TB, Berg KM, Adsit RT, Skora AD, Swedlund MP, Zehner ME, et al. Closed‑loop electronic referral from primary care clinics to a state tobacco cessation quitline: effects using real‑world implementation train‑ ing. Am J Prev Med. 2021;60(3):S113–22. Fiore MC, D’Angelo H, Baker T. Effective cessation treatment for patients with cancer who smoke—the fourth pillar of cancer care. JAMA Netw Open. 2019;2(9):e1912264. 7. 8. Croyle RT, Morgan GD, Fiore MC. Addressing a Core Gap in Cancer Care — The NCI Moonshot Program to help oncology patients stop smoking. N Engl J Med. 2019;380(6):512–5. 9. National Cancer Institute. Cancer Center Cessation Initiative. 2022. 10. D’Angelo H, Rolland B, Adsit R, Baker TB, Rosenblum M, Pauk D, et al. Tobacco treatment program implementation at NCI cancer centers: progress of the NCI Cancer Moonshot‑funded cancer center cessation initiative. Cancer Prev Res. 2019;12(11):735–40. 11. D’Angelo H, Ramsey AT, Rolland B, Chen LS, Bernstein SL, Fucito LM, et al. Pragmatic application of the RE‑AIM framework to evaluate the imple‑ mentation of tobacco cessation programs within NCI‑designated cancer centers. Front Public Health. 2020:221. 12. Glasgow RE, Vogt TM, Boles SM. Evaluating the public health impact of health promotion interventions: the RE‑AIM framework. Am J Public Health. 1999;89(9):1322–7. 13. Gaglio B, Shoup JA, Glasgow RE. The RE‑AIM framework: a systematic review of use over time. Am J Public Health. 2013;103(6):e38‑46. 14. Salloum RG, LeLaurin JH, Dallery J, Childs K, Huo J, Shenkman EA, et al. Cost evaluation of tobacco control interventions in clinical settings: a systematic review. Prev Med (Baltim). 2021;146:106469. 15. The Cancer Center Cessation Initiative Sustainability Working Group. Sus‑ tainability of tobacco treatment programs in the cancer center cessation initiative. J Natl Comprehens Cancer Netw. 2021;19(Suppl_1):S16‑20. 16. Hohl SD, Matulewicz RS, Salloum RG, Ostroff JS, Baker TB, Schnoll R, et al. Integrating Tobacco Treatment Into Oncology Care: Reach and Effective‑ ness of Evidence‑Based Tobacco Treatment Across National Cancer Institute–Designated Cancer Centers. J Clin Oncol. 2022:JCO‑22. 17. Malone S, Prewitt K, Hackett R, Lin JC, McKay V, Walsh‑Bailey C, et al. The Clinical Sustainability Assessment Tool: measuring organizational capac‑ ity to promote sustainability in healthcare. Implement Sci Commun. 2021;2(1):77. 18. Salloum RG, D’Angelo H, Theis RP, Rolland B, Hohl S, Pauk D, et al. Mixed‑ methods economic evaluation of the implementation of tobacco treat‑ ment programs in National Cancer Institute‑designated cancer centers. Implement Sci Commun. 2021;2(1):41. 19. Charnes A, Cooper WW, Rhodes E. Measuring the efficiency of decision making units. Eur J Oper Res. 1978;2(6):429–44. 20. Cooper WW, Seiford LM, Tone K. Data envelopment analysis. In: Cooper WW, Seiford LM, Zhu J, editors. Handbook on data envelopment analysis. 1st ed. Springer; 2000. p. 1–40. 21. Sickles RC, Zelenyuk V. Measurement of productivity and efficiency. Cambridge University Press; 2019. Pluta et al. Implementation Science Communications (2023) 4:50 Page 13 of 13 22. Goldstein AO, Ripley‑Moffitt CE, Pathman DE, Patsakham KM. Tobacco use treatment at the U.S. National Cancer Institute’s designated cancer cent‑ ers. Nicot Tobacco Res. 2013;15(1):52–8. 23. Craig EJ, Ramsey AT, Baker TB, James AS, Luke DA, Malone S, et al. Point of care tobacco treatment sustains during COVID‑19, a global pandemic. Cancer Epidemiol. 2022;78:102005. 24. Ramsey AT, Baker TB, Stoneking F, Smock N, Chen J, Pham G, et al. Increased reach and effectiveness with a low‑burden point‑of‑care tobacco treatment program in cancer clinics. J Natl Compr Canc Netw. 2022;20(5):488‑495.e4. 25. Wang L, Grignon M, Perry S, Chen X, Ytsma A, Allin S, et al. The determi‑ nants of the technical efficiency of acute inpatient care in Canada. Health Serv Res. 2018;53(6):4829–47. 26. Varabyova Y, Schreyögg J. International comparisons of the technical efficiency of the hospital sector: panel data analysis of OECD countries using parametric and non‑parametric approaches. Health Policy (New York). 2013;112(1–2):70–9. 27. Hollingsworth B. The measurement of efficiency and productivity of health care delivery. Health Econ. 2008;17(10):1107–28. 28. Banker RD, Thrall RM. Estimation of returns to scale using data envelop‑ ment analysis. Eur J Oper Res. 1992;62(1):74–84. 29. Alatawi AD, Niessen LW, Khan JAM. Efficiency evaluation of public hos‑ pitals in Saudi Arabia: an application of data envelopment analysis. BMJ Open. 2020;10(1):e031924. 30. Zakowska I, Godycki‑Cwirko M. Data envelopment analysis applications in primary health care: a systematic review. Fam Pract. 2020;37(2):147–53. 31. Raith A, Ehrgott M, Fauzi F, Lin KM, Macann A, Rouse P, et al. Integrating Data Envelopment Analysis into radiotherapy treatment planning for head and neck cancer patients. Eur J Oper Res. 2022;296(1):289–303. 32. Ostroff JS, Copeland A, Borderud SP, Li Y, Shelley DR, Henschke CI. Readi‑ ness of lung cancer screening sites to deliver smoking cessation treat‑ ment: current practices, organizational priority, and perceived barriers. Nicotine Tob Res. 2016;18(5):1067–75. 33. Lancaster T, Stead LF. Individual behavioural counselling for smoking ces‑ sation. Cochrane Database Syst Rev. 2017;3:CD001292. 34. Glasgow RE, Huebschmann AG, Brownson RC. Expanding the CONSORT figure: increasing transparency in reporting on external validity. Am J Prev Med. 2018;55(3):422–30. 35. Salloum RG, Huo J, Lee JH, Lee J, Dallery J, George T, Warren G. Tobacco and E‑cigarette use among cancer survivors in the United States. PloS One. 2019;14(12):e0226110. 36. Gritz ER, Talluri R, Domgue JF, Tami‑Maury I, Shete S. Smoking behaviors in survivors of smoking‑related and non–smoking‑related cancers. JAMA Netw Open. 2020;3(7):e209072. 37. Westmaas JL, Alcaraz KI, Berg CJ, Stein KD. Prevalence and correlates of smoking and cessation‑related behavior among survivors of ten cancers: findings from a nationwide survey nine years after diagnosis. Cancer Epidemiol Biomarkers Prev. 2014;23(9):1783–92. 38. Roth PL. Missing data: A conceptual review for applied psychologists. Pers Psychol. 1994;47(3):537–60. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub‑ lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12909-023-04193-5	Deng et al. BMC Medical Education (2023) 23:203 https://doi.org/10.1186/s12909-023-04193-5 BMC Medical Education RESEARCH Open Access The impact of COVID-19 on online medical education: a knowledge graph analysis based on co-term analysis Huijiao Deng1†, Yi Jiang1†, Qinrong Han1†, Xingyu Zhou1†, Siyang Zhong1, Kai Hu1* and Lin Yang2* Abstract Background This study aims to identify the characteristics and future directions of online medical education in the context of the novel coronavirus outbreak new through visual analytics using CiteSpace and VOSviewer bibliometric methods. Method From Web of Science, we searched for articles published between 2020 and 2022 using the terms online education, medical education and COVID-19, ended up with 2555 eligible papers, and the articles published between 2010 and 2019 using the terms online education, medical education and COVID-19, and we ended up with 4313 eligible papers. Results Before the COVID-19 outbreak, Medical students and care were the most frequent keywords and the most cited author was BRENT THOMA with 18 times. The United States is the country with the greatest involvement and research impact in the field of online medical education. The most cited journal is ACAD MED with 1326 citations. After the COVID-19 outbreak, a surge in the number of research results in related fields, and ANXIETY and four second- ary keywords were identified. In addition, the concentration of authors of these publications in the USA and China is a strong indication that local epidemics and communication technologies have influenced the development of online medical education research. Regarding the centrality of research institutions, the most influential co-author network is Harvard Medical School in the United States; and regarding the centrality of references, the most representative journal to which it belongs is VACCINE. Conclusion This study found that hey information such as keywords, major institutions and authors, and countries differ in the papers before and after the COVID-19 outbreak. The novel coronavirus outbreak had a significant impact on the online education aspect. For non-medical and medical students, the pandemic has led to home isolation, making it difficult to offer face-to-face classes such as laboratory operations. Students have lost urgency and control over the specifics of face-to-face instruction, which has reduced the quality of teaching. Therefore, we should improve †Huijiao Deng, Yi Jiang, Qinrong Han and Xingyu Zhou contributed equally to this work. †Huijiao Deng, Yi Jiang, Qinrong Han and Xingyu Zhou are co-first authors. Correspondence: Kai Hu 196875734@qq.com Lin Yang aiyzwll@aliyun.com Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Deng et al. BMC Medical Education (2023) 23:203 Page 2 of 17 our education model according to the actual situation to ensure the quality of teaching while taking into account the physical and psychological health of students. Keywords Online education, COVID-19, Medical education, Co-word analyse, Knowledge spectrum, Visual analysis Background Two thousand nineteen Coronavirus disease is a disease caused by a novel coronavirus called severe acute respira- tory syndrome coronavirus [1].On December 31, 2019, WHO first learned of a cluster of pneumonia cases of the novel coronavirus reported in the city of Wuhan, People’s Republic of China. This major outbreak has had a pro- found impact on all walks of life [2]. In particular, in the education sector, the decision to close, partially close or reopen schools should be guided by a risk management mindset, more so to maximize the education, well-being and health benefits for students, teachers, staff and the broader community, and to help prevent another out- break of COVID-19 in the community. The decision to resume classes should be based on a careful assessment of the situation and consultation with various stakehold- ers, including health and education decision makers, teachers and other school staff, parents, and medical and community workers. Thus, changing the current class- room approach becomes a new and viable measure. The World Health Organization states, "If children are unable to attend school, they should be supported to ensure that students have continued access to educational materials and technologies" [3]. Online education, the Internet-based model of educa- tion in which teachers’ curricula and materials are shared remotely with students through the Internet, enables the separation of teaching and learning, facilitates the flow of high-quality teaching resources, and further improves educational equity in society as a whole [4]. Since 2000, online education has become an important research direction in the field of education as an important supple- ment to the traditional education model. As the Internet industry continues to upgrade, the software and hard- ware equipment required for online teaching continues to change, making the effect of online education medical education’s networking process has been ahead of other disciplines [5]. Online education in medicine has played an extremely important role for the majority of under- graduate and postgraduate teaching work and even for the continuing education of clinicians [6]. Since the novel coronavirus outbreak, social factors have contributed to the flourishing of online education, and the hotspots and directions of research have changed [7, 8]. To elucidate the impact COVID-19 has had on medi- cal online education and to provide a systematic and objective overview of the development of medical online education, this study used Citespace, VOSviewer to separately identify, through scientometric methods, the bibliographic data published in the 10 years prior to the COVID-19 outbreak (2010–2019) and between the COVID-19 outbreak and the present day in Web of Sci- ence (WoS) journals published in the field and visualized their relationships. The analysis was based on CiteSpace. Specifically, the study was guided by four key objectives: (1) to understand the characteristics of the keywords of research topics before and after the outbreak; (2) to iden- tify the most prominent researchers in the field before and after the outbreak and the associations between them; (3) to describe the main research institutions and the evolution of the links between geographic regions in the field before and after the outbreak; and (4) the jour- nals with the highest number of citations and publica- tions in the field before and after the outbreak. Each part of this paper is organized as follows. The "Materials and Methods" section describes the data collection and CiteS- pace and VOSviewer, and the "Results" section provides a comprehensive analysis of the research results and their categories: research hotspots, collaborative networks, co- citation networks, etc. The "Discussion" section summa- rizes the history of the field of online medical education and suggests future research directions, demonstrating the broad applicability of this work. Finally, "Conclu- sion" section summarizes each section of the article and explains the focus of the research. Method Research overview Research tools In this study, the literature visualization and analysis tools used were CiteSpace 5.8.R3 and VOS viewer 1.6.18. CiteSpace is a Java language-based visual literature analysis software developed by Professor Chaomei Chen of Drexel University [9]. It is able to analyze the research frontiers and research hotspots of a topic in the con- text of scientific metrology and data visualization [10]. Through its powerful visualization processing capability, CiteSpace has many functions such as posting volume statistics, author and institution collaboration mapping, time zone map, keyword co-occurrence map, and key- word clustering map. In CiteSpace, it counts the num- ber of occurrences by extracting words or terms that co-occur in the same or several documents. The more co-occurrence, the closer the relationship between two Deng et al. BMC Medical Education (2023) 23:203 Page 3 of 17 subject terms or other noun terms in terms of subject content, and thus forms a co-occurrence relationship network [11]. VOS viewer is a JAVA-based software tool for building and visualizing bibliometric networks, developed in 2009 by van Eck and Waltman of The Centre for Science and Technology Studies at CWTS [12]. Its basic function and implementation principle are similar to that of citation space, but VOS viewer has better performance in specific common network mapping through co-occurrence clus- tering with its unique text mining techniques. In CiteSpace, time chosen to build the dataset is selected as the time window, and the time slice is 1 year, and the text processing options of Title, Abstract, Author Keywords, and Keywords Plus are selected, and Author, Institution, and Keyword are selected as the nodes respectively The text processing options of Title, Abstract, Author Keywords, and Keywords Plus are selected, and Author, Institution, and Keyword are selected as nodes, respectively. In the VOS viewer, you can create clusters and density diagrams by creating a bibliographic dataset based map to read the data in bibliographic database file, then filter- ing in the data from the scientific network and selecting co-existing, co-authorship, and other small cells in the entries to create clusters and density map that are more intuitive and clear. Data sources and processing The literature included in this paper was obtained from the Web of Science, written in English, and the Web of Science Core Collection was used as the database. In this paper, in order to better reflect the research results at different time points, the datasets were created based on the pre-COVID-19 and post-COVID-19 outbreaks, respectively. For Early Access papers published in Web of Science, we manually categorized them according to their topics, and the rest of the papers were categorized according to their publication dates. For the literature before the COVID-19 outbreak, we chose the publication period from January 1, 2010 to December 31, 2019, 2022; the search formula was TS = online education AND medical. 4313 documents were obtained in total. For the literature after the COVID-19 outbreak, we chose the publication dates of January 1, 2020 to July 16, 2022; the search formula was TS = online education AND medical AND COVID-19. 2555 papers were obtained in total. The obtained documents were exported to Plain text file Full record with cited references format for analysis. Results Time distribution of the literature The number of papers and the trend of changes can measure the level and dynamics of scientific research results in the field, which is very important for predicting future development trends. As shown in the Fig. 1, within the field of medical education, the number of online education-related papers published surged between 2019–2020, and since the literature was retrieved in October 2022 and some research papers in 2022 have not yet been published, we can reasonably speculate that the number of publications in 2022 should be significantly higher than that in 2021. Although this is influenced to some extent by information technology advances such as 5G, the role of COVID-19 in this should not be over- looked. We believe that the COVID-19 pandemic has led to a huge impact on offline teaching due to blockades of varying degrees around the world, and the huge practical Fig.1 Chronological distribution of medical online education literature before and after the outbreak Deng et al. BMC Medical Education (2023) 23:203 Page 4 of 17 demand for technological advances in online educa- tion cannot be ignored. It is foreseeable that research on online medical education will continue to increase in the future. Visualization of important keyword clusters analysis An analysis of keywords in the literature for the decade prior to the epidemic (January 2010 to December 2019) and the post-epidemic period (January 2020 to June 2022) led to the following conclusions. Pre‑epidemic keyword cluster analysis Scholars around the world before and after the epidemic maintained a certain level of attention to online teaching of clinical research, but after the epidemic, with the rapid development of Internet technology and the trend of the current social situation, online teaching that can com- municate remotely has received extensive attention from schools [13, 14]. A total of 10,942 keywords from the decade before the epidemic constructed a keyword network for clinical research teaching research in the decade before the epi- demic. Based on the keywords in the 21 main thematic clusters, keywords with a frequency of ≥ 3 were filtered and keywords with "covid-19", "education" and their related synonyms, i.e. the terms used in the literature search for this study, were removed. We obtained 1822 keywords that were the most prominent, dominant and relevant. The visualization of the keywords drawn using VOSviewer is shown in Fig. 2. Based on the "total link strength" indicator, it is easy to see that the following ten keywords are among the key- words with a strong link strength, which are not listed in this paper due to space, but are shown in Table 1. The term "medical student" was the most relevant term in the online teaching field during the new epidemic, with a total link strength of 2420. In addition, the keywords "care", "students" and "impact" all ranked in the top two, three and four in terms of frequency of occurrence with Fig. 2 Visual clustering analysis of popular research keywords before the outbreak Deng et al. BMC Medical Education (2023) 23:203 Page 5 of 17 Table 1 10 keywords based on total link strength before the epidemic Keyword Occurrences Total link strength Medical student Care Students Impact Curriculum Knowledge Attitude Quality Outcomes Health 338 321 292 248 233 213 209 174 162 163 2420 2257 1984 1835 1634 1578 1522 1334 1244 1159 an association intensity of more than 1900, respectively. The phrase "curriculum" is the most frequently used term in the field of teaching and learning. The phrase "curricu- lum" was the fifth most frequently associated term with a strength of 1634.In terms of subject clusters, the subject clusters are ranked according to the strength of relevance and frequency of occurrence of the keywords they con- tain, with the first largest cluster containing 235 items, shown in red. The most frequent phrase in this cluster is "medical students", with 338 occurrences and 2420 links. The second largest cluster is shown in blue in the graph and contains 157 items. The most frequent phrase in this cluster is "care", with 321 occurrences and 2257 links, making it the most frequent phrase in this cluster. The third largest cluster contains 132 items and is shown in pink in the graph. The most frequent phrase in this cluster is "impact", with 248 occurrences and 1835 links. The fourth largest cluster contains 117 items, high- lighted in brown in the graph. Knowledge" is the most frequent phrase in this cluster, with 213 occurrences and 1578 links. The fifth largest cluster, containing 109 items, is high- lighted in light green in the graph. The most frequent phrase in this cluster is "attitudes", with 209 occurrences and 1522 links. Post‑epidemic keyword cluster analysis A total of 6283 keywords were used to construct a key- word network for clinical research and teaching research during the novel coronavirus outbreak. Based on the key- words in the 14 main thematic clusters, keywords with a frequency of ≥ 3 were filtered and keywords with "covid- 19", "education" and their related synonyms, i.e. the terms used in the literature search for this study, were removed. We obtained 1004 of the most prominent, dominant and relevant keywords. The visualization of the keywords drawn using VOSviewer is shown in Fig. 3. Based on the "total link strength" indicator, it is easy to see that the following ten keywords are among the keywords with strong linkage, which are not listed in this paper due to space, but are shown in (Table 2). The term "Anxiety" was the most relevant term in the online teaching field during the new epidemic with a total link strength of 2138. In addition, the terms "Depression", "Impact" and "Mental health" all ranked as the most rele- vant terms with a total association intensity of over 1500. The keywords "Depression", "Impact" and "Mental health" are in the top two, three and four places respectively in terms of frequency of occurrence. The phrase "Stress" is in fifth place with an association strength of 1460. In terms of subject clusters, the subject clusters were ranked according to the strength of relevance and fre- quency of the keywords included. Anxiety" is the most frequent phrase in this cluster, with 252 occurrences and 2138 links. The second largest cluster is shown in red in the graph and contains 282 items. The most frequent phrase in this cluster is "Impact", which has 224 occurrences and 1685 links. The third largest cluster contains 57 items and is shown in brown in the graph. Pandemic" is the most frequent phrase in this cluster, with 216 occurrences and 1441 links. The fourth largest cluster contains 141 items, high- lighted in green in the graph. Health" is the most frequent phrase in this cluster, with 175 occurrences and 1210 links. The fifth largest cluster, containing 93 items, is shown in blue. Knowledge" is the most frequent phrase in this cluster, with 140 occurrences and 861 links. Comparative analysis of keyword clusters before and after the outbreak By analyzing the top ten keywords before and after the epidemic, it is easy to find that "Students", "Impact" and "Health" are the main keywords before and after the epi- demic. However, the ranking of "Knowledge" has dropped and the position of "Impact" has increased. After the epi- demic, keywords such as "Anxiety" and "Stress" gradually emerged to describe psychological stress states, and the focus on students’ health was gradually refined to include mental health. Visualization of important authors analysis In the author coupling analysis before and after the epi- demic using VOSviewer, the threshold was set to ≥ 7. 17 authors were obtained from 16,989 authors screened before the epidemic, and 18 authors were obtained from Deng et al. BMC Medical Education (2023) 23:203 Page 6 of 17 Fig.3 Visual clustering analysis of popular research keywords after the outbreak 15,676 authors screened after the epidemic. The deeper the curve in the figure indicates the stronger the asso- ciation. From Figs. 4 and 5, we can see that although the final screening results are similar, some authors are not even shown due to clustering, indicating that some authors are not strongly associated. In general, the authors posting after the epidemic are more closely connected, and for the same topic medicine and online education, the coupling network graph of that author changes greatly after adding the topic of the epidemic, indicating the impact of the change of the research topic on the authors in the related fields. Then using Citespace to summarize the top 10 authors of the cited articles published, as shown in the figure, before the epidemic, Brent Thoma’s article was the most cited with 18 times; after the epidemic, Chun- gying Lin’s article was the most cited with 11 times, so Deng et al. BMC Medical Education (2023) 23:203 Page 7 of 17 Table 2 10 keywords based on total link strength after the epidemic Keyword Occurrences Total link strength Anxiety Depression Impact Mental health Stress Pandemic Health Knowledge Students China 252 242 224 185 171 216 175 140 130 90 2138 2122 1685 1506 1460 1441 1210 861 859 667 for a quick understanding of the field, the articles of these two are more readable (Tables 3 and 4). In terms of author centrality, pre-epidemic A Bullock, E Barnes, A Kavadella and A Liepa ranked first with a cen- trality of 17, while post-epidemic Aimen Khacharem and Achim Jerg ranked first with a centrality of 54, indicating the strong influence of these authors in the field (Tables 5 and 6). Visualising the country density map In this country density map showing each country’s involvement in online teaching and learning during the novel coronavirus outbreak. Each point in the item den- sity visualisation has a colour that indicates the density of the item at that point. The colours range from blue to green to yellow [15]. The greater the number of items near a point visualized in this density view, the higher the weight of neighboring items, and the closer the color of the point is to yellow. Conversely, the lower the num- ber of items near a point, the lower the weight of neigh- boring items, and the closer the color of the point is to blue. From this, the most important countries in terms of online teaching and learning research engagement during the novel coronavirus outbreak period can be observed concisely and directly by plotting the visualized country densities. Prior to the COVID-19 outbreak, most countries had not conducted extensive research in this area (Fig. 6). Among the major countries conducting research, the United States is the country with the greatest involve- ment and research impact in the field of medical online education, and researchers from the United States have closer ties with Ireland and England, China and Australia, and Canada and Latvia, but most countries are still con- ducting research independently without forming. The research is still conducted independently in most coun- tries without a cooperative system. After the outbreak of COVID-19, more countries have conducted research on online medical education, but the United States is still the most influential country and has developed close ties with China and England, and countries such Fig.4 Visual clustering analysis of author coupling network before the outbreak Deng et al. BMC Medical Education (2023) 23:203 Page 8 of 17 Fig.5 Visual clustering analysis of author coupling network after the outbreak Table 3 Top 10 citations by authors of articles published before the outbreak Table 5 Top 10 posting author centrality before the outbreak Centrality References Citation Counts References 18 17 15 13 8 7 7 7 7 7 BRENT THOMA, 2015, SO, 0, 0 TERESA M CHAN, 2015, SO, 0, 0 NITIN AGA RWA L, 2013, SO, 0, 0 DAVID R HANSBERRY, 2014, SO, 0, 0 MICHELLE LIN, 2015, SO, 0, 0 JOANNA GOTLIB, 2012, SO, 0, 0 B PRICE KERFOOT, 2010, SO, 0, 0 LORAINNE TUDOR CAR, 2019, SO, 0, 0 BERNARD T LEE, 2014, SO, 0, 0 CHRISTINA R VARGAS, 2014, SO, 0, 0 17 17 17 17 15 14 14 14 14 14 A BULLOCK, 2013, SO, 0, 0 E BARNES, 2013, SO, 0, 0 A KAVADELLA, 2013, SO, 0, 0 A LIEPA, 2013, SO, 0, 0 E POVEL, 2013, SO, 0, 0 I AKOTA, 2013, SO, 0, 0 H KERSTEN, 2013, SO, 0, 0 R THOMAS, 2013, SO, 0, 0 S BAILEY, 2013, SO, 0, 0 J COWPE, 2013, SO, 0, 0 Table 4 Top 10 citations by authors of articles published after the outbreak Table 6 Top 10 author centrality after the outbreak Centrality References Citation Counts References 11 8 8 8 7 7 7 7 7 7 CHUNGYING LIN, 2021, SO, 0, 0 FERNANDO BARBOSA, 2020, SO, 0, 0 LEE SMITH, 2020, SO, 0, 0 IHUA CHEN, 2021, SO, 0, 0 OMAR BOUKHRIS, 2020, SO, 0, 0 CHRISTIAN WREDE, 2020, SO, 0, 0 KHALED TRABELSI, 2020, SO, 0, 0 LIWA MASMOUDI, 2020, SO, 0, 0 ACHRAF AMMAR, 2020, SO, 0, 0 ASEEM MEHRA, 2020, SO, 0, 0 54 54 50 50 36 21 21 21 21 21 AIMEN KHACHAREM, 2020, SO, 0, 0 ACHIM JERG, 2020, SO, 0, 0 ACHRAF AMMAR, 2020, SO, 0, 0 ANDREA GAGGIOLI, 2020, SO, 0, 0 ANITA HOEKELMAMN, 2020, SO, 0, 0 KHADIJEH IRANDOUST, 2020, SO, 0, 0 ELLEN BENTLAGE, 2020, SO, 0, 0 KARIM CHAMARI, 2020, SO, 0, 0 HADJ BATATIA, 2020, SO, 0, 0 FAIEZ GARGOURI, 2020, SO, 0, 0 Deng et al. BMC Medical Education (2023) 23:203 Page 9 of 17 Fig. 6 Visual clustering analysis of country density maps before the outbreak as Spain, Germany, Canada, Brazil, and France have also conducted a lot of research in medical online education (Fig. 7). 84, followed by "McMaster University" with 38. It is not difficult to find that the average number of "ducuments" of the top ten comprehensive literatures is about 67. Collaborative network of organizations that visualize co‑authors The visualization of the institutional collaboration net- work (nodes are the names of institutions that connect institutions that have collaborative relationships) pro- vides a clear visual representation of the collaborative relationships and key institutions. (Figs. 8 and 9). The size of the nodes provides a visual representation of the cen- trality of each institution and the density of connections is also directly related to the closeness of the cooperation relationship. Pre‑epidemic organizations cluster analysis Based on the centrality index, it is easy to conclude that "University of Toronto" is in the first place with a cen- trality of 2450. "Harvard University" is the second most centralised institution in the online inter-institutional collaboration network. The No. 3 institution for central- ity is “University of California, San Francisco”. According to the "Ducuments" metric, the highest- ranking institution in the visual agency collaboration network is "University of Toronto" with 115. "Univer- sity of California, San Francisco" ranked second with "Total link strength" refers to the Total co-occur- rence times of keywords and other keywords (includ- ing repeated co-occurrence times). In the organization cooperation network of the visual co-author, it can indicate the close cooperation degree between the organization and different institutions. By analyzing the data, it’s easy to know that "University of Toronto" ranks first with 20,123, "McMaster University" came in second with 17,301, and "Johns Hopkins University" came in third with 11,631. Not surprisingly, the top three are all over 10,000. Based on the centrality index, it is easy to conclude that "National University of Singapore" is in the first place with a centrality of 3324. "Harvard Medical School" is the second most centralised institution in the online inter-institutional collaboration network. The No. 3 institution for centrality is “Huazhong University of Science and Technology”. According to the "Ducuments" metric, the highest- ranking institution in the visual agency collaboration network is "Harvard Medical School" with 47. " University of Toronto "ranked second with 39, fol- lowed by "The University of Hong Kong" with 38. It is not difficult to find that the average number of Deng et al. BMC Medical Education (2023) 23:203 Page 10 of 17 Fig. 7 Visual clustering analysis of country density maps after the outbreak "ducuments" of the top ten comprehensive literatures is about 34. "Total link strength" refers to the Total co-occur- rence times of keywords and other keywords (includ- ing repeated co-occurrence times). In the organization cooperation network of the visual co-author, it can indi- cate the close cooperation degree between the organiza- tion and different institutions. By analyzing the data, it’s easy to know that "Harvard Medical School" ranks first with 10,080. "Johns Hopkins University" came in second with 8431, and "University of Toronto" came in third with 8062. However, considering the three indicators of "Ducu- ments", "Citations" and "Total link strength", it is easy to find that the development of "Harvard Medical School" in terms of publications, centrality, and total link strength with other institutions is more balanced and better. In addition, it is easy to find that the institution "Harvard Medical School" has direct links with many other insti- tutions, which further reflects that this is a very influen- tial institution and it maintains close cooperation with many other institutions. We realize that the outbreak of COVID-19 has produced a dramatic change in the research community in terms of online medical educa- tion. Taking this change into account in a timely manner when selecting a partner institution can go a long way in helping researchers find the right partner institution. Visualization of collaborative journal clustering networks The clustering network drawn by the journals to which the visual references in this study belong (the nodes are the names of the journals, and the associated journals are connected by curves), and the clustering analysis is per- formed according to the main research keywords of the journals. By looking at the graphs and analyzing the data, you can explore the interconnections between journals. Specifically, the following conclusions are obtained: The journals included in the study before the epidemic were divided into 21 clusters through cluster analysis. (Fig. 10) The journal clustering network contains a total of 770 nodes and 2647 lines. There are about 20 jour- nals with citations above 238. By sorting the number of Deng et al. BMC Medical Education (2023) 23:203 Page 11 of 17 Fig. 8 Visual clustering analysis of the institutions of co-authors before the outbreak (2) Post-epidemic organizations cluster analysis journal citations, the following graph (Table 7) can be obtained. This study found that the journal ACAD MED ranked first with 1326 citations, JAMA-J AM MED ASSOC was a little behind in second place with 1119 citations, and MED TEACH was third with 1080 applica- tions. In addition, by observing the data, it can be seen that the top three citations have more than 1000 cita- tions. In addition, MED EDUC and BMC MED EDUC are ranked fourth and fifth, respectively, with very excel- lent journal citations. The journals included in the study after the epidemic were divided into 19 clusters through cluster analysis. (Fig. 11) By ranking the number of journal citations, the following graph can be obtained (Table 8). The journal THE LANCET was found to be in the first place with 784 citations, Public Library of Science (PLOS ONE) was slightly behind the first place in the second place with 782 citations, International Journal of Environmental Research and Public Health (INT J ENV RES PUB HE) was in the third place with 747 applications. In addition, by looking at the data it can be seen that the top three citation numbers are above 700. However, a large drop in the number of citations occurs from the fourth position. A comprehensive comparison between before and after the epidemic showed that 60% of the journals remained in the top 10 citation numbers after the epidemic. The analysis included "JAMA-J AM MED ASSOC" "BMC MED EDUC" "NEW ENGL J MED" "BMJ-BRIT MED J" "J MED INTERNET RES" "LANCET". Discussion The essence of online education is to use Internet tech- nology to transfer knowledge to learners through a network platform in order to achieve the purpose of cross-distance education and teaching. In addition to not being restricted by time and space, online education mode also saves time and money costs for learning to a certain extent. What’s more, as a new way of education that has emerged in recent years, online education can effectively exercise students’ independent learning ability Deng et al. BMC Medical Education (2023) 23:203 Page 12 of 17 Fig. 9 Visual clustering analysis of the institutions of co-authors after the outbreak and flexibility, so that learners can arrange their learning contents according to their own learning progress and interests. Medical students, because of the special nature of their profession, need to learn knowledge and accumu- late experience over a long period of time, and medical education is very focused on clinical practice and hands- on. In order to meet this demand, online education com- bines modern intelligent visual technology to build online virtual simulation experiment simulation courses that can meet the needs of medical education. Obviously, the teaching mode of teachers and students is also facilitated by online education. In recent years, medical education and online education have shown a good trend of com- plementing and promoting each other. However, at the same time, online education also faces some challenges. Students’ mental health issues With the development of Internet and computer tech- nologies, the technical difficulties of online teaching methods have been largely overcome, but more issues regarding students’ mental health have been revealed [16]. The outbreak of Newcastle Pneumonia in early 2020 has severely impacted the normal school day. As a result of the epidemic preparedness requirements, schools have been teaching online, and it cannot be ruled out that a significant number of students will need to be educated during the quarantine period due to the Newcastle pneumonia infection. As a result, the education sector will see a massive integration of online and offline teaching and learning as the epidemic con- tinues to progress [17]. In a study by Aidos K. Bolatov, Telman Z. Seisembekov [18], the anxiety level of medi- cal students was somewhat increased when online edu- cation was first introduced, but was lower when fully adapted to online education. A study by Basema Sad- dik,1,2 Amal Hussein,1 et al. [19] on anxiety levels of clinical medicine/dental medicine students during hos- pital visits, before and after the introduction of online learning, noted that clinical medicine students reported higher levels of anxiety during clinical rotations, which decreased with the introduction of online learning, but non-medical students’ anxiety levels increased with online learning. Similarly, a study by Jessica García- González, Wei Ruqiong, and Raquel Alarcon-Rodriguez [20] noted that the main reasons for anxiety among medical students who received online instruction were online education during a period close to graduation and being forced to live in poorer (or medically iso- lated) conditions for self-study. In this context, certain demands are placed on students’ learning: they need to plan their studies in accordance with the requirements of the school’s educational management and teachers’ classroom teaching, etc., to increase their self-aware- ness and to overcome their fear of difficulty. On top Deng et al. BMC Medical Education (2023) 23:203 Page 13 of 17 Fig. 10 Visualizing the clustering network of journals before the epidemic Table 7 Top 10 journals in citation counts in the clinical learning before the epidemic Citation Counts References 1326 1119 1080 995 697 658 655 512 512 501 ACAD MED, 2010, ACED MED, 0, 0 JAMA-J AM MED ASSOC, 2010, JAMA-J AM MED ASSOC, 0, 0 MED TEACH, 2010, MED TEACH, 0, 0 MED EDUC, 2010, MED EDUC, 0, 0 BMC MED EDUC, 2010, BMC MED EDUC, 0, 0 NEW ENGL J MED, 2010, NEW ENGL J MED, 0, 0 J GEN INTERN MED, 2010, J GEN INTERN MED, 0, 0 BMJ-BRIT MED J, 2010, BMJ-BRIT MED J, 0, 0 J MED INTERNET RES, 2010, J MED INTERNET RES, 0, 0 LANCET, 2010, LANCET, 0, 0 of synchronous online learning, asynchronous online learning and independent learning are combined to achieve the best possible learning results and to allevi- ate anxiety [21]. At the same time, the anxiety associ- ated with online teaching is difficult to remedy through conventional means [22]. According to one study, even with online to offline transition activities, full imple- mentation of regular lesson plans, adequate discussion and communication in the classroom, and well-devel- oped student–teacher feedback mechanisms, the psy- chological changes in students are not promising, with about a quarter of the students in the survey facing anxiety problems during online instruction [23]. Deng et al. BMC Medical Education (2023) 23:203 Page 14 of 17 Fig. 11 Visualizing the clustering network of journals after the epidemic Teaching quality during online teaching From the visual keyword clustering, it is easy to find that students/knowledge is also a hot direction of research. When students receive online education, it is difficult to generate direct communication with teachers, and it is even more difficult to realize the learning of experiments and clinical operations [24]. Meganne N. Ferrel, John J. Ryan [25] argue that replac- ing face-to-face courses with online courses during the COVID-19 epidemic, while the best solution for now, may also cause significant damage to education. The main impact factors are the loss of interactive and collaborative teaching models, difficulties in starting clinical apprenticeships in hospitals, and the postpone- ment and cancellation of academic conferences [26]. In addition, this may also be related to the inadequate guarantee of teaching operation, such as the frequent "network lag", "system crash" and a series of network signal and server capacity problems during the online teaching at the beginning of the epidemic, especially in remote mountainous and less developed areas.In third world countries, not only the quality of online Deng et al. BMC Medical Education (2023) 23:203 Page 15 of 17 Table 8 Top 10 journals in citation counts in the clinical learning after the epidemic Citation Counts References to further explore and practice around teaching qual- ity improvement has become an important research theme. 784 782 747 559 457 423 419 394 389 351 LANCET, 2020, LANCET, 0, 0 PLOS ONE, 2020, PLOS ONE, 0, 0 INT J ENV RES PUB HE, 2020, INT J RES PUB HE, 0, 0 JAMA-J AM MED ASSOC, 2020, JAMA-J AM MED ASSOC, 0, 0 NEW ENGL J MED, 2020, NEW ENGL J MED, 0, 0 PSYCHIAT RES,2020, PSYCHIAT RES, 0, 0 J MED INTERNET RES, 2020, J MED INTERNET RES, 0, 0 BMJ-BRIT MED J,2020, BMJ-BRIT MED J, 0, 0 BMC PUBLIC HEALTH, 2020, BMC PUBLIC HEALTH, 0, 0 BMC MED EDUC, 2020, BMC MED EDUC, 0, 0 teaching is not promising due to these factors, but also the popularity of online teaching is not satisfactory [27]. Ahmed Alsoufi, Ali Alsuyihili, Ahmed Msherghi, Ahmed Elhadi, [28] showed that most medical stu- dents in Libya do not participate in online teaching and that the prevalence of COVID-19 significantly hinders local medical education. A study by Michał Bączek, MD, Michalina Zagańczyk-Bączek, MD [29], pointed out that low activity in traditional courses compared to online courses and lack of interaction with patients and inadequate computer equipment were the main draw- backs of online teaching. A study by Motte-Signoret, Antoine Labbé, Grégoire Benoist [30], indicated that medical students generally believe that online teaching is not a substitute for offline free teaching and clearly oppose the continuation of online teaching after the crisis is resolved. The traditional offline teaching model is more popular and adaptable to students than the new online teaching model. Inevitably, the constraints of time and distance affect students’ motivation to learn. In the future, the flexible use of online teaching and learning may make a difference. How to get students to accept and adapt to online teaching, how to make use of the advantages of online teaching, and how to strengthen the monitoring mechanism of online teach- ing are all things that should be taken into account in the future of online teaching [31]. As online teaching is affected by many factors and there are many problems to be solved, such as the ineffective guarantee of online teaching quality, the simple transplantation of the tra- ditional offline classroom teaching model under the constraints of inertia, and the insufficient correlation and synergy of the core subjects and elements of online education [32]. Therefore, in the subsequent process of online education teaching reform, how to combine the advantages of traditional offline teaching mode Limitations A bibliometric analysis cannot present a complete picture of the current and future state of the research field, and this study is no exception; therefore, the results of this paper are limited to its scope—that is, this paper focuses on the development of online education programs for medical students before and after the COVID-19 out- break from 2010 to 2022 [33, 34]. However, this decade was also the period when online education developed most rapidly, especially after the outbreak of the new crown epidemic, and therefore the time frame of the data in this paper is representative; second, our study only used data obtained from Web of Science, and therefore, may have missed data that existed only in other databases (e.g., PubMed, Scopus, MEDLINE or Google Scholar) [35]. Third, due to the differences in medical education systems in different countries, the authors neglected to describe the synonymy of relevant subject terms in the construction of the dataset, resulting in the omission of some literature, which may lead to incomplete analysis results. Therefore, there is still some room for improve- ment in this paper, and more in-depth research can be conducted. Conclusion With the development of the epidemic, the education industry, especially face-to-face medical education, which requires practice, has taken a huge hit. Online teaching is still lacking as the first method to deal with the educational problems of students affected by the epi- demic. After the COVID-19 outbreak, there has been a surge in the number of relevant studies, with changes in research hotspots and in the leading researchers, institu- tions, countries, and the most cited journals. Analyzing a large number of academic papers can help to filter more valuable content so that the right method can be used to mitigate the negative effects of the epidemic on people. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12909- 023- 04193-5. Additional file 1. Acknowledgements Not applicable. Authors’ contributions L.Y and K.H.: Study supervision and funding acquisition. HJ.D: Conduct of the research and investigation process, scrubbed data and maintained research data. QR.H: Analyzed the study data and writing of the initial draft. Y.J: Writing Deng et al. BMC Medical Education (2023) 23:203 Page 16 of 17 of the initial draft and provision of analysis tools. XY.Z: Graphics and data visualization, article revision and retouching. SY.Z: Article revision and retouch- ing. The author(s) read and approved the final manuscript. Funding The study was funded by the Cooperative Education Project of Produc- tion and Education of the Higher Education Department of the Ministry of Education (No.202101250012), National Natural Science Foundation of China (No.82260456), The Undergraduate education and teaching reform plan project of Zunyi Medical University (No.XJJG2022-95, FB-2018-3, ZH202110), National Innovation and Entrepreneurship Training Program for College Students (No.202210661029) and the College Students Innovation and Entre- preneurship Training Program of Zunyi Medical University (Nos. ZHCX2022002, ZHCX2022003). Availability of data and materials All data generated or analysed during this study are included in this published article and its supplementary information files. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Author details 1 Zhuhai Campus of Zunyi Medical University, Zhuhai 519000, China. 2 Key Laboratory of Human-Machine Intelligence-Synergy Systems, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences (CAS), Shenzhen 518055, China. Received: 27 December 2022 Accepted: 23 March 2023 References 1. Priftis K, Velardo V, Vascello MGF, Villella S, Galeri S, Spada MS, et al. Lim- ited evidence for neuropsychological dysfunction in patients initially affected by severe COVID-19. Neurol Sci. 2022. https:// doi. org/ 10. 1007/ s10072- 022- 06373-5. 2. Katz M, Nandi N. Social media and medical education in the con- text of the COVID-19 pandemic: scoping review. JMIR Med Educ. 2021;7:e25892. Sheikh A, Sheikh A, Sheikh Z, Dhami S. Reopening schools after the COVID-19 lockdown. J Glob Health. 2020;10:010376. 3. 4. Masonbrink AR, Hurley E. Advocating for children during the COVID-19 school closures. Pediatrics. 2020;146:e20201440. 5. World Health Organization. https:// www. who. int/ zh/ emerg encies/ disea ses/ novel- coron avirus- 2019/ quest ion- and- answe rs- hub/q- a- detail/ q-a- schoo ls- and- covid- 19. 6. Gao J, Yang L, Zou J, Fan X. Comparison of the influence of massive open online courses and traditional teaching methods in medical education in China: a meta-analysis. Biochem Mol Biol Educ. 2021;49:639–51. Ebrahimi A, Ebrahimi S, Ashkani Esfahani S. How COVID-19 pandemic can lead to promotion of remote medical education and democratization of education? J Adv Med Educ Prof. 2020;8(3):144-145. https:// doi. org/ 10. 30476/ jamp. 2020. 86090. 1217. 7. 8. O’Doherty D, Dromey M, Lougheed J, Hannigan A, Last J, McGrath D. Bar- riers and solutions to online learning in medical education – an integra- tive review. BMC Med Educ. 2018;18:130. Liu JW, Huang LC. Detecting and Visualizing Emerging Trends and Tran- sient Patterns in Fuel Cell Scientific Literature. In: 2008 4th International Conference on Wireless Communications, Networking and Mobile Computing. Dalian: IEEE; 2008. p. 1–4. 9. 10. Ma Y, Zhou R, Wu Q. Global research hotspots and research trends on idi- opathic pulmonary fibrosis: a bibliometric and visualization analysis. Ann Palliat Med. 2021;10:9057–68. 11. Liu S, Sun Y-P, Gao X-L, Sui Y. Knowledge domain and emerging trends in Alzheimer’s disease: a scientometric review based on CiteSpace analysis. Neural Regen Res. 2019;14:1643. 12. Ma D, Guan B, Song L, Liu Q, Fan Y, Zhao L, et al. A bibliometric analysis of exosomes in cardiovascular diseases from 2001 to 2021. Front Cardiovasc Med. 2021;8:734514. 13. Goh P-S, Sandars J. A vision of the use of technology in medical educa- tion after the COVID-19 pandemic. MedEdPublish. 2020;9:49. 14. Dwivedi YK, Hughes DL, Coombs C, Constantiou I, Duan Y, Edwards JS, et al. Impact of COVID-19 pandemic on information management research and practice: Transforming education, work and life. Int J Inf Manage. 2020;55:102211. 15. van Eck NJ, Waltman L. Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics. 2010;84:523–38. 16. Ferri F, Grifoni P, Guzzo T. Online Learning and Emergency Remote Teach- ing: Opportunities and Challenges in Emergency Situations. Societies. 2020;10:86. 17. Zhang W, Wang Y, Yang L, Wang C. Suspending classes without stopping learning: China’s education emergency management policy in the COVID-19 outbreak. JRFM. 2020;13:55. 18. Bolatov AK, Seisembekov TZ, Askarova AZh, Baikanova RK, Smailova DS, Fabbro E. Online-learning due to COVID-19 improved mental health among medical students. MedSciEduc. 2021;31:183–92. 19. Saddik B, Hussein A, Sharif-Askari FS, Kheder W, Temsah M-H, Koutaich RA, et al. Increased levels of anxiety among medical and non-Medical University students during the COVID-19 pandemic in the United Arab Emirates. RMHP. 2020;13:2395–406. 20. García-González J, Ruqiong W, Alarcon-Rodriguez R, Requena-Mullor M, Ding C, Ventura-Miranda MI. Analysis of anxiety levels of nursing students because of e-learning during the COVID-19 pandemic. Healthcare. 2021;9:252. 21. Shamir-Inbal T, Blau I. Facilitating emergency remote K-12 teaching in computing-enhanced virtual learning environments During COVID-19 pandemic - blessing or curse? J Educa Comput Res. 2021;59:1243–71. 22. Dhawan S. Online learning: a Panacea in the time of COVID-19 crisis. J Educ Technol Syst. 2020;49:5–22. 23. Srivastava S, Jacob J, Charles AS, Daniel P, Mathew JK, Shanthi P, et al. Emergency remote learning in anatomy during the COVID-19 pandemic: a study evaluating academic factors contributing to anxiety among first year medical students. Med J Armed Forces India. 2021;77:S90–8. 24. Hodges C, Moore S, Lockee B, Trust T, Bond A. The difference between emergency remote teaching and online learning. :12. https:// vtech works. lib. vt. edu/ bitst ream/ handle/ 10919/ 104648/ facdev- artic le. pdf? seque nce 1& isAll owed y. 25. Morreale SP, Thorpe J, Westwick JN. Online teaching: challenge or = = opportunity for communication education scholars? Commun Educ. 2021;70:117–9. 26. Young S. Student views of effective online teaching in higher education. Am J Distance Educ. 2006;20:65–77. 27. Coman C, Țîru LG, Meseșan-Schmitz L, Stanciu C, Bularca MC. Online teaching and learning in higher education during the coronavirus pan- demic: students’ perspective. Sustainability. 2020;12:10367. 28. Alsoufi A, Alsuyihili A, Msherghi A, Elhadi A, Atiyah H, Ashini A, et al. Impact of the COVID-19 pandemic on medical education: medical stu- dents’ knowledge, attitudes, and practices regarding electronic learning. PLoS ONE. 2020;15:e24821. 29. Bączek M, Zagańczyk-Bączek M, Szpringer M, Jaroszyński A, Wożakowska- Kapłon B. Students’ perception of online learning during the COVID- 19 pandemic: a survey study of polish medical students. Medicine. 2021;100:e24821. 30. Motte-Signoret E, Labbé A, Benoist G, Linglart A, Gajdos V, Lapillonne A. Perception of medical education by learners and teachers during the COVID-19 pandemic: a cross-sectional survey of online teaching. Med Educ Online. 2021;26:1919042. 31. Dost S, Hossain A, Shehab M, Abdelwahed A, Al-Nusair L. Perceptions of medical students towards online teaching during the COVID-19 pandemic: a national cross-sectional survey of 2721 UK medical students. BMJ Open. 2020;10:e042378. Deng et al. BMC Medical Education (2023) 23:203 Page 17 of 17 32. Yu S, Liu Y, Yang B, Chen Z. Disciplinary differences in the experience of online education among teachers and students in Chinese universities during COVID-19. Front Psychol. 2022;13:909269. 33. Küçük-Avci Ş, Topal M, İstanbullu A. The effects of the Covid-19 pandemic on distance education in higher education: a bibliometric analysis study. Croat J Educ. 2022;24:457–88. 34. Topal M, İstanbullu A, Kucuk-avcı S. Covid-19 effects on distance educa- tion in higher education: a comparison with bibliometric analysis before Covid-19 pandemic and the Covid-19 pandemic periods. J Educ Technol Online Learn. 2021;4(4): 864-879. https:// doi. org/ 10. 31681/ jetol. 10167 05. 35. Wei X, Li X, Song S, et al. Trends and focuses of hantavirus researches: a global bibliometric analysis and visualization from 1980 to 2020. Arch Public Health. 2022;80:218. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12916-023-02895-7	Zhang et al. BMC Medicine (2023) 21:199 https://doi.org/10.1186/s12916-023-02895-7 RESEARCH ARTICLE BMC Medicine Open Access MPC-n (IgG) improves long-term cognitive impairment in the mouse model of repetitive mild traumatic brain injury Chaonan Zhang1†, Cheng Wei1†, Xingqi Huang1†, Changxin Hou2, Chuan Liu1, Shu Zhang1, Zilong Zhao1, Yafan Liu1, Ruiguang Zhang1, Lei Zhou1, Ying Li1, Xubo Yuan2* and Jianning Zhang1* Abstract Background Contact sports athletes and military personnel who suffered a repetitive mild traumatic brain injury (rmTBI) are at high risk of neurodegenerative diseases such as advanced dementia and chronic traumatic encephalop- athy (CTE). However, due to the lack of specific biological indicators in clinical practice, the diagnosis and treatment of rmTBI are quite limited. Methods We used 2-methacryloyloxyethyl phosphorylcholine (MPC)-nanocapsules to deliver immunoglobulins (IgG), which can increase the delivery efficiency and specific target of IgG while reducing the effective therapeutic dose of the drug. Results Our results demonstrated that MPC-capsuled immunoglobulins (MPC-n (IgG)) significantly alleviated cogni- tive impairment, hippocampal atrophy, p-Tau deposition, and myelin injury in rmTBI mice compared with free IgG. Furthermore, MPC-n (IgG) can also effectively inhibit the activation of microglia and the release of inflammatory factors. Conclusions In the present study, we put forward an efficient strategy for the treatment of rmTBI-related cognitive impairment and provide evidence for the administration of low-dose IgG. Keywords Repetitive mild traumatic brain injury, Immunoglobulin, Cognitive impairment, Neuro-inflammation †Chaonan Zhang, Cheng Wei and Xingqi Huang contributed equally to this work. *Correspondence: Xubo Yuan xbyuan@tju.edu.cn Jianning Zhang jianningzhang@hotmail.com 1 Key Laboratory of Post-Neurotrauma Neuro-Repair and Regeneration in Central Nervous System, Ministry of Education and Tianjin City, Department of Neurosurgery, Tianjin Neurological Institute, Tianjin Medical University General Hospital, Tianjin 300052, China 2 Tianjin Key Laboratory of Composite and Functional Materials, School of Materials Science and Engineering, Tianjin University, Tianjin 300072, China Background Traumatic brain injury (TBI) is a common and seri- ous neurological disease, which contributes to approxi- mately $400 billion annually [1]. The number of new TBI patients is as high as 50 million to 60 million each year, and 80–90% of them are mild TBI (mTBI) [2, 3]. In par- ticular, repetitive mild traumatic brain injury (rmTBI) can contribute to chronic traumatic encephalopathy (CTE) due to the continuous accumulation of damage. CTE is mainly characterized by phosphorylated-Tau (p-Tau) deposition, microglial activation, and white mat- ter rarefaction [4–6]. The highest incidence of rmTBI patients are among contact sports athletes, military veterans, and elderly © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Zhang et al. BMC Medicine (2023) 21:199 Page 2 of 16 people who have fallen over the years [4, 7]. Because of the mild symptoms and the late onset, the consultation rate of rmTBI is low and the treatment is always delayed. Immunoglobulin (IgG) is a polyclonal product puri- fied from human serum and has been used as an effective first-line drug for various neurological diseases, such as Guillain Barre syndrome, chronic inflammatory demyeli- nating polyneuropathy, and multifocal motor neuropathy [8]. Since IgG can directly target the immune system and neurons, IgG has also shown a potential in the treatment of ischemic stroke [9, 10]. In addition, in preclinical and clinical trials of mild to moderate Alzheimer’s disease (AD), IgG was reported to efficiently reduce amyloido- sis (Aβ) and modulated the neuroimmune response, as well as remit brain atrophy in patients [11, 12]. In a study of closed cranial trauma, high doses of IgG (600 mg/kg) were shown benefits in improving behavioral and cogni- tive function in mice with a single impact [13]. However, high dose use of IgG was limited consider- ing safety and feasibility. Many clinical trials have found that patients treated with high-dosage IgG tend to suf- fer side effects, such as thrombosis and anaphylaxis [8, 14]. In recent years, nanocapsules were used to improve the delivery efficiency and reduce the effective therapeu- tic dose of drugs due to their excellent biocompatibility and blood–brain barrier (BBB) permeability [15]. We have previously demonstrated that 2-methacryloyloxy- ethyl phosphorylcholine (MPC) synthesized with the MPC monomer and ethylene glycol dimethyl acrylate (EGDMA) crosslinker, as a choline and acetylcholine analog, can be taken up and transferred by high-affinity choline transporter (ChTs) receptors of endothelial cells from the blood into the brain parenchyma [16]. Although IgG has been extensively studied in a variety of neurological disorders, its efficacy in cognitive impair- ment in rmTBI has been rarely studied. In the present study, MPC-capsuled IgG (MPC-n (IgG)) was used to treat long-term cognitive impairment in rmTBI mice. We first demonstrated the specific accumulation of MPC-n (IgG) in the cortex and hippocampus of rmTBI mice, which confirmed that MPC-n (IgG) increased the BBB penetration and drug delivery efficiency. Next, we deter- mined p-Tau deposition, hippocampal atrophy, cognitive function, and microglial activation in rmTBI mice during the chronic phase. Our study indicates that MPC-n (IgG) can effectively ameliorate cognitive dysfunction and neu- roinflammation in rmTBI mice, which provides a poten- tial strategy for the application with a low dose of IgG. Methods Materials IgG was purchased from Solarbio. N-(3-Aminopro- (APM) was pyl) methacrylamide hydro-chloride from Macklin. purchased 2-Methacryloyloxyethyl phosphorylcholine (MPC), ethylene glycol dimethyl acrylate (EGDMA), fluorescein isothiocyanate (FITC), and 2-(1E, 3E, 5E)-5-(3-(6-(2, 5-dioxopyrrolidin-1-yl) oxy)-6-oxohexyl)-1 from Sigma-Aldrich. Ammonium persulfate (APS), N, N, N′, N′-tetramethylethylenediamine (TEMED) were obtained from Alfa Aesar. All reagents were used without purification. (Cy5.5) were purchased Synthesis of MPC‑n (IgG) The preparation principle of nano-microcapsules is to carry out in situ free radical polymerization on the sur- face of IgG. The specific process is as follows: dissolv- ing IgG in 1 mL phosphate buffer (PBS, pH 7.4), adding APM, and MPC to mix evenly, and then adding cross- linking agent EGDMA, molar ratios of free IgG to APM to MPC to EGDMA = 1: 300: 7000: 600. After mixing for 10 min, the catalyst TEMED and initiator APS (APS: IgG = 300, n/n; TEMED: APS = 2:1, w/w) [16, 17]. The reaction will last 4 h. At the end of the reaction, the nano- capsules with MPC as monomers were obtained. After that, we use dialysis bags (MWCO: 8000–14,000, Solar- bio) to remove the free monomer. It was used for dialysis in phosphate buffer saline (10 mM PBS, pH = 7.4). The fresh PBS solution was replaced every 6 h, and the nano- capsules solution was obtained after 48 h of dialysis. And then, to remove unencapsulated proteins, we passed the solution through a hydrophobic interaction column (Phe- nyl-Sepharose CL-4B, Solarbio, laboratory reagent). The purified nanocapsules were stored at 4 °C. Characterization of MPC‑n (IgG) The particle size of the nanoparticle solution after 1 mL (1 mg/mL) dialysis was measured by Brookhaven’s BI-90 Plus Zeta PALS analyzer [18]. The morphology of nano- capsules was characterized by using the JEM-2100F field transmission electron microscope (TEM) with an accel- eration voltage of 200 kV. The solution of nanocapsules (0.01 mg/mL) was dripped onto the copper mesh and it was stained with 2% (w/v) phosphotungstic acid solu- tion, washed with deionized water, fully dried, and then observed under TEM. The chemical groups on the sur- face of nano-microcapsules were analyzed by the Fourier transform infrared (FT-IR). The freeze-dried sample was mixed with potassium bromide and fully ground, and IgG, MPC-n (IgG) and gel without IgG were prepared for scanning [19]. The scanning range is 400–4000 cm−1. The freeze-dried sample MPC, IgG, and MPC-n (IgG) were dissolved in 0.6 mL D2O, and proton nuclear mag- netic resonance (1H NMR) spectra were measured using AVANCE IIITM HD 400 MHz NanoBAY (Bruker). We can also scan them with a UV–vis spectrophotometer Zhang et al. BMC Medicine (2023) 21:199 Page 3 of 16 (AOE instruments, A360 spectrophotometer). The range is 200−500 nm. The freeze-dried MPC-n (IgG) was added into the solution of PBS (pH = 7.4) to form 1 mg/ mL, and the same in H2O, DMEM, and serum at 37 °C for 48 h. The stability of MPC-n (IgG) was quantified using the scattering light intensity ratio I/I0 by dynamic light scattering (DLS) [20]. Because the crosslinker has an ester group with pH response, we set up two groups of nanocapsules under different pH (pH = 6.5, pH = 7.4) and measured the light scattering intensity I0 of the two groups of solutions respectively. After the determina- tion, both groups of samples were incubated at 37 °C for 60 min. In the process of incubation, the DLS intensity I was measured by the solution of nanocapsules at differ- ent time points, and the enzymatic degradation kinetics of nanocapsules were detected by I/I0. Animals Adult male C57BL/6 J mice (aged 8–10 weeks old, weigh- ing 20–25 g) purchased from Hfk Bioscience company (Beijing, China) were housed at the Experimental Animal Laboratories of Tianjin Neurological Institute. They were randomly fed food and water with a 12-h light/dark cycle. All experimental operations were allowed by the Animal Ethics Committee of Tianjin Medical University. Experimental rmTBI model The mice were anesthetized with 4.6% isoflurane and fixed to an acrylic mold. A concave metal disc was placed caudal to bregma on the shaved head of mice, and the impounder tip of controlled cortical impact (electronic CCI model 6.3, American Instruments, Richmond, VA, USA) was positioned at the center of the disc surface, which is discharged at 5 m/s with a head displacement of 5 mm. Mice were divided into four groups: sham, rmTBI, rmTBI + IgG, and rmTBI + MPC-n (IgG) group. Injured mice were impacted 4 times with a 48-h interval. The sham group underwent the same operating procedures without any impact. After the last impact, the rmTBI mice were respectively administrated IgG (600 mg/kg) and MPC-n (IgG) (120 mg/kg) through the tail veil. In vivo distribution, imaging, and quantification The distribution of Cy5.5-labeled MPC-n (IgG) and IgG in mice was observed using the in vivo imaging system (IVIS Lumina II, PerkinElmer, USA) 2, 6, and 24 h post- injection. Mouse brain tissues of all groups were col- lected 24 h post-injection. Fluorescence intensity values were acquired and analyzed using the Living Image soft- ware version 3.1 (Caliper Life Sciences) [17]. Immunofluorescence staining Mouse brains were harvested following cardiac perfu- sion and fixed in 4% PFA for 24 h. The dehydrated tis- sues were frozen rapidly in liquid nitrogen and sliced into 6-μm-thick sections with the freezing microtome (Leica CM 1950). Sections were blocked with 0.3% Tri- tonX and 5% Albumin Bovine V for 1 h. Sections were incubated with primary antibodies overnight at 4 °C (Iba1, 1:500, Abcam; Phospho-Tau, 1:200, Cell Signal- ing Technology; Phospho-p38 MAPK, 1:200, Cell Sign- aling Technology). After being washed in PBS, sections were incubated with Alexa Fluor conjugated secondary antibodies (AlexaFluor 488/555, 1:500, Life Technolo- gies) and stained with 4′,6-diamidino-2-phenylindole (DAPI, Sigma). Western Blot Brain tissues were obtained at 28 DPI and total pro- tein was lysed with RIPA lysis buffer (Solarbio), PMSF (Solarbio), and phosphatase inhibitor (Sigma). Equally loaded proteins were electrophoretically separated on 10% and 15% SDS-PAGE gels and then transferred to PVDF membranes (Millipore). After being blocked by silk milk, membranes were incubated with primary antibodies for Phospho-Tau, Iba1, Phospho-p38 MAPK (1:1000, Cell Signaling Technology), and GAPDH (1:1000, Abcam). Horseradish peroxidase-conjugated secondary antibodies were used as a chromogenic rea- gent (1:5000, Zhongshanxinqiao). Transmission electron microscope (TEM) The cortex and hippocampus of mice were fixed in 2.5% glutaraldehyde for 24 h and in 1% osmium tetroxide for 1.5 h at 4 °C. After dehydrated, the tissues were cut into ultra-thin slices and observed under transmission elec- tron microscopy as previously described (TEM, HIT CHI-HT7700) [21]. Magnetic resonance imaging (MRI) Mice were anesthetized with isoflurane in a 9.4 T small-bore animal scanner (Bruker bio spec 94/30 USR) at 28 DPI and 42 DPI. The body temperature of mice was monitored and maintained at 37 ± 1° C. T2-weighted images were captured according to the time = 2500 ms, following parameters: repetition echo time = 33 ms, rare factor = 8, the field of vision = 20 × 20 mm, matrix size = 256 × 256, slice thick- ness = 0.5 mm, scanning time: 2 min 40 s [22]. Morris water maze test Morris water maze test was used to evaluate the learn- ing ability and spatial memory of mice. The mice at Zhang et al. BMC Medicine (2023) 21:199 Page 4 of 16 28 DPI were placed into the pool from four quadrants to search for the underwater platform in the 90 s. After 90 s, mice that had not found the platform were guided to the platform and stayed for 20 s. Four trials per day for six consecutive days later, the platform was removed on the testing day, and the computer recorded the swimming track, dwelling time, and path length of mice [23]. Cytokine quantification by Array Mouse brains of all groups were used for cytokine pro- filing. The relative levels of different cytokines were assessed by Proteome Profiler Mouse Cytokine Array Panel A (R&D Systems). The densitometry of each spot was measured using the ChemiDo XRS + imaging system (Bio-Rad, CA, USA), and the pixel density was evaluated by ImageJ Software. Sampling and preparation of mouse brain tissue samples RNA samples were sent to Shanghai Bohao Biotech- nology Co., Ltd., China for cRNA library preparation and RNA sequencing. Total RNA from the samples was extracted by the Animal Total RNA Extraction Kit (Mag- netic Bead method, MJYHIVD). Among them, RNA quantity and quality were assessed using Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, US) and RNAClean XP Kit (Cat A63987, Beckman Coulter, Inc. Kraemer Boulevard Brea, CA, USA) and RNase-Free DNase Set (Cat#79,254, QIAGEN, GmBH, Germany) were used for purification. All samples used Illumina NovaSeq6000 sequencer, model: PE150. Gene expression and differential gene acquisition FASTQ data were processed using Seqtk (https:// github. com/ lh3/ seqtk) to extract qualified raw read sub-data (the amount of data is about 6G/sample, and the ratio of base quality greater than 20 (Q20) in each direction is no less than 90%). We used the spliced mapping algo- rithm of Hisat2 (version: 2.0.4) [24] to perform genome mapping on the preprocessed reads, where the mapped genome is “GRCm38.p4 (mm10)”(ftp:// ftp. ensem bl. org/ pub/ relea se83/ fasta/ mus_ muscu lus/ dna/ Mus_ muscu lus. GRCm38. dna. prima ry_ assem bly. fa. gz), the param- eter defaults. Then “Stringtie” (version: 1.3.0) was used to count the “Fragments” of the mapped genes [25, 26], and after normalization using the TMM method [27], the FPKM value of each gene was calculated using the perl script. All downstream analyses were performed in R version 3.6.3. “edgeR” was used to screen for differential genes between samples. Genes with log2\|FC\|> 0.2 and p < 0.05 identified were identified as differentially expressed genes (DEGs) [28]. Using “umsp” (version 0.2.7.0) to observe heterogeneity between data samples, DEG volcano plots and heatmaps were visualized using the “ggplot2” and “ComplexHeatmap” packages, respectively. Functional enrichment analysis Gene Ontology (GO) [29] and KEGG enrichment analy- sis [30] was conducted to detect those DEGs’ function. The “GOplot” package and “cluster profiler” are used to visualize the enrichment results. Statistical analysis All the data were analyzed by Prism 9 (GraphPad Soft- ware, San Diego, USA. One-way analysis of variance (ANOVA) followed by Tukey’s post hoc test. All data were presented as mean ± standard error of the mean, and P-values less than 0.05 were deemed statistically significant. Results Characterization of MPC‑n (IgG) Nanocapsules are IgG encapsulated by MPC monomer, and then MPC-n (IgG) is formed by adding pH-respon- sive cross-linking agent EGDMA under the principle of in situ radical polymerization (Fig. 1A). The particle size of the nanocapsules is about 18 nm (± 2 nm). Compared with the particle size of IgG, the particle size of nanocap- sules is about 18.5 nm, which increases obviously. The reason for the increase in particle size is that a polymer shell coated with protein is formed on the surface of IgG by in situ free radical reaction, which makes the whole spherical particles larger, and the experimental results are consistent with the analysis, which proves the success- ful preparation of nanocapsules. This is consistent with the results shown in the TEM photos that the nanocap- sules are more regularly spherical (Fig. 1B, C). The sur- face morphology of the dialyzed nano-microcapsules was analyzed by TEM. From the TEM photos of nanocap- sules, we can see that the morphology of MPC-n (IgG) is mostly spherical particles. Compared with the scale, we can see that the particle size is about 20 nm, which is consistent with the particle size measured by our par- ticle size meter. The formation of spherical particles is mainly due to the negative charge of protein in phosphate buffer, while the positively charged monomer APM is a kind of acrylic hydrochloride. Through the interaction of static electricity and hydrogen bond, APM, MPC mono- mer, and cross-linking agent carry out in situ free radical polymerization on the protein surface under the action of the initiator to form a three-dimensional network structure. The research group proved earlier that if the monomer does not form a shell structure on the protein surface, only the polymerization between monomers will Zhang et al. BMC Medicine (2023) 21:199 Page 5 of 16 Fig. 1 Characterization of MPC-n (IgG). A Schematic for the synthesis of MPC-n (IgG). B Size of IgG and MPC-n (IgG) measured by dynamic light scattering. C Representative transmission electron micrographs of MPC-n (IgG). Scale bar = 50 nm. D FT-IR spectra of free IgG, polymer gels without IgG, and MPC-n (IgG) after lyophilization. E Release of MPC-n (IgG) of different pH (6.5 and 7.4) measured by dynamic light scattering. F Flow chart showing IgG and MPC-n (IgG) administration and experimental design not form a spherical structure, which further illustrates the formation of spherical nanocapsules. From the FT-IR (Fig. 1D), the absorption peaks identified by 1726 cm−1, 1240 cm−1, 1084 cm−1, and 960 cm−1 belong to the ester absorption peak of- COOCH2-, the absorption peak of P = O, P-O-C, and the antisymmetric stretching vibration peak of C-N. Because the polymer material on the surface of MPC-n (IgG) is mainly PMPC [31], the appearance of these four peaks verifies its existence, while the composition of other monomers and cross-linking agents is difficult to see, mainly due to the small proportion added. In the 1H NMR spectra (Additional file 1: Fig. S1A), the main proton peaks of MPC-n (IgG) were located at 1.8, 3.1, and 4.1–4.3 ppm, which corresponded to the characteristic peaks of poly 2-methacryloyloxy ethyl Zhang et al. BMC Medicine (2023) 21:199 Page 6 of 16 phosphorylcholine (PMPC). Because the protein has a characteristic peak in 280 nm, in the ultraviolet spectrum (Additional file 1: Fig. S1B), IgG and MPC-n (IgG) will have a characteristic peak in 280 nm, while gel without IgG will not have a characteristic peak, which proves that nanocapsules do contain IgG. We further confirmed the degradation behavior of MPC-n (IgG) by DLS (Fig. 1E). Compared with pH = 7.4, in the environment of pH = 6.5, with the extension of culture time, I/I0 of MPC-n (IgG) gradually decreased, indicating that the particle size of MPC-n (IgG) was decreasing, which further confirmed that it could be degraded. To investigate the stability of MPC-n (IgG), it was dis- solved in H2O, PBS, DMEM, and serum at 37 °C [17], and the size of MPC-n (IgG) was measured by DLS for 48 h. As shown in Additional file 1: Fig. S1C, the MPC-n (IgG) remained stable in H2O, PBS, DMEM, and serum at 37 °C for 48 h. These findings suggested that the encapsulation of IgG enabled controlled release and enhanced its sta- bility. The modeling process and disease progression of rmTBI are shown in Fig. 1F. MPC‑n (IgG) promotes IgG accumulation in rmTBI mice Given the good stability and BBB targeting of microcap- sules, we first assessed the delivery capacity of MPC-n (IgG) in rmTBI mice. IgG and MPC-n (IgG) coupled with Cy5.5 fluorescence were intravenously injected into rmTBI mice after the last impact. In a vivo imaging sys- tem, it exhibited a higher Cy5.5 signal in the mice brains of the MPC-n (IgG) group compared to the IgG group at 2 h, 6 h, and 24 h post-injection (Fig. 2A). In addition, we obtained the mouse brain tissue at 24 h for fluores- cence detection, which displayed a consistent result. To further investigate the spatial differences in IgG distribu- tion, the distribution of Cy5.5-coupled IgG and MPC-n (IgG) 24 h post-injection was observed under a confocal microscope, which showed that MPC-n (IgG) was more significantly distributed in the cortex and hippocampus compared to the free IgG (Fig. 2B). These results illus- trated that MPC-n (IgG) significantly enhanced BBB per- meability of IgG and increased the specific accumulation in the cortex and hippocampus of rmTBI mice. increasing number of platform-site crossovers (Fig. 3C). These results suggested that MPC-n (IgG) can signifi- cantly improve long-term cognitive and memory func- tion more than free IgG. Due to the sensibility of MRI in anatomical detail, it was performed to further measure the hippocampal volume changes. Although there was no significant sta- tistical difference among the groups at 28 DPI (Fig. 3D), MPC-n (IgG) showed a therapeutic effect in hippocam- pal atrophy at 42 DPI (Fig. 3E), while IgG had no dis- tinct benefit at each time point. These results indicated that the hippocampal lesions were still progressing after rmTBI in the long term, and MPC-n (IgG) could amelio- rate cognitive and memory dysfunction through mitigat- ing hippocampal atrophy. Administration of MPC‑n(IgG) alleviates Tau deposition and myelin injury Tau protein is an axonal protein that can stabilize and bundle microtubules [33]. However, excessive phospho- rylation of Tau protein results in its shedding from the axon and formation of neurofilament tangles, which was considered a key factor in rmTBI-related dementia [34, 35]. The expression of p-Tau protein was observed by immunofluorescence staining, which suggested that p-Tau protein was mainly deposited in the hippocampus (Fig. 4A). Meanwhile, MPC-n (IgG) markedly reduced the deposition of p-Tau in the hippocampus of mice (Fig. 3A, B), which was in line with the result of Western Blot (Fig. 3C, D). In mTBI, shearing and strain forces can directly contribute to multifocal axonal injury [36]. Mye- lin surrounding the axons of neurons acts as an insulator and increases the conduction speed of nerve impulses, as well as protecting the axons. TEM was used to observe the changes in the ultrastructure of the myelin sheath. As the TEM displayed, the lamellar structure of the myelin sheath was destroyed and the cytoplasm of the neuron was dissolved in both cortex and hippocampus of rmTBI, while MPC-n (IgG) obviously amended the myelin injury compared with free IgG (Fig. 4E). Taken together, MPC-n (IgG) can more effectively remove p-Tau deposition and remit myelin degeneration than free IgG. MPC‑n (IgG) improves cognitive impairment and hippocampal atrophy The main pathological manifestation of rmTBI is long- term cognitive dysfunction. Thus, Morris water maze was performed to determine spatial cognition and mem- ory function in rmTBI mice at 28 DPI [32]. On the test- ing day, compared with the IgG group, the MPC-n (IgG) group had longer dwelling time in the target quadrant (Fig. 3A), shorter latency to first target-site (Fig. 3B), and MPC‑n(IgG) suppresses the neuro‑inflammatory response Continuous chronic neuro-inflammation often leads to secondary injury and neurodegeneration such as mild cognitive impairment [37, 38]. Microglia acts as the first line of defense in the immune response in the central nervous system, and its activation can last for several months [39]. To evaluate the state of microglia, immunofluorescence was used to detect the expres- sion of activated microglia marker Iba1 in the cortex Zhang et al. BMC Medicine (2023) 21:199 Page 7 of 16 Fig. 2 The selective accumulation of MPC-n (IgG) in rmTBI mice. A In vivo imaging exhibited the fluorescence distribution of cy5.5 signal in the sham, rmTBI + IgG, rmTBI + MPC-n (IgG) group at 2, 6, and 24 h after the injection in rmTBI mice (n = 3). B The confocal microscope displayed the spatial distribution of Cy5.5 in the cortex and hippocampus 24 h post injection. Scale bar = 20 μm Zhang et al. BMC Medicine (2023) 21:199 Page 8 of 16 Fig. 3 MPC-n (IgG) improves cognitive impairment and hippocampal atrophy. A Heat map of the trajectory in the Morris water maze test (n = 8). B The first latency duration to pass over the platform and C the numbers of crossing the platform. D The statistics of hippocampal volume change fold at 28 DPI and E 42 DPI (n = 5–6). F The representative magnetic resonance images at 28 and 42 DPI and hippocampus at 28 DPI. The results exhibited that MPC-n (IgG) was more effective in restraining micro- glia activation in both the cortex (Fig. 5A, B) and hip- pocampus (Fig. 5C, D). In addition, cytokines were quantified by Array at 42 DPI, which demonstrated the descent of pro-inflammatory factors under the treat- ment of MPC-n (IgG), including M-CSF, CXCL12, CCL2/MCP-1, IFN-γ, CCL12, and TNF-α (Fig. 5E, F). All these results proved that MPC-n (IgG) showed a higher capacity of anti-inflammatory compared with free IgG. Zhang et al. BMC Medicine (2023) 21:199 Page 9 of 16 Fig. 4 Administration of MPC-n (IgG) alleviates p-Tau deposition and myelin injury. A, B Immunofluorescence staining of p-Tau in the hippocampus. Scale bar = 50 μm. (n = 5–6). C, D Western blot quantification of p-Tau in the hippocampus. (n = 5–6). E The ultrastructure of myelin of the cortex and hippocampus was observed at 28 DPI under TEM. Scale bar = 200 nm Zhang et al. BMC Medicine (2023) 21:199 Page 10 of 16 Fig. 5 MPC-n (IgG) suppresses the activation of microglia and the release of pro-inflammatory factors. A, B Immunofluorescence staining of activated microglia in the cortex and C, D hippocampus. Scale bar = 50 μm (n = 5–6). E, F Array quantification of cytokine quantification indicated the decrease of M-CSF, CXCL12, CCL2/MCP-1, IFN-γ, CCL12, and TNF-α after the treatment of MPC-n (IgG) RNA sequencing reveals the mechanism of MPC‑n (IgG) treatment in rmTBI In the differential analysis, the results showed the dif- ferential genes between the rmTBI group and the sham group (Fig. 6A, C), including 3,574 differential genes, 1794 upregulated genes, and 1780 downregulated genes. Compared with the MPC-n (IgG) group and the rmTBI group, a total of 1625 differential genes were found, including 811 upregulated genes and 814 downregulated genes (Fig. 6B, D). The completed differential gene pro- files can be found in Supplementary Tables S1 and S2. Functional enrichment analysis demonstrated that the changed pathways in the rmTBI group were mainly MAPK signaling pathway, PI3K-Akt pathway, and cell matrix-related pathways compared with the sham group (Fig. 6E). Interestingly, MAPK signaling pathway was also enriched in the MPC-n (IgG) and rmTBI groups (Fig. 6F). Previous studies have shown that MAPK Zhang et al. BMC Medicine (2023) 21:199 Page 11 of 16 Fig. 6 RNA Sequencing reveals the mechanism of MPC-n (IgG) treatment in rmTBI. A Heatmap of differential genes, C differential gene volcano plot, E KEGG circle diagram of differential genes in MAKP pathway between the rmTBI and sham groups. B Heatmap of differential genes, D differential gene volcano plot, F KEGG circle diagram of differential genes between the MPC-n (IgG) and rmTBI groups Zhang et al. BMC Medicine (2023) 21:199 Page 12 of 16 pathway could play an important role in rmTBI [40, 41]; thus, we marked the differential genes in MAPK pathway. As displayed in Additional file 2: Fig. S2A and B, p-p38 MAPK pathway was the most critical gene which upregu- lated in the rmTBI group compared with the sham group and was lessened in the MPC-n (IgG) group, suggesting that p-p38 MAPK was an underlying treatment target of MPC-n (IgG). MPC‑n (IgG) downregulates the p‑p38 MAPK pathway in rmTBI mice It was reported that the p38-MAPK pathway contributed to the phosphorylation of Tau protein and synaptic dam- age [42, 43], and a clinical study showed that the p38α- MAPK inhibitor could improve memory function in AD patients [44]. In addition, the phosphorylation of MAPK in neurons may be a kind of stress response, which turns to promote the activation of microglia and the release of pro-inflammatory factors [45]. According to the sequencing results, we further verified the expression of MAPK pathway in neurons. Western Blot showed that MPC-n (IgG) could effectively reduce the expression of p-p38 MAPK in the hippocampus of rmTBI mice (Fig. 7A, B), and immunofluorescence illus- trated that p-p38 MAPK co-localized with neurons in the hippocampus (Fig. 7C), suggesting MPC-n (IgG) can act directly on neurons to exert therapeutic effects (Fig. 8). Discussion Mild traumatic brain injury (mTBI) is the most com- mon subtype of TBI, often presenting with dizziness, headache, and cognitive deficits. Numerous studies have shown that military personnel and contact sports athletes who have undergone rmTBI are at high risk of CTE and advanced dementia [46]. Nevertheless, the lack of effec- tive biological indicators and diagnostic criteria for CTE in clinical practice gaps the difficulty in the treatment. IgG has been widely used in clinical for more than a century as a superior immunomodulatory agent. High-dosage IgG is approved by the Food and Drug Administration (FDA) as an anti-inflammatory and immune-modulator for several autoimmune diseases such as chronic inflammatory demyelinating poly- neuropathy (CIDP) and multifocal motor neuropathy (MMN) [8]. Besides, some experimental studies have shown that IgG exhibits a prominent neuronal protec- tive potential in the treatment of TBI and stroke by removing the C3 complement and downregulating the toll-like receptor of neurons [9, 10, 47]. In recent years, phase II and III clinical trials have been carried out in AD patients, but the effect is not satisfactory. The results suggested that IgG can only remit brain atro- phy and cognitive impairment in mild AD patients in a short course, and there was no significant improvement in the long term [12]. In another trial, the AD patients who received IgG did not show a beneficial outcome on cognition but occurred with more systemic responses such as chills and rashes. What is noteworthy is that focused ultrasound (FUS) can temporarily open the BBB and contribute to the specific target of IgG to the hippocampus to reduce amyloid plaque pathology and pro-inflammatory factors [48]. These promising results indicated the significance of facilitating the targeting efficiency of IgG in the treatment of cognitive disorders. MPC-nanocapsules are composed of PMPC polymer shells and protein cargo. Due to the protection of nano- capsules, the protein can avoid rapid degradation and improve the transportation efficiency of BBB [49, 50]. Our previous studies have clarified that MPC-n (IgG) can be taken up from the peripheral blood by ChT1 receptors of endothelial cells and transported to the brain parenchyma, thereby improving cerebral infarc- tion, neurological deficits and neuro-inflammation after stroke. As previously mentioned, the association between TBI and cognitive deficits has been expounded. Two main hypotheses have been proposed regarding the mechanism of increased risk, one that TBI decreases cognitive reserve and the second that TBI directly initi- ates Tau and Aβ pathophysiology processes in dementia [51]. In this study, we administrated low-dosage MPC-n (IgG) (120 mg/kg) and high-dosage free IgG (600 mg/ kg) respectively after rmTBI and found that MPC-n (IgG) more significantly improved the cognitive function, p-Tau deposition, and myelin damage of rmTBI mice. It is also notable that the efficacy at42 DPI was more obvious than that at 28 DPI, suggesting that MPC-n (IgG) had a continuous therapeutic effect on the long-term progno- sis of rmTBI. In addition, the inflammatory response is an important pathophysiological event in TBI, which may affect outcomes in various ways. Microglia are rapidly activated after brain injury and can persist for months [52, 53]. Our study demonstrated that MPC-n (IgG) also was effective in inhibiting microglial activation and the release of inflammatory factors. Finally, RNA from the hippocampus at 28 DPI was sequenced to further explore the therapeutic mecha- nism of MPC-n (IgG). The sequencing results indicated that the mitogen activated protein kinase (MAPK) sig- nal pathway could be an effective target for MPC-n (IgG) treatment. MAPK is an important transmitter of sig- nals from the cell surface to the interior of the nucleus, which can be divided into four subtypes: ERK, P38, JNK, and ERK5. A clinical study of AD showed that the expression MAPK/ metabolism pathway was negatively related to cognitive performance [54]. Another study Zhang et al. BMC Medicine (2023) 21:199 Page 13 of 16 Fig. 7 MPC-n (IgG) downregulates p-p38 MAPK pathway in rmTBI mice. A Western blot quantification showed that MPC-n (IgG) reduced the expression of p-p38 MAPK in the hippocampus in rmTBI mice (n = 5). B Immunofluorescence staining illustrated that p-p38 MAPK co-localized with hippocampal neurons. Scale bar = 10 μm confirmed that knockdown of p38α-MAPK in AD mice reduced p-Tau load, enhanced synaptic plasticity, and improved cognitive function [55]. IgG has been reported to improve cognition by modulating intracranial and peripheral immunity as well as Aβ pathology [11], but the main mechanism of IgG in the rmTBI was only involved in anti-inflammatory therapy [56]. In our study, we deter- mined that p-p38 MAPK was significantly upregulated after rmTBI and obviously diminished by MPC-n (IgG) using Western blot. Meanwhile, immunofluorescence staining exhibited that p-p38 MAPK could co-localize with neurons, supporting that MPC-n (IgG) may play a neuroprotective role by suppressing the expression of p-p38 MAPK in neurons. In summary, we proposed an efficient drug delivery system that significantly promoted IgG accumulation in Zhang et al. BMC Medicine (2023) 21:199 Page 14 of 16 Fig. 8 The mechanism diagram of MPC-n (IgG) treatment in rmTBI the brain parenchyma of rmTBI mice. Compared with free IgG, MPC-n (IgG) improved cognitive impairment and alleviated chronic inflammation after rmTBI. Conclusion In this study, we applied MPC-capsuled IgG for the treat- ment of rmTBI-associated cognitive impairment and obtained promising results. We verified that MPC-n (IgG) has a superior ability of BBB penetration and spe- cific target and distinctly improved cognitive function recovery and inhibited neuro-inflammatory response compared with free IgG. Moreover, RNA sequencing revealed that MAPK was a potential target in the clini- cal therapy of rmTBI-related cognitive deficits. Our study may shed a light on the treatment of cognitive dysfunc- tion and provide evidence for the application of low dosage-IgG. Abbreviations AD Aβ APM BBB CTE Alzheimer’s disease Amyloidosis N-(3-Aminopropyl) methacrylamide hydro-chloride Blood-brain barrier Chronic traumatic encephalopathy Choline transporter Dynamic light scattering Ethylene glycol dimethyl acrylate Food and Drug Administration Fluorescein isothiocyanate Fourier transform infrared Focused ultrasound Immunoglobulins Mitogen activated protein kinase Multifocal motor neuropathy 2-Methacryloyloxyethyl phosphorylcholine ChTs DLS EGDMA FDA FITC FT-IR FUS IgG MAPK MMN MPC MPC-n (IgG) MPC-capsuled immunoglobulins mTBI PMPC p-Tau rmTBI TBI TEM TEMED Mild traumatic brain injury Poly 2-methacryloyloxy ethyl phosphorylcholine Phosphorylated-Tau Repetitive mild traumatic brain injury Traumatic brain injury Transmission electron microscope Tetramethylethylenediamine Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12916- 023- 02895-7. Additional file 1: Figure S1. Characteristics of MPC-n.1H NMR spectra of, IgG, polymer gels without IgG, and MPC-nrecorded in D2O at a con- centration of 10 mg/mL.UV-vis spectra of IgG, polymer gels without IgG and MPC-n.The stability of MPC-nin H2O, PBS, DMEM, and serum at 37°C, determined by monitoring particle sizefor 48 h. Zhang et al. BMC Medicine (2023) 21:199 Page 15 of 16 Additional file 2: Figure S2. Differential genes altered in MAPK pathway in all groupsVisualization of differential genes in MAKP pathway between the rmTBI and sham groups,visualization of differential genes in MAKP pathway between the MPC-nand rmTBI groups. Additional file 3: Table S1. The number of experimental animals. Additional file 4: Table S2. All differentially expressed genes between the rmTBI group and the sham group. Additional file 5: Table S3. All differentially expressed genes between the MPC-ngroup and the rmTBI group. Additional file 6. Original uncropped Western blots of GAPDH 1. Additional file 7. Original uncropped Western blots of GAPDH 2. Additional file 8. Original uncropped Western blots of p-Tau. Additional file 9. Original uncropped Western blots of p-p38 MAPK. Acknowledgements Not applicable. Authors’ contributions JZ and XY conceived and designed the study. CH designed, engineered, and optimized the targeted delivery system used in this publication. CZ, CW and XH performed the experiments and analyzed the results. CL, ZZ, and SZ helped interpret the data. LZ, YL, RZ, and YL provided technical support. CZ wrote and revised the manuscript, which was subsequently edited by XY and JZ. All authors read and approved the final manuscript. Funding This study was financially supported by Tianjin Key Medical Discipline (Spe- cialty) Construction Project, the National Natural Science Foundation of China (No. 81930031, No. 81971176 and No. 81901525), and the Young Scientists Award of the National Natural Science Foundation of China (No.82022020). Availability of data and materials The data that support the findings of this study are available from the cor- responding author upon reasonable request. Declarations Ethics approval and consent to participate Animal protocols were performed with approval of the Experimental Animals. Ethics Committee of Tianjin Medical University (SYXK: 2016–0012). Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Received: 9 October 2022 Accepted: 9 May 2023 References 1. Maas AIR, et al. Traumatic brain injury: integrated approaches to improve prevention, clinical care, and research. Lancet Neurol. 2017;16(12):987–1048. 2. Xu X, et al. Repetitive mild traumatic brain injury in mice triggers a slowly developing cascade of long-term and persistent behavioral deficits and pathological changes. Acta Neuropathol Commun. 2021;9(1):60. 3. McCrory P, et al. Consensus statement on concussion in sport: the 4th 4. Thompson HJ, McCormick WC, Kagan SH. Traumatic brain injury in older adults: epidemiology, outcomes, and future implications. J Am Geriatr Soc. 2006;54(10):1590–5. 5. Blennow K, et al. Traumatic brain injuries. Nat Rev Dis Primers. 2016;2:16084. Smith C, et al. The neuroinflammatory response in humans after trau- 6. matic brain injury. Neuropathol Appl Neurobiol. 2013;39(6):654–66. 7. Bieniek KF, et al. Chronic traumatic encephalopathy pathology 8. in a neurodegenerative disorders brain bank. Acta Neuropathol. 2015;130(6):877–89. Lünemann JD, Nimmerjahn F, Dalakas MC. Intravenous immunoglobu- lin in neurology–mode of action and clinical efficacy. Nat Rev Neurol. 2015;11(2):80–9. 9. Arumugam TV, et al. Intravenous immunoglobulin (IVIG) protects the brain against experimental stroke by preventing complement-mediated neuronal cell death. Proc Natl Acad Sci U S A. 2007;104(35):14104–9. 10. Lok KZ, et al. Intravenous immunoglobulin (IVIg) dampens neuronal toll-like receptor-mediated responses in ischemia. J Neuroinflammation. 2015;12:73. 11. St-Amour I, et al. IVIg protects the 3xTg-AD mouse model of Alzheimer’s disease from memory deficit and Aβ pathology. J Neuroinflammation. 2014;11:54. 12. Kile S, et al. IVIG treatment of mild cognitive impairment due to Alzhei- mer’s disease: a randomised double-blinded exploratory study of the effect on brain atrophy, cognition and conversion to dementia. J Neurol Neurosurg Psychiatry. 2017;88(2):106–12. 13. Jeong S, et al. Intravenous immunoglobulin G improves neurobehavioral and histological outcomes after traumatic brain injury in mice. J Neuro- immunol. 2014;276(1–2):112–8. 14. Dalakas MC, Clark WM. Strokes, thromboembolic events, and IVIg: rare incidents blemish an excellent safety record. Neurology. 2003;60(11):1736–7. 15. von Roemeling C, et al. Breaking down the barriers to precision cancer nanomedicine. Trends Biotechnol. 2017;35(2):159–71. 16. Han L, et al. Systemic delivery of monoclonal antibodies to the central nerv- ous system for brain tumor therapy. Adv Mater. 2019;31(19): e1805697. 17. Jin W, et al. Early administration of MPC-n(IVIg) selectively accumulates in ischemic areas to protect inflammation-induced brain damage from ischemic stroke. Theranostics. 2021;11(17):8197–217. 18. Xing H, et al. “Sustained irrigation” effect enhanced the accumulation and retention of ultra-long circulating nanoparticles in tumor. Biochem Biophys Res Commun. 2022;595:82–8. 19. Qi H, et al. Systemic administration of enzyme-responsive growth factor nanocapsules for promoting bone repair. Biomater Sci. 2019;7(4):1675–85. 20. Zhan Q, et al. Phosphatidylcholine-engineered exosomes for enhanced tumor cell uptake and intracellular antitumor drug delivery. Macromol Biosci. 2021;21(8): e2100042. 21. Zhao Z, et al. Cardiolipin-mediated procoagulant activity of mitochondria contributes to traumatic brain injury-associated coagulopathy in mice. Blood. 2016;127(22):2763–72. 22. Shultz SR, et al. Tibial fracture exacerbates traumatic brain injury out- comes and neuroinflammation in a novel mouse model of multitrauma. J Cereb Blood Flow Metab. 2015;35(8):1339–47. 23. Zhong J, et al. Bexarotene protects against traumatic brain injury in mice partially through apolipoprotein E. Neuroscience. 2017;343:434–48. 24. Kim D, Langmead B, Salzberg SL. HISAT: a fast spliced aligner with low memory requirements. Nat Methods. 2015;12(4):357–60. 25. Pertea M, et al. StringTie enables improved reconstruction of a transcrip- tome from RNA-seq reads. Nat Biotechnol. 2015;33(3):290–5. 26. Pertea M, et al. Transcript-level expression analysis of RNA-seq experi- ments with HISAT. StringTie and Ballgown Nat Protoc. 2016;11(9):1650–67. 27. Risso D, et al. Normalization of RNA-seq data using factor analysis of control genes or samples. Nat Biotechnol. 2014;32(9):896–902. 28. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioin- formatics. 2010;26(1):139–40. 29. The Gene Ontology Consortium. Gene Ontology Consortium: going forward. Nucleic Acids Res. 2015;43(D1):D1049-D1056. https:// doi. org/ 10. 1093/ nar/ gku11 79. International Conference on Concussion in Sport held in Zurich, Novem- ber 2012. Br J Sports Med. 2013;47(5):250–8. 30. Kanehisa M, et al. KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res. 2017;45(D1):D353-d361. Zhang et al. BMC Medicine (2023) 21:199 Page 16 of 16 31. Wu D, et al. A bioinspired platform for effective delivery of protein thera- peutics to the central nervous system. Adv Mater. 2019;31(18): e1807557. 32. Vorhees CV, Williams MT. Morris water maze: procedures for assessing spatial and related forms of learning and memory. Nat Protoc. 2006;1(2):848–58. 33. Kellogg EH, et al. Near-atomic model of microtubule-tau interactions. Science. 2018;360(6394):1242–6. 34. Omalu B, et al. Emerging histomorphologic phenotypes of chronic traumatic encephalopathy in American athletes. Neurosurgery. 2011;69(1):173–83 discussion 183. 35. Corsellis JA, Bruton CJ, Freeman-Browne D. The aftermath of boxing. Psychol Med. 1973;3(3):270–303. 36. King AI. Fundamentals of impact biomechanics: part I-Biomechanics of the head, neck, and thorax. Annu Rev Biomed Eng. 2000;2:55–81. 37. Johnson VE, Stewart W, Smith DH. Traumatic brain injury and 38. amyloid-β pathology: a link to Alzheimer’s disease? Nat Rev Neurosci. 2010;11(5):361–70. Ikram M, et al. Melatonin as a potential regulator of oxidative stress, and neuroinflammation: mechanisms and implications for the management of brain injury-induced neurodegeneration. J Inflamm Res. 2021;14:6251–64. 39. Eyolfson E, et al. Microglia dynamics in adolescent traumatic brain injury. J Neuroinflammation. 2020;17(1):326. 40. Li G, et al. Curcumin suppress inflammatory response in trau- matic brain injury via p38/MAPK signaling pathway. Phytother Res. 2022;36(3):1326–37. 41. Zhang L, et al. Inhibition of Epac2 attenuates neural cell apoptosis and improves neurological deficits in a rat model of traumatic brain injury. Front Neurosci. 2018;12:263. 42. Reynolds CH, et al. Reactivating kinase/p38 phosphorylates tau protein in vitro. J Neurochem. 1997;69(1):191–8. 43. Li S, et al. Soluble Aβ oligomers inhibit long-term potentiation through a mechanism involving excessive activation of extrasynaptic NR2B-contain- ing NMDA receptors. J Neurosci. 2011;31(18):6627–38. 44. Scheltens P, et al. An exploratory clinical study of p38α kinase inhibition in Alzheimer’s disease. Ann Clin Transl Neurol. 2018;5(4):464–73. 45. Hanisch UK. Microglia as a source and target of cytokines. Glia. 2002;40(2):140–55. 46. Baugh CM, et al. Chronic traumatic encephalopathy: neurodegeneration following repetitive concussive and subconcussive brain trauma. Brain Imaging Behav. 2012;6(2):244–54. 47. Thom V, et al. Therapeutic potential of intravenous immunoglobulin in acute brain injury. Front Immunol. 2017;8:875. 48. Dubey S, et al. Clinically approved IVIg delivered to the hippocampus with focused ultrasound promotes neurogenesis in a model of Alzhei- mer’s disease. Proc Natl Acad Sci U S A. 2020;117(51):32691–700. 49. Lewis A, et al. Poly(2-methacryloyloxyethyl phosphorylcholine) for protein conjugation. Bioconjug Chem. 2008;19(11):2144–55. 50. Gaitzsch J, et al. Synthesis of an amphiphilic miktoarm star terpoly- mer for self-assembly into patchy polymersomes. ACS Macro Lett. 2016;5(3):351–4. 51. Magnoni S, Brody DL. New perspectives on amyloid-beta dynamics after acute brain injury: moving between experimental approaches and studies in the human brain. Arch Neurol. 2010;67(9):1068–73. 52. Smith C. Review: the long-term consequences of microglial activation following acute traumatic brain injury. Neuropathol Appl Neurobiol. 2013;39(1):35–44. 53. Simon DW, et al. The far-reaching scope of neuroinflammation after traumatic brain injury. Nat Rev Neurol. 2017;13(3):171–91. 54. Johnson ECB, et al. Large-scale deep multi-layer analysis of Alzheimer’s disease brain reveals strong proteomic disease-related changes not observed at the RNA level. Nat Neurosci. 2022;25(2):213–25. 55. Schnöder L, et al. Neuronal deficiency of p38α-MAPK ameliorates symp- toms and pathology of APP or Tau-transgenic Alzheimer’s mouse models. Faseb j. 2020;34(7):9628–49. 56. Sankar SB, et al. Low cerebral blood flow is a non-invasive biomarker of neuroinflammation after repetitive mild traumatic brain injury. Neurobiol Dis. 2019;124:544–54. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s40249-021-00834-3	Dong et al. Infect Dis Poverty (2021) 10:46 https://doi.org/10.1186/s40249-021-00834-3 RESEARCH ARTICLE Open Access Fluorescence polarization assay improves the rapid detection of human brucellosis in China Shuai‑Bing Dong1,2, Di Xiao1, Jing‑Yao Liu3, Hui‑Mei Bi3, Zun‑Rong Zheng3, Li‑Da Wang3, Xiao‑Wen Yang1, Guo‑Zhong Tian1, Hong‑Yan Zhao1, Dong‑Ri Piao1, Zhi‑Feng Xing4 and Hai Jiang1* Abstract Background: Brucellosis is an infectious‑allergic zoonotic disease caused by bacteria of the genus Brucella. Early diagnosis is the key to preventing, treating, and controlling brucellosis. Fluorescence polarization immunoassay (FPA) is a new immunoassay for relatively rapid and accurate detection of antibodies or antigens based on antigen–anti‑ body interaction. However, there is no report on FPA‑based detection of human brucellosis in China. Therefore, this study is to evaluate the value of FPA for the diagnosis of human brucellosis in China. Methods: We recruited 320 suspected brucellosis cases who had the clinical symptoms and epidemiological risk fac‑ tors between January and December, 2019. According to China Guideline for Human Brucellosis Diagnosis, the Rose Bengal test (RBT) was used for the screening test, and the serum agglutination test (SAT) was used as the confirma‑ tory test. Brucellosis was confirmed only if the results of both tests were positive. Additionally, FPA and enzyme linked immune sorbent assay (ELISA) were compared with SAT, and their sensitivity, specificity, coincidence rate and consist‑ ency coefficient (Kappa value) as diagnostic tests were analyzed individually and in combination. The optimal cut‑off value of FPA was also determined using the receiver operator characteristic (ROC) curve. Results: The optimum cut‑off value of FPA was determined to be 88.5 millipolarization (mP) units, with a sensitivity of 94.5% and specificity of 100.0%. Additionally, the coincidence rate with the SAT test was 96.6%, and the Kappa value (0.9) showed excellent consistency. The sensitivity and specificity of FPA and ELISA combined were higher at 98.0% and 100.0% respectively. Conclusions: When the cut‑off value of FPA test is set at 88.5 mP, it has high value for the diagnosis of brucellosis. Additionally, when FPA and ELISA are combined, the sensitivity of diagnosis is significantly improved. Thus, FPA may have potential in the future as a diagnostic method for human brucellosis in China. Keywords: Human brucellosis, Fluorescence polarization assay, Diagnosis *Correspondence: jianghai@icdc.cn 1 State Key Laboratory for Infectious Disease Prevention and Control, Collaborative Innovation Center for Diagnosis and Treatment of Infectious Diseases, National Institute for Communicable Disease Control and Prevention, Chinese Center for Disease Control and Prevention, Beijing, China Full list of author information is available at the end of the article Background Brucellosis is an infectious-allergic zoonotic disease caused by bacteria of the genus Brucella [1]. The dis- ease is transmitted to humans mainly by contact with infected animals and the ingestion of infected meat or unpasteurized dairy product [2]. Currently, more than 170 countries have reported human cases of brucellosis, and approximately 500 000 new cases are reported each © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Dong et al. Infect Dis Poverty (2021) 10:46 Page 2 of 6 year [3]. Despite this, brucellosis is a highly neglected zoonotic disease, according to the World Health Organi- zation [4]. Brucellosis is especially prevalent in several low-income and middle-low-income countries [5]. In China, brucellosis has been recognized as an epidemic on account of its high incidence and wide spread since the mid-1990s, and it is an important public health problem in the country [6]. Early diagnosis is the key to preventing, treating, and controlling brucellosis. At present, there are many tech- niques for the detection of human brucellosis, but bac- terial isolation and culture is still the gold standard [7]. However, this method has a low clinical isolation rate and is time consuming. Serological diagnosis of human bru- cellosis is usually easier and faster than bacterial isolation and culture, so serological diagnosis methods are widely used [8]. The Rose Bengal plate agglutination test (RBT) is a screening test, but its efficiency is greatly affected by the test conditions. Therefore, in China, an antiglobulin test (Coomb’s test) and serum agglutination test (SAT) are used as confirmatory tests, along with RBT for the diagnosis of brucellosis [9–11]. However, the procedures for these confirmatory tests are a bit complicated and time consuming, and the interpretation of the results is easily affected by subjective factors as, occasionally, false negatives occur due to the prozone phenomenon [12]. Based on the current situation, there is a clear need for more reliable tests for the diagnosis of brucellosis. Fluorescence polarization immunoassay (FPA) is a new immunoassay for relatively rapid and accurate detec- tion of antibodies or antigens based on antigen–anti- body interaction [13]. FPA meets the standards of the World Organization for Animal Health, and, therefore, it has been adopted as a laboratory testing method for animal brucellosis. The advantages of FPA are that the reaction time is only 5 min and it can be used for both individual detection and large-scale field screening [14]. Further, unlike the conventional tests, data are obtained electronically. Therefore, any subjectivity is eliminated, and instead, rapid analysis, a permanent record, and easy data dispersal are possible. Some studies have reported that FPA is widely used to detect Brucella spp. antibody in the serum, whole blood, and milk of cattle [15], sheep [16], pigs [17], deer [18], camel [19], and other animals. There are also a few reports on the detection of human brucellosis with FPA [14, 20]. However, there is no report on FPA-based detection of human brucellosis in China. Therefore, in order to explore the possibility of applying FPA for human brucellosis diagnosis in China, in this paper, FPA was evaluated for its efficiency and compared with a variety of standard laboratory detection methods, including RBT, SAT, and enzyme linked immune sorbent assay (ELISA). Materials and methods Serum samples This study included 320 patients with suspected brucel- losis, who had the clinical symptoms of the disease and epidemiological risk factors. These patients were admit- ted to the Heilongjiang Provincial General Administra- tion of Agriculture and Reclamation General Hospital between January 1, 2019, and December 31, 2019. Fasting venous blood (4 ml) was collected for brucellosis serolog- ical testing, and the diagnosis of brucellosis was based on the Diagnostic Criteria for Brucellosis WS269-2019 [21]. Suspected cases of brucellosis were defined as people with clinical symptoms [fever (≥ 37.5 ℃), fatigue, night sweats, and joint pain] and epidemiologic risk factors for infection. Confirmed cases were defined as suspected cases with an antibody titer of ≥ 1:100 (+ +) in SAT or positive Brucella isolate. If the antibody titer for SAT is 1:50, Coomb’s test is generally used as an additional con- firmatory test. However, culture and isolation can be only performed at a few provincial-level laboratories and the National Brucellosis Laboratory in Beijing. All subjects provided informed consent to participate in the study. Instruments The following instruments were used: fluorescence pola- rimeter (FLUPO®, Peace River research Institute, Hei- longjiang Province), fluorescence polarimetry test tube antibody detection kit (Peace River ®, 921,021, Peace River research Institute, Heilongjiang Province), ELISA antibody detection kit for brucellosis (Peace River®, 650,112, Peace River Research Institute, Heilongjiang Province), and RBT antigen and SAT antigen detection kits (BLSH-01 and BLSS-02, National Institute for Com- municable Disease Control and Prevention, Chinese Center for Disease Control and Prevention). Detection methods SAT, RBT, ELISA and Coombs were performed, and the data interpreted, according to the Diagnostic Criteria for Brucellosis WS269-2019 in China [21]. ELISA is used as a quantitative screening test for the diagnosis of brucel- losis. If the serum OD ratio/positive control OD value is ≥ 24%, the patient is considered to be positive, and if it is < 24%, the patient is considered to be negative. In FPA, the titer of antibody bound to the antigen directly is determined with the help of a fluorescent dye attached to a small antigen fragment, which is excited by plane polarized light of a specific wavelength. In the absence of an antibody, the molecular size of the antigen remains unchanged, and therefore, the rate of rotation and the extent of light polarization remains constant. On the other hand, when an antigen–antibody complex is Dong et al. Infect Dis Poverty (2021) 10:46 Page 3 of 6 formed, the molecular size increases. As a result, the rate of rotation is reduced and the extent of light polarization is high. This change can be measured by a fluorescence polarization analyzer, and the result is expressed in mil- lipolarization (mP) units. According to the fluorescence polarization test tube antibody detection kit for brucel- losis, the test result is negative when the FPA value is ≤ 72 mP; a value ≥ 93 mP indicates a positive result and a value between 72 and 93 mP indicates that there is a sus- picion of brucellosis. Data analysis Microsoft Excel software 2010 (Microsoft Office, CA, USA) and IBM SPSS Statistics 22.0 (IBM Corp; Armonk NY, USA) were used for data analysis. SAT was used as the confirmatory test. ELISA and FPA (individually and in combination) were compared with SAT, based on their sensitivity, specificity, coincidence rate, and consistency coefficient (Kappa value). A Kappa value ≤ 0.4 indicates poor consistency; 0.4 < Kappa < 0.75, medium and high consistency; and Kappa ≥ 0.75, excellent consistency. The optimal cut-off value of FPA was also determined using the receiver operator characteristic (ROC) curve. Results Demographic data and grouping Of the 320 cases of suspected brucellosis, the results of both RBT and SAT were positive in 200 cases, which formed the brucellosis group. This group of 200 patients included 149 males and 51 females (mean age, 45.5 ± 13.4 years). The remaining 120 patients had nega- tive results on both the RBT and SAT tests and served as the control group. This group included 70 males and 50 females, with a mean age of 39.8 ± 16.6 years (Table 1). Table 1 Demographic features of brucellosis group and control group Demographic feature Brucellosis group n (%) Control group n (%) Sex Male Female Age group (years) < 20 20–30 31–40 41–50 51–60 > 60 Age (years), mean ± SD Standard deviation SD 149 (74.5%) 51 (25.5%) 4 (2.0%) 26 (13.0%) 33 (16.5%) 64 (32.0%) 46 (23.0%) 27 (13.5%) 45.5 ± 13.4 70 (58.3%) 50 (41.7%) 7 (5.9%) 40 (33.3%) 18 (15.0%) 16 (13.3%) 24 (20.0%) 15 (12.5%) 39.8 ± 16.6 FPA results The FPA results showed that 180 patients were positive, 75 patients were suspicious for brucellosis, and 65 were negative. Among the 75 cases of suspected brucellosis, 20 were from the brucellosis group and 55 were from the control group (Table 2). The maximum FPA value of the negative group was 88 mP; the minimum FPA value of the positive group was 80 mP; the FPA values were between 80 and 88 mP in 20 out of 120 cases (16.7%) of the control group, and in 11 out of 200 cases (5.5%) in the brucellosis group (Fig. 1). Optimal cut‑off value for FPA At present, the FPA test is not included in the diagnos- tic criteria for brucellosis in China. Therefore, ROC curve analysis of the FPA test results was used to deter- mine the optimal cut-off value. The results showed that the area under the curve was 0.997 [95% confidence interval (CI): 0.994–1.000, standard error: 0.002]. The optimal cut-off was determined as 88.5 mP, because it provided the maximum sum of sensitivity and Table 2 Results of SAT and FPA summarized according to the reagent reference standard SAT + − Total FPA + 180 0 180 Suspicious 20 55 75 − 0 65 65 SAT Serum agglutination test, FPA Fluorescence polarization immunoassay 250 200 A 150F P 100 50 Nega(cid:24)ve Posi(cid:24)ve Result Total 200 120 320 n1 n2 n3 n4 Fig. 1 Interactive dot diagram of fluorescence polarization immunoassay (FPA) results for 200 confirmed (positive) cases of brucellosis and 120 control (negative) cases. n1 standard upper limit for reagents). n2 maximum). n3 (reference standard lower limit for reagents) 88 mP (negative group 72 mP 80 mP (positive group minimum). n4 = = = = 93 mP (reference Dong et al. Infect Dis Poverty (2021) 10:46 Page 4 of 6 1.0 0.8 y t i v (cid:19) i s n e S 0.6 0.4 0.2 0.0 0.0 0.2 0.4 0.6 1−speciﬁcity 0.8 1.0 Fig. 2 Receiver operator characteristic (ROC) analysis of sensitivity (%) plotted against 1 cut‑off value of fluorescence polarization immunoassay (FPA) for the detection of antibodies against Brucella spp specificity (%) to determine the optimal − Table 3 Serological results of SAT for patients with an antibody titer of 1:50 (n 10) = Case RBT SAT ( + ) + Coomb’s ( ) + + ELISA (%) 12 25 52 72 78 104 105 121 133 199 + + + + + + + + + + 50 50 50 50 50 50 50 50 50 50 200 200 200 200 200 200 200 200 200 200 17 24 23 71 18 22 15 13 20 42 FPA 85 128 109 144 90 142 113 96 98 145 SAT Serum agglutination test, RBT Rose bengal test, ELISA Enzyme linked immunesorbent assay, FPA Fluorescence polarization immunoassay, Antibody titer Positive; + + + specificity (194.5), with the individual sensitivity and specificity values being 94.5% and 100.0%, respectively (Fig. 2). In 10 patients who were positive according to RBT and had a titer of 1:50 according to SAT, Coomb’s test was used to confirm the diagnosis of brucellosis. When the cut-off of 88.5 mP for the FPA test was applied in this group of patients, only one of the ten patients was found to be negative (as indicated by an FPA value less than 88.5 mP) (Table 3). Thus, there was only one false- negative result and no false-positive results with FPA. Comparison and combination of FPA and ELISA For the FPA test results, based on the determined cut-off value, an FPA value of ≥ 88.5 mP was considered to be positive, and a value of < 88.5 mP was considered to be negative. With SAT as the reference test, the sensitivity, specificity, and coincidence rate of FPA, ELISA, and FPA combined with ELISA were analyzed. The sensitivity, coincidence rate, negative predictive value, and Kappa value of the FPA test were higher than those of the ELISA test. The sensitivity and specificity of the FPA and ELISA tests combined were as high as 98.0 and 100.0%, respec- tively (Table 4). Discussion In this study, we have assessed the efficiency of FPA for the diagnosis of human brucellosis in China. FPA was compared to other tests and also combined with ELISA to determine its efficiency. Additionally, the optimal cut- off value of FPA for this population was also determined. In the present study, the ROC curve analysis for the FPA results showed that when the cut-off value is 88.5 mP, the sensitivity and specificity of the FPA test are at an optimum, at 94.5% and 100.0%, respectively. Cut-off values for FPA have also been reported by Konstantinidis in a Greek population [20]. The optimum cut-off value reported was 99 mP, and the sensitivity and specificity were 93.5% (95% CI: 89.5–96.3) and 96.1% (95% CI: 93.2– 97.9) respectively. In a similar study in Argentina, Lucero [14] reported an optimal cut-off value of 72 mP, with a sensitivity and specificity of 96.1% and 97.9% respectively. The difference in sensitivity across these studies and the present one might be related to differences in the study populations tested. The dada from Dr. Lucero showed that the positive cases were culture-proven population, while in this study, the positive of both RBT and SAT tests were used for confirmed cases of the human brucel- losis, according to the Diagnostic Criteria for Brucellosis WS269-2019 in China [21]. In the present study, FPA was found to have excellent consistency (Kappa value = 0.93), and the coincidence rate with the SAT test was 96.6%. However, there were 11 false-negative cases. This might have been caused by poor affinity of the antigen with the antibody or a low titer of serum antibodies [20]. If the cut-off value is decreased, the sensitivity may increase, but the specific- ity will certainly decrease. Alternatively, FPA could be combined with ELISA, as the sensitivity of brucellosis diagnosis was improved with FPA and ELISA combined, according to the present findings. When FPA and ELISA were compared in this study, the sensitivity, coincidence rate, negative predictive value, and Kappa value of FPA were found to be higher Dong et al. Infect Dis Poverty (2021) 10:46 Page 5 of 6 l a t o T 0 0 2 0 2 1 − 4 0 2 1 I A S L E h t i i w d e n b m o c A P F + 6 9 1 0 − 9 3 0 2 1 A S L E I + 1 6 1 0 − 1 1 0 2 1 I A S L E d n a , A P F , T A S f o s t l u s e r e h t . 0 8 9 . 0 0 0 1 . 8 8 9 . 0 0 0 1 . 8 6 9 0 1 . . 5 0 8 . 0 0 0 1 . 8 7 8 . 0 0 0 1 . 5 5 7 8 0 . . 5 4 9 . 0 0 0 1 . 6 6 9 . 0 0 0 1 . 6 1 9 9 0 . A P F + 9 8 1 0 ) f o s i s y a n A l 4 e l b a T T A S + e t a r e c n e d c n o C i i ) % ( y t i c fi c e p S i ) % ( ) % ( V P P ) % ( V P N a p p a K % ( y t i v i t i s n e S − than those of ELISA. Studies have shown that ELISA has a sensitivity of 83.3% for IgM and 41.7% for IgG, while the combined specificity for IgG and IgM is 92.3% [22]. Therefore, the present comparison might have been affected by the disease course of the patient. Nonethe- less, there was no false-positive result with FPA. The lack of false-positive results indicates that there was no cross-reaction of antibodies produced by bacteria with structurally similar antigens (such as Yersinia enterocol- itis O:9, Escherichia coli O:157, Salmonella serotypes of Kaufmann-White group N, and S. maltophilia) [23]. The finding is in basically agreement with the report of Nielsen about cross-reaction [13]. Therefore, given that Coomb’s test is complicated and time consuming, the FPA test could potentially replace Coomb’s test based on the present findings. Limitation of this study is that bacterial culture was not performed for confirmation of the results, even though isolation of Brucella spp. from blood, tissue or bone mar- row cultures is known to be the only means of definitively diagnosing brucellosis [21]. Despite this, at the cut-off value of 88.5 mP, FPA has high sensitivity and specific- ity. Additionally, FPA is rapid, convenient, and reliable as a quantitative test. The results also show that when FPA is combined with ELISA, the sensitivity of brucellosis diagnosis can be significantly improved. In the future, the reproducibility of FPA test should be determined, so that it can be used for screening and confirmation of brucel- losis in China. Conclusions FPA is a new immunoassay for relatively rapid and accurate detection of antibodies or antigens. When the cut-off value of FPA test is set at 88.5 mP, it has high value for the diagnosis of brucellosis. Additionally, when FPA and ELISA are combined, the sensitivity of diagnosis is significantly improved. Thus, FPA may have potential in the future as a diagnostic method for human brucellosis in China. Abbreviations CI: Confidence interval; ELISA: Enzyme linked immune sorbent assay; FPA: Fluorescence polarization immunoassay; mP: Millipolarization; NPV: Negative predictive value; PPV: Positive predictive value; RBT: Rose bengal test; ROC: Receiver operator characteristic; SAT: Serum agglutination test; SD: Standard deviation. Acknowledgements Not applicable. Authors’ contributions S‑BD and HJ designed and supervised this study; S‑BD wrote the manuscript; DX, J‑YL, H‑MB, Z‑RZ, L‑DW, X‑WY, G‑ZT, H‑YZ, D‑RP and Z‑FX prepared and cleaned the data; all the authors interpreted the data and critically revised the e v i t a g e N − , e v i t i s o P + , y a s s a o n u m m i l n o i t a z i r a o p e c n e c s e r o u F A P F l , y a s s a t n e b r o s e n u m m i d e k n i l e m y z n E A S I L E l , t s e t n o i t a n i t u g g a m u r e S T A S , l e u a v e v i t c i d e r p e v i t a g e n V P N , l e u a v e v i t c i d e r p e v i t i s o p V P P Dong et al. Infect Dis Poverty (2021) 10:46 Page 6 of 6 manuscript for important intellectual content. All authors read and approved the final manuscript. Funding This study was supported by National Key R&D Program of China(Grant No.2020YFA0907101) and Major Infectious Diseases such as AIDS and Viral Hepatitis Prevention and Control Technology Major Projects ( 2018ZX10201002). Availability of data and materials All original (de‑identified) data and materials are available upon request from the corresponding author. Declarations Ethical approval and consent to participate The ethics committee of the hospital approved the study. Written informed consent has been obtained from the patients in accordance with the Declaration of Helsinki. We confirmed that the identification information of all participants (including patient names, ID numbers, home addresses and telephone numbers) would not be included in recordings, written descrip‑ tions, or publications. Consent for publication Not applicable. Competing interests All authors declare that they have no competing interest. Author details 1 State Key Laboratory for Infectious Disease Prevention and Control, Collabo‑ rative Innovation Center for Diagnosis and Treatment of Infectious Diseases, National Institute for Communicable Disease Control and Prevention, Chinese Center for Disease Control and Prevention, Beijing, China. 2 Beijing Center for Disease Prevention and Control, Beijing Research Center for Preventive Medicine, Beijing, China. 3 General Hospital of Heilongjiang Province Land Rec‑ lamation Bureau, Harbin, China. 4 Heilongjiang Provincial Centre for Disease Control and Prevention, Harbin, China. Received: 13 November 2020 Accepted: 24 March 2021 References 1. Franco MP, Mulder M, Gilman RH, Smits HL. Human brucellosis. Lancet Infect Dis. 2007;7(12):775–86. 2. Dong SB, Jiang H, Wang LP. Progress in research and practice of brucello‑ sis surveillance in China. Chin J Epidemiol. 2019;40(7):870–4. (in Chinese). 3. Pappas G, Papadimitriou P, Akritidis N, Christou L, Tsianos EV. The new global map of human brucellosis. Lancet Infect Dis. 2006;6(2):91–9. 4. Mableson HE, Okello A, Picozzi K, Welburn SC. Neglected zoonotic diseases‑the long and winding road to advocacy. PLoS Negl Trop Dis. 2014;8(6):e2800. 5. Rubach MP, Halliday JE, Cleaveland S, Crump JA. Brucellosis in 6. low‑income and middle‑income countries. Curr Opin Infect Dis. 2013;26(5):404–12. Lai S, Zhou H, Xiong W, Gilbert M, Huang Z, Yu J, et al. Changing epi‑ demiology of human brucellosis, China, 1955–2014. Emerg Infect Dis. 2017;23(2):184–94. 7. Matope G, Muma JB, Toft N, Gori E, Lund A, Nielsen K, et al. Evaluation of sensitivity and specificity of RBT, c‑ELISA and fluorescence polarisation assay for diagnosis of brucellosis in cattle using latent class analysis. Vet Immunol Immunopathol. 2011;141(1):58–63. 8. Patra S, Tellapragada C, Vandana KE, Mukhopadhyay C. Diagnostic utility of in‑house loop‑mediated isothermal amplification and real‑time PCR targeting virB gene for direct detection of Brucella melitensis from clinical specimens. J Appl Microbiol. 2019;27(1):230–6. 9. Xiao Y, Zou G, Yin J, Tan W, Zhou J, Zhang H. Seroepidemiology of human Brucella infection in Yixing. China Trop Doct. 2017;47(2):165–7. 10. Corbel MJ. The relationship between the protective and cross‑reacting antigens of Brucella spp., Yersinia enterocolitica 0:9 and Salmonella serotypes of Kauffmann‑White group N. Contrib Microbiol Immunol. 1979;5(8):50–63. 11. Memish ZA, Almuneef M, Mah MW, Qassem LA, Osoba AO. Comparison of the Brucella Standard Agglutination Test with the ELISA IgG and IgM in patients with Brucella bacteremia. Diagn Microbiol Infect Dis. 2002;44(2):129–32. 12. Nielsen K, Yu WL. Serological diagnosis of brucellosis. Prilozi. 2010;31(1):65–89. 13. Nielsen K, Lin M, Gall D, Jolley M. Fluorescence polarization immunoassay: detection of antibody to Brucella abortus. Methods. 2000;22(1):71–6. 14. Lucero NE, Escobar GI, Ayala SM, Paulo PS, Nielsen K. Fluorescence polarization assay for diagnosis of human brucellosis. J Med Microbiol. 2003;52(10):883–7. 15. Dajer A, Luna‑Martínez E, Zapata D, Villegas S, Gutiérrez E, Peña G, et al. Evaluation of a fluorescence‑polarization assay for the diagnosis of bovine brucellosis in México. Prev Vet Med. 1999;40(1):67–73. 16. Minas A, Stournara A, Minas M, Papaioannou A, Krikelis V, Tselepidis S. Validation of fluorescence polarization assay (FPA) and comparison with other tests used for diagnosis of B. melitensis infection in sheep. Vet Microbiol. 2005;111(3–4):211–21. 17. Nielsen K, Gall D, Smith P, Vigliocco A, Perez B, Samartino L, et al. Validation of the fluorescence polarization assay as a serological test for the presumptive diagnosis of porcine brucellosis. Vet Microbiol. 1999;68(3–4):245–53. 18. Gall D, Nielsen K, Forbes L, Cook W, Leclair D, Balsevicius S, et al. Evalu‑ ation of the fluorescence polarization assay and comparison to other serological assays for detection of brucellosis in cervids. J Wildl Dis. 2001;37(1):110–8. 19. Gwida MM, ElGohary AH, Melzer F, Tomaso H, Rösler U, Wernery U, et al. Comparison of diagnostic tests for the detection of Brucella spp. in camel sera. BMC Res Notes. 2011;4:525. 20. Konstantinidis A, Minas A, Pournaras S, Kansouzidou A, Papastergiou P, Maniatis A, et al. Evaluation and comparison of fluorescence polarization assay with three of the currently used serological tests in diagnosis of human brucellosis. Eur J Clin Microbiol Infect Dis. 2007;26(10):715–21. 21. National Health Commission of the People’s Republic of China. The Notice of the WS 269–2019 Diagnosis for Brucellosis. http:// www. nhc. gov. cn/ wjw/ s9491/ 201905/ b109b 71e7a 62425 6985b 57394 4b5d2 92. shtml. Accessed 15 Sept 2019. (in Chinese) 22. Avijgan M, Rostamnezhad M, Jahanbani‑Ardakani H. Clinical and serologi‑ cal approach to patients with brucellosis: a common diagnostic dilemma and a worldwide perspective. Microb Pathog. 2019;129:125–30. 23. Friedland JS. Topley & Wilson’s microbiology and microbial infections. Trans R Soc Trop Med Hyg. 1998;92(4):470.
10.1186_s13018-021-02887-4	Dong et al. Journal of Orthopaedic Surgery and Research (2022) 17:5 https://doi.org/10.1186/s13018-021-02887-4 RESEARCH ARTICLE Open Access miR-653-5p suppresses the viability and migration of fibroblast-like synoviocytes by targeting FGF2 and inactivation of the Wnt/ beta-catenin pathway Peilong Dong, Xiaobo Tang, Jian Wang, Botao Zhu and Zhiyun Li* Abstract Background: Rheumatoid arthritis (RA) is a chronic systemic autoimmune disease. Several studies reported that fibroblast-like synoviocytes (FLSs) and miRNAs are associated with RA pathogenesis. This study explored the function of miR-653-5p in the regulation of human fibroblast-like synoviocytes-rheumatoid arthritis (HFLS-RA) cells. Methods: The mRNA and protein levels of genes were measured by RT-qPCR and western blot, respectively. MTT, wound healing, and invasion assays were used to evaluate the viability and metastasis of FLSs. Luciferase reporter and RNA pull-down assays were employed to determine the interaction between miR-653-5p and FGF2. Results: RT-qPCR results demonstrated that miR-653-5p expression was decreased and FGF2 level was increased in synovial tissues and FLSs of RA. Moreover, the viability and metastasis of FLSs were accelerated by miR-653-5p addi- tion, which was restrained by miR-653-5p suppression. Furthermore, we demonstrated that levels of Rac1, Cdc42, and RhoA were decreased after miR-653-5p addition. Besides, luciferase reporter and RNA pull-down assays implied that miR-653-5p targeted the 3′-UTR of FGF2. Functional assays showed that FGF2 overexpression neutralized the suppres- sive effects of miR-653-5p addition on HFLS-RA cell viability, metastasis, and the levels of Rho family proteins. Mean- while, the levels of β-catenin, cyclin D1, and c-myc were declined by miR-653-5p supplementation, but enhanced by FGF2 addition. Conclusion: In sum, we manifested that miR-653-5p restrained HFLS-RA cell viability and metastasis via targeting FGF2 and repressing the Wnt/beta-Catenin pathway. Keywords: miR-653-5p, FGF2, Rheumatoid arthritis, Fibroblast-like synoviocytes Introduction Rheumatoid arthritis (RA) is a heterogeneous and sys- temic autoimmune disease characterized by synovial cell inflammation and subsequent damage primarily to joint structures [1]. Fibroblast-like synovial cells (FLSs) are a key component in RA development [2, 3]. HFLS-RA cells Correspondence: li_zhiyun000@163.com Department of Orthopedics, Affiliated Jianhu Hospital of Nantong University, No. 666 Nanhuan Road, Jianhu, Yancheng 224700, Jiangsu, People’s Republic of China present a series of invasive features, including enhanced proliferation, increased aggressiveness, and production of inflammatory mediators [4]. Previous studies reported that the lifetime risk of RA is 3.6% for women and 1.7% for men [5]. The current main clinical treatment strat- egy of RA is drug therapy, including immunosuppressive drugs and biologics. However, resistance to these thera- pies increases the risk of infectious diseases and cancer [6, 7]. Thus, it is necessary to look for new strategies for RA treatment. © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Dong et al. Journal of Orthopaedic Surgery and Research (2022) 17:5 Page 2 of 8 MiRNAs are small non-coding RNA molecules (~ 20 nt) that bind to 3’-UTR of mRNA and modulate gene level at the post-transcriptional level, which have essen- tial functions in musculoskeletal conditions, such as osteoarthritis and tendon injuries [8–11]. Moreover, increasing reports demonstrated that miRNAs play cru- cial roles in RA development [12]. For instance, miR- 140-3p restrained the cell viability and induced apoptosis of synovial fibroblasts in RA via modulating sirtuin 3 [13]. Knockdown of PVT1 repressed the viability and inflammation of HFLS-RA cells through targeting miR- 145-5p [14]. In a recent study, it was shown that upregu- lated miR-653-5p impeded the viability and promoted the apoptosis and inflammatory response of CHON-001 cells in osteoarthritis [15]. However, the role of miR- 653-5p in RA progression is unclear. Fibroblast growth factor 2 (FGF2), a member of the fibroblast growth factor (FGF) family, is implicated in multiple biological processes, such as cell proliferation, differentiation, and cell growth [16, 17]. A recent study reported that FGF2 was upregulated in chondrogenic differentiation and miR-23c repressed marrow stromal cell differentiation to chondrocytes through modulat- ing FGF2 expression [17]. miR-105/Runx2 axis mediates FGF2-induced ADAMTS expression in osteoarthritis cartilage [18]. miR-16 modulated MgCl2-induced accel- eration of osteogenic differentiation through regulating FGF2-mediated ERK/MAPK pathway activation [19]. Nonetheless, the mechanism of FGF2 in RA development remains elusive. This study exhibited aberrant levels of miR-653-5p and FGF2 in RA patients and investigated the influence of miR-653-5p on RA development. Materials and methods Samples Thirty-two synovial tissues were collected from RA patients (15 males, 17 females, 43–74 years old) after knee replacement surgeries at Affiliated Jianhu Hospital of Nantong University. Normal synovial biopsies from 32 patients with traumatic knee injuries served as healthy controls (18 males, 14 females, 42–72 years old). RA was diagnosed according to the previous reference standards [20]. The study was permitted by the Ethics Committee of Affiliated Jianhu Hospital of Nantong University, and written consent was gained from all patients. negative controls (NC mimics/inhibitor), shFGF2, shNC, pcDNA3.1/FGF2, and pcDNA3.1 were generated by GenePhama (Shanghai, China). The plasmid vectors were transfected using Lipofectamine 2000 (Thermo Fisher Scientific). RT‑qPCR Total RNA was extracted from synovial tissues and HFLS-RA cells using Trizol reagent kits (Invitrogen). Then, 1 μg of total RNA was reverse transcribed to cDNA using the Revert Aid™ First Strand cDNA Synthesis kit (Takara) at 37˚C for 15 min. RT-qPCR was performed by SYBR Premix Ex Taq II (TaKaRa) on ABI 7500 real-time PCR system (Applied Biosystems). GAPDH (for mRNA) and U6 (for miR-653-5p) were used as an internal refer- ence. Gene level was quantified by 2−ΔΔCT method. MTT assay MTT was conducted to assess HFLS-RA cell viability. In brief, 200 μl HFLS-RA cells (6 × 103 cells/well) were seeded into 96-well plates and incubated with 10 μl MTT solution (Sigma) for 4 h at 37 °C. Then, the medium was removed and 150 μl DMSO was added to each well. Finally, the absorbance was determined at 490 nm using a spectrophotometric plate reader. Wound healing assay Cell migration ability was performed using wound heal- ing assay. HFLS-RA cells were seeded in 6-well plates and grown to full confluence. A 200 μl pipette tip was applied to generate artificial scratches. The wounded areas were observed and imaged by a microscope (Nikon, Japan). Transwell assays HFLS-RA cell invasion was assessed using 8-μm-pore transwell chambers (BD Biosciences). HFLS-RA cells were seeded on the Matrigel chambers pre-coated with Matrigel. The lower chambers were filled with DMEM medium, and the upper chambers were filled with serum- free DMEM. After incubation for 24 h, cells were invaded to the lower chambers and stained with 0.1% crystal violet. Then, the cells were counted with a microscope (Olympus Corporation). Luciferase reporter assay Cell culture and transfection HFLS and HFLS-RA cells were purchased from Jennio Biotech Co., Ltd. (Guangzhou, China) and were cultured in DMEM containing 10% FBS (GBICO), 100 U/mL peni- cillin, and 100 μg/mL streptomycin in an incubator with 5% CO2 at 37 °C. miR-653-5p mimics/inhibitor and their FGF2-(wild-type) wt and its mutant (FGF2-mut) were inserted into the pGL3 luciferase reporter vector (Pro- mega,). Then, the above reporters were co-transfected miR-653-5p mimics or NC mimics into HFLS-RA cells for 48 h. The activities were measured by dual-luciferase reporter assay system (Promega). Dong et al. Journal of Orthopaedic Surgery and Research (2022) 17:5 Page 3 of 8 RNA pull‑down RNA pull-down was performed using a Magnetic RNA Pull-Down Kit (Thermo Fisher Scientific). The bioti- nylated miR-653-5p (Bio-miR-653-5p) and Bio-miR- NC were generated by RiboBio (Guangzhou, China). Then HFLS-RA cells were lysed and the protein lysates were incubated with M-280 streptavidin-coated with magnetic beads (Sigma-Aldrich) and then washed with buffer. Finally, RT-qPCRs were performed to determine gene expression. Western blot assay Cells were lysed in RIPA buffer to isolate total proteins, and BCA protein kit (Sigma-Aldrich) measured the protein concentration. Protein lysates were separated by 10% SDS–PAGE and transferred to PVDF mem- branes (Bio-Rad, USA). After blocked with 5% skimmed milk, the membrane was incubated with primary anti- bodies against Rac1, Cdc42, RhoA, and β-catenin, cyc- lin D1, c-myc, or GAPDH, and then interacted with HRP-conjugated secondary antibodies. The protein bands were visualized with an ECL detection system (Millipore, USA). Statistical analysis The data were exhibited as mean ± SD. Statistical analy- sis was conducted using SPSS 16.0 software (SPSS, IL, USA.). The student’s t test was applied for comparisons between two groups. One-way analysis of variance was carried out for comparisons among multiple groups. P < 0.05 indicated statistically significant. Results miR‑653‑5p level is decreased in synovial tissue and RA‑FLSs Initially, RT-qPCR assay was adopted to determine miR-653-5p level in synovial tissues. We uncovered that miR-653-5p level was reduced in synovial tissues of RA patients (Fig. 1A). Consistently, we also identified that miR-653-5p level was decreased in HFLS-RA cells compared with the normal HFLS cells (Fig. 1B). Hence, these results manifested that miR-653-5p might act as a vital role in RA pathogenesis. The addition of miR‑653‑5p restrains the viability, migration, and invasion in HFLS‑RAs Subsequently, we explored the function of miR-653-5p in RA. RT-qPCR results elaborated the miR-653-5p level was heightened by miR-653-5p addition and was reduced by miR-653-5p silencing (Fig. 2A). Moreover, cell viability was inhibited in HFLS-RA cells transfected with miR-653-5p mimics, whereas miR-653-5p deletion Fig. 1 miR-653-5p level is decreased in synovial tissue and RA-FLSs. A RT-qPCR was used to determine the expression levels of miR-653-5p in the synovial tissue of RA patients versus that of healthy synovial tissue. B RT-qPCR analysis showed the expression of miR-653-5p in HFLS-RA cells compared to that of healthy FLSs. P < 0.05, *P < 0.01 reversed this effect (Fig. 2B, C). Meanwhile, we found that miR-653-5p addition restrained the migration and invasion of HFLS-RA cells; however, miR-653-5p knockdown exhibited a contrary effect (Fig. 2D, E). Also, we uncovered that miR-653-5p mimics decreased Rac1, Cdc42, and RhoA levels, which was increased by miR-653-5p deletion (Fig. 2F, G). Above all, these data suggested that miR-653-5p restrained HFLS-RA cell viability, migration, and invasion. FGF2 is the direct target of miR‑653‑5p We then predicted miR-653-5p target genes with star- Base and uncovered that FGF2 was a potential tar- get of miR-653-5p. The binding sites are shown in Fig. 3A. Then, luciferase reporter assay elaborated that miR-653-5p mimics repressed the luciferase activity of FGF2-wt, but had no effect on FGF2-mut (Fig. 3B). Moreover, pull-down assay implied that FGF2 could bind with miR-653-5p (Fig. 3C). Furthermore, the FGF2 level was decreased by miR-653-5p supplementation, which was heightened by miR-653-5p inhibition (Fig. 3D). Collectively, we determined that FGF2 was a target of miR-653-5p. FGF2 deletion suppresses RA development To further verify the function of FGF2 in RA, HFLS- RA cells were transfected with shFGF2. Results elabo- rated that FGF2 was downregulated by FGF2 deficiency (Fig. 4A). Meanwhile, cell viability, migration, and inva- sion were suppressed after FGF2 depletion (Fig. 4B–D). Similarly, we uncovered that the mRNA and protein levels of Rac1, Cdc42, and RhoA were both decreased by FGF2 silence (Fig. 4E, F). As a result, FGF2 deletion diminished HFLS-RA cell viability and metastasis. Dong et al. Journal of Orthopaedic Surgery and Research (2022) 17:5 Page 4 of 8 Fig. 2 The addition of miR-653-5p restrains the viability, migration, and invasion in HFLS-RA cells. A RT-qPCR analysis showed the expression of miR-653-5p in HFLS-RA cells transfection of miR-653-5p mimics or NC mimics and miR-653-5p inhibitor or NC inhibitor. B and C MTT assay showed cell viability in HFLS-RA cells transfection of miR-653-5p mimics or NC mimics and miR-653-5p inhibitor or NC inhibitor. D and E HFLS-RA cell migration and invasion were measured using wound healing and transwell assays after transfection with miR-653-5p mimics or NC mimics and miR-653-5p inhibitor or NC inhibitor. F and G The protein and mRNA expression of Rac1, Cdc42, and RhoA in HFLS-RA cells transfection of miR-653-5p mimics or NC mimics and miR-653-5p inhibitor or NC inhibitor was detected by western blot and RT-qPCR assay. P < 0.05, P < 0.01, P < 0.001 Dong et al. Journal of Orthopaedic Surgery and Research (2022) 17:5 Page 5 of 8 Fig. 3 FGF2 is the direct target of miR-653-5p. A The complementary sequences of miR-653-5p and corresponding sequence of the 3’-UTR of FGF2 was predicted by starBase website. B Luciferase reporter assay showed luciferase activity of FGF2-WT or FGF2-MUT in HFLS-RA cells transfected with NC mimics or miR-653-5p mimics. C Association between miR-653-5p and FGF2 was determined using an RNA pull-down assay. D RT-qPCR analysis was used to evaluate the expression of FGF2 in HFLS-RA cells transfection of miR-653-5p mimics or NC mimics and miR-653-5p inhibitor or NC inhibitor. P < 0.01, P < 0.001 Fig. 4 FGF2 deletion suppresses RA development. A RT-qPCR was used to determine FGF2 expression in HFLS-RA cells transfected with shFGF2 or shNC. B–D MTT, wound healing, and transwell assays showed cell viability, migration, and invasion in HFLS-RA cells transfection of shFGF2 or shNC. E and F Western blot and RT-qPCR assays showed the protein and mRNA expression of Rac1, Cdc42, and RhoA in HFLS-RA cells transfected with shFGF2 or shNC. P < 0.01 miR‑653‑5p regulates HFLS‑RA cell viability and metastasis via targeting FGF2 To further determine whether miR-653-5p exerted its function via FGF2, HFLS-RA cells were transfected with NC mimics, miR-653-5p mimics, miR-653-5p mimics + pcDNA3.1/FGF2. Figure 5A displays that the FGF2 level was inhibited by miR-653-5p addition, while FGF2 overexpression reversed this effect. Func- tional assays revealed that upregulated FGF2 rescued the repression effects of miR-653-5p supplementation on HFLS-RA cell viability, migration, and invasion (Fig. 5B–D). Moreover, FGF2 overexpression rescued the suppressive effect on the levels of Rac1, Cdc42, and RhoA caused by miR-653-5p addition (Fig. 5E, F). In Dong et al. Journal of Orthopaedic Surgery and Research (2022) 17:5 Page 6 of 8 Fig. 5 miR-653-5p regulates HFLS-RA cell viability and metastasis via targeting FGF2. A RT-qPCR determined FGF2 expression in HFLS-RA cells transfected with NC mimics, miR-653-5p mimics, miR-653-5p mimics pcDNA3.1/FGF2. E and F cell viability, migration, and invasion in HFLS-RA cells transfection of NC mimics, miR-653-5p mimics, miR-653-5p mimics Western blot and RT-qPCR assays showed the protein and mRNA expression of Rac1, Cdc42, and RhoA in HFLS-RA cells transfected with NC mimics, miR-653-5p mimics, miR-653-5p mimics pcDNA3.1/FGF2. B–D MTT, wound healing, and transwell assays showed pcDNA3.1/FGF2. P < 0.05, P < 0.01 + + + sum, miR-653-5p modulated RA progression via target- ing FGF2. miR‑653‑5p/FGF2 restrains the Wnt/β‑catenin pathway in HFLS‑RA cells Wnt/β-catenin signaling can be mediators in RA bio- logical processes [21, 22]. Previous studies reported that FGF2 is a modulator of Wnt/β-catenin pathway. Thus, we speculated that miR-653-5p/FGF2 regulated RA progres- sion via regulating the Wnt/β-catenin pathway. Results elaborated that β-catenin, cyclin D1, and c-myc levels were reduced by miR-653-5p addition, while FGF2 over- expression reversed these effects (Fig. 6A, B). These data manifested that miR-653-5p targeted FGF2 and inhibited the Wnt/β-catenin pathway in HFLS-RA cells. Conclusion Accumulating evidence has reported that miRNAs are closely related to RA occurrence and develop- ment, such as miR-140-3p [23], miRNA-146a [24], and miR-17-5p [25]. Therefore, miRNAs were considered as a potential therapeutic target for RA. For example, it was reported that miR-6089 restrained HFLS-RA cell viability and promoted apoptosis through regulat- ing CCR4 [26]. MiR-421 accelerated the inflammatory response of FLS s in RA via targeting SPRY1 [27]. This research explored the effect of miR-653-5p on the via- bility, migration, and invasion of HFLS-RA cells. HFLS-RA cells are the main cell population involved in RA progression of synovial tissues. Previous researches exhibited that inhibiting FLS migration and Dong et al. Journal of Orthopaedic Surgery and Research (2022) 17:5 Page 7 of 8 Fig. 6 miR-653-5p/FGF2 restrains the Wnt/β-Catenin pathway in HFLS-RA cells. A and B RT-qPCR and western blot analysis showed the levels of β-catenin, cyclin D1 and c-myc in HFLS-RA cells transfected with NC mimics, miR-653-5p mimics, miR-653-5p mimics pcDNA3.1/FGF2. P < 0.05, **P < 0.01 + invasion may protect RA joint destruction [28]. Thus, regulating the migration and invasion of HFLS-RAs may be a new strategy for RA treatment. Moreover, miR-653-5p was identified to impede cell migrative and invasive ability in breast cancer [29]. Our work sug- gested that miR-653-5p restrained cell migration and invasion in HFLS-RA cells. Rho family proteins participated in regulating HFLS-RA cell viability and invasion. For instance, RhoA is considered a new target for modulating HFLS-RA cell invasion [30]. Rac1 activation facilitated proliferation and mediates IL- 17A-induced HFLS-RA cell migration [31, 32]. This led us to propose that miR-653-5p played a vital role through Rho proteins in HFLS-RA cells. This study elucidated that miR- 653-5p addition inhibited RhoA, Rac1, and Cdc42 protein levels in HFLS-RA cells. Multiple miRNAs exerted their biological functions via modulating their target molecules. miR-653-5p was iden- tified to target different genes, such as EMSY [33], and RAI14 [34], which participate in many tumor develop- ments. In this work, we determined that FGF2 was the target of miR-653-5p in HFLS-RA cells. FGF2 is corre- lated with multiple biological processes including tumor growth, apoptosis, and angiogenesis [35, 36]. Importantly, FGF2 was found to be upregulated in RA patients and its level was closely related to Larsen’s grade of bone erosion [37, 38]. Herein, we revealed that FGF2 addition alleviated the repressive influence of miR-653-5p supplementation on the viability and metastasis of HFLS-RA cells as well as on Rho family protein levels. Moreover, β-catenin, cyclin D1, and c-myc levels were decreased by miR-653-5p addition, which was increased by FGF2 addition. Taken together, miR-653-5p/FGF2 axis inhibited the Wnt/β-catenin path- way of HFLS-RA cells. study Conclusion The current that miR-653-5p restrained RA progression through targeting FGF2 and inactivation of the Wnt/β-catenin pathway, indicating that miR-653-5p may be an effective treatment target for RA. illustrated Abbreviations RA: Rheumatoid arthritis; FLSs: Fibroblast-like synovial cells; HFLS-RA: Human fibroblast-like synoviocytes-rheumatoid arthritis; FGF2: Fibroblast growth fac- tor 2; Rac1: Rac family small GTPase 1; Cdc42: Cell division cycle 42; RhoA: Ras homolog family member A. Acknowledgements Not applicable. Authors’ contributions PD and XT designed this study. PD and JW performed all the experiments. JW and BZ analyzed the data and prepared the figures. PD, XT, and ZL drafted the initial manuscript. ZL reviewed and revised the manuscript. All authors read and approved the final manuscript. Funding Not applicable. Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This study was approved by Affiliated Jianhu Hospital of Nantong University. All the patients signed written informed consent. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Dong et al. Journal of Orthopaedic Surgery and Research (2022) 17:5 Page 8 of 8 Received: 16 August 2021 Accepted: 14 December 2021 References 1. Mateen S, Zafar A, Moin S, Khan AQ, Zubair S. Understanding the role of cytokines in the pathogenesis of rheumatoid arthritis. Clin Chim Acta. 2016;455:161–71. 2. Rockel JS, Kapoor M. Autophagy: controlling cell fate in rheumatic dis- eases. Nat Rev Rheumatol. 2017;13(3):193. 3. Bartok B, Hammaker D, Firestein GS. Phosphoinositide 3-kinase delta regulates migration and invasion of synoviocytes in rheumatoid arthritis. J Immunol. 2014;192(5):2063–70. Turner JD, Filer A. The role of the synovial fibroblast in rheumatoid arthri- tis pathogenesis. Curr Opin Rheumatol. 2015;27(2):175–82. 4. 6. 5. Crowson CS, Matteson EL, Myasoedova E, Michet CJ, Ernste FC, War- rington KJ, Davis JM 3rd, Hunder GG, Therneau TM, Gabriel SE. The lifetime risk of adult-onset rheumatoid arthritis and other inflammatory autoimmune rheumatic diseases. Arthritis Rheum. 2011;63(3):633–9. Flores-Borja F, Mauri C, Ehrenstein MR. Restoring the balance: harness- ing regulatory T cells for therapy in rheumatoid arthritis. Eur J Immunol. 2008;38(4):934–7. Imperato AK, Bingham CO 3rd, Abramson SB. Overview of benefit/risk of biological agents. Clin Exp Rheumatol. 2004;22(5 Suppl 35):S108-114. 8. He L, Hannon GJ. MicroRNAs: small RNAs with a big role in gene regula- 7. tion. Nat Rev Genet. 2004;5(7):522–31. 9. Giordano L, Porta GD, Peretti GM, Maffulli N. Therapeutic potential of microRNA in tendon injuries. Br Med Bull. 2020;133(1):79–94. 10. Oliviero A, Della Porta G, Peretti GM, Maffulli N. MicroRNA in osteoarthri- tis: physiopathology, diagnosis and therapeutic challenge. Br Med Bull. 2019;130(1):137–47. 11. Gargano G, Oliviero A, Oliva F, Maffulli N. Small interfering RNAs in tendon homeostasis. Br Med Bull. 2021;138(1):58–67. 12. Filkova M, Jungel A, Gay RE, Gay S. MicroRNAs in rheumatoid arthritis: potential role in diagnosis and therapy. BioDrugs. 2012;26(3):131–41. 13. Zu B, Liu L, Wang J, Li M, Yang J. MiR-140-3p inhibits the cell viability and promotes apoptosis of synovial fibroblasts in rheumatoid arthritis through targeting sirtuin 3. J Orthop Surg Res. 2021;16(1):105. 14. Tang J, Yi S, Liu Y. Long non-coding RNA PVT1 can regulate the proliferation and inflammatory responses of rheumatoid arthritis fibroblast-like synoviocytes by targeting microRNA-145-5p. Hum Cell. 2020;33(4):1081–90. 15. Lian LP, Xi XY. Long non-coding RNA XIST protects chondrocytes ATDC5 and CHON-001 from IL-1beta-induced injury via regulating miR-653-5p/ SIRT1 axis. J Biol Regul Homeost Agents. 2020;34(2):379–91. 16. Chen F, Qi S, Zhang X, Wu J, Yang X, Wang R. miR-23a-3p suppresses cell proliferation in oral squamous cell carcinomas by targeting FGF2 and correlates with a better prognosis: miR-23a-3p inhibits OSCC growth by targeting FGF2. Pathol Res Pract. 2019;215(4):660–7. 17. Shen PF, Wang B, Qu YX, Zheng C, Xu JD, Xie ZK, Ma Y. MicroRNA-23c inhibits articular cartilage damage recovery by regulating MSCs differen- tiation to chondrocytes via reducing FGF2. Eur Rev Med Pharmacol Sci. 2019;23(3):941–8. 18. Ji Q, Xu X, Xu Y, Fan Z, Kang L, Li L, Liang Y, Guo J, Hong T, Li Z, et al. miR-105/Runx2 axis mediates FGF2-induced ADAMTS expression in osteoarthritis cartilage. J Mol Med (Berl). 2016;94(6):681–94. 19. Qi H, Liu Y, Wu L, Ni S, Sun J, Xue J, Liu Q, Ni X, Fan W. MicroRNA-16, via FGF2 regulation of the ERK/MAPK pathway, is involved in the magne- sium-promoted osteogenic differentiation of mesenchymal stem cells. Oxid Med Cell Longev. 2020;2020:3894926. 20. Aggarwal R, Rider LG, Ruperto N, Bayat N, Erman B, Feldman BM, Oddis CV, Amato AA, Chinoy H, Cooper RG, et al. 2016 American College of Rheumatology/European League Against Rheumatism criteria for mini- mal, moderate, and major clinical response in adult dermatomyositis and polymyositis: an International Myositis Assessment and Clinical Studies Group/Paediatric Rheumatology International Trials Organisation Col- laborative Initiative. Ann Rheum Dis. 2017;76(5):792–801. 21. Li GQ, Fang YX, Liu Y, Meng FR, Wu X, Zhang CW, Zhang Y, Liu D, Gao B. MALAT1-driven inhibition of Wnt signal impedes proliferation and inflammation in fibroblast-like synoviocytes through CTNNB1 promoter methylation in rheumatoid arthritis. Hum Gene Ther. 2019;30(8):1008–22. 22. Wu J, Fan W, Ma L, Geng X. miR-708-5p promotes fibroblast-like synoviocytes’ cell apoptosis and ameliorates rheumatoid arthritis by the inhibition of Wnt3a/beta-catenin pathway. Drug Des Devel Ther. 2018;12:3439–47. 23. Wangzhou K, Lai Z, Lu Z, Fu W, Liu C, Liang Z, Tan Y, Li C, Hao C. MiR- 143–3p Inhibits osteogenic differentiation of human periodontal ligament cells by targeting KLF5 and inactivating the Wnt/beta-catenin pathway. Front Physiol. 2020;11:606967. 24. Tavasolian F, Hosseini AZ, Soudi S, Naderi M. miRNA-146a improves immunomodulatory effects of MSC-derived exosomes in rheumatoid arthritis. Curr Gene Ther. 2020;20(4):297–312. 25. Najm A, Masson FM, Preuss P, Georges S, Ory B, Quillard T, Sood S, Good- year CS, Veale DJ, Fearon U, et al. MicroRNA-17-5p reduces inflammation and bone erosions in mice with collagen-induced arthritis and directly targets the JAK/STAT pathway in rheumatoid arthritis fibroblast-like synoviocytes. Arthritis Rheumatol. 2020;72(12):2030–9. 26. Lin S, Wang S, Zhang Z, Lu Y, Yang M, Chen P, Chen L, Wang M. MiRNA- 6089 inhibits rheumatoid arthritis fibroblast-like synoviocytes prolifera- tion and induces apoptosis by targeting CCR4. Arch Physiol Biochem. 2020. https:// doi. org/ 10. 1080/ 13813 455. 2020. 17738 62. 27. Jiang F, Zhou HY, Zhou LF, Wen YH, Gai HH, Wu GM. MicroRNA-421 promotes inflammatory response of fibroblast-like synoviocytes in rheumatoid arthritis by targeting SPRY1. Eur Rev Med Pharmacol Sci. 2019;23(19):8186–93. 28. Lao M, Shi M, Zou Y, Huang M, Ye Y, Qiu Q, Xiao Y, Zeng S, Liang L, Yang X, et al. Protein inhibitor of activated STAT3 regulates migration, invasion, and activation of fibroblast-like synoviocytes in rheumatoid arthritis. J Immunol. 2016;196(2):596–606. 29. Zhang M, Wang H, Zhang X, Liu F. miR6535p suppresses the growth and migration of breast cancer cells by targeting MAPK6. Mol Med Rep. 2021;23(3):1. 30. Xiao Y, Liang L, Pan Y, Lian F, Li L, Lin H, Fu D, Fan J, Yang X, Sun L, et al. Inhibitory effects of simvastatin on migration and invasion of rheumatoid fibroblast-like synoviocytes by preventing geranylgeranylation of RhoA. Rheumatol Int. 2013;33(2):389–99. 31. Chan A, Akhtar M, Brenner M, Zheng Y, Gulko PS, Symons M. The GTPase Rac regulates the proliferation and invasion of fibroblast-like synoviocytes from rheumatoid arthritis patients. Mol Med. 2007;13(5–6):297–304. 32. Kanai Y, Dohmae N, Hirokawa N. Kinesin transports RNA: isola- tion and characterization of an RNA-transporting granule. Neuron. 2004;43(4):513–25. 33. Yan A, Chen G, Nie J. DGUOK-AS1 promotes the proliferation cervical cancer through regulating miR-653-5p/EMSY. Cancer Biol Ther 2020:1–9. 34. Liu F, Hu L, Pei Y, Zheng K, Wang W, Li S, Qiu E, Shang G, Zhang J, Zhang X. Long non-coding RNA AFAP1-AS1 accelerates the progression of mela- noma by targeting miR-653-5p/RAI14 axis. BMC Cancer. 2020;20(1):258. 35. Sun LL, Lei FR, Jiang XD, Du XL, Xiao L, Li WD, Li XQ. LncRNA GUSBP5-AS promotes EPC migration and angiogenesis and deep vein thrombosis resolution by regulating FGF2 and MMP2/9 through the miR-223-3p/ FOXO1/Akt pathway. Aging (Albany NY). 2020;12(5):4506–26. 36. Li R, Wang Y, Xu Y, He X, Li Y. Silencing the long noncoding RNA, TINCR, a molecular sponge of miR335, inhibits the malignant phe- notype of epithelial ovarian cancer via FGF2 suppression. Int J Oncol. 2019;55(5):1110–24. 37. Manabe N, Oda H, Nakamura K, Kuga Y, Uchida S, Kawaguchi H. Involve- ment of fibroblast growth factor-2 in joint destruction of rheumatoid arthritis patients. Rheumatology (Oxford). 1999;38(8):714–20. 38. Qu Z, Huang XN, Ahmadi P, Andresevic J, Planck SR, Hart CE, Rosenbaum JT. Expression of basic fibroblast growth factor in synovial tissue from patients with rheumatoid arthritis and degenerative joint disease. Lab Invest. 1995;73(3):339–46. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations.
10.1186_s12974-023-02805-x	Kelly et al. Journal of Neuroinflammation (2023) 20:124 https://doi.org/10.1186/s12974-023-02805-x Journal of Neuroinflammation RESEARCH Open Access Progressive inflammation reduces high-frequency EEG activity and cortical dendritic arborisation in late gestation fetal sheep Sharmony B. Kelly1,2, Justin M. Dean3, Valerie A. Zahra1, Ingrid Dudink1,2, Alison Thiel1, Graeme R. Polglase1,2, Suzanne L. Miller1,2, Stuart B. Hooper1,2, Laura Bennet3, Alistair J. Gunn3 and Robert Galinsky1,2* Abstract Background Antenatal infection/inflammation is associated with disturbances in neuronal connectivity, impaired cortical growth and poor neurodevelopmental outcomes. The pathophysiological substrate that underpins these changes is poorly understood. We tested the hypothesis that progressive inflammation in late gestation fetal sheep would alter cortical neuronal microstructure and neural function assessed using electroencephalogram band power analysis. Methods Fetal sheep (0.85 of gestation) were surgically instrumented for continuous electroencephalogram (EEG) recording and randomly assigned to repeated saline (control; n = 9) or LPS (0 h = 300 ng, 24 h = 600 ng, 48 h = 1200 ng; n = 8) infusions to induce inflammation. Sheep were euthanised 4 days after the first LPS infusion for assessment of inflammatory gene expression, histopathology and neuronal dendritic morphology in the somatosen- sory cortex. Results LPS infusions increased delta power between 8 and 50 h, with reduced beta power from 18 to 96 h (P < 0.05 vs. control). Basal dendritic length, numbers of dendritic terminals, dendritic arborisation and numbers of dendritic spines were reduced in LPS-exposed fetuses (P < 0.05 vs. control) within the somatosensory cortex. Numbers of micro- glia and interleukin (IL)-1β immunoreactivity were increased in LPS-exposed fetuses compared with controls (P < 0.05). There were no differences in total numbers of cortical NeuN + neurons or cortical area between the groups. Conclusions Exposure to antenatal infection/inflammation was associated with impaired dendritic arborisation, spine number and loss of high-frequency EEG activity, despite normal numbers of neurons, that may contribute to disturbed cortical development and connectivity. Keywords Neuroinflammation, Neurodevelopment, Fetal infection/inflammation, Neurons Correspondence: Robert Galinsky robert.galinsky@hudson.org.au Full list of author information is available at the end of the article © Crown 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the mate- rial. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 2 of 16 Introduction Exposure to antenatal infection/inflammation is a leading cause of brain injury and neurodevelopmental impair- ment [1]. For example, exposure to infection/inflamma- tion increases the risk of cerebral palsy by 2- to 12-fold in near-term/term infants [2, 3]. Improving our under- standing of the pathogenesis of brain injury is essential for better injury detection and development of interven- tions to improve outcomes, and reduce the socioeco- nomic burdens on the affected individuals, their families and society [4]. Perinatal infection/inflammation is common in low, middle and high-income countries [5], and is associated with grey matter abnormalities. For example, neuroimag- ing studies have shown that chorioamnionitis is indepen- dently associated with reduced sulcal depth and cortical volume in the temporal lobe, without overt grey matter injury [6, 7]. In rodents, chorioamnionitis induced with a single intrauterine injection of lipopolysaccharide (LPS) was associated with reduced dendritic arborisation in corti- cal neuronal cultures [8]. In pregnant rat dams, twelve hourly intraperitoneal (i.p.) injections of inactivated group B Streptococcus (109 CFU) between G19 and G22 (term) led to cortical thinning and neuromotor impair- ments in the offspring at postnatal day 40 [9]. Similarly, mild-to-moderate inflammation in neonatal rats using daily i.p. LPS (0.3 mg/kg) injections from P1–3 reduced dendritic arborisation and spine formation in cortical pyramidal neurons without overt neuronal loss [10]. While these preclinical data provide a compelling link between acute inflammation and impaired cortical devel- opment, the most common clinical fetal and neonatal scenario involves a progressive inflammatory response without severe fetal compromise [11]. Consistent with this, in fetal sheep live bacterial inoculation is associated with progressive systemic inflammation [12]. Similarly, recent clinical evidence strongly suggests that antenatal inflammation is a progressive process that persists into the newborn period [13] and supports the concept that sustained inflammation is associated with long-term impairments in brain development [14–17]. It is important to find ways to rapidly and non-inva- sively identify such progressive inflammation-induced injury at the bedside. Electroencephalography (EEG) is widely used to identify functional impairments in cases of preterm and term encephalopathy [18, 19]. However, the relationship between EEG and abnormal develop- ment of the neuronal microstructure is poorly under- stood. Thus, in the present study we tested the hypothesis that progressive antenatal inflammation induced by repeated, increasing-dose infusions of Gramnegative LPS would be associated with cortical inflammation and altered neuronal microstructural development. Further- more, we hypothesised that pathological changes asso- ciated with altered cortical neuronal microstructural development could be detected using EEG assessment of neural function. Materials and methods All procedures were approved by the Hudson Institute of Medical Research Animal Ethics committee and were conducted in accordance with the National Health and Medical Research Council Code of Practice for the Care and Use of Animals for Scientific Purposes (Eighth Edi- tion). The experiments are reported in accordance with the ARRIVE guidelines for reporting animal research [20]. In this study, we compared two groups of interest: (i) vehicle controls and (ii) antenatal inflammation. The key outcome measures were: (i) cortical inflammation, (ii) neuronal microstructural development, assessed using dendritic length, numbers of dendritic terminals, neuronal arborisation and numbers of dendritic spines and (iii) EEG measures of neuronal function. Seven- teen pregnant Border-Leicester ewes bearing singleton or twin fetuses underwent aseptic surgery at either 124 or 125 days of gestation. Food but not water was with- drawn approximately 18 h before surgery. Anaesthe- sia was induced by i.v injection of sodium thiopentone (20 mL) and maintained using 2–3% isoflurane in oxygen (Bomac Animal Health, New South Wales, Australia). Ewes received prophylactic antibiotics (ampicillin: 1 g i.v; Austrapen, Lennon Healthcare, St. Leonards, NSW, Australia, and engemycin: 500 mg i.v; Schering-Plough, Upper Hutt, New Zealand) immediately before surgery. Isoflurane levels, heart rate, oxygen saturation, and res- piratory rate were continuously monitored throughout surgery by trained anaesthetic staff. Fetal instrumentation A midline maternal laparotomy was performed, the fetus was exposed and partially removed from the uterus for implantation of polyvinyl catheters into the right bra- chiocephalic artery and amniotic cavity. In twin preg- nancies, only one twin was instrumented. Two pairs of electroencephalograph (EEG) electrodes (AS633- 7SSF; Cooner Wire, Chatsworth, CA, USA) were placed through burr holes onto the dura over the parasagittal parietal (somatosensory) cortex (10 and 20 mm anterior to bregma, and 10 mm lateral) and secured using surgi- cal bone wax and cyanoacrylate glue. A catheter was inserted into the left fetal axillary vein for administra- tion of post-operative antibiotics and lipopolysaccharide (LPS) or vehicle (saline). The fetus was returned to the uterus in its original orientation and all fetal leads were exteriorised through the maternal flank. A catheter was Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 3 of 16 inserted into the maternal jugular vein for administration of post-operative antibiotics and euthanasia at the end of the experimental period. At the completion of surgery, ewes received fentanyl for 3 days via a transdermal patch placed on the left hind leg (75 μg/h; Janssen Cilag, North Ryde, NSW, USA). Ewes were housed together in separate pens in a tem- perature controlled (20 ± 2 °C and relative humidity of 50 ± 10%) room with a 12-h light–dark cycle with access to food and water ad libitum. Four to five days of post- operative recovery were allowed before experiments commenced. Ewes and fetuses received daily i.v. infusions of ampicillin (800 mg, maternal i.v. and 200 mg, fetal i.v.) and engemycin (500 mg, maternal i.v.) for three consecu- tive days after surgery. Catheters were maintained patent with a continuous infusion of heparinised saline (25 IU/ mL) at a rate of (0.2 mL/h). Experimental recordings Fetal EEG was continuously recorded from 24 h prior to the first saline or LPS infusion (129 days of gestation) until the end of the experiment (134 days of gestation). The analogue fetal EEG signal was bandpass-filtered with a cut-off frequency set at 1 and 22 Hz and digitised at a sampling frequency of 400 Hz. EEG power was derived from the analogue signal, whilst spectral edge was cal- culated as the frequency below which 90% of the inten- sity was present. Relative (%) spectral power in the Δ (0–3.9 Hz), θ (4–7.9 Hz), ɑ (8–12.9 Hz), and β (13–22 Hz) frequency bands were quantified. This involved calculat- ing power spectra, by fast Fourier transform, of the EEG on sequential epochs using a 10-s Hanning window to minimise spectral leakage, as previously described [21, 22]. Experimental protocol Experiments started at 129 days of gestation (term is ~ 147 days). This study examined 0.85 gestation fetal sheep, at an age when brain development is broadly equivalent to that of a near term/term human infant [23]. Fetuses were randomly allocated to two groups: vehi- cle (saline, n = 9 [6 males, 3 females]) or LPS (Escheri- chia coli, O55:B5, MilliporeSigma, MO, USA; n = 8 [6 males, 2 females]). Fetuses received 300 ng, 600 ng, and 1200 ng infusions of LPS diluted in 2 mL of saline i.v. (infusion rate: 1 mL/min) at 0, 24 and 48 h, respectively, along with a 3 mL vehicle (saline) infusion at a rate of 0.75 mL/h starting 1 h after the saline infusion, as pre- viously described [24]. Control and LPS groups received the 3 mL vehicle infusion. Inflammation was confirmed based on increased plasma cytokine concentrations relative to baseline after LPS infusion. Serial cytokine measurements have been previously reported in Kelly et al. [24]. This model is relevant to the fetal inflamma- tory response syndrome caused by chorioamnionitis and reproduces the acute inflammatory exacerbations associ- ated with adverse neurodevelopment [25, 27, 27]. Con- trols received an equivalent volume of saline at the same infusion rate. Randomisation was stratified by cohort to control for the time of year and twin pregnancy. Fetal preductal arterial blood samples were collected every morning (0900 h) starting from 30 min before the start of the experiment until the day of post-mortem for pH, blood gases, and glucose and lactate concentrations (ABL 90 Flex Plus analyser; Radiometer, Brønshøj, Denmark). Four days after the start of infusions, sheep were euthan- ised by intravenous injection of pentobarbitone sodium (100 mg/kg, Lethabarb, Virbac, NSW, Australia). Brain collection and processing At post-mortem the right hemisphere was immersion- fixed with 10% phosphate-buffered formalin for 3 days before processing and embedding using a standard par- affin tissue preparation. Using a brain mould, the right hemisphere was cut with a blocking blade into 5-mm- thick coronal blocks. Blocks from the forebrain, approxi- mately 23 mm anterior to stereotaxic zero, with a clearly visible lateral gyrus containing the somatosensory cortex were sectioned into 8-μm-thick coronal sections using a microtome (Leica Microsystems, Victoria, Australia). Region matched tissue sections from the left hemi- sphere were rinsed in distilled water and immersion- fixed using a commercially available FD Rapid Golgi Stain Kit (FD Neurotechnologies Inc., MD, USA). Tissue containing a clearly visible lateral gyrus containing the somatosensory cortex was frozen and sectioned with a Leica VT1200S vibratome at 100 μm. The sections were mounted onto coverslips, processed for Golgi visualisa- tion, dehydrated in a graded series of alcohol solutions and cover slipped. Gene expression analysis Remaining grey matter tissue from the lateral gyrus, adjacent to the section collected for Golgi staining, was dissected, snap-frozen in liquid nitrogen and stored at − 80 °C for mRNA analysis of inflammatory genes. The tissue was homogenised and total mRNA was isolated using an RNeasy Midi Kit (QIAGEN, Venlo, Nether- lands) and reverse transcribed into single stranded cDNA (SuperScript III First-Strand Synthesis System, Invit- rogen, MA, USA). Relative mRNA expression levels of interleukin (IL)1A, IL1B, and IL6 were measured by qRT- PCR using an Applied Biosystems Quantstudio 6 Real- Time PCR system. Relative mRNA levels of the genes of interest were normalised to the 18S RNA for each sam- ple by subtracting the cycle threshold (Ct) value for 18S Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 4 of 16 from the Ct value for the gene of interest (ΔCt). mRNA levels of genes of interest were normalised using the for- mula 2 − ΔCt and the results expressed as a fold change from control. A threshold value (Ct) for each sample was measured in triplicate and a control sample containing no cDNA template was included in each run. Details of the primers used are presented in Table 1. Immunohistochemistry Slides were baked at 60 °C for 1 h then dewaxed in xylene, rehydrated in increasing concentrations of etha- nol and washed in 0.1 mol/L phosphate buffered saline (PBS). Antigen retrieval was performed in citrate buffer (pH 6) using a microwave for 15 min. Endogenous per- oxide quenching was performed by incubating slides in 0.1% H2O2 in methanol. Non-specific antigen blocking was performed using 3% normal goat serum. Sections were labelled with 1:250 rabbit anti-IL-1β (cat#: NB600- 633, Novus, CO, USA), 1:200 rabbit anti-ionised cal- cium binding adaptor molecule 1 (Iba-1, cat#: ab153696, Abcam, Cambridge, UK) 1:200 rabbit anti-glial fibrillary acidic protein (GFAP; cat#: ab68428, Abcam) 1:350 rab- bit anti-neuronal nuclei (NeuN, cat#: ab177487, Abcam), and 1:800 rabbit anti-Caspase3 (cat#: AF835, R&D Sys- tems, MS, USA), overnight at 4 °C. Sections were incu- bated in biotin conjugated IgG goat anti-rabbit (1:200; Dako, Victoria, Australia) for 3 h at room temperature before being incubated in avidin–biotin complex (Mil- liporeSigma) for 45 min at room temperature. Sections were reacted with 3,3’-diaminobenzidine tetrahydrochlo- ride (MilliporeSigma). The reaction was stopped in PBS and slides were then dehydrated in xylene followed by increasing concentrations of ethanol, mounted in dibutyl phthalate polystyrene xylene and cover slipped. ApopTag was used to detect single and double-stranded breaks in DNA associated with apoptosis [28]. Staining was carried out according to manufacturer’s instruc- tions (MilliporeSigma, s7100, ApopTag Peroxidase in Situ Apoptosis Detection Kit). In brief, tissue was dewaxed in xylene, rehydrated in increasing concentrations of ethanol, and washed in PBS. The tissue was then pre- treated with proteinase K for 15 min, washed in PBS, and background peroxidase activity quenched in 3.0% hydro- gen peroxide for 5 min. The equilibration buffer was added for 10 s, before the TdT enzyme was added and left for 1 h at 37 °C. The reaction was stopped in stop buffer for 10 min, then washed before adding the anti-digoxi- genin conjugate for 30 min at room temperature. Finally, peroxidase substrate was added for 6 min before the tis- sue was counterstained in 50% haematoxylin (cat#MH- 1NPR, Trajan Scientific, VIC, Australia), dehydrated in xylene followed by increasing concentrations of ethanol, mounted in dibutyl phthalate polystyrene xylene and cover slipped. Immunohistochemistry analysis Cortical areas were quantified using QuPath imaging software (version 0.2.3) [29]. Microglia (Iba-1 + cells), neuronal nuclei (NeuN + cells), and caspase 3 + cells were visualised using light microscopy (Olympus, Tokyo, Japan) at 40 × magnification using CellSens imaging soft- ware (version 2.3; Olympus). Caspase 3 + cells display- ing both immunostaining and apoptotic bodies were counted. NeuN + cells were counted only if they were morphologically normal, while cells displaying con- densed or fragmented nuclei were not counted [30]. IL-1β-stained sections were scored at 20 × magnification using an immunoreactivity scoring system adapted from Girard et al. [31]. Scoring was based on the intensity of staining whereby 1 = light, 2 = moderate, 3 = moderate to intense and 4 = intense, as previously reported [32]. The area fraction of GFAP (astrocyte) staining was quantified in ImageJ software (v2.0, LOCI, University of Wiscon- sin) using a standard intensity threshold. ApopTag + cells were quantified using QuPath imaging software. Total numbers of positive cells were counted within the lateral gyrus from the parietal lobe between cortical layers 3 and 5. For all other immunohistochemical analyses, positive cells or immunoreactivity were quantified for each field of view (1 field from the base of the gyrus and 1 field from Table 1 Primer sequences for qPCR Gene Species Accession number Primer sequence 18S IL1A IL1B IL6 Rat Sheep Sheep Sheep NR_046237.1 NM_001009808.1 NM_001009465.2 NM_001009392 5ʹ-GTA ACC CGT TGA ACC CCA TT-3ʹ 3ʹ-CCA TCC AAT CGG TAG TAG CG-5ʹ 5ʹ-GTC CAT ACA TGA CGG CTG CTA-3ʹ 3ʹ-GGT GTC TCA GGC ATC TCC TTAT-5ʹ 5ʹ-CGA TGA GCT TCT GTG TGA TG-3ʹ 3ʹ-CTG TGA GAG GAG GTG GAG AG-5ʹ 5ʹ-CGC AAA GGT TAT CAT CAT CC-3ʹ 3ʹ CCC AGG AAC TAC CAC AAT CA-5ʹ Amplicon length, nt 151 184 121 108 Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 5 of 16 the top of the gyrus; Fig. 1) from two sections per subject using ImageJ. For each field of view, average scores from two slides from the right hemisphere were calculated. The size of the field of view was the same for all assess- ments (20 x magnification = 0.489 mm2 and 40 x magni- fication = 0.130 mm2). All imaging and cell counts were performed by an assessor who was blinded to the treat- ment group by independent coding of slides and data files. Golgi analysis Coded region-matched Golgi-stained tissue sections (10 serial sections per subject) were used to assess basal den- drites from pyramidal neurons in layers 3 and 5 of the lateral cortex as described previously [33, 34]. The Golgi staining tended to produce less complete filling of the apical dendrites relative to the basal dendrites. Thus, api- cal dendrites were not analysed. Basal dendrites were vis- ualised using an Olympus BX61 stereology microscope (Olympus) equipped with an Olympus DP73 colour cam- era (× 0.5 lens) at 60 × magnification and CellSens imag- ing software (version 2.3; Olympus). From each subject, basal dendrites from a total of 20 pyramidal neurons selected from 10 serial sections of the lateral gyrus met the inclusion criteria for imaging. A total of 10 neurons were selected from the base of the gyrus and 10 were selected from the top of the gyrus (Fig. 1). We found that 20 was the maximum number of neurons per subject that could be selected based on the pre-defined selection criteria. Neurons were selected based on morphological criteria [34]: triangular shaped soma and apical dendrites perpendicular to the pial surface, complete Golgi impreg- nation of the cell that permitted visualisation of the entire dendritic arbour and spines, neuronal soma and processes not obscured by other neurons, glia or blood vessels, and neurons exhibiting a complete basilar den- dritic tree without truncated or cut processes. Pyramidal neuronal subtypes were not distinguished. A preliminary analysis of n = 4 subjects per group showed no significant differences between neurons selected from the base of the gyrus and neurons selected from the top of the gyrus for dendritic length (control: gyrus base = 2874 ± 495, gyrus top = 3261 ± 517; LPS: gyrus base = 1663 ± 137, gyrus top = 1945 ± 173) and numbers of terminals (con- trol: gyrus base = 42 ± 2, gyrus top = 48 ± 3; LPS: gyrus base = 30 ± 3, gyrus top = 32 ± 2). Nevertheless, equal numbers of neurons from the gyrus base (n = 10) and top (n = 10) were analysed from each subject to ensure our data were not confounded by potential regional dif- ferences in neuronal development. Images were cropped and separated into individual channels using ImageJ. The images were then imported into Imaris (version 9.2.1, Bitplane, Oxford Instruments Company, Abington, UK). Measures of dendritic complexity including sum- mated dendritic length, numbers of dendritic terminals and Sholl analysis (numbers of dendrite intersections Fig. 1 Schematic outlining the study design. The study consisted of two groups: control (vehicle, n = 9) and LPS (n = 8). The solid lines show the timing of the lipopolysaccharide (LPS)/vehicle infusions which were given over 2 min at increasing doses (300 ng, 600 ng, and 1200 ng). Controls received an equivalent volume of vehicle (saline) during the infusion period. Continuous electroencephalogram (EEG) recordings were performed throughout the experiential period. At 96 h, brains were collected for Golgi staining to examine neuronal arborisation and numbers of dendritic spines, immunohistochemistry to assess neuroinflammation, neuronal numbers and cortical area and mRNA assessment of proinflammatory proteins. Boxes indicate regions of interest used for immunohistochemical and Golgi analysis. Image source:[86] Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 6 of 16 per Sholl ring) were assessed using the Imaris filament tracer tool. Sholl intersections were analysed using 5 μm interval concentric rings centred on the soma. Numbers of dendritic spines were quantified using the Imaris fila- ment tracer and dendritic spine classification (filopo- dia, long thin, stubby, mushroom) was performed using the MATLAB spine classification extension (MATLAB, R2019a, Mathworks Inc., CA, USA). our laboratory. Blood gas data and EEG power have been previously published in Kelly et al. [24]. Confirmation of systemic inflammation All LPS-exposed fetuses had increased plasma cytokine levels relative to baseline after LPS infusions [24], confirming the induction of a systemic inflammatory response. Data analysis and statistics Offline analysis of physiological data was performed using LabChart Pro software (v8.1.3; ADInstruments, Sydney, NSW, Australia). EEG data were processed as hourly averages and presented from 24 h before the first saline or LPS infusion until the end of the experiment. Due to small differences in baseline spectral edge fre- quency between the group, spectral edge frequency was normalised by subtracting the baseline values (average of 24 h before the first saline/LPS infusion) from the abso- lute value. EEG data during the baseline, 24 h period after LPS/saline infusion, and recovery (from 24 h after the final LPS infusion until the end of the study, i.e. 72–96 h) periods were analysed separately. Sleep stage cycling was assessed using the raw EEG spectral edge frequency trace during the last 5 h of the experimental period (91–96 h). Sleep stage cycling was defined as a repetitive alternat- ing pattern of high and low-frequency activity, with each phase lasting approximately 20 min, as previously described [35]. Data were tested for normality using the Shapiro–Wilk test. Histological and PCR data were ana- lysed using an unpaired t-test. Mann–Whitney U-tests were used for testing non-parametric data. For EEG data and Sholl analysis of dendritic morphology, when statis- tical significance was found between groups, group and time (EEG) or group and radius (Sholl analysis) post hoc comparisons were made using the Fisher’s least signifi- cant difference test [36]. Linear and non-linear regres- sion were used to assess the relationship between EEG spectral edge frequency and band power with neuronal microstructure. Post hoc power analysis for summated dendritic length showed 85% power to detect a minimum difference of 970 µm. Statistical significance was accepted when P < 0.05. Data are presented as scatter plots with mean ± standard error of the mean (SEM). Results Baseline period Before LPS exposure, fetal arterial blood gases, pH, glu- cose, lactate and EEG power and frequency did not dif- fer between groups and were within the normal range for Spectral edge frequency (SEF) After the first LPS infusion, SEF was lower in the LPS group at 7 h (P < 0.05 vs. controls, Fig. 2A). After the second LPS infusion, SEF was reduced in the LPS group from 28–31 h (i.e., 4–7 h after the second infusion; P < 0.05 vs. controls). After the third LPS infusion, SEF was reduced in the LPS group from 50–52, 57–58 and at 65 h (i.e., at 2–4 h, 9–10 h and 17 h after the third infusion, P < 0.05 vs. controls). Qualitative assessment of EEG spectral edge frequency during the last 5 h of the recording period (91–96 h) showed sleep stage cycling was present in all control and LPS-exposed fetuses. Spectral band power analysis After the first LPS infusion, delta band power increased in the LPS group from 7–15 h (P < 0.05 vs. controls, Fig. 2B). After the third LPS infusion, %delta band power was higher in the LPS group from 48–50 h (i.e., 0–2 h after the third infusion, P < 0.05 vs con- trols, Fig. 2B). There were no differences in %theta and %alpha band powers after LPS infusions between groups (Fig. 2C, D). After the first LPS infusion, %beta band power decreased in the LPS group from 7–9 h and 16–18 h and at 22 h (P < 0.05 vs. controls, Fig. 2E). After the second LPS infusion, %beta band power was lower in the LPS group from 31–33 h and 38–40 h and at 47 h (i.e., at 7–9 h, 14–16 h, and 23 h after the second LPS infusion; P < 0.05 vs. controls, Fig. 2E). After the third LPS infusion, %beta band power was lower at 48 h and between 55–66 h, 73–86 h, and 92–94 h (i.e., at 0 h, 7–18 h, 25–38 h, and 44–46 h after the third LPS infu- sion; P < 0.05 vs. controls, Fig. 2E). Gene expression analysis There were no significant changes in mRNA expression of IL1A, IL1B (P = 0.06) and IL6 (P = 0.09) in the soma- tosensory cortex in the LPS group compared with con- trols (Fig. 3). Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 7 of 16 A B C D E Fig. 2 Neurophysiological changes over time. From the top down, the figure shows spectral edge frequency (A), %delta activity (B), %theta activity (C), %alpha activity (D), and %beta activity (E) in the control (black, n = 9) and LPS (red n = 8) groups. Vertical lines indicate the timings of LPS administration. Data are hourly means ± standard error (SE). P < 0.05 vs. control Neuronal basal dendrite morphology The summated basal dendritic length of cortical Fig. 3 Interleukin (IL)1A, IL1B, and IL6 mRNA levels in the somatosensory cortex in tissue sections adjacent to samples processed for Golgi staining from controls (white circles, n = 7, two subjects had undetectable values), and LPS (black circles; IL1B, n = 7 [one subject had undetectable values], IL1A, n = 8, and IL6, n = 8) groups. Data are means ± SE and are expressed as the fold change from the mean control values pyramidal neurons was significantly reduced in the LPS group (by ~ 36%) compared to control (control: 2716 ± 314 µm vs. LPS: 1738 ± 92 µm; P < 0.05, Figs. 4A, 6A). The number of basal dendritic terminals was sig- nificantly reduced in the LPS group (by ~ 31%) com- pared with controls (control: 46 ± 3 vs. LPS: 32 ± 3, P < 0.01, Figs. 4B, 6B). Sholl analysis of pyramidal neu- ron complexity showed reduced dendritic arborisation in the LPS group at 40–230 μm away from the soma (P < 0.05 vs. controls; Fig. 4C). Neuronal basal dendritic spine number and morphology The total number of basal dendritic spines was reduced in the LPS group compared with controls (P < 0.05, Figs. 5E, 6C). Spine morphology classification showed reduced numbers of long thin spines in the LPS group compared to control (P < 0.05; Fig. 5B). There were no differences in numbers of filopodia, stubby or mushroom spines between the groups (Fig. 5A, C, D, respectively). Histopathology Immunoreactivity of IL-1β was increased in the soma- tosensory cortex of the LPS group (P < 0.05 vs. controls; Figs. 7A, 8A). Numbers of Iba-1 + microglia were signifi- cantly increased in the LPS group (P < 0.01 vs. controls, Figs. 7B, 8B). There was no difference in area fraction of GFAP + astrocyte staining between the groups (Figs. 7C, 8C). Numbers of caspase3 + cells were increased in the LPS group compared with controls (P < 0.05; Figs. 7D, 8E). There were no differences in the total numbers of ApopTag + cells between the groups (Figs. 7E, 8F). Finally, there were no differences in either numbers of NeuN + neurons (Figs. 7F, 8D) or cortical area (control: 2.7 × 107 ± 2.4 × 106 µm2 vs LPS: 2.6 × 107 ± 3.4 × 106 µm2) in the lateral gyrus between the groups. Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 8 of 16 Discussion This study demonstrates that antenatal inflammation reduced the length, number and complexity (arbori- sation) of basal dendrites, and reduced the numbers of dendritic spines on pyramidal neurons within the somatosensory cortex in late gestation fetal sheep. The reduction in dendritic length and complexity was associ- ated with an increased number of microglia and IL-1β- positive staining but was not associated with changes in the number of cortical neurons or cortical area. Func- tionally, the inflammation-induced reduction in dendritic length and complexity were associated with transient increases in delta (slow wave) activity, reduced beta (fast wave) activity and an overall reduction in the spec- tral edge frequency of the EEG. Given the association between exposure to perinatal inflammation and reduc- tions in cortical growth and connectivity [6, 7, 37], the present study provide critical new insight into the deficits in neuronal structure and function arising from antenatal inflammation. Clinically, Gram-negative infections, including E. coli, continue to be among the most common pathogens linked to perinatal infection/inflammation and increased risk of perinatal brain injury [38, 39]. We sought to repro- duce features of Gram-negative infection/inflammation using repeated increasing doses of LPS infusions to pro- mote a chronic progressive fetal inflammatory response that is commonly associated with adverse neurodevel- opmental outcomes [24, 40]. By contrast, most of the previous preclinical studies have focused on the patho- physiological consequences of single or repeated bolus doses of LPS or other infectious/inflammatory stimuli [41]. This study showed that inflammation induced by increasing doses of LPS infusions was associated with increased numbers of microglia and greater immunore- activity of IL-1β in the somatosensory cortex but no sig- nificant increases in cortical IL1B (p = 0.06), IL6 (p = 0.09) or IL1A mRNA expression. Consistent with this obser- vation, human post-mortem studies have reported increased numbers of microglia and IL-1β immuno- reactivity in areas of white and grey matter inflamma- tion and injury [42–44]. Indeed, cerebral recognition of pathogen-associated molecular patterns such as LPS by innate immune receptors, including toll like receptor 4, on microglia and other immune cells leads to glial cell activation and nuclear factor kappa B induced transcrip- tion of bioactive IL-1β [45]. Furthermore, circulating cytokines, including IL-1β, can penetrate the blood brain barrier [46–48] to recruit and activate microglia within the central nervous system [45]. Thus, the concomitant increase in numbers of microglia and IL-1β immunoreac- tivity observed in this study strongly supports sustained Fig. 4 Summated dendritic length (μm) (A), number of dendritic terminals (B) and Sholl analysis (C) showing the number of dendritic intersections (dendritic arborisation) indicated by the number of intersections at 5 μm intervals away from the soma in the control (white circles, n = 7; two subjects had limited Golgi penetration) and LPS (black circles, n = 7; 1 subject had limited Golgi penetration) groups. Data are means ± SE, P < 0.05 vs control Correlative analysis Mean %beta band power during the final 12 h of the experi- ment was positively correlated with dendritic length (linear regression: R2 = 0.30, p = 0.0417) and neuronal arborisation (linear regression: R2 = 0.30, P = 0.0423). There were no significant correlations between beta band power and numbers of dendritic terminals and numbers of dendritic spines. There were no significant correlations between spectral edge frequency, or the %delta, %theta and %alpha spectral bands and markers of neuronal microstructure. Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 9 of 16 Fig. 5 Numbers of filopodia, long thin, stubby, mushroom and total dendritic spines in control (white circles, n = 7; two subjects had limited Golgi penetration) and LPS (black circles, n = 7; 1 subject had limited Golgi penetration) groups. Data are means ± SE, P < 0.05 vs control inflammation in the somatosensory cortex 4 days after beginning intravenous LPS infusions. In this experimental model of antenatal inflammation, we have previously reported progressive increases in concentrations of systemic pro- and anti-inflammatory cytokines (IL-1β, tumour necrosis factor [TNF], IL-6 and IL-10), in addition to diffuse white matter gliosis and reduced numbers of precursor oligodendrocytes [24]. Clinically, in large prospective studies of preterm infants, increased concentrations of these inflamma- tory proteins in cord blood and postnatal blood sam- ples have been associated with perinatal brain injury and impaired neurodevelopment in childhood [49]. The present study shows that exposure to inflammation did Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 10 of 16 Fig. 6 Representative Golgi stained (A) and traced (B) images of basal dendrites and dendritic spines (C) from pyramidal neurons in the somatosensory cortex from control and LPS-exposed subjects. Scale bar panel A = 100 μm, panel B = 50 μm and panel C = 15 μm not affect overall numbers of neurons (NeuN +) or neu- ronal density in the areas of the somatosensory cortex evaluated in this study, suggesting a lack of overt cortical injury. This is further confirmed by the similar numbers of TUNEL + cells between groups, suggesting no effect of LPS-exposure on acute cell death in the somatosen- sory cortex. This observation is consistent with neonatal experimental and clinical studies showing limited or no neuronal cell death in cases of perinatal encephalopa- thy, including after systemic inflammation [10, 49, 50]. By contrast, we observed increased numbers of caspase 3 + cells in LPS-exposed fetuses compared to controls. This finding of increased numbers of caspase-3-positive cells without increased cell death has been reported in the adult and perinatal brain [51–54] and is likely to be linked to other roles played by caspases, which include immune/microglial activation and cell differentiation [54–56]. We observed a reduction in neuronal dendritic com- plexity in LPS-exposed fetuses as shown by reduced Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 11 of 16 Fig. 7 Interleukin 1(IL)-1β immunoreactivity scores (A), numbers of ionised calcium binding adaptor molecule (IBA-1) + microglia (B), glial fibrillary acidic protein (GFAP +) area fraction staining (C), numbers of caspase 3 + cells (D), numbers of ApopTag (TUNEL) + cells (E) numbers of NeuN + neurons (F), in the lateral gyrus (LG) in control (open circles, n = 9) and LPS groups (black circles, n = 8) groups. Data are means ± SE, *P < 0.05 vs control dendritic length, numbers of dendritic terminals, den- dritic arborisation, and numbers of dendritic spines. In humans, the marked cortical expansion that occurs during the last trimester primarily reflects the prolific increase in neuronal dendritic growth and complexity during this stage of development [57, 58]. Our observa- tions suggest that at this period in late gestation, neu- ronal development within the somatosensory cortex in the developing fetus is highly vulnerable to inflamma- tion-induced impairments of dendritic arborisation and spine formation. These data are consistent with previ- ous studies that reported reduced neuronal arborisation in the frontal cortex of fetal sheep after acute cerebral ischaemia at mid-gestation, and reduced dendritic num- ber and spine density in the retrosplenial cortex of new- born rabbits (P1) after a single bolus of intra-amniotic LPS (20 µg/kg) [59]. Similarly, in separate rodent studies examining the long-term effects of prenatal and early postnatal LPS-induced inflammation, reduced dendritic arborisation was seen in the motor cortex and medial prefrontal cortex on postnatal days 21 and 60 [10, 60]. Furthermore, these data support a link between diffuse white matter injury, which we have previously reported in the same experimental paradigm [24], and impaired neuronal development. For example, human case series have shown reduced dendritic length in cases of both dif- fuse and necrotic white matter injury [61, 62]. Similarly, moderate LPS-induced inflammation in neonatal rodents (from P1-P3) was associated with reduced dendritic arborisation in the motor cortex, diffuse white matter injury, and impaired myelination and motor function on postnatal day 21 [10]. The somatosensory cortex has been shown to syn- apse with cervical excitatory neurons and modulate Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 12 of 16 did not evaluate neuronal complexity within the motor cortex in this study, our data raise the possibility that impaired neuronal development in the somatosensory cortex may contribute to the inflammation-induced reduction in fetal body movements observed in this preclinical model of antenatal infection/inflammation. Taken together, these data support a close link between impaired neuronal development in the somatosensory cortex and inhibition of motor function. Although the precise mechanism/s underpinning the inflammation-induced impairment in neuronal devel- opment are yet to be identified, it is likely to include a direct effect of inflammation on the central nervous system. For example, in vitro studies have shown that cortical neurons exposed to inflammatory cytokines, including IL-1β, IL-6, TNF, and interferon gamma, show reduced dendritic branching and synapse for- mation [64, 65]. This is supported by our findings of increased numbers of cortical microglia, which are known to secrete proinflammatory cytokines, along with increased IL-1β immunoreactivity observed in the LPS-exposed fetuses. Microglial processes have also been shown to interact with synapses to eliminate spines, suggesting a direct effect of microglial activation on spine density [66, 67]. Furthermore, reduced circu- lating concentrations of neural growth factors, includ- ing nerve growth factor and brain derived neurotrophic factor, have been reported in human and animal studies of perinatal infection/inflammation [59, 68, 69]. Consistent with the inflammation-induced reduc- tion in neuronal complexity in the present study, neu- ronal activity in LPS-exposed fetuses was impaired, as shown by an overall reduction in the spectral edge frequency of EEG activity along with an increase in the proportion of EEG activity in the delta band and a reduced proportion of activity in the beta band. Col- lectively, these data indicate loss of high-frequency activity after LPS-exposure with a shift to lower fre- quency activity. The inflammation-induced change in EEG spectra could reflect alterations to fetal behaviour/ sleep stages. The fetus is never awake, but rather cycles between low voltage (high frequency) and high voltage (low frequency) sleep [70]. Qualitative analysis of EEG frequency signals over the last 5 h of the experimental period showed sleep stage cycling was present in all control and LPS-exposed fetuses. Consistent with these data, pre- and post-natal immune challenges in mice have shown inflammation-induced increases in slow wave sleep in association with a shift in spectral band power (increased low-frequency and reduced high-fre- quency activity) that was consistent with our analysis [71]. Although sleep stage cycling was observed in all fetuses in the present study at the end of the recording Fig. 8 Representative photomicrographs showing immunohistochemical staining of IL-1β (A), IBA-1 (B), GFAP (C), NeuN (D), caspase3 (E) and ApopTag (TUNEL, F) in the lateral gyrus. Arrowheads point to Caspase 3 + cells and arrows point to ApopTag (TUNEL +) cells. Scale bar = 100 μm locomotion independently of the motor cortex [63]. Consistent with this observation, we have previously reported reduced fetal movements in the same fetal sheep paradigm, as shown by reduced nuchal electro- myographic activity from 3 days after starting LPS infu- sions until the time of post-mortem [24]. Although we Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 13 of 16 period, immediately prior to post-mortem, it is possible that the inflammation-induced changes in EEG spectra are associated with subtle changes in the proportions of high and low-frequency sleep. We have previously reported that there were no differ- ences in myelin density or numbers of mature myelinat- ing oligodendrocytes at this timepoint in this cohort of LPS-exposed fetuses [24]. This suggests that the changes in EEG frequency in the present study are not related to altered myelination. Alternatively, the inflammation- induced reduction in high-frequency activity may reflect inhibition of synaptic activity. This could be due to the reduced neuronal arborisation and numbers of dendritic spines on cortical neurons directly underlying the EEG electrodes (i.e., a direct functional consequence of inflam- mation-induced changes in neuronal pathology). Alter- natively, elevated central levels of IL-1β have been shown to induce NMDA-mediated suppression of synaptic function [72]. Consistent with this, TNF inhibition using the soluble TNF receptor antagonist, etanercept, reduced the magnitude of EEG suppression in fetal sheep exposed to LPS [73], possibly due to reduced NMDA receptor activation [74]. Furthermore, in vivo and in vitro stud- ies have shown that both LPS- and IL-1β-induced central inflammation actively mediate EEG suppression through the release of inhibitory neuromodulators, such as allo- pregnanolone and adenosine [75, 76]. Taken together, the present data suggest that inflammation-induced suppres- sion of EEG activity is mediated by a combination of acti- vation of anti-excitotoxic mediators as well as reduced complexity of the neuronal microstructure. Clinical studies have shown that reduced EEG fre- quency strongly predicts subsequent brain injury and neurodevelopmental impairment in preterm and term infants. For example, in a cohort study, reduced EEG frequency was associated with the severity of neona- tal white matter injury [77]. Similarly, depression of the EEG background pattern was associated with both motor and cognitive impairment in preterm and term infants with evidence of central inflammation [78, 79]. Furthermore, increased latency of somatosensory evoked potentials was reported in children with bilat- eral spastic cerebral palsy. Notably, the latency of soma- tosensory evoked potentials correlated with a history of exposure to perinatal infection/inflammation [80]. Functional MRI studies have shown reduced cortical functional connectivity in fetuses exposed to inflamma- tion before birth [37]. Similarly, reduced cortical func- tional connectivity was observed in preterm infants without evidence of overt cortical injury [81]. Subse- quent investigation of infants with moderate to severe white matter injury, but without overt cortical injury, showed a reduction in cortical functional connectivity that correlated with the severity of white matter injury [82]. Our data suggest that in the absence of overt neu- ronal injury or white matter loss, reduced neuronal complexity and synaptic density may contribute to reduced functional connectivity within and between major grey matter structures after exposure to perinatal inflammation. This study was not designed to test the effect of sex and is not large enough to determine whether there were sex-specific effects of antenatal inflammation on neuronal microstructure or EEG activity. Previous studies in perinatal rodents reported sex-dependant effects in central and peripheral immune activation, as well as neuronal development [83, 84]. Further studies evaluating the impact of sex on inflammation-induced changes to neuronal structure and function in large animals are needed. In conclusion, this study demonstrates that an inflam- mation-induced reduction in high-frequency spec- tral band power was associated with reduced cortical neuronal arborisation and dendritic spine density. Collectively, these data support the concept that inflam- mation-induced impairments in neuronal maturation and function, rather than overt neuronal loss, make a key contribution to disturbed cortical development and con- nectivity, and subsequent impaired neurodevelopmental outcomes, in infants exposed to perinatal inflammation. We propose that early EEG monitoring combined with neuroimaging modalities that enable more sensitive assessment of brain microstructure [10, 85] and thera- peutics designed to mitigate systemic and central inflam- mation, could provide an effective approach for early detection and therapeutic intervention. Author contributions SBK, AJG, JMD and RG conceptualised and designed the study. SBK, VZ, AM, ID, SLM, SBH, JMD, LB, AJG, GRP, RG undertook experiments, formal analysis and interpretation of the data. SBK and RG undertook the Golgi analysis, immunohistochemistry, cell quantification, analysis and preparation of figures. RG provided overall oversight of the research. All authors critically reviewed the manuscript and approved the final manuscript as submitted and agree to be accountable for all aspects of the work. Funding This study was supported by project grants from the National Health and Medical Research Council of Australia (R.G.; 1090890 and 1164954), the Cerebral Palsy Alliance, Harold and Cora Brennen Benevolent Trust, Health Research Council of New Zealand (17/601, 22/559) and the Victorian Govern- ment’s Operational Infrastructure Support Program. Availability of data and materials The datasets used during the current study are available from the correspond- ing author upon reasonable request. Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 14 of 16 Declarations Ethics approval and consent to participate All procedures were approved by the Hudson Institute of Medical Research Animal Ethics committee and were conducted in accordance with the National Health and Medical Research Council Code of Practice for the Care and Use of Animals for Scientific Purposes (Eighth Edition). Consent for publications Not applicable. Competing interests The authors declare they have no competing interests. Author details 1 The Ritchie Centre, Hudson Institute of Medical Research, 27-31 Wright Street, Melbourne, VIC 3168, Australia. 2 Department of Obstetrics and Gynaecology, Monash University, Melbourne, VIC, Australia. 3 Department of Physiology, The University of Auckland, Auckland, New Zealand. Received: 10 February 2023 Accepted: 15 May 2023 References 1. Martinello K, Hart AR, Yap S, Mitra S, Robertson NJ. Management and 3. 2. investigation of neonatal encephalopathy: 2017 update. Arch Dis Child Fetal Neonatal Ed. 2017;102:F346-f358. Eunson P. The long-term health, social, and financial burden of hypoxic– ischaemic encephalopathy. Dev Med Child Neurol. 2015;57:48–50. Economic costs associated with mental retardation. cerebral palsy, hearing loss, and vision impairment—United States, 2003. MMWR Morb Mortal Wkly Rep. 2004;53:57–9. Shih STF, Tonmukayakul U, Imms C, Reddihough D, Graham HK, Cox L, Carter R. Economic evaluation and cost of interventions for cerebral palsy: a systematic review. Dev Med Child Neurol. 2018;60:543–58. 5. Becroft DM, Thompson JM, Mitchell EA. Placental chorioamnionitis at 4. 6. term: epidemiology and follow-up in childhood. Pediatr Dev Pathol. 2010;13:282–90. Jain VG, Kline JE, He L, Kline-Fath BM, Altaye M, Muglia LJ, DeFranco EA, Ambalavanan N, Parikh NA. Acute histologic chorioamnionitis independently and directly increases the risk for brain abnormalities seen on magnetic resonance imaging in very preterm infants. Am J Obstet Gynecol. 2022;227:623.e621-623.e613. 7. Hatfield T, Wing DA, Buss C, Head K, Muftuler LT, Davis EP. Magnetic resonance imaging demonstrates long-term changes in brain structure in children born preterm and exposed to chorioamnionitis. Am J Obstet Gynecol. 2011;205(384):e381-388. 8. Burd I, Brown A, Gonzalez JM, Chai J, Elovitz MA. A mouse model of term chorioamnionitis: unraveling causes of adverse neurological outcomes. Reprod Sci. 2011;18:900–7. 9. Allard MJ, Brochu ME, Bergeron JD, Segura M, Sébire G. Causal role of group B Streptococcus-induced acute chorioamnionitis in intrauterine growth retardation and cerebral palsy-like impairments. J Dev Orig Health Dis. 2019;10:595–602. 10. Prasad JD, van de Looij Y, Gunn KC, Ranchhod SM, White PB, Berry MJ, Bennet L, Sizonenko SV, Gunn AJ, Dean JM. Long-term coordinated microstructural disruptions of the developing neocortex and subcortical white matter after early postnatal systemic inflammation. Brain Behav Immun. 2021;94:338–56. 11. Sweeney EL, Dando SJ, Kallapur SG, Knox CL. The human ureaplasma species as causative agents of chorioamnionitis. Clin Microbiol Rev. 2017;30:349–79. 12. Moss TJ, Nitsos I, Ikegami M, Jobe AH, Newnham JP. Experimental intrauterine ureaplasma infection in sheep. Am J Obstet Gynecol. 2005;192:1179–86. 13. Oh JW, Park CW, Moon KC, Park JS, Jun JK. The relationship among the progression of inflammation in umbilical cord, fetal inflammatory response, early-onset neonatal sepsis, and chorioamnionitis. PLoS ONE. 2019;14: e0225328. 14. Spann MN, Monk C, Scheinost D, Peterson BS. Maternal immune activa- tion during the third trimester is associated with neonatal functional con- nectivity of the salience network and fetal to toddler behavior. J Neurosci. 2018;38:2877–86. 15. Brown AS, Sourander A, Hinkka-Yli-Salomäki S, McKeague IW, Sundvall J, Surcel HM. Elevated maternal C-reactive protein and autism in a national birth cohort. Mol Psychiatry. 2014;19:259–64. 16. Jiang NM, Tofail F, Ma JZ, Haque R, Kirkpatrick B, Nelson CA, Petri WA. Early life inflammation and neurodevelopmental outcome in Bangladeshi infants growing up in adversity. Am J Trop Med Hyg. 2017;97:974–9. 17. Jiang NM, Tofail F, Moonah SN, Scharf RJ, Taniuchi M, Ma JZ, Hamadani JD, Gurley ES, Houpt ER, Azziz-Baumgartner E, et al. Febrile illness and pro- inflammatory cytokines are associated with lower neurodevelopmental scores in Bangladeshi infants living in poverty. BMC Pediatr. 2014;14:50. 18. Tsuchida TN. EEG background patterns and prognostication of neo- natal encephalopathy in the era of hypothermia. J Clin Neurophysiol. 2013;30:122–5. 19. Al Naqeeb N, Edwards AD, Cowan FM, Azzopardi D. Assessment of neo- natal encephalopathy by amplitude-integrated electroencephalography. Pediatrics. 1999;103:1263–71. 20. Percie du Sert N, Hurst V, Ahluwalia A, Alam S, Avey MT, Baker M, Browne WJ, Clark A, Cuthill IC, Dirnagl U, et al. The ARRIVE guidelines 2.0: updated guidelines for reporting animal research. J Physiol. 2020;598:3793–801. 21. Keogh MJ, Bennet L, Drury PP, Booth LC, Mathai S, Naylor AS, Fraser M, Gunn AJ. Subclinical exposure to low-dose endotoxin impairs EEG matu- ration in preterm fetal sheep. Am J Physiol Regul Integr Comp Physiol. 2012;303:R270-278. 22. Galinsky R, Dhillon SK, Lear CA, Yamaguchi K, Wassink G, Gunn AJ, Bennet L. Magnesium sulfate and sex differences in cardiovascular and neural adaptations during normoxia and asphyxia in preterm fetal sheep. Am J Physiol Regul Integr Comp Physiol. 2018;315:R205-r217. 23. McIntosh TK, Vink R, Yamakami I, Faden AI. Magnesium protects against neurological deficit after brain injury. Brain Res. 1989;482:252–60. 24. Kelly SB, Stojanovska V, Zahra VA, Moxham A, Miller SL, Moss TJM, Hooper SB, Nold MF, Nold-Petry CA, Dean JM, et al. Interleukin-1 blockade attenu- ates white matter inflammation and oligodendrocyte loss after progres- sive systemic lipopolysaccharide exposure in near-term fetal sheep. J Neuroinflamm. 2021;18:189. 25. Gotsch F, Romero R, Kusanovic JP, Mazaki-Tovi S, Pineles BL, Erez O, Espi- noza J, Hassan SS. The fetal inflammatory response syndrome. Clin Obstet Gynecol. 2007;50:652–83. 26. Grether JK, Nelson KB. Maternal infection and cerebral palsy in infants of normal birth weight. [Erratum appears in JAMA 1998;279(2):118]. JAMA. 1997;278:207–11. 27. Yanowitz TD, Jordan JA, Gilmour CH, Towbin R, Bowen A, Roberts JM, Brozanski BS. Hemodynamic disturbances in premature infants born after chorioamnionitis: association with cord blood cytokine concentrations. Pediatr Res. 2002;51:310–6. 28. Mohan C, Long K, Mutneja M, Ma J. Detection of end-stage apoptosis by ApopTag® TUNEL technique. Methods Mol Biol. 2015;1219:43–56. 29. Bankhead P, Loughrey MB, Fernández JA, Dombrowski Y, McArt DG, Dunne PD, McQuaid S, Gray RT, Murray LJ, Coleman HG, et al. QuPath: open source software for digital pathology image analysis. Sci Rep. 2017;7:16878. 30. Pozo Devoto VM, Chavez JC, Fiszer de Plazas S. Acute hypoxia and pro- grammed cell death in developing CNS: differential vulnerability of chick optic tectum layers. Neuroscience. 2006;142:645–53. 31. Girard S, Kadhim H, Larouche A, Roy M, Gobeil F, Sébire G. Pro-inflamma- tory disequilibrium of the IL-1 beta/IL-1ra ratio in an experimental model of perinatal brain damages induced by lipopolysaccharide and hypoxia- ischemia. Cytokine. 2008;43:54–62. 32. Kelly SB, Stojanovska V, Zahra VA, Moxham A, Miller SL, Moss TJM, Hooper SB, Nold MF, Nold-Petry CA, Dean JM, et al. Interleukin-1 blockade attenu- ates white matter inflammation and oligodendrocyte loss after progres- sive systemic lipopolysaccharide exposure in near-term fetal sheep. J Neuroinflamm. 2021; 18. 33. Dudink I, White TA, Ardalan M, Mallard C, Ballerin G, Creed SJ, Pham Y, Sutherland AE, Castillo-Melendez M, Allison BJ, Miller SL. An opti- mized and detailed step-by-step protocol for the analysis of neuronal Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 15 of 16 morphology in golgi-stained fetal sheep brain. Dev Neurosci. 2022;44:344–62. 34. Dean JM, McClendon E, Hansen K, Azimi-Zonooz A, Chen K, Riddle A, Gong X, Sharifnia E, Hagen M, Ahmad T, et al. Prenatal cerebral ischemia disrupts MRI-defined cortical microstructure through disturbances in neuronal arborization. Sci Transl Med. 2013;5:168ra167. 53. Villapol S, Bonnin P, Fau S, Baud O, Renolleau S, Charriaut-Marlangue C. Unilateral blood flow decrease induces bilateral and symmetric responses in the immature brain. Am J Pathol. 2009;175:2111–20. 54. Northington FJ, Chavez-Valdez R, Martin LJ. Neuronal cell death in neonatal hypoxia-ischemia. Ann Neurol. 2011;69:743–58. 55. Abraham MC, Shaham S. Death without caspases, caspases without 35. Galinsky R, Dhillon SK, Kelly SB, Wassink G, Davidson JO, Lear CA, van death. Trends Cell Biol. 2004;14:184–93. den Heuij LG, Bennet L, Gunn AJ. Magnesium sulphate reduces tertiary gliosis but does not improve EEG recovery or white or grey matter cell survival after asphyxia in preterm fetal sheep. J Physiol. 2023. https:// doi. org/ 10. 1113/ JP284 381. 36. Levin JR, Serlin RC, Seaman MA. A controlled, powerful multiple-com- parison strategy for several situations. Psychol Bull. 1994;115:153–9. 37. Thomason ME, Scheinost D, Manning JH, Grove LE, Hect J, Marshall N, Hernandez-Andrade E, Berman S, Pappas A, Yeo L, et al. Weak func- tional connectivity in the human fetal brain prior to preterm birth. Sci Rep. 2017;7:39286. 38. Stoll BJ, Puopolo KM, Hansen NI, Sánchez PJ, Bell EF, Carlo WA, Cotten CM, D’Angio CT, Kazzi SNJ, Poindexter BB, et al. Early-onset neonatal sepsis 2015 to 2017, the rise of Escherichia coli, and the need for novel prevention strategies. JAMA Pediatr. 2020;174: e200593. 39. Flannery DD, Edwards EM, Puopolo KM, Horbar JD. Early-onset sepsis among very preterm infants. Pediatrics. 2021; 148. 40. Galinsky R, Polglase GR, Hooper SB, Black MJ, Moss TJ. The conse- quences of chorioamnionitis: preterm birth and effects on develop- ment. J Pregnancy. 2013;2013: 412831. 41. Dean JM, Shi Z, Fleiss B, Gunn KC, Groenendaal F, van Bel F, Derrick M, Juul SE, Tan S, Gressens P, et al. A critical review of models of perinatal infection. Dev Neurosci. 2015;37:289–304. 42. Girard S, Sébire G, Kadhim H. Proinflammatory orientation of the interleukin 1 system and downstream induction of matrix metallo- proteinase 9 in the pathophysiology of human perinatal white matter damage. J Neuropathol Exp Neurol. 2010;69:1116–29. 43. Kadhim H, Tabarki B, De Prez C, Sebire G. Cytokine immunoreactivity in cortical and subcortical neurons in periventricular leukomalacia: are cytokines implicated in neuronal dysfunction in cerebral palsy? Acta Neuropathol. 2003;105:209–16. 44. Kadhim H, Tabarki B, Verellen G, De Prez C, Rona AM, Sebire G. Inflam- matory cytokines in the pathogenesis of periventricular leukomalacia. Neurology. 2001;56:1278–84. 45. Kelly SB, Green E, Hunt RW, Nold-Petry CA, Gunn AJ, Nold MF, Galinsky R. Interleukin-1: an important target for perinatal neuroprotection? Neural Regen Res. 2023;18:47–50. 46. Sadowska GB, Chen X, Zhang J, Lim YP, Cummings EE, Makeyev O, Besio WG, Gaitanis J, Padbury JF, Banks WA, Stonestreet BS. Interleukin-1β transfer across the blood-brain barrier in the ovine fetus. J Cereb Blood Flow Metab. 2015;35:1388–95. 47. Threlkeld SW, Lynch JL, Lynch KM, Sadowska GB, Banks WA, Stonestreet BS. Ovine proinflammatory cytokines cross the murine blood-brain barrier by a common saturable transport mechanism. NeuroImmu- noModulation. 2010;17:405–10. 48. Skinner RA, Gibson RM, Rothwell NJ, Pinteaux E, Penny JI. Transport of interleukin-1 across cerebromicrovascular endothelial cells. Br J Pharmacol. 2009;156:1115–23. 56. Kesavardhana S, Malireddi RKS, Kanneganti TD. Caspases in cell death, inflammation, and pyroptosis. Annu Rev Immunol. 2020;38:567–95. 57. Mrzljak L, Uylings HB, Kostovic I, van Eden CG. Prenatal development of neurons in the human prefrontal cortex. II. A quantitative Golgi study. J Comp Neurol. 1992;316:485–96. 58. Lu D, He L, Xiang W, Ai WM, Cao Y, Wang XS, Pan A, Luo XG, Li Z, Yan XX. Somal and dendritic development of human CA3 pyramidal neurons from midgestation to middle childhood: a quantitative Golgi study. Anat Rec (Hoboken). 2013;296:123–32. 59. Balakrishnan B, Dai H, Janisse J, Romero R, Kannan S. Maternal endotoxin exposure results in abnormal neuronal architecture in the newborn rab- bit. Dev Neurosci. 2013;35:396–405. 60. Baharnoori M, Brake WG, Srivastava LK. Prenatal immune challenge induces developmental changes in the morphology of pyramidal neurons of the prefrontal cortex and hippocampus in rats. Schizophr Res. 2009;107:99–109. 61. Stolp HB, Fleiss B, Arai Y, Supramaniam V, Vontell R, Birtles S, Yates AG, Baburamani AA, Thornton C, Rutherford M, et al. Interneuron develop- ment is disrupted in preterm brains with diffuse white matter injury: observations in mouse and human. Front Physiol. 2019;10:955. 62. Takashima S, Becker LE, Chan FW. Retardation of neuronal maturation in premature infants compared with term infants of the same postconcep- tional age. Pediatrics. 1982;69:33–9. 63. Karadimas SK, Satkunendrarajah K, Laliberte AM, Ringuette D, Weisspapir I, Li L, Gosgnach S, Fehlings MG. Sensory cortical control of movement. Nat Neurosci. 2020;23:75–84. 64. Gilmore JH, Jarskog LF, Vadlamudi S, Lauder J. Prenatal infection and risk for schizophrenia: IL-I beta, IL-6, and TNF alpha inhibit cortical neuron dendrite development. Neuropsychopharmacology. 2004;29:1221–9. 65. Neumann H, Schweigreiter R, Yamashita T, Rosenkranz K, Wekerle H, Barde YA. Tumor necrosis factor inhibits neurite outgrowth and branching of hippocampal neurons by a rho-dependent mechanism. J Neurosci. 2002;22:854–62. 66. Wake H, Moorhouse AJ, Jinno S, Kohsaka S, Nabekura J. Resting microglia directly monitor the functional state of synapses in vivo and determine the fate of ischemic terminals. J Neurosci. 2009;29:3974–80. 67. Davalos D, Grutzendler J, Yang G, Kim JV, Zuo Y, Jung S, Littman DR, Dustin ML, Gan WB. ATP mediates rapid microglial response to local brain injury in vivo. Nat Neurosci. 2005;8:752–8. 68. Skogstrand K, Hagen CM, Borbye-Lorenzen N, Christiansen M, Bybjerg- Grauholm J, Bækvad-Hansen M, Werge T, Børglum A, Mors O, Nordentoft M, et al. Reduced neonatal brain-derived neurotrophic factor is associ- ated with autism spectrum disorders. Transl Psychiatry. 2019;9:252. 69. Holm M, Morken TS, Fichorova RN, VanderVeen DK, Allred EN, Dammann O, Leviton A. Systemic inflammation-associated proteins and retinopathy of prematurity in infants born before the 28th week of gestation. Invest Ophthalmol Vis Sci. 2017;58:6419–28. 49. Galinsky R, Lear CA, Dean JM, Wassink G, Dhillon SK, Fraser M, Davidson 70. Mellor DJ, Diesch TJ, Gunn AJ, Bennet L. The importance of “awareness” for JO, Bennet L, Gunn AJ. Complex interactions between hypoxia- ischemia and inflammation in preterm brain injury. Dev Med Child Neurol. 2018;60:126–33. 50. Rathbone R, Counsell SJ, Kapellou O, Dyet L, Kennea N, Hajnal J, Allsop JM, Cowan F, Edwards AD. Perinatal cortical growth and childhood neurocognitive abilities. Neurology. 2011;77:1510–7. 51. Martinello KA, Meehan C, Avdic-Belltheus A, Lingam I, Ragab S, Hristova M, Tann CJ, Peebles D, Hagberg H, Wolfs T, et al. Acute LPS sensitization and continuous infusion exacerbates hypoxic brain injury in a piglet model of neonatal encephalopathy. Sci Rep. 2019;9:10184. 52. Burguillos MA, Deierborg T, Kavanagh E, Persson A, Hajji N, Garcia- Quintanilla A, Cano J, Brundin P, Englund E, Venero JL, Joseph B. Cas- pase signalling controls microglia activation and neurotoxicity. Nature. 2011;472:319–24. understanding fetal pain. Brain Res Brain Res Rev. 2005;49:455–71. 71. Missig G, Mokler EL, Robbins JO, Alexander AJ, McDougle CJ, Carlezon WA Jr. Perinatal immune activation produces persistent sleep alterations and epileptiform activity in male mice. Neuropsychopharmacology. 2018;43:482–91. 72. Coogan A, O’Connor JJ. Inhibition of NMDA receptor-mediated synaptic transmission in the rat dentate gyrus in vitro by IL-1 beta. NeuroReport. 1997;8:2107–10. 73. Galinsky R, Dhillon SK, Dean JM, Davidson JO, Lear CA, Wassink G, Nott F, Kelly SB, Fraser M, Yuill C, et al. Tumor necrosis factor inhibition attenuates white matter gliosis after systemic inflammation in preterm fetal sheep. J Neuroinflammation. 2020;17:92. 74. Jara JH, Singh BB, Floden AM, Combs CK. Tumor necrosis factor alpha stimulates NMDA receptor activity in mouse cortical neurons resulting in ERK-dependent death. J Neurochem. 2007;100:1407–20. Kelly et al. Journal of Neuroinflammation (2023) 20:124 Page 16 of 16 75. Billiards SS, Walker DW, Canny BJ, Hirst JJ. Endotoxin increases sleep and brain allopregnanolone concentrations in newborn lambs. Pediatr Res. 2002;52:892–9. 76. Luk WP, Zhang Y, White TD, Lue FA, Wu C, Jiang CG, Zhang L, Moldofsky 77. H. Adenosine: a mediator of interleukin-1beta-induced hippocampal synaptic inhibition. J Neurosci. 1999;19:4238–44. Inder TE, Buckland L, Williams CE, Spencer C, Gunning MI, Darlow BA, Volpe JJ, Gluckman PD. Lowered electroencephalographic spectral edge frequency predicts the presence of cerebral white matter injury in premature infants. Pediatrics. 2003;111:27–33. 78. Hayakawa F, Okumura A, Kato T, Kuno K, Watanabe K. Dysmature EEG pattern in EEGs of preterm infants with cognitive impairment: matura- tion arrest caused by prolonged mild CNS depression. Brain Dev. 1997;19:122–5. 79. Sävman K, Blennow M, Gustafson K, Tarkowski E, Hagberg H. Cytokine response in cerebrospinal fluid after birth asphyxia. Pediatr Res. 1998;43:746–51. 80. Teflioudi EP, Zafeiriou DI, Vargiami E, Kontopoulos E, Tsikoulas I. Soma- tosensory evoked potentials in children with bilateral spastic cerebral palsy. Pediatr Neurol. 2011;44:177–82. 81. Smyser CD, Inder TE, Shimony JS, Hill JE, Degnan AJ, Snyder AZ, Neil JJ. Longitudinal analysis of neural network development in preterm infants. Cereb Cortex. 2010;20:2852–62. 82. Smyser CD, Snyder AZ, Shimony JS, Blazey TM, Inder TE, Neil JJ. Effects of white matter injury on resting state fMRI measures in prematurely born infants. PLoS ONE. 2013;8: e68098. 83. Smith PLP, Mottahedin A, Svedin P, Mohn CJ, Hagberg H, Ek J, Mallard C. Peripheral myeloid cells contribute to brain injury in male neonatal mice. J Neuroinflamm. 2018;15:301. 84. Ardalan M, Chumak T, Vexler Z, Mallard C. Sex-dependent effects of perinatal inflammation on the brain: implication for neuro-psychiatric disorders. Int J Mol Sci. 2019. https:// doi. org/ 10. 3390/ ijms2 00922 70. 85. Galinsky R, van de Looij Y, Mitchell N, Dean JM, Dhillon SK, Yamaguchi K, Lear CA, Wassink G, Davidson JO, Nott F, et al. Magnetic resonance imaging correlates of white matter gliosis and injury in preterm fetal sheep exposed to progressive systemic inflammation. Int J Mol Sci. 2020;21:8891. 86. Johnson JI, Sudheimer KD, Davis KK, Kerndt GM, Winn BM. The sheep brain atlas. https:// msu. edu/ ~brains/ brains/ sheep/ index. html. Accessed 14 May 2023. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1192_j.eurpsy.2023.2418	European Psychiatry www.cambridge.org/epa Research Article Cite this article: Rømer V, Sivapalan P, Eklöf J, Nielsen SD, Harboe ZB, Biering- Sørensen T, Itenov T, Jensen J-US (2023). SARS-CoV-2 and risk of psychiatric hospital admission and use of psychopharmaceuticals: A nationwide registry study of 4,585,083 adult Danish citizens. European Psychiatry, 66(1), e50, 1–10 https://doi.org/10.1192/j.eurpsy.2023.2418 Received: 22 October 2022 Revised: 10 April 2023 Accepted: 04 May 2023 Keywords: Long COVID; psychiatry; psychopharmaceuticals; SARS-CoV-2 Corresponding author: Jens-Ulrik S. Jensen; Email: jens.ulrik.jensen@regionh.dk V.R. and P.S. contributed equally. © The Author(s), 2023. Published by Cambridge University Press on behalf of the European Psychiatric Association. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http:// creativecommons.org/licenses/by/4.0), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited. SARS-CoV-2 and risk of psychiatric hospital admission and use of psychopharmaceuticals: A nationwide registry study of 4,585,083 adult Danish citizens Valdemar Rømer1 Susanne D. Nielsen2 Theis Itenov5,6 , Pradeesh Sivapalan1 , Zitta B. Harboe3 and Jens-Ulrik S. Jensen1,5,7 , Josefin Eklöf1 , , Tor Biering-Sørensen4 , 1Section of Respiratory Medicine, Herlev-Gentofte University Hospital, Hellerup, Denmark; 2Department of Infectious Diseases, University Hospital of Copenhagen, Copenhagen, Denmark; 3Department of Pulmonary and Infectious Diseases, University Hospital of Copenhagen, North Zealand, Denmark; 4Department of Cardiology, Herlev-Gentofte University Hospital, Hellerup, Denmark; 5Centre for Health and Infectious Diseases Research (CHIP), University Hospital of Copenhagen, Copenhagen, Denmark; 6Department of Anaesthesiology and Intensive Care, University Hospital of Copenhagen, Copenhagen, Denmark and 7Department of Clinical Medicine, Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark Abstract Background. Current evidence on the risk of admission- or medication-requiring psychiatric sequelae of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection is limited to selected populations, short durations, and loss to follow-up. This study examined if SARS-CoV-2 infection was associated with increased long-term risk of psychiatric admis- sions and de novo prescription of psychoactive medication in the general population of Denmark. Methods. Adults (≥18 years) were assigned to either the control or SARS-CoV-2 group based on polymerase chain reaction (PCR) tests between 1 January 2020 and 27 November 2021. Infected subjects were matched 1:5 to control subjects by propensity score. Incidence rate ratios (IRRs) were calculated. Adjusted Cox regression was applied to the unmatched population with SARS- CoV-2 infection as a time-dependent covariate. Follow-up time was 12 months or until the end of the study. Results. A total of 4,585,083 adults were included in the study. Approximately 342,084 had a PCR-confirmed SARS-CoV-2 infection and were matched 1:5 with 1,697,680 controls. The IRR for psychiatric admission was 0.79 in the matched population (95% confidence interval [CI]: 0.73–0.85, p < 0.001). In the unmatched population, the adjusted hazard ratios (aHR) for psychiatric admission were either below 1.00 or with a 95% CI lower limit of 1.01. SARS- CoV-2 infection was associated with an increased risk of de novo prescription of psychoactive medication in both the matched (IRR 1.06, 95% CI: 1.02–1.11, p < 0.01) and unmatched population (HR 1.31, 95% CI: 1.28–1.34, p < 0.001). Conclusions. We found a signal of increased use of psychoactive medication, specifically benzodiazepines, among SARS-CoV-2-positive persons, but the risk of psychiatric admissions did not increase. Introduction It is estimated that more than 1 in 10 with acute COVID-19 experience symptoms persisting after the primary severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection – commonly known as long COVID or post-COVID conditions [1, 2]. The symptoms reported vary from chest pain, fatigue, dyspnea, coughing, cognitive impairment, psychological distress, and several others [3–5]. A retrospective cohort study from the United States found an association between coronavirus disease-2019 (COVID-19) and an increased risk of first psychiatric diagnosis (anxiety was the most prevalent) within 14 to 90 days post-infection compared to the risk of psychiatric diagnoses after other infections such as influenza [6]. Other studies also reported an increased occurrence of neurological and psychiatric disorders or altered mental state, where most psychiatric diagnoses were new/first diagnoses [7–9]. For the aforementioned studies, the follow-up time was relatively short, ranging from 2 to 24 weeks, which could have been a weakness, since psychiatric sequelae may take some time to develop [10]. There are little data on mental health in adults recovering from COVID-19, especially in those with symptoms weeks to months after their initial infection or Long COVID [6]. 2 Rømer et al. It has been speculated whether SARS-CoV-2 infection could influence mental outcomes via a biological link: the first case of suspected SARS-CoV-2 meningitis was seen in February 2020 [11]. Since then, it has been established that SARS-CoV-2 infection is associated with several neurobiological outcomes, resulting from either hyperinflammatory or hypercoagulable states, direct central nervous system (CNS) infection, and postinfectious immune- mediated processes [12–15]. Other coronaviruses have been shown to have such potential [16, 17]. Furthermore, it has been suggested that these impacts on the CNS could be linked to psychiatric sequelae of SARS-CoV-2 infection [14], although conflicting evi- dence on inflammation and symptom severity of some psychiatric disorders has been reported [18]. Outcomes of SARS-CoV-2 infection on mental health are poorly investigated [19–21]. Understanding potential implications of SARS-CoV-2 infection on mental health to preempt psychiatric events is important. Based on the diverging results from small studies, with somewhat selected populations and with considerable loss to follow-up, more solid data are needed from studies taking these limitations into account. In this study, which is a nationwide cohort study with all adults in our country, with accurate information on the timing of poly- merase chain reaction (PCR) positivity of SARS-CoV-2 virus, and complete follow-up, we sought to investigate whether a first positive SARS-CoV-2 PCR test is associated with an increased risk of psychiatric admission or prescription of psychoactive medication as compared to uninfected controls. Methods We conducted a nationwide retrospective population-based regis- try study utilizing the National Danish registries. The study was approved by the Danish Data Protection Agency (j.no. P-2021- 360). In Denmark, informed consent is not required for retrospect- ive studies. Hypotheses (1) (2) SARS-CoV-2 infection increases the risk of being admitted to a psychiatric hospital department. SARS-CoV-2 infection increases the risk of de novo prescrip- tion of psychoactive medication. Data sources All persons in Denmark are assigned a unique personal identifica- tion number, which is used in all national registers, enabling accurate linkage between them [22]. The following registries were used: (1) The Danish Central Personal Registry containing person identification number, sex, vital status, and death date of all persons residing in Denmark since 1968 [22] (2) The Danish National Patient Registry containing data on all somatic hospital contacts nationwide since 1977, including all psychiatric hospital contacts since 1994 [23] (3) The National Prescription Registry containing data on dis- pensed prescriptions since 2004 [24] (4) The Danish Microbiology Database containing data on PCR- confirmed SARS-CoV-2 infection since February 2020 [25]. (5) The Danish Vaccination Registry containing data on vaccin- ations since 1996 [26]. We did not have data from general practitioners; thus, diagnoses were derived from hospital registries. Reliable data on the indica- tions for prescribed medications were not available. We did not have data on the specialization of the prescriber. Data on income and postal code were not obtained. During the study period, frequent PCR test for COVID-19 was encouraged and freely avail- able regardless of symptoms. The PCR laboratory data did only contain one instance per person being either the first SARS-CoV- 2-positive test, if any, or the last SARS-CoV-2 negative test, and did not contain information on symptoms or severity of disease. Study population The population was defined as all adults (18 years or older) residing in Denmark (except Greenland and Faroe Islands) by January 1, 2020. The only exclusion criteria were invalid SARS- CoV-2 PCR test date data (e.g., test registered postmortem) to prevent persons at risk from not being eligible – thus including persons with an ongoing psychiatric hospitalization at the begin- ning of the study. SARS-CoV-2 infection was defined by a regis- tered positive PCR test nonregarding symptoms, severity of disease, and hospitalization. The first registered SARS-CoV-2 infection in Denmark was on February 26, 2020 [27]. The study period was defined as January 1, 2020 to November 27, 2021 to report pre-pandemic baseline values and omit interference with the omicron-variant of which the first registered case in Denmark was November 28, 2021 [27]. Comorbidities were defined by a hospital contact or outpatient clinic contact with an active diag- nosis within 5 years before baseline and classified using the Charlson Comorbidity Index (CCI) [28]. Medication at baseline was defined by ≥1 collected prescription within 1 year prior to baseline. A propensity score-matched model was used in the primary analysis, comparing the SARS-CoV-2 infected to matched con- trols to assess the risk of reaching an endpoint between similar groups. Follow-up for SARS-CoV-2-infected subjects started by the date of infection. Follow-up for controls was started by the same date as their matched infected subject, or they were excluded if dead at the time of starting follow-up. Follow-up for all indi- viduals in the primary analysis was 12 months or until the end of the study. To further test the findings of the primary analysis, and to also provide a true nationwide perspective, a secondary analysis of the complete, unmatched population was followed by January 1, 2020. All persons started as controls; SARS-CoV-2 infection was treated as a time-dependent variable allowing persons to change status during the study from noninfected to infected. Subjects were fol- lowed until the end of the study or 12 months postinfection. Outcomes The primary outcome was psychiatric hospital admissions, defined as psychiatric hospital contacts of more than 24 hours initiated at least one calendar-date after inclusion with International Classifi- cation of Diseases, Tenth Revision (ICD-10) codes F-20 to F-50 as the primary diagnosis. Transfers from somatic to psychiatric units were also considered for this outcome. This selection of diagnoses includes affective disorders, anxiety disorders, and psychotic dis- orders, but not, e.g., personality disorders, attention deficit hyper- activity disorder, and mental retardation, as we could not reasonably suspect a causal association between SARS-CoV-2 infection and these. European Psychiatry 3 The secondary outcome was the de novo prescription of any psychoactive medication, regardless of indication, as we did not have data on this. However, assuming the somatic indications are infrequent compared to the psychiatric indications when not hospitalized, the risk ratio estimates should still be reliable as an approximate measure for medication-requiring psychiatric con- ditions. De novo psychoactive medication was considered a cat- egorical variable (either none or at least one prescription). It was assumed that most persons in active psychopharmacological treatment would collect their prescribed medication in intervals of less than 3 months. As we did not have data on whether a prescription collection was due to the commencement or con- tinuation of treatment, individuals with a prescription collection within 90 days were excluded from the analysis to increase the likelihood that events were true de novo prescriptions. Psycho- active medication was defined by the Anatomical Therapeutic Chemical (ATC) codes for antidepressants (N06A), benzodiazep- ines and benzodiazepine-like drugs (N03AE, N05BA, N05CD, N05CF), antipsychotics (N05A except N05AN01), and lithium (N05AN01). In the case of a positive SARS-CoV-2 PCR test on the calendar date of admission in the time-dependent Cox-regression on the unmatched population, this was considered an event happening while noninfected, as to avoid the outcome psychiatric hospital admission itself increasing the risk of an event and a short time-to- event in the SARS-COV-2-infected group. The proportion of persons vaccinated against COVID-19 in the different groups was reported, including the proportion in the SARS-CoV-2 group vaccinated ≥14 days before infection. Statistical analysis Continuous variables were presented as mean values with 95% confidence intervals (CIs) or medians with interquartile range (IQR). Categorical variables were presented as proportions. In the propensity score-matched population, the control sample was created utilizing a parallel balanced propensity score-matched model using the nearest neighbor greedy-match algorithm from the MatchIt package (R v. 4.1.3) [29]. SARS-CoV-2 infected subjects were matched 1:5 to controls. Matching was performed on age, sex, CCI score, baseline history of psychoactive medication, and base- line history of hospital contacts with registered psychiatric diag- noses in order to achieve presumed equal-risk groups at baseline. Outcomes were reported as incidence rates (IR) and incidence rate ratios (IRRs) with 95% CI. We analyzed the unmatched population using a Cox proportional hazard regression model with SARS-CoV-2 infection as a time- dependent covariate. Calendar time was accounted for in the regres- sion model as all participants were followed from the same calendar date (1 January 2020). The Cox model was tested for the proportional hazards assumption and linearity. In addition, the model was adjusted for age, sex, and CCI score. In this multivariate regression on the total unmatched population, it was decided not to adjust for baseline psychiatric history to avoid overfitting and removal of relevant dif- ferences between groups. Outcomes were reported as hazard ratios (HR) and adjusted hazard ratios (aHR) with 95% CI and p-values. Interaction analyses on psychiatric admission was performed using the two-way analysis of variance (ANOVA) test on Cox regressions comparing additive to multiplicative models [30]. Inter- action was tested between [SARS-CoV-2 infection status] (time- dependent) and independent variables with suspected interaction and highest main effect: (i) (ii) (iii) (iv) [SARS-CoV-2 Alpha versus Delta variant community- domination] (time-dependent) [age ≥ 40 years versus age < 40 years] (baseline) [male versus female] [COVID-19 vaccination ≥14 days before infection yes ver- sus no] Variant community-domination time periods were chosen for interaction-testing as different social restrictions and risk of severe infection if infected in this time periods differed and could plausibly explain more than infection with SARS-CoV-2 during any period itself, as disease severity was different between variants [27, 31]. SARS-CoV-2 Alpha variant community domination was defined as the period from the beginning of study until 4 July 2021. The delta variant dominated from 5 July 2021 until the end of the study (the study was terminated before any registered SARS-CoV- 2 Omicron variant infections in Denmark) based on evidence from the Danish health authorities (Statens Serum Institut) [27]. Age was suspected to increase the risk of psychiatric sequelae following infection. A review of COVID-19 sequalae reports mean ages of patients with symptoms post-infection from 43 to 63 years [32] and a study on self-reported sequelae of SARS-CoV-2 in different age groups found that all groups with age ≥ 40 years was associated with increased odds of sequelae compared to all groups with age < 40 years [33], why this was chosen as the cut-off value in the interaction-test. Further, it can be suspected that the effect of the SARS-CoV-2 infection and the associated induced social isolation and disease on the risk of psychiatric admission differs between this group and people of older age in a non- additive manner. Sex was included in the interaction analysis as characteristics between males and females admitted to a psychi- atric department varies [34], and as was sex suspected to influence the susceptibility to psychiatric sequelae of COVID-19 [34, 35]. It was decided to test interaction for vaccination against COVID-19 as this suspected interaction seems relevant and plausible however unsure; infections ≥14 days post-vaccination have been associated with a reduced risk of psychotic disorder (but not mood and anxiety disorders) [36] and reduced odds of long COVID (how- ever with an upper limit of the 95% CI of 0.99) [37]. When an interaction was found, outcomes were presented separately for interacting variables. Results A total of 4,585,094 individuals aged ≥18 years were identified. In total, 11 individuals were excluded due to incorrect registration date of SARS-CoV-2 PCR test, leaving a total study population of 4,585,083 subjects. Of these, 342,084 were registered with a con- firmed positive SARS-CoV-2 PCR test. In the propensity score- matched population, all the SARS-CoV-2 positive individuals were matched 1:5 with 1,697,680 controls (12,740 controls were removed due to death before infection of match) (Figure 1). No subjects were lost to follow-up. The baseline values of those infected with SARS-CoV-2 differed from the total population. The infected were generally younger, less medicated with psychoactive medication, and had fewer comorbid- ities. Baseline values were overall similar in the SARS-CoV-2 infected and controls (Table 1). 4 Rømer et al. Figure 1. Study flowchart. In the propensity score-matched population, subjects in the SARS-CoV-2 Group and their corresponding matches were observed from the date of the registered positive PCR-test, or removed if dead by then. In the unmatched population, all subjects were observed from 1 January 2020, and SARS-CoV-2 infection was treated as a time dependent variable. Table 1. Baseline patient demographic and clinical characteristics by 1 January 2020 Characteristics Age, median (IQR) Male sex, n (%) History of psychiatric admission ≥1 psychiatric admission within 1 year, n (%) Depression Anxiety disorder Bipolar or mania Schizophrenia or psychosis Medication Prescribed psychoactive medication, n (%) Antidepressants BZD and BZD-like Antipsychotics Lithium Comorbidities Specialist treated psychiatric illness, n (%) Depression Anxiety disorder Bipolar Schizophrenia COPD, n (%) Diabetes Mellitus, n (%) Stroke and transient cerebral ischemia, n (%) Unmatched Propensity score-matched population Total adult population (N = 4,585,083) SARS-CoV-2 infected (N = 342,084) 49 (33 to 65) 2,256,635 (49.22) 40 (27 to 54) 167,674 (49.02) Matched controls (N = 1,697,680) 40 (27 to 53) 832,363 (49.03) 18,561 (0.40) 3,701 (0.08) 1,419 (0.03) 2,011 (0.04) 7,491 (0.16) 594,093 (12.96) 400,905 (8.74) 230,009 (5.02) 115,765 (2.52) 8,962 (0.20) 233,253 (5.09) 81,448 (1.78) 64,487 (1.41) 15,550 (0.34) 27,603 (0.60) 148,711 (3.24) 140,826 (3.07) 90,266 (1.97) 878 (0.26) 202 (0.06) 99 (0.03) 98 (0.03) 267 (0.08) 31,827 (9.30) 22,067 (6.45) 11,365 (3.32) 5,673 (1.66) 417 (0.12) 16,123 (4.71) 5,529 (1.62) 4,709 (1.38) 776 (0.23) 1,162 (0.34) 9,635 (2.82) 8,297 (2.43) 3,925 (1.15) 5,451 (0.32) 986 (0.06) 443 (0.03) 530 (0.03) 2,197 (0.13) 155,203 (9.14) 108,315 (6.38) 52,811 (3.11) 32,175 (1.90) 2,479 (0.15) 77,806 (4.58) 24,835 (1.46) 22,953 (1.35) 4,682 (0.28) 8,750 (0.52) 39,529 (2.33) 34,145 (2.01) 18,744 (1.10) Charlson Comorbidity Index, mean (CI) 1.18 (1.18 to 1.18) 0.68 (0.67 to 0.68) 0.64 (0.64 to 0.64) Abbreviations: CI, 95% confidence interval; COPD, chronic obstructive lung disease; IQR, interquartile range. European Psychiatry 5 The proportion receiving at least one vaccine against COVID-19 varied between groups. In the total population, 4,096,800 (89.25%) received at least one vaccine during the study. Among the SARS- CoV-2 infected, this number was 273,095 (79.83%), and 1,515,930 (89.29%) for the matched controls. In the SARS-CoV-2 group, 82,921 (24.24%) were vaccinated ≥14 days before infection. Propensity score-matched population The incidence rate of psychiatric admission was 360 per 100,000 subjects in the SARS-CoV-2 infected group, and 460 per 100,000 person-years among the matched controls. This corresponds to an IRR of 0.79 (95% CI 0.73 to 0.85, p < 0.001), thus SARS-CoV-2 infection was not associated with an increased risk of psychiatric admission. Stratifying for vaccination against COVID-19 ≥ 14 days before infection or start of observation of controls, the IRR of psychiatric admission did not relevantly differ between the vaccinated (IRR 0.77, 95% CI 0.55 to 1.08, p = 0.13) and the unvaccinated (IRR 0.79, 95% CI 0.72 to 0.86, p < 0.001). Subdividing admissions by primary diagnosis of admission, reduced incidence of admission with schizophrenia or psychosis seems to be the main driver of the overall reduced IRR, whereas admissions coded with either depression, anxiety disorders, and bipolar/mania were neutral (Table 2). Furthermore, the relative risk of psychiatric admission was lowest during the second-month postinfection (Figure 2B). SARS-CoV-2 infection was associated with an increased risk of de novo prescription of psychoactive medication was found (IRR 1.06, CI 1.02 to 1.11, p < 0.001). When granulating the results for type of psychoactive medication, use of benzodiazepines and benzodiazepine-like medication seemed to be a driver of the signal: IRR of 1.17 (CI 1.10 to 1.25, p < 0.001) (Table 2). Unmatched population Interaction analysis Interaction analysis was performed on psychiatric hospital admis- sion for variables [SARS-CoV-2 infection status] and, respectively, (i) [age < 40 years versus age ≥ 40 years] (p < 0.001), (ii) sex [male versus female] (p = 0.12), (iii) [SARS-CoV-2 Alpha versus Delta variant community-domination] (p = 0.01) and (iv) [COVID-19 vaccination ≥14 days before infection yes versus no] (p = 0.06). As a result of this, HR were presented separately for all combinations of interacting variables (i) and (iii). Hazard estimates SARS-CoV-2 infection was not associated with an increased risk of psychiatric hospital admission in the age group <40 years, neither in the Alpha variant community-dominance period (aHR 0.79, CI 0.75 to 0.82, p < 0.001) nor in the Delta variant community- dominance period (aHR 0.67, CI 0.56 to 0.80, p < 0.001). In the age group ≥40 years, SARS-CoV-2 infection was not associated with an increased risk of psychiatric hospital admission in the Alpha variant community-dominance period (aHR 0.96, CI 0.87 to 1.06, p = 0.38). However, for this age group, the Delta variant community-dominance period was associated with an increased risk of psychiatric hospital admission (aHR 1.34, 1.01 to 1.77, p = 0.04). Refer to Figure 3 for all adjusted and unadjusted HR estimates on psychiatric admission. Table 2. Incidence rates and incidence rate ratios of psychiatric outcomes after SARS-CoV-2 infection compared to propensity score-matched controls observed from the same date as the infected match Follow-up Lost to follow-up, n (%) Psychiatric admission Admitted within 12 months, IR × 105 Admitted within 12 months, IRR (CI) Depression Anxiety disorders Bipolar or manias Schizophrenia or psychosis Psychoactive medication De novo prescription within 12 months, IR × 105 De novo prescription within 12 months, IRR (CI) Antidepressants BZD and BZD-like Antipsychotics Lithium Matched controls (N = 1,697,680) SARS-CoV-2 infected (N = 342,084) p-value 0/2,687,120 (0%) 0/342,084 (0%) 459.63 Ref. Ref. Ref. Ref. Ref. 1,424 Ref. Ref. Ref. Ref. Ref. 360.42 0.79 (0.73 to 0.85) 0.92 (0.78 to 1.09) 1.03 (0.81 to 1.31) 1.07 (0.86 to 1.32) 0.64 (0.56 to 0.74) 1,512 1.06 (1.02 to 1.11) 1.04 (0.99 to 1.09) 1.17 (1.10 to 1.25) 0.92 (0.82 to 1.03) 0.97 (0.53 to 1.75) <0.001 0.36 0.82 0.55 <0.001 <0.01 0.17 <0.001 0.14 0.91 Note: De novo prescription of psychoactive medication was defined as the first new prescription of any psychoactive medication, excluding subjects with a history of psychoactive medication within 3 months prior to start of observation. Abbreviations: BZD, benzodiazepines; CI, 95% confidence interval; IR, incidence rate; IRR, incidence rate ratio. 6 Rømer et al. Figure 2. Psychiatric hospital admission incidence comparison after SARS-CoV-2 infection compared to propensity score-matched controls observed from the same date as the infected match. (A) Daily psychiatric admission incidence rates. (B) Psychiatric admission incidence rate ratios (IRRs) with 95% confidence intervals. SARS-CoV-2 was associated with an increased risk of a de novo prescription of any psychoactive pharmaceutical (aHR 1.31, CI 1.28 to 1.34, p < 0.001). An association with increased risk of de novo prescription of all subgroups of psychoactive medication, except lithium, was found (Table 3:). Except for the apparent increased risk of de novo prescription of antidepressant and antipsychotic drugs for the unmatched population, all other estimates were similar to the risk estimates obtained from the propensity score-matched population. Discussion In this nationwide registry-based cohort study investigating whether PCR-confirmed SARS-CoV-2 infection is associated with European Psychiatry 7 Figure 3. Hazard ratio estimates of psychiatric hospital admission following SARS-CoV-2 infection. Infection and alpha/delta variant dominance were treated as time-dependent covariates. During the study 342,084 of 4,585,083 subjects were infected with SARS-CoV-2 (7.46%). No patients were lost to follow-up. psychiatric hospital admission and prescription of psychoactive medication, we found no clinically relevant associations between SARS-CoV-2 infection and increased risk of psychiatric admission, neither in the propensity score-matched population nor in the unmatched population. Vaccination against COVID-19 did not seem to alter this risk. Generally, SARS-CoV-2 infection was not or only slightly asso- ciated with an increased risk of de novo prescription of any psy- choactive medication. We found an increased risk of de novo prescription of benzodiazepines and benzodiazepine-like pharma- ceuticals in the propensity score matched model and the adjusted Cox regression, but no other clinically relevant elevated risks were found analyzing sub-groups of psychoactive medication. Benzodi- azepines are used for a wide range of conditions, including acute anxiety, alcohol- and substance-withdrawal, insomnia, delirium, and prior to medical, dental, or surgical procedures. According to the Danish national guidelines, the first choice for pharmacological treatment of anxiety disorders is antidepressants (selective sero- tonin reuptake inhibitors specifically) [38]; however, benzodiazep- ines still seem to be considered a good quick-onset short-term treatment option by many clinicians [39] why this association with SARS-CoV-2 and increased risk of benzodiazepines/benzodiazep- ine like pharmaceuticals could be interpreted as an increased long- term risk of acute pathological non-hospital-requiring anxiety possibly in combination with insomnia and alcohol- or substance-withdrawal. As the prescriber and indication for pre- scribed psychoactive medication was not known, it is not possible to conclude neither if the increased benzodiazepine treatment was specialist-requiring nor the proportions of the risk estimate driven by psychiatric and somatic indication. Risk estimates should not be overinterpreted, as they are likely influenced by unmeasured confounding, possibly that individuals 8 Rømer et al. Table 3. Hazard ratio estimates for de novo prescription of psychoactive medication after SARS-CoV-2 infection in the unmatched population Total population (N = 4,585,083) p-value Follow-up Lost to follow-up, n/N (%) 0/4,585,083 (0.00) SARS-CoV-2 infected during study, n/N (%) Psychoactive medication De novo prescription within 12 months, unadjusted HR (CI) Adjusteda HR (CI) Antidepressants BZD and BZD-like Antipsychotics Lithium 342,084/4,585,083 (7.46) 1.12 (1.10 to 1.15) <0.001 1.31 (1.28 to 1.34) 1.09 (1.06 to 1.12) 1.65 (1.60 to 1.71) 1.05 (1.00 to 1.11) 1.20 (0.88 to 1.63) <0.001 <0.01 <0.001 <0.05 0.24 Note: Infection was treated as a time-dependent covariate. Abbreviations: BZD, benzodiazepines; CI, 95% confidence interval; HR, hazard ratio. aAdjusted for age, sex, and Charlson’s Comorbidity Index score. infected with SARS-CoV-2 could have a less “anxious” mindset resulting in exposure to the infection. Schizophrenia was the most common primary diagnosis of psychiatric admission; however, these patients were underrepresented in the SARS-CoV-2 group, and propensity score matching did not fully account for this difference. This could explain the unexpected protective risk esti- mate of SARS-CoV-2 on hospital admission with schizophrenia or psychosis as the primary diagnosis. Also, as there was a general reduction in psychiatric hospitalizations in Denmark during the pandemic [40], it is likely that psychiatric hospital admission- requiring individuals were either not admitted at all or hospitalized for severe COVID-19 in a somatic hospital to ensure isolation until recovered from SARS-CoV-2 infection, thus not registered as a psychiatric admission by diagnosis code, unless transferred after discontinuation of isolation and COVID-19 treatment. Only a few studies on long-term risk of prescription of psycho- active medication after SARS-CoV-2 infection exist; a retrospective registry-based study of American veterans with a similar study design (prescription based, propensity score matching and syn- chronized inclusion dates of infected and controls) showed an increased risk of antidepressant use (HR 1.55) and use of benzodi- azepines (HR 1.65) [41]. However, the higher HR reported could be explained at least in part by the homogeneity (all military veterans, 89% males, 81% overweight or obese, 59% current or former smokers) and high mean age (63 years) of the American veteran cohort. Therefore, it seems very plausible that the selected cohort was more vulnerable to psychiatric events. This study better rep- resents a general adult population as we included all Danish adults with no exceptions or further selection. The outcomes of this study not signaling increased risk of psychiatric sequelae challenge the signals and interpretation of other studies. A large systematic review on hospitalized COVID- 19 patients reported both short and long-term (up to 7 months after discharge) psychiatric sequelae such as depression and anxiety [42]. However, most included studies did not have a control group or compared only hospitalized COVID-19 patients to non- hospitalized healthy controls. These studies do not provide know- ledge on mental health sequelae of SARS-CoV-2 infection but only of severe COVID-19 requiring hospitalization. The ANCHOHVID study, an Andalusian prospective cohort study, did not detect any difference in mental health-related sequelae after discharge between patients hospitalized for COVID-19 and patients hospi- talized for other causes [43]. This supports the findings of the present study that SARS-CoV-2 infection itself does not seem to be associated with an increased long-term risk of severe psychiatric sequelae. It can be speculated that psychiatric sequelae are instead a more general characteristic of severe disease and hospital admission. A major strength of this study is that we followed the total Danish adult population above 18 years of age (4.5 million indi- viduals). This is important, as previous studies based on hospital- data and smaller selected databases are not representative of the general population; opposite, this study very precisely represents reality of the broad general population not excluding non-hospital requiring SARS-CoV-2 infection. We had a longer follow-up than most studies investigating psychiatric sequalae of SARS-CoV-2 infection (12 months). No subjects were lost to follow-up and we had access to validated and complete data on hospital admissions and prescriptions. PCR tests were free of charge and widely avail- able. Regular testing for SARS-CoV-2 was encouraged in periods with high transmission rates, and a negative test was required in many settings such as restaurant visits and museums. Access to treatment and hospitalization is free of charge for all Danish residents, thus bias due to avoidance of proper psychiatric treat- ment for economic reasons can be ruled out. Targeted treatment of COVID-19 with monoclonal antibodies was not standard of care during the study period, so effective treatment of COVID-19 can- not be assumed to have influenced the outcomes. There are some limitations to this study. There were major differences in baseline between the uninfected and the infected in the total population. Although we did perform both adjusting and matching, in sequence, there may have been some residual con- founding, possibly including income, social class, and geography. The lower vaccination rate in the SARS-CoV-2 infected group compared to the matched controls could indicate differences in socio-demographic factors such as a lower household income and lower level of education among the unvaccinated [44]. Further, in a similar population, a lower COVID-19 vaccination rate has been reported among the psychiatric vulnerable – despite being offered vaccination earlier than the general population [45], thus possibly affecting risk estimates in an unknown direction. Despite PCR confirmation being encouraged after a positive SARS-CoV-2 anti- gen test, there is a risk some might not have PCR-confirmed the diagnosis, and thus be included in the non-infected group although actually being infected. Different test patterns in vulnerable groups could affect risk estimates: according to a study on an overlapping population and time period, psychiatric vulnerable persons had lower odds of PCR testing than the general population [46]. We did not have access to SARS-CoV-2 variant data; hence the distribution of variants among the infected is unknown, which could have affected the outcomes and comparability of the findings to other studies. It is however known that only the alpha and delta variants have been community dominant during the study period [27], and as we stratified for this on the primary outcome in the Cox regres- sion, we do not consider this a major limitation. As the delta variant only dominated for approximately 4 months, this could explain the wide confidence interval of the hazard estimate for this strata, and possibly part of the apparent increase in hazard among the infected during this period if the ratio of hazards was not truly constant over time. We did not have information on the indication for the European Psychiatry 9 prescriptions of psychoactive medication; psychoactive medication is prescribed for a variety of conditions like insomnia, delirium, and pain. Also, it cannot be ruled out that some events of de novo prescription of psychoactive medication were not true commence- ments of treatment, as the period chosen was arbitrary and could not account for all discontinuations before the beginning of the study – neither that some events were actually continued treatment in persons with collection pick-up frequency of more than 90 days. In conclusion, this is the first study to investigate the impact of SARS-CoV-2 infection on psychiatric outcomes in an entire coun- try, with access to precise data on test positivity and complete follow-up. Psychiatric admissions did not increase among the SARS-CoV-2 infected. However, the risk of prescription of psycho- active medication, specifically benzodiazepines, seemed to increase. Possibly, our results reflect a moderately increased risk of mild- to-moderate anxiety-spectrum disorders among the SARS-CoV-2 positive. We believe our results should lead to close monitoring of psychiatric sequalae in future pandemics of SARS-CoV-2 and other infectious viruses in the general population, as it seems post-acute treatment-requiring anxiety-related psychiatric manifestations ori- gin may occur in a broader population regardless of symptom severity. Data availability statement. We believe that knowledge sharing increases the quantity and quality of scientific results. Sharing of relevant data will be discussed within the study group upon reasonable request. Acknowledgments. We thank the COP:TRIN (coptrin.dk) Steering Commit- tee and the CURE group for providing input into this study at meetings. Author contribution. Conceptualization: P.S., J.E., J-U.S.J.; Data acquisition and validation: P.S., J.E.; Formal analysis: V.R.; Funding acquisition: J-U.S.J.; Methodology: P.S., J.E., J-U.S.J.; Project administration: J-U.S.J.; Supervision: P.S., J-U.S.J.; Visualization: V.R.; Writing—original draft preparation: V.R; all authors participated in Writing—review and editing; All authors have read and agreed to the published version of the manuscript. Financial support. This study was supported by the Novo Nordisk Founda- tion (Grant No. NNF20SA0062834). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interest. The authors declare none. References [1] Altmann DM, Boyton RJ. Decoding the unknowns in long covid. BMJ. 2021;372:n132. doi:10.1136/bmj.n132. [2] Carfi A, Bernabei R, Landi F, Gemelli Against COVID-19 Post-Acute Care Study Group. Persistent symptoms in patients after acute COVID-19. JAMA. 2020;324(6):603–5. doi:10.1001/jama.2020.12603. [3] Cabrera Martimbianco AL, Pacheco RL, Bagattini AM, Riera R. Fre- quency, signs and symptoms, and criteria adopted for long COVID: a J Clin Pract. 2021;75:e14357. doi:10.1111/ systematic review. ijcp.14357. Int [4] Lopez-Leon S, Wegman-Ostrosky T, Perelman C, Sepulveda R, Rebolledo P, Cuapio A, et al. More than 50 long-term effects of COVID-19: A systematic review and meta-analysis. Sci Rep. 2021;11:16144. doi: 10.1038/s41598-021-95565-8. [5] SeyedAlinaghi S, Afsahi AM, MohsseniPour M, Behnezhad F, Salehi MA, Barzegary A, et al. Late complications of COVID-19; a systematic review of current evidence. Arch Acad Emerg Med. 2021;9(1):e14. [6] Taquet M, Luciano S, Geddes JR, Harrison PJ. Bidirectional associations between COVID-19 and psychiatric disorder: retrospective cohort studies of 62 354 COVID-19 cases in the USA. Lancet Psychiatry. 2021;8(2):130– 40. doi:10.1016/S2215-0366(20)30462-4. [7] Varatharaj A, Thomas N, Ellul MA, Davies NWS, Pollak TA, Tenorio EL, et al. Neurological and neuropsychiatric complications of COVID-19 in 153 patients: a UK-wide surveillance study. Lancet Psychiatry. 2020;7(10): 875–82. doi:10.1016/S2215-0366(20)30287-X. [8] Nalleballe K, Reddy Onteddu S, Sharma R, Dandu V, Brown A, Jasti M, et al. Spectrum of neuropsychiatric manifestations in COVID-19. Brain Behav Immun. 2020;88:71–4. doi:10.1016/j.bbi.2020.06.020. [9] Taquet M, Geddes JR, Husain M, Luciano S, Harrison PJ. 6-month neurological and psychiatric outcomes in 236 379 survivors of COVID- 19: a retrospective cohort study using electronic health records. Lancet Psychiatry. 2021;8(5):416–27. doi:10.1016/S2215-0366(21)00084-5. [10] Premraj L, Kannapadi NV, Briggs J, Seal SM, Battaglini D, Fanning J, et al. Mid and long-term neurological and neuropsychiatric manifestations of post-COVID-19 syndrome: a meta-analysis. J Neurol Sci. 2022;434: 120162. doi:10.1016/j.jns.2022.120162. [11] Moriguchi T, Harii N, Goto J, Harada D, Sugawara H, Takamino J, et al. A first case of meningitis/encephalitis associated with SARS-Coronavirus-2. Int J Infect Dis. 2020;94:55–8. doi:10.1016/j.ijid.2020.03.062. [12] Roy D, Ghosh R, Dubey S, Dubey MJ, Benito-Leon J, Kanti Ray B. Neuro- logical and neuropsychiatric impacts of COVID-19 pandemic. Can J Neurol Sci. 2021;48(1):9–24. doi:https://doi.org/10.1017/cjn.2020.173. [13] Koralnik IJ, Tyler KL. COVID-19: a global threat to the nervous system. Ann Neurol. 2020;88(1):1–11. doi:10.1002/ana.25807. [14] Fotuhi M, Mian A, Meysami S, Raji CA. Neurobiology of COVID-19. J Alzheimers Dis. 2020;76(1):3–19. doi:10.3233/JAD-200581. [15] Roman GC, Spencer PS, Reis J, Buguet A, Faris MEA, Katrak SM, et al. The neurology of COVID-19 revisited: a proposal from the environmental neurology specialty Group of the World Federation of neurology to implement international neurological registries. J Neurol Sci. 2020;414: 116884. doi:10.1016/j.jns.2020.116884. [16] Correia AO, Feitosa PWG, Moreira JLS, Nogueira SAR, Fonseca RB, Nobre MEP. Neurological manifestations of COVID-19 and other cor- onaviruses: a systematic review. Neurol Psychiatry Brain Res. 2020;37: 27–32. doi:10.1016/j.npbr.2020.05.008. [17] Cataldi M, Pignataro G, Taglialatela M. Neurobiology of coronaviruses: potential relevance for COVID-19. Neurobiol Dis. 2020;143:105007. doi: 10.1016/j.nbd.2020.105007. [18] Renaud-Charest O, Lui LMW, Eskander S, Ceban F, Ho R, Di Vincenzo JD, et al. Onset and frequency of depression in post-COVID-19 syndrome: a systematic review. J Psychiatr Res. 2021;144:129–37. doi:10.1016/j.jpsy- chires.2021.09.054. [19] Iwu CJ, Iwu CD, Wiysonge CS. The occurrence of long COVID: a rapid review. Pan Afr Med J. 2021;38:65. doi:10.11604/pamj.2021.38.65.27366. [20] Holmes EA, O’Connor RC, Perry VH, Tracey I, Wessely S, Arseneault L, et al. Multidisciplinary research priorities for the COVID-19 pandemic: a call for action for mental health science. Lancet Psychiatry. 2020;7(6):547– 60. doi:10.1016/S2215-0366(20)30168-1. [21] Vindegaard N, Benros ME. COVID-19 pandemic and mental health consequences: systematic review of the current evidence. Brain Behav Immun. 2020;89:531–42. doi:10.1016/j.bbi.2020.05.048. [22] Pedersen CB. The Danish Civil Registration System. Scand J Public Health. 2011;39(7 Suppl):22–5. doi:10.1177/1403494810387965. [23] Schmidt M, Schmidt SA, Sandegaard JL, Ehrenstein V, Pedersen L, Sor- ensen HT. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol. 2015;7:449–90. doi: 10.2147/CLEP.S91125. [24] Johannesdottir SA, Horvath-Puho E, Ehrenstein V, Schmidt M, Pedersen L, Sorensen HT. Existing data sources for clinical epidemiology: The Danish National Database of reimbursed prescriptions. Clin Epidemiol. 2012;4:303–13. doi:10.2147/CLEP.S37587. [25] Schonning K, Dessau RB, Jensen TG, Thorsen NM, Wiuff C, Nielsen L, et al. Electronic reporting of diagnostic laboratory test results from all healthcare sectors is a cornerstone of national preparedness and control of COVID-19 in Denmark. APMIS 2021;129(7):438–51. doi:10.1111/ apm.13140. 10 Rømer et al. [26] Grove Krause T, Jakobsen S, Haarh M, Molbak K. The Danish vaccination register. Euro Surveill. 2012;17(17):20155. doi:10.2807/ese.17.17.20155-en. [27] (SSI) SSI. Tidslinje for covid-19, https://covid19.ssi.dk/-/media/arkiv/sub sites/covid19/presse/tidslinje-over-covid-19/covid-19-tidslinje-for-2020- 2022-lang-version—version-1—april-2022.pdf; 2022 [accessed 21 Sep- tember 2022]. treatment of long COVID: a systematic review and meta-analysis. Brain Behav Immun. 2023;111: 211–29. doi:10.1016/j.bbi.2023.03.022. [38] (Sundhedsstyrelsen) DHA. National klinisk retningslinje for behandling af angstlidelser hos voksne, https://www.sst.dk/-/media/Udgivelser/2021/ NKR_Angstlidelser/Endelig-rapport—NKR-for-behandling-af-angslidel ser-hos-voksne.ashx; 2021 [accessed 21 September 2022]. [28] Glasheen WP, Cordier T, Gumpina R, Haugh G, Davis J, Renda A. Charlson comorbidity index: ICD-9 update and ICD-10 translation. Am Health Drug Benefits. 2019;12(4):188–97. [39] Cloos JM, Ferreira V. Current use of benzodiazepines in anxiety disorders. Curr Opin Psychiatry. 2009;22(1):90–5. doi:10.1097/YCO.0b013e32831 a473d. [29] Ho D, Imai K, King G, Stuart EA. MatchIt: nonparametric preprocessing inference. J Stat Softw. 2011;42(8):1–28. doi: for parametric causal 10.18637/jss.v042.i08. [30] Rouder JN, Engelhardt CR, McCabe S, Morey RD. Model comparison in ANOVA. Psychon Bull Rev. 2016;23(6):1779–86. doi:10.3758/s13423- 016-1026-5. [31] Lin L, Liu Y, Tang X, He D. The disease severity and clinical outcomes of the SARS-CoV-2 variants of concern. Front Public Health. 2021;9:775224. doi:10.3389/fpubh.2021.775224. [32] Lopez-Sampalo A, Bernal-Lopez MR, Gomez-Huelgas R, Persistent COVID-19 syndrome. A narrative review. Rev Clin Esp (Barc). 2022; 222(4):241–50. doi:10.1016/j.rceng.2021.10.001. [33] Yomogida K, Zhu S, Rubino F, Figueroa W, Balanji N, Holman E. Post- acute sequelae of SARS-CoV-2 infection among adults aged >/=18 years - long beach, California, April 1–December 10, 2020. MMWR Morb Mortal Wkly Rep. 2021;70(37):1274–7. doi:10.15585/mmwr.mm7037a2. [34] Vigod SN, Kurdyak P, Fung K, Gruneir A, Herrmann N, Hussain-Shamsy N, et al. Psychiatric hospitalizations: a comparison by gender, sociodemo- graphics, clinical profile, and postdischarge outcomes. Psychiatr Serv. 2016;67(12):1376–9. doi:10.1176/appi.ps.201500547. [35] Tsukamoto R, Kataoka Y, Mino K, Ishibashi N, Shibata M, Matsuo H, et al. Gender differences in anxiety among COVID-19 inpatients under isola- tion: a questionnaire survey during the first and second waves of the COVID-19 pandemic in Japan. Front Public Health. 2021;9:708965. doi: 10.3389/fpubh.2021.708965. [36] Taquet M, Dercon Q, Harrison PJ. Six-month sequelae of post-vaccination SARS-CoV-2 infection: a retrospective cohort study of 10,024 break- through infections. Brain Behav Immun. 2022;103:154–62. doi:10.1016/j. bbi.2022.04.013. [37] Ceban F, Kulzhabayeva D, Rodrigues NB, Di Vincenzo JD, Gill H, Sub- ramaniapillai M, et al. COVID-19 vaccination for the prevention and [40] Romer TB, Christensen RHB, Blomberg SN, Folke F, Christensen HC, Benros ME. Psychiatric admissions, referrals, and suicidal behavior before and during the COVID-19 pandemic in Denmark: a time-trend study. Acta Psychiatr Scand. 2021;144(6):553–62. doi:10.1111/acps.13369. [41] Xie Y, Xu E, Al-Aly Z. Risks of mental health outcomes in people with covid-19: cohort study. BMJ. 2022;376:e068993. doi:10.1136/bmj-2021- 068993. [42] Schou TM, Joca S, Wegener G, Bay-Richter C. Psychiatric and neuro- psychiatric sequelae of COVID-19 - a systematic review. Brain Behav Immun. 2021;97:328–48. doi:10.1016/j.bbi.2021.07.018. [43] Rivera-Izquierdo M, Lainez-Ramos-Bossini AJ, de Alba IG, Ortiz- Gonzalez-Serna R, Serrano-Ortiz A, Fernandez-Martinez NF, et al. Long COVID 12 months after discharge: persistent symptoms in patients hospitalised due to COVID-19 and patients hospitalised due to other causes-a multicentre cohort study. BMC Med. 2022;20(1):92. doi:10.1186/ s12916-022-02292-6. [44] Roberts HA, Clark DA, Kalina C, Sherman C, Brislin S, Heitzeg MM, et al. To vax or not to vax: predictors of anti-vax attitudes and COVID-19 vaccine hesitancy prior to widespread vaccine availability. PLoS One. 2022; 17(2):e0264019. doi:10.1371/journal.pone.0264019. [45] Nilsson SF, Laursen TM, Osler M, Hjorthoj C, Benros ME, Ethelberg S, et al. Vaccination against SARS-CoV-2 infection among vulnerable and marginalised population groups in Denmark: a nationwide population- based study. Lancet Reg Health Eur. 2022;16:100355. doi:10.1016/j. lanepe.2022.100355. [46] Nilsson SF, Laursen TM, Osler M, Hjorthoj C, Benros ME, Ethelberg S, et al. Adverse SARS-CoV-2-associated outcomes among people experien- cing social marginalisation and psychiatric vulnerability: a population- based cohort study among 4,4 million people. Lancet Reg Health Eur. 2022;20:100421. doi:10.1016/j.lanepe.2022.100421.
10.1186_s13018-021-02394-6	Feichtinger et al. Journal of Orthopaedic Surgery and Research (2021) 16:254 https://doi.org/10.1186/s13018-021-02394-6 R E S E A R C H A R T I C L E Open Access Lugol’s solution but not formaldehyde affects bone microstructure and bone mineral density parameters at the insertion site of the rotator cuff in rats Xaver Feichtinger1,2,3,4* Roland Kocijan6,7, Heinz Redl1,4, Johannes Grillari1,4, Christian Fialka2,7 and Rainer Mittermayr1,2,4,7 , Patrick Heimel1,4,5, Claudia Keibl1,4, David Hercher1,4, Jakob Emanuel Schanda1,2,4, Abstract Background: This study aimed to investigate whether rodent shoulder specimens fixed in formaldehyde for histological and histomorphometric investigations and specimens stained using Lugol’s solution for soft tissue visualization by micro-computed tomography (microCT) are still eligible to be used for bone architecture analysis by microCT. Methods: In this controlled laboratory study, 11 male Sprague-Dawley rats were used. After sacrifice and exarticulation both shoulders of healthy rats were assigned into three groups: (A) control group (n = 2); (B) formaldehyde group (n = 4); (C) Lugol group (n = 5). Half of the specimens of groups B and C were placed in a 4% buffered formaldehyde or Lugol’s solution for 24 h, whereas the contralateral sides and all specimens of group A were stored without any additives. MicroCT of both sides performed in all specimens focused on bone mineral density (BMD) and bone microstructure parameters. Results: BMD measurements revealed higher values in specimens after placement in Lugol’s solution (p < 0.05). Bone microstructure analyses showed increased BV/TV and Tb.Th values in group C (p < 0.05). Specimens of group C resulted in clearly decreased Tb.Sp values (p < 0.05) in comparison to the control group. Formaldehyde fixation showed minimally altered BMD and bone microstructure measurements without reaching any significance. Conclusions: MicroCT scans of bone structures are recommended to be conducted natively and immediately after euthanizing rats. MicroCT scans of formaldehyde-fixed specimens must be performed with caution due to a possible slight shift of absolute values of BMD and bone microstructure. Bone analysis of specimens stained by Lugol’s solution cannot be recommended. Keywords: Rotator cuff tear, Bone-tendon interface, Lugol, Formaldehyde, MicroCT * Correspondence: xaver.feichtinger@gmail.com 1Ludwig Boltzmann Institute for Experimental and Clinical Traumatology, Vienna, Austria 2AUVA Trauma Center Vienna - Meidling, Vienna, Austria Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Feichtinger et al. Journal of Orthopaedic Surgery and Research (2021) 16:254 Page 2 of 7 Introduction Depending on tear size, healing failure rates of rotator cuff reconstructions are reported up to 94% [1, 2]. The reasons for the high failure rate are multifactorial. Bony changes at the tendon insertion site in patients suffering from rotator cuff tears have been previously described [3–5]. Decreased bone mineral density and diminished bone microstructure were shown to have an important influence on the risk of re-ruptures [6, 7]. Characteristics of the reconstructed ten- don itself, namely its organization and structure in addition to muscle quality are important factors to con- sider [8]. Experimental animal models have been developed to treatment possibilities to avoid investigate additional these problems associated with re-tears [9, 10]. As bone, tendon, and muscle structures constitute a these functional unit, simultaneous investigations of structures in one sample would be of great interest. To investigate bone structure and bone mineral density (BMD), specimens are routinely placed in saline solu- tion, 4% paraformaldehyde, or are frozen natively at − 80 °C until micro-computed tomography (microCT) testing [8, 11, 12]. Recently, a staining method for soft tissue visualization (muscle and tendon structures as well as nervous tissue) by microCT analysis was described [13–15]. Thereby, Lugol’s solution, defined as a mixture of two parts potas- sium iodide in water and one part iodine, is used for staining of specimens [13, 16, 17]. Until now, little is known about the behavior of Lugol’s solution on bone structure parameters. To histologically investigate ten- don and muscle structures, 4% buffered formaldehyde solution is commonly used for fixation [18]. Afterwards, different staining methods are performed to visualize the respective structures of interest. Consequently, the aim of this study was to investigate the influence of Lugol’s solution and formaldehyde solution on bone microstructure and BMD during microCT analysis, as they may have a profound influence in interpreting microCT results in biomechanical rotator cuff studies. The primary hypothesis was that Lugol’s solution has no influence on bone microstructure and BMD in microCT analysis compared to native scans. Secondly, we hypothesized that 4% buffered formalde- hyde solution has no influence on bone microarchitec- ture using microCT analysis compared to native scans. Methods The study was authorized by the Institutional Animal Care and Use Committee (No. 504113/2016/16). Eleven male Sprague-Dawley rats were used for this study. Weight and age of the rats were homogeneous in the 3 groups (390–410 g). The rats (2/cage) were housed in a light- and temperature-controlled room. After 1 week of acclimatization period, rats were euthanized under deep anesthesia by an intracardially overdose of thiopental. Subsequently, both humeri were exarticulated. Supraspi- natus tendon and muscle structures were preserved in- cluding the enthesis. Rats were divided into three groups: (A) control group (n = 2); (B) formaldehyde group (n = 4); (C) Lugol group (n = 5). MicroCT scans of the control group were performed immediately after exarticulation. Right shoulders of the Lugol group were placed in Lugol’s solution for 24 h at 4 °C [13, 16, 17]. Right shoulders from the formaldehyde group were placed in 4% buffered formaldehyde solution for 24 h at 4 °C. After 24 h, microCT scans of the Lugol and for- maldehyde group were performed in the residual stain- ing solution. The contralateral shoulders were scanned natively without any additional substances and were compared to the corresponding staining method within the group. MicroCT (μCT 50, SCANCO Medical AG, Brüttisellen, Switzerland) scanning and segmentation was conducted by a blinded examiner [19]. The speci- mens were placed in 15 ml centrifugation tubes. MicroCT scans were conducted at 200 μA, 90 kVp with a field-of-view of 20.48 mm and reconstructed to a reso- lution of 10 μm. The SCANCO calibration phantom was used for calibration. Fiji was used to transform the humeri to an upright orientation along the Z axis [20, 21]. The scans were then rigidly registered using Amira 6.2 (FEI, Thermo Fisher Scientific, Hillsboro, OR, USA) to a single speci- men. Regions of interest were manually drawn in the trabecular space of the epiphysis equally in all groups. A sub-volume of the epiphysis space at tendon attachment site was manually selected in all specimens as region of interest (ROI) (Fig. 1). To prevent measurement errors due to differences in the geometry of the specimen, the regions were then corrected using Definiens Developer XD 2.0 (Definiens AG, Munich, Germany). In Definiens, a ruleset was developed to segment the cortical portion of the specimen, based on local volumetric bone density. The cortical portion was not considered for measure- ment. Two measurements were performed with different thresholds for bone segmentation. Firstly, a measure- ment using a fixed threshold of 500 mgHA/cm3 for all native and formalin-stained specimen. Secondly, due to obvious differences in brightness in the Lugol stained specimen, a measurement was performed using a thresh- old calculated via Otsu’s automatic threshold selection in the ROI of each Lugol stained specimen and their na- tive contralateral controls [22]. Affected and contralat- eral control were always thresholded and measured with the same method. Bone mineral density and bone volume fraction were measured in Definiens. For measurement of bone min- eral density, the outer 20 μm of bone structures were Feichtinger et al. Journal of Orthopaedic Surgery and Research (2021) 16:254 Page 3 of 7 Fig. 1 Computer tomography analysis in sagittal plane (left), frontal plane (middle), and transverse plane (right). The epiphysis is marked with a blue star. Green area marks subvolume of the epiphysis space at tendon attachment as region of interest (ROI) excluded. The segmented trabecular structures were exported as binary image stacks and analyzed in Fiji using the BoneJ plugin for measurement of trabecular thickness and spacing [23]. BMD (mgHA/cm3) as well as bone microstructure pa- rameters including mean bone volume fraction (BV/TV; %), mean trabecular thickness (Tb.Th, μm), and mean trabecular spacing (Tb.Sp, μm) were then calculated. Analyses of the ipsilateral and contralateral side were then compared and a ratio was generated for statistical analyses. Statistical analyses Testing for normal distribution was conducted using the D’Agostino and Pearson omnibus normality test. Ac- cordingly, Kruskal-Wallis test and Mann-Whitney test were performed. In graphs, values are presented as mean values and associated standard error of mean. For calcu- lation, a ratio of the affected to the non-affected native side for each parameter was created. A p value < 0.05 was considered statistically significant. GraphPad Prism version 8.3.1 (GraphPad Software, La Jolla, CA, USA) was used for statistical calculations. Results Bone mineral density (BMD) BMD measurements revealed higher values in specimens storage in Lugol’s solution (Fig. 2). Analyses after showed significantly higher (p < 0.05) ratios of the af- fected side compared to the native contralateral side within group C. Group B revealed minimally increased ratios in comparison to the control group without reach- ing significance (Fig. 2). Fig. 2 Bone mineral density (BMD): left: absolute values. right: relative difference (%) of affected side and native contralateral side. Values are presented as mean values and associated standard error of mean Feichtinger et al. Journal of Orthopaedic Surgery and Research (2021) 16:254 Page 4 of 7 Fig. 3 Bone microstructure: relative difference (%) for bone volume fraction (BV/TV); trabecular thickness (Tb.Th); trabecular spacing (Tb.Sp). Values are presented as mean values and associated standard error of mean Bone microarchitecture Shown are the relative differences between the affected and native contralateral control as a ratio affected/native. Bone microstructure analyses showed increased BV/TV and Tb.Th ratios in group C (p < 0.05) (Fig. 3). Speci- mens from group C resulted in significantly decreased (p < 0.05) Tb.Sp ratios compared to the control group (Fig. 3). Formaldehyde fixation showed minimally increased BV/TV and Tb.Th ratios and slightly decreased Tb.Sp measurements without reaching significance. Discussion This study aimed to investigate the influence of formalde- hyde fixation and staining using Lugol’s solution on BMD and bone microstructure of the rotator cuff insertion site in rats. Results showed significantly altered values after immersion in Lugol’s solution for 24 h. Comparisons of bone parameters between formaldehyde-fixed specimens and native controls did not differ. The importance of bone structures in rotator cuff tears and their influence on success after reconstruction was shown in earlier studies. Meyer et al. described signifi- cant differences regarding BMD in human cadavers with rotator cuff tears in comparison to their contralateral side without tears [4]. Oh et al. described seven different regions of bone density in the proximal humerus in pa- tients with unilateral rotator cuff repairs stating in the posterolateral portion the highest volumetric BMD [5]. Chung et al. presented the influence on success of re- construction and stated that lower BMD and fatty infil- tration are associated with lower healing results after reconstruction [6]. Also bone microstructure was deteri- orated in patients with rotator cuff tears. Kirchhoff et al. showed differences in BV/TV values at the proximal hu- merus by high-resolution quantitative computed tomog- raphy [7]. These findings underline the importance of bone structure evaluation in rotator cuff tear models and require further investigations of treatment options for improvement of bone deficiencies after rotator cuff tears [10, 24]. Not only bone quality evaluated by microCT investiga- tions are of high scientific interest. Also histomorpho- metric analysis of bone, tendon, and muscle are essential methods of analysis [25, 26]. Myocellular and intramus- cular fat infiltration, atrophy, and fibrosis of muscle and tendon structures are associated with high rates of heal- ing failure [25, 27]. Kim et al. stated that tears at the an- terior part of the supraspinatus tendon particularly need to be treated early due to the high risk of fatty infiltra- tion associated with inferior clinical outcome [27]. These findings require techniques that allow for inves- tigations in both bone structures and soft tissue struc- tures at a high-quality level. Until now, it was unclear if specimens fixed with formaldehyde for histological and histomorphometric investigations or specimens stained by Lugol’s solution can be used for reliable bone struc- ture evaluations by microCT as well. the musculo-tendinous For histological investigation of soft tissue structures, a 4% buffered formaldehyde solution is commonly used for fixation [18]. Afterwards, different staining methods are used to investigate structures of interest [9]. Particu- larly in rotator cuff reconstruction animal models, inves- tigations of transition zone including vascularization and collagen types relations (Fig. 4) are of great interest. In this study, microCT scans of shoulders from the formaldehyde group showed minimally increased ratios in comparison to the control group without reaching significance for BMD measure- ments. Bone microarchitecture assessment showed in the formaldehyde group minimally increased BV/TV and Tb.Th ratios and slightly decreased Tb.Sp measurements without reaching significance. Earlier studies showed the effect of formaldehyde on the biomechanical properties of bone [28, 29]. Elements in the hydroxyapatite of bone as Ca, P, and Mg, can dissolve in formaldehyde and alter the biomechanics [28, 29]. In the present study, the Feichtinger et al. Journal of Orthopaedic Surgery and Research (2021) 16:254 Page 5 of 7 Fig. 4 Stained sections of musculo-tendinous transition zones: left: Martius, Scarlet, and Blue (MSB) for muscle and tendon quality assessment, middle: CD31 stained section for vascularization analysis, right: collagen III stained section for muscle and tendon regeneration analysis minimal changes of BMD and microarchitecture param- eters in the formaldehyde group reflect ongoing pro- cesses of the bone but minimize its relevance for microCT analyses. The duration of preservation in for- maldehyde seems to play an important role, as other studies investigated biomechanics after long-term pres- ervation, in this study shoulders were placed in 4% buff- ered formaldehyde solution for 24 h. Due to minimal changes after formaldehyde preservation in microCT analyses in this study, native scans are recommended to avoid incorrect measurements. Recent studies describe staining methods with Lugol’s solution enabling visualization and analyses of soft tis- sues such as muscle and tendon structures as well as nerve structures by microCT analysis [13, 14, 16, 17, 30] (Fig. 5). Consequently, 3D reconstructions of soft tissue structures including nerve visualization enable new out- come measurement options in experimental studies. The importance of neurologic deficiencies have been shown earlier and are of high scientific interest in experimental rotator cuff tear animal models [31, 32]. In this study, significantly higher BMD values were mea- sured in specimens after storage in Lugol’s solution. Bone microarchitecture evaluation resulted in clearly increased BV/TV and Tb.Th ratios as well as decreased Tb.Sp mea- surements after preservation in Lugol’s solution. Despite rigidity and the low permeability of bone tissue, the effect of Lugol’s solution showed significant changes in microCT analysis. The effect most probably depends on time of staining, as the acidity changes during staining with Lugol’s solution and the pH begins to decrease. As a result of the lowered pH, decalcification of bone increases simi- larly to other acidic stains [13]. In the present study, the main interest was in soft tissues on the outer surface of the specimen. The specimens were thus only stained for a relatively short time, resulting in little stain reaching the trabecular space of the epiphysis. Despite this minimal staining, the effect on the bone measurements was signifi- cant. If soft tissues in the trabecular space of the epiphysis or medullary cavity are of interest, longer staining dura- tions and a higher amount of stain are necessary, com- pounding the negative effect on bone measurements. The Fig. 5 Sagittal cut of a humerus specimen in Lugol’s solution. Blue star: epiphysis. Red star: supraspinatus muscle. Green area marks supraspinatus tendon structure Feichtinger et al. Journal of Orthopaedic Surgery and Research (2021) 16:254 Page 6 of 7 exact amount of stain taken up by the tissue is difficult to control, resulting in a degree of variability in the final at- tenuation [13]. Measurements performed using a fixed threshold may therefore show differences between sam- ples based on the individual staining. We attempted to correct for this using Otsu’s automatic threshold selection in the region of interest. In conclusion, this study describes significant BMD and bone microstructure deteriorations in microCT ana- lysis after staining of proximal humeri in Lugol’s solu- tion. Formaldehyde fixation may have a slight influence on bone evaluation values in comparison to native microCT scans. Consequently, microCT scans of bone structures are recommended to be conducted natively and immediately after sacrifice of rats. MicroCT scans of formaldehyde-fixed specimens can be performed with caution of interpretation due to a possible shift of abso- lute values of BMD and bone structure. Bone analyses of specimens stained by Lugol’s solution are not recom- mended due to significant deteriorations. Abbreviations microCT: Micro-computed tomography; BMD: Bone mineral density; BV/ TV: Mean bone volume fraction; Tb.Th: Mean trabecular thickness; Tb.Sp: Mean trabecular spacing Acknowledgements The authors thank N. Swiadek, MSc, for her support throughout the study. The authors thank T. Vacca for proofreading. Authors’ contributions Individual contributions include the following: XF, PH, CK, DH, RK, JES, HR, JG, CF and RM for the study concept and design; XF, PH, CK, and DH for data collection; XF, PH, CK, DH, JES and RK for data analysis; XF, HR, JG, CF and RM for data interpretation; XF, PH, CK, DH, RK and JES for drafting the manuscript and literature research; XF, PH and RM for the figures; XF, PH, CK, DH, RK, JES, HR, JG, CF and RM for the extensive revision of the manuscript. All authors read and approved the final manuscript. Funding This work was supported by the Medical Scientific Fund of the Mayor of the City of Vienna. Availability of data and materials The datasets used and analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The study was authorized by the Institutional Animal Care and Use Committee (No. 504113/2016/16). Consent for publication Not applicable. Competing interests All authors declare that they have no competing interests. Author details 1Ludwig Boltzmann Institute for Experimental and Clinical Traumatology, Vienna, Austria. 2AUVA Trauma Center Vienna - Meidling, Vienna, Austria. 3Department of Orthopedic Surgery II, Herz-Jesu Hospital, Vienna, Austria. 4Austrian Cluster for Tissue Regeneration, Vienna, Austria. 5Karl Donath Laboratory for Hard Tissue and Biomaterial Research, Department of Oral Surgery, University Clinic of Dentistry, Medical University of Vienna, Vienna, Austria. 6Ludwig Boltzmann Institute of Osteology, 1st Medical Department at Hanusch Hospital, Vienna, Austria. 7Center for the Musculoskeletal System, Medical Faculty, Sigmund Freud University, Vienna, Austria. Received: 18 January 2021 Accepted: 5 April 2021 References 1. 2. Chona DV, Lakomkin N, Lott A, Workman AD, Henry AC, Kuntz AF, et al. The timing of retears after arthroscopic rotator cuff repair. J Shoulder Elbow Surg. 2017;26:2054–9. Galatz LM, Ball CM, Teefey SA, Middleton WD, Yamaguchi K. The outcome and repair integrity of completely arthroscopically repaired large and massive rotator cuff tears. J Bone Joint Surg Am. 2004;86-A: 219–24. Kannus P, Leppälä J, Lehto M, Sievänen H, Heinonen A, Järvinen M. A rotator cuff rupture produces permanent osteoporosis in the affected extremity, but not in those with whom shoulder function has returned to normal. J Bone Miner Res Off J Am Soc Bone Miner Res. 1995;10:1263–71. 4. Meyer DC, Fucentese SF, Koller B, Gerber C. Association of osteopenia of the 3. 5. 6. 7. 8. 9. 10. humeral head with full-thickness rotator cuff tears. J Shoulder Elbow Surg. 2004;13(3):333–7. https://doi.org/10.1016/j.jse.2003.12.016. Oh JH, Song BW, Lee YS. Measurement of volumetric bone mineral density in proximal humerus using quantitative computed tomography in patients with unilateral rotator cuff tear. J Shoulder Elbow Surg. 2014;23(7):993–1002. https://doi.org/10.1016/j.jse.2013.09.024. Chung SW, Oh JH, Gong HS, Kim JY, Kim SH. Factors affecting rotator cuff healing after arthroscopic repair: osteoporosis as one of the independent risk factors. Am J Sports Med. 2011;39(10):2099–107. https://doi.org/10.1177/ 0363546511415659. Kirchhoff C, Braunstein V, Milz S, Sprecher CM, Fischer F, Tami A, et al. Assessment of bone quality within the tuberosities of the osteoporotic humeral head: relevance for anchor positioning in rotator cuff repair. Am J Sports Med. 2010;38(3):564–9. https://doi.org/10.1177/0363546509354989. Killian ML, Cavinatto LM, Ward SR, Havlioglu N, Thomopoulos S, Galatz LM. Chronic degeneration leads to poor healing of repaired massive rotator cuff tears in rats. Am J Sports Med. 2015;43(10):2401–10. https://doi.org/10.1177/ 0363546515596408. Feichtinger X, Monforte X, Keibl C, Hercher D, Schanda J, Teuschl AH, et al. Substantial biomechanical improvement by extracorporeal shockwave therapy after surgical repair of rodent chronic rotator cuff tears. Am J Sports Med. 2019;47(9):2158–66. https://doi.org/10.1177/0363546519854760. Shah SA, Kormpakis I, Havlioglu N, Ominsky MS, Galatz LM, Thomopoulos S. Sclerostin antibody treatment enhances rotator cuff tendon-to-bone healing in an animal model. J Bone Joint Surg Am. 2017;99(10):855–64. https://doi.org/10.2106/JBJS.16.01019. 11. Hettrich CM, Gasinu S, Beamer BS, Stasiak M, Fox A, Birmingham P, et al. The 12. effect of mechanical load on tendon-to-bone healing in a rat model. Am J Sports Med. 2014;42(5):1233–41. https://doi.org/10.1177/0363546514526138. Tanaka K, Kanazawa T, Gotoh M, Tanesue R, Nakamura H, Ohzono H, et al. Effects of estrogen-deficient state on rotator cuff healing. Am J Sports Med. 2019;47(2):389–97. https://doi.org/10.1177/0363546518815869. 13. Heimel P, Swiadek NV, Slezak P, Kerbl M, Schneider C, Nürnberger S, et al. Iodine-enhanced micro-CT imaging of soft tissue on the example of peripheral nerve regeneration. Contrast Media Mol Imaging. 2019;2019: 7483745. 14. Hopkins TM, Heilman AM, Liggett JA, LaSance K, Little KJ, Hom DB, et al. Combining micro-computed tomography with histology to analyze biomedical implants for peripheral nerve repair. J Neurosci Methods. 2015; 255:122–30. https://doi.org/10.1016/j.jneumeth.2015.08.016. 15. Wu J, Yin N. Anatomy research of nasolabial muscle structure in fetus with cleft lip: an iodine staining technique based on microcomputed tomography. J Craniofac Surg. 2014;25(3):1056–61. https://doi.org/10.1097/ SCS.0000000000000651. 16. Metscher BD. MicroCT for developmental biology: a versatile tool for high- contrast 3D imaging at histological resolutions. Dev Dyn. 2009;238:632–40. 17. Metscher BD. MicroCT for comparative morphology: simple staining methods allow high-contrast 3D imaging of diverse non-mineralized animal tissues. BMC Physiol. 2009;9(1):11. https://doi.org/10.1186/1472-6793-9-11. Feichtinger et al. Journal of Orthopaedic Surgery and Research (2021) 16:254 Page 7 of 7 18. 19. 20. 21. Schneider KH, Aigner P, Holnthoner W, Monforte X, Nürnberger S, Rünzler D, et al. Decellularized human placenta chorion matrix as a favorable source of small-diameter vascular grafts. Acta Biomater. 2016;29:125–34. https://doi. org/10.1016/j.actbio.2015.09.038. Feichtinger X, Muschitz C, Heimel P, Baierl A, Fahrleitner-Pammer A, Redl H, et al. Bone-related circulating microRNAs miR-29b-3p, miR-550a-3p, and miR-324-3p and their association to bone microstructure and histomorphometry. Sci Rep. 2018;8(1):4867. https://doi.org/10.1038/s41598- 018-22844-2. Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, et al. Fiji: an open-source platform for biological-image analysis. Nat Methods. 2012;9(7):676–82. https://doi.org/10.1038/nmeth.2019. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. Nat Methods. 2012;9(7):671–5. https://doi.org/10.1038/ nmeth.2089. 22. Otsu N. A Threshold selection method from gray-level histograms. IEEE Trans Syst Man Cybern. 1979;9(1):62–6. https://doi.org/10.1109/TSMC.1 979.4310076. 23. Doube M, Kłosowski MM, Arganda-Carreras I, Cordelières FP, Dougherty RP, Jackson JS, et al. BoneJ: Free and extensible bone image analysis in ImageJ. Bone. 2010;47(6):1076–9. https://doi.org/10.1016/j.bone.2010.08.023. 24. Oh JH, Kim DH, Jeong HJ, Park JH, Rhee S-M. Effect of recombinant human parathyroid hormone on rotator cuff healing after arthroscopic repair. Arthroscopy. 2019;35(4):1064–71. https://doi.org/10.1016/j.arthro.2018.11.038. 25. Goutallier D, Postel JM, Bernageau J, Lavau L, Voisin MC. Fatty muscle 26. 27. degeneration in cuff ruptures. Pre- and postoperative evaluation by CT scan. Clin Orthop. 1994;304:78–83. Kovacevic D, Fox AJ, Bedi A, Ying L, Deng X-H, Warren RF, et al. Calcium- phosphate matrix with or without TGF-β3 improves tendon-bone healing after rotator cuff repair. Am J Sports Med. 2011;39(4):811–9. https://doi.org/1 0.1177/0363546511399378. Kim HM, Dahiya N, Teefey SA, Keener JD, Galatz LM, Yamaguchi K. Relationship of tear size and location to fatty degeneration of the rotator cuff. J Bone Joint Surg Am. 2010;92(4):829–39. https://doi.org/10.2106/JBJS. H.01746. 28. Zhang G, Wang S, Xu S, Guan F, Bai Z, Mao H. The effect of formalin preservation time and temperature on the material properties of bovine femoral cortical bone tissue. Ann Biomed Eng. 2019;47(4):937–52. https:// doi.org/10.1007/s10439-019-02197-1. 29. Boskey AL, Cohen ML, Bullough PG. Hard tissue biochemistry: a comparison of fresh-frozen and formalin-fixed tissue samples. Calcif Tissue Int. 1982; 34(4):328–31. https://doi.org/10.1007/BF02411262. 30. Alike Y, Yushan M, Keremu A, Abulaiti A, Liu Z-H, Fu W, et al. Application of custom anatomy-based nerve conduits on rabbit sciatic nerve defects: in vitro and in vivo evaluations. Neural Regen Res. 2019;14(12):2173–82. https://doi.org/10.4103/1673-5374.262601. 31. Gayton JC, Rubino LJ, Rich MM, Stouffer MH, Wang Q, Boivin GP. Rabbit 32. supraspinatus motor endplates are unaffected by a rotator cuff tear. J Orthop Res. 2013;31(1):99–104. https://doi.org/10.1002/jor.22192. Sasaki Y, Ochiai N, Hashimoto E, Sasaki Y, Yamaguchi T, Kijima T, et al. Relationship between neuropathy proximal to the suprascapular nerve and rotator cuff tear in a rodent model. J Orthop Sci. 2018;23:414–9. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
10.1186_s40359-023-01196-1	Mathew et al. BMC Psychology (2023) 11:151 https://doi.org/10.1186/s40359-023-01196-1 RESEARCH BMC Psychology Open Access Comparison of quality of life in patients with advanced chronic kidney disease undergoing haemodialysis, peritoneal dialysis and conservative management in Johannesburg, South Africa: a cross-sectional, descriptive study , Malcolm Davies1,2 Neelu Mathew1 and Zaheera Cassimjee1,2* , Feroza Kaldine1,3 Abstract Introduction Mental health and quality of life are under-appreciated clinical targets which affect patient and modality survival. Lack of dialysis availability in the resource-constrained public health sector in South Africa results in assignment to treatment modalities without regard to effects on these parameters. We assessed the effect of dialysis modality, demographic and laboratory parameters on mental health and quality of life measurements. Methods Size-matched cohorts were recruited from patients on haemodialysis (HD), peritoneal dialysis (PD), and patients on conservative management (CM) between September 2020 and March 2021. Responses to the Hospital Anxiety and Depression Scale (HADS) and Kidney Disease Quality of Life Short Form 36 (KDQOL-SF36) questionnaires and demographic and baseline laboratory parameters were compared between modalities. Multivariate linear regres- sion was used to evaluate independent effect of baseline characteristics on HADS and KDQOL-SF36 scores between treatment groups where significant difference was observed. Results Anxiety, depression, and reduced KDQOL measures were widespread amongst respondents. Dialyzed patients reported higher anxiety and depression scores than those on CM (p = 0.040 and p = 0.028). Physical compos- ite (PCS), role–physical (RP), vitality (VS), and emotional well-being (EWB) KDQOL-SF36 scores were poorer in dialyzed patients (p < 0.001 for all). PCS (p = 0.005), pain (p = 0.030), vitality (p = 0.005), and social functioning KDQOL scores were poorer in PD compared to HD; HADS anxiety (p < 0.001) and KDQOL-SF36 EWB scores (p < 0.001) were better in PD. PD patients were more likely to be employed (p = 0.008). Increasing haemoglobin concentration reduced anxiety (p < 0.001) and depression scores (p = 0.004), and improved PCS (p < 0.001), and pain scores (p < 0.001). Higher serum albumin improved PCS (p < 0.001) and vitality (p < 0.001) scores. Conclusion Advanced chronic kidney disease increases anxiety and depression and limits quality of life. PD improves mental health and emotional wellbeing and preserves the ability to undertake economic activity but limits social Correspondence: Zaheera Cassimjee zaheeracassimjee@gmail.com Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Mathew et al. BMC Psychology (2023) 11:151 Page 2 of 11 functioning and causes greater physical discomfort. Targeting haemoglobin may ameliorate modality effects on men- tal health and quality of life. Keywords Health related quality of life, Haemodialysis, Peritoneal dialysis, Chronic kidney disease Introduction A significant burden of communicable and non-commu- nicable diseases, combined with poor access to nephrol- ogy services, render chronic kidney disease (CKD) not uncommon in low-to-middle income countries (LMIC); in Africa genetic risks exacerbate this pattern [1, 2]. South Africans are at increased risk of CKD due to a combination of these genetic and environmental risk fac- tors [1]. Socioeconomic improvements since the advent of democracy are likely to exacerbate this risk as urbani- zation and lifestyle change increase the prevalence of hypertension and diabetes [2]. Although the majority of South Africans are reliant on the public sector for management of kidney failure (KF), long-term under-resourcing of state-sponsored dialysis units has resulted in disparities in access to treat- ment, with the dialysis treatment rate in the private sec- tor being 787 per million population (pmp) as compared to 57 pmp in the state sector [3]. Patients with KF in the state sector may therefore not be able to access dialy- sis and may instead receive conservative management. Those patients fortunate enough to receive dialysis often do not have the luxury of choosing a preferred modality; instead, many state facilities pursue a policy of “perito- neal dialysis first”. Lack of participation in treatment planning may adversely affect treatment satisfaction and quality of life (QOL) [4]. Increased survival rates on dialysis, combined with declining rates of transplantation in South Africa, [5] contribute to prolonged waitlisting, in turn leading to accrual of aging and dialysis vintage- associated pathologies which may compromise QOL. In addition, dialysis treatments themselves carry the poten- tial for significant morbidity which may offset improve- ments in QOL attained through amelioration of uraemic symptoms. QOL is known to affect future risk of hospitalisation and mortality [6, 7], as well as compliance with ther- apy [8]. In addition, QOL is itself a significant outcome measurement of kidney replacement therapy (KRT) [9]. Despite this, QOL remains an underappreciated and under-researched component of KF management, and there is a paucity of data on the quality of life experi- enced by dialysis patients in the South African context [10–12]. In particular, the effect of non-directed dialytic therapy on QOL has not been contextualised against the QOL of patients receiving conservative management for KF. This study compares the quality of life of patients on peritoneal and haemodialysis against those with advanced CKD not receiving dialysis and considers the effect of comorbidities on quality-of-life scores. Methods Study design and settings Helen Joseph Hospital (HJH) is a tertiary-level facility which provides specialised nephrology services to resi- dents of the western areas of Johannesburg. The dialysis unit at HJH has the capacity to provide haemodialysis to 74 patients and peritoneal dialysis to 80 patients. Entrance onto the dialysis programme is regulated by local policy in consideration of transplant eligibility. Patients excluded from dialysis are referred to the Renal Outpatients Department for further treatment. Size- matched cohorts (n = 50, reflecting the minimum num- ber of patients attending dialysis clinics) were voluntarily recruited from the outpatient haemodialysis, peritoneal dialysis, and Renal Outpatients Department at HJH using convenience sampling between September 2020 and March 2021. Non-dialyzed patients were only considered for inclusion if baseline eGFR was consistently below 20 mL/min/1.73 m2 for at least 3 months. Patients with a history of hospitalization within the antecedent month were excluded. Study instruments Participants were administered the Hospital Anxiety and Depression Scale (HADS) and the Kidney Disease Quality of Life Short-Form 36 (KDQOL-SF36) question- naires during routine clinical appointments. Participants who were unable to complete the questionnaires inde- pendently or who required translation of the question- naires into their vernacular language were assisted by nursing staff or the primary investigator. KDQOL-SF36 responses were transformed to domain scores using the online RAND health group KDQOL-SFTM 1.3 calcula- tor. Respondent demographics, HADS and transformed KDQOL-SF36 domain scores were entered with most recent laboratory results in an anonymized database which was used for statistical analysis using Statistica version 14.0.1 (TIBCO Software, Palo Alto, California). Both the HADS and the KDQOL-SF36 surveys are vali- dated for use in patients with advanced CKD [13, 14]. The KDQOL-SF36 consists of 36 questions which provide information on quality of life across a number of sub- domains, including physical functioning, role-physical, Mathew et al. BMC Psychology (2023) 11:151 Page 3 of 11 pain, vitality, emotional well-being, role-emotional, and social functioning; scoring is from 0 to 100 with higher scores indicating better function. Physical function and mental health subdomain scores are used to calculate composite physical and mental score, reflective of over- all limitation in these respective areas. The HADS scale questionnaire produces a score of 0–21 for both anxiety and depression symptoms, a score of more than 8 corre- lates with overt clinical disorder, whilst a score of 4–8 is indicative of borderline symptomatology. Statistical analysis Distribution of continuous variables was analysed through Shapiro Wilk W testing and visual inspection of the histogram plot. Baseline laboratory and demographic characteristics were compared between treatment groups using ANOVA and Pearson Chi-square testing as indi- cated; the Kruskall Wallis ANOVA and Mann Whitney U test were used for non-parametric continuous data. HADS and KDQOL-SF36 scores were compared between treatment groups using the ANOVA, Student t-, Kruskall Wallis ANOVA, or Mann–Whitney U test as indicated. Where statistically significant difference was observed between treatment groups for quality-of-life measure- ment, subsequent analysis using multivariate stepwise sigma-restricted linear regression was undertaken to test the independence of treatment modality from confound- ing effect using a model comprising a priori factors (eth- nicity, sex, source of income, and relationship status) and factors showing variation between treatment modalities (age, comorbid diabetes, dialysis vintage, haemoglobin, phosphate and albumin concentration, and parathyroid hormone level); significant multicollinearity of param- eters was excluded with all parameters having a variance inflation factor below 2 and tolerance above 0.5 (Addi- tional file 1: Table S1). Ethics approval Approval for this study was obtained from the Human Research Ethics Committee of the University of the Wit- watersrand (protocol number M200635). Institutional approval was obtained from Helen Joseph Hospital. The study was conducted in accordance with the principles of the Declaration of Helsinki. Results One hundred and fifty patient participants were recruited to this study, comprising 50 patients on haemodialysis (HD) and peritoneal dialysis (PD) each, and 50 patients with advanced kidney dysfunction (eGFR less than 20 mL/min/1.73 m2) undergoing conservative outpa- tient management (CM). Amongst CM participants, 31 (62%) were in stage 5 of CKD (eGFR consistently below 15 mL/min/1.73 m2). Baseline characteristics are shown in Table 1 below. Significant differences in age (p < 0.001), haemoglobin (p = 0.007), phosphate (p = 0.002) concentration, para- thyroid hormone (PTH) (p < 0.001), and albumin con- centration (p < 0.001) were detected between the CM and dialysis treatment groups; diabetes mellitus was more frequent in patients on the CM programme than those receiving dialysis (p < 0.001). Haemoglobin concentration was higher (p = 0.001) and PTH control better (p = 0.027) in patients prescribed PD compared to those receiving HD; albumin concentration was, however, poorer in the former (p < 0.001). Dialysis vintage was greater in patients receiving HD compared to those on PD (p = 0.017). Sex, level of education, relationship status, prevalence of HIV infection and cardiovascular disease, recombinant erythropoietin (rEPO) dose, and serum urea and calcium showed no significant difference between groups; preva- lence of diabetes was similar between patients prescribed PD and those prescribed HD (p = 0.611). No significant difference was detected for age between the PD and HD cohorts (p = 0.494). Unemployment was more frequent in the HD treat- ment group (p = 0.044). Patients on PD were less fre- quently supported by a social grant than other treatment groups (p = 0.008), whilst patients on HD were more frequent recipients of a social grant than other groups (p = 0.005). Features of anxiety and depression were widespread amongst all respondents (Fig. 1, Table 2). Anxiety symp- toms of significance (HADS anxiety score greater than 4) were significantly more frequent in the HD group (22, 44%) than in the PD (8, 16%) or CM (11, 22%) groups (p = 0.004). HADS anxiety score was higher in patients on dialysis compared to those on CM (p = 0.040), although this difference was ameliorated with progression to CKD stage 5 in the latter group (p = 0.284). Patients receiv- ing HD reported higher anxiety scores than either those receiving PD (p < 0.001) or those on CM with CKD stage 5 (p = 0.045). Thirty-four percent (17 respondents) in the HD treat- ment group reported symptoms suggestive of depres- sion (HADS depression score greater than 4) compared to 24% (12 respondents) in the PD treatment group and 22% (11 respondents) of those on CM (p = 0.348). HADS depression score was higher in patients on dialysis com- pared to those receiving CM (p = 0.028) although this sig- nificance was lost when analysis was restricted to the CM group with CKD stage 5 (p = 0.664). Depression scores were higher in patients prescribed HD, although statisti- cally significant difference was not shown in comparison to those receiving PD (p = 0.092) or CKD stage 5 patients on CM (p = 0.124). Mathew et al. BMC Psychology (2023) 11:151 Page 4 of 11 Table 1 Baseline characteristics of participants Age (years) Sex Race Relationship status Haemodialysis (n = 50) Peritoneal dialysis (n = 50) 45.5 ± 12.3 Male: 23 (46%) Female: 27 (54%) African: 37 (75%) Asian: 3 (6%) Mixed race: 9 (18%) Caucasian: 1 (2%) Married: 14 (28%) Partner: 9 (18%) None: 27 (54%) 44.1 ± 11.6 Male: 25 (50%) Female: 25 (50%) African: 34 (68%) Asian: 2 (4%) Mixed race: 8 (16%) Cau- casian: 6(12%) Married: 19 (38%) Partner: 5 (10%) None: 26 (52%) Conservative management (n = 50) 59.0 ± 13.5 Male: 27 (54%) Female: 23 (46%) African: 33 (66%) Asian: 0 (0%) Conservative management stage 5 (n = 31) 59.9 ± 14.1 Male: 14 (45.2%) Female: 17 (54.8%) African: 19 (61.1%) Asian: 2 (6.5%) Mixed race: 14 (28%) Mixed race: 6 (19.4%) Caucasian: 3 (6%) Married: 21 (42%) Partner: 3 (6%) None: 26 (52%) Caucasian: 4 (12.9%) Married: 10 (32.3%) Partner: 2 (6.4%) None: 19 (61.3%) Source of income Unemployed: 47 (94%) Unemployed: 42 (84%) Unemployed: 38 (76%) Unemployed: 25 (80.6%) Social grant: 37 (74%) Social grant: 21 (42%) Social grant: 29 (58%) Social grant: 17 (54.8%) Dialysis vintage (months) 47 [15–82] Access in HD patients / PD modality Tunneled cuffed catheter: 31 (62%) Arteriovenous Fistula: 14 (28%) Arteriovenous graft: 5(10%) 24 [13–40] CAPD: 45 (90%) CCPD: 5 (10%) – – – – Comorbidities Diabetic: 11 (22%) Diabetic: 8 (16%) Diabetic: 25 (50%) Diabetic: 14 (45.2%) HIV positive: 11 (22%) HIV positive: 10 (20%) HIV positive: 11 (22%) HIV positive: 7 (22.6%) Cardiovascular disease: 48 (96%) Cardiovascular disease: 43 (86%) Cardiovascular disease: 47 (94%) Cardiovascular disease: 31 (100%) Hemoglobin (g/dL) 9.5 [8.7–10.0] 10.9 [9.3–12.6] 11.3 [9.6–12.8] 11.1 [9.4–12.8] rEPO dose (units/month) 40,000 [24,000–64,000] 32,000 [0–48,000] – Urea (mmol/L) Calcium (mmol/L) Phosphate (mmol/L) Albumin (g/L) Parathyroid hormone (ng/ mL) 19.0 [11.8–26.1] 2.22 [2.13–2.40] 1.71 [1.26–2.40] 40 [36–42] 34.4 [14.8–59.8] 20.8 [15.9–25.5] 2.20 [2.09–2.30] 1.76 [1.15–2.6]) 35 [30–38] 48.2 [29.6–79] 20.1 [15.8–27.3] 2.27 [2.14–2.37] 1.38 [1.07–1.66] 40 [37–44] 15.1 [8.7–21.5] – 21.6 [16.9–30.5] 2.28 [2.10–2.36] 1.48 [1.20–1.72] 40.0 [37–44] 18.4 [10.8–21.5] Values are mean ± SD or median [interquartile range] *Patients on CM are precluded from rEPO prescription by local resource constraints Patients receiving either dialysis modality reported poorer KDQOL-SF36 physical composite score (PCS) than those on CM (p < 0.001), a finding which persisted when analysis was restricted to those CM patients with CKD stage 5 (p = 0.027) (Table 2, Fig. 2). Subjective assessment of ability to meet expected physical activ- ity (role–physical, RP) was similarly poorer in patients on dialysis compared to these CM groups (p < 0.001 and p = 0.003, respectively). Dialyzed patients addition- ally reported lower vitality (VS) (p < 0.001 and p = 0.027, respectively) and emotional well-being (EWB) (p < 0.001 and p = 0.002, respectively) compared to CM groups. No significant differences were detected between patients receiving dialysis and those on conservative management for the domains of physical functioning (p = 0.232), pain (p = 0.212), general health (p = 0.181), mental composite score (p = 0.960), social functioning (p = 0.683), or role- emotional (p = 0.110). Amongst dialysis recipients, PCS was significantly lower in those on PD compared to those prescribed HD (p = 0.005). Patients receiving PD additionally reported a lower mental composite score (MCS) compared to those prescribed HD (p = 0.003). Pain scores (PS) were poorer in patients prescribed PD compared to those receiving HD (p = 0.030). Vitality scores were lower in patients receiving PD compared to those on HD (p = 0.005) Social functioning score (SF) was similarly poorer in patients on PD compared to those on HD Mathew et al. BMC Psychology (2023) 11:151 Page 5 of 11 Fig. 1 HADS anxiety and depression scores between treatment groups Table 2 HADS and KDQOL-SF36 scores across treatment groups All dialysis HD PD CM CM stage 5 Hospital anxiety and depression scale Anxiety Depression 5 [3–8] 5 [2 – 8.5] Kidney disease quality of life—short form 36 Physical composite (PCS) Physical functioning (PF) Role – physical (RP) Pain (PS) General health (GH) Mental composite (MCS) Vitality (VS) Social functioning (SF) Role – emotional (RE) Emotional well-being (EWB) 60.2 ± 10.7 49.4 ± 10.4 58 ± 35.3 81.5 ± 21.2 54.2 ± 10.6 51.1 ± 8.4 60.3 ± 10.7 54.5 ± 18.9 40.7 ± 3.7 47.7 ± 12.3 Values are median [interquartile range] or mean ± SD 7 [4–10] 6 [3–10] 63.2 ± 9.6 48.7 ± 11.6 60 ± 35.4 86.1 ± 16.7 54.4 ± 10.6 53.6 ± 9.3 63.2 ± 9.6 59.3 ± 17.8 46.7 ± 32.3 40.4 ± 10.0 4 [2–6] 4 [2–7] 57.3 ± 10.9 50.0 ± 9.0 56.0 ± 35.6 76.9 ± 24.3 53.9 ± 10.6 48.6 ± 6.6 57.3 ± 10.9 49.8 ± 18.9 34.7 ± 32.3 55.0 ± 9.8 4 [2–7] 3 [1–7] 66.2 ± 7.7 51.5 ± 10.3 82.5 ± 26.8 85.8 ± 17.0 50.1 ± 9.3 50.7 ± 7.6 66.2 ± 7.7 55.8 ± 14.8 32.0 ± 27.7 54.6 ± 7.5 5 [3–8] 4 [1–7] 64.9 ± 8.4 51.5 ± 10.1 79.0 ± 29.6 84.0 ± 17.2 51.3 ± 9.6 51.2 ± 7.5 64.9 ± 8.4 53.6 ± 15.5 36.6 ± 27.7 55.3 ± 7.8 (p = 0.011); EWB was however better in the PD cohort compared to those on HD (p < 0.001). No significant difference was observed between dialysis modalities for the subdomains of physical function (p = 0.533), RP (p = 0.574), general health (p = 0.814), or role-emotional (p = 0.066). Taken together, these findings suggest poorer quality of life amongst patients receiving dialysis across measure- ments of composite physical health (PCS), self-assessed ability to meet expectations of physical activity (RP), subjective feelings of fatigue (VS), and emotional well- being (EWB). In addition, a possible effect for prescribed Mathew et al. BMC Psychology (2023) 11:151 Page 6 of 11 e r o c s 6 3 F S - L O Q D K n a e M 90 80 70 60 50 40 30 HD Fig. 2 KDQOL-SF36 scores between treatment groups PD CM Stage 5 CM dialysis modality was observed for HADS anxiety score, PCS, pain (PS), general mental health (MCS), VS, and EWB. Stepwise sigma-restricted multivariate linear regres- sion modelling was used to evaluate the confounding effect of baseline characteristics on scores which showed significant difference between treatment modalities (Tables 3, 4, 5). Improving haemoglobin concentration was associated with an ameliorated anxiety and lower depression scores across all treatment groups. Long-term relationships appeared to exert a salutatory effect on anxiety, with the effect being more significant for patients on dialysis. In this group, prescription of PD in preference to HD fur- ther reduced anxiety. Younger age was associated with increased anxiety across the cohorts as a whole. Acceptance onto dialysis was associated with reduced perceived ability to perform expected physical activi- ties (RP) and poorer emotional well-being. Improving haemoglobin concentration was associated with bet- ter role-physical score; higher albumin concentration was associated with better PCS and vitality scores, but poorer EWB. In patients receiving dialysis, prescription of PD was associated with poorer PCS, worse pain score, lower vitality, and poorer social functioning; the modality was, however, associated with better EWB. Haemoglobin concentration in dialyzed patients showed positive asso- ciations with PCS, pain, and vitality scores; an inverse association was observed for MCS. Black African ethnic- ity was associated with better reported pain and social functioning scores. Improving albumin concentration showed negative association with EWB. Age, patient sex, relationship status, comorbidity with diabetes, dialysis vintage, and phosphate and parathyroid hormone con- centration showed no association with any of the ana- lysed domains. Discussion This study emphasises the significant psychosocial chal- lenges faced by patients living with advanced chronic kidney disease. Underpinning disordered psychological health, patients experience significant physical limita- tion resulting in distress regarding their ability to meet expected levels of physical function. Whilst disease- dependent physiological factors such as anaemia may contribute to these symptoms, choice of dialysis modality Mathew et al. BMC Psychology (2023) 11:151 Page 7 of 11 Table 3 Effect of baseline characteristics on HADS anxiety and depression scores HADS anxiety score HADS depression score β ± SE β p β ± SE β Patients living with advanced chronic kidney disease Age Black African ethnicity Relationship status Income source Haemoglobin Patients receiving dialysis Relationship status Prescribed peritoneal dialysis Haemoglobin R2adj 0.18, p < 0.001 − 0.21 ± 0.08 − 0.19 ± 0.07 − 0.17 ± 0.07 – − 0.27 ± 0.07 R2adj 0.22, p < 0.001 − 0.22 ± 0.09 − 0.20 ± 0.09 − 0.31 ± 0.10 0.008 0.012 0.021 – < 0.001 0.018 0.037 0.001 R2adj 0.08, p < 0.001 – – – − 0.19 ± 0.08 − 0.23 ± 0.08 p – – – 0.015 0.004 Table 4 Effect of baseline characteristics on KDQOL-SF36 physical health domains Physical composite Role–physical β ± SE β p β ± SE β Patients living with advanced chronic kidney disease Black African ethnicity Income source On dialysis programme Haemoglobin Albumin Patients receiving dialysis Black African ethnicity Prescribed peritoneal dialysis Haemoglobin R2adj 0.12, p < 0.001 0.18 ± 0.08 0.17 ± 0.08 – – 0.27 ± 0.08 R2adj 0.16, p < 0.001 – − 0.40 ± 0.10 0.34 ± 0.10 0.024 0.029 – – < 0.001 – < 0.001 < 0.001 R2adj 0.14, p < 0.001 – – − 0.29 ± 0.08 0.21 ± 0.08 – p – – < 0.001 0.008 – Pain β ± SE β p R2adj 0.24, p < 0.001 0.39 ± 0.09 − 0.30 ± 0.09 0.32 ± 0.09 < 0.001 0.002 < 0.001 may play an important role in the severity of depression and anxiety and in individual emotional well-being. Population studies estimate the prevalence of depres- sion amongst South Africans to be 9.8% and that of anxi- ety disorder to be 15.8% [15]; prevalence of symptoms of these disorders as evidenced by HADS scoring was higher in this cohort of patients living with advanced CKD. Prevalence rates of depression and anxiety disor- ders have been reported to be higher in patients living with CKD than in either the general population [16–18] or in patients living with other chronic diseases [17], and to increase with progression of chronic kidney disease stage [19]. HRQOL scores deteriorate with advancing CKD as uraemic symptoms mount [20, 21]. Physical function is prominently affected by CKD due to the limitations imposed on activity by ensuing anaemia, fluid retention, and acidosis [21]. Reflecting this, physical functioning was reduced in all treatment groups in this study. Con- sistent with the high prevalence of mood and anxiety disorders amongst patients in this study, emotional well- being was generally poor in all treatment groups, and patient subjective assessment of ability to meet expected emotional function (role–emotion) was uniformly poor in all respondents. Amelioration of the physical symp- toms of uraemia by kidney replacement therapies may not necessarily bring about parallel improvements in emotional wellbeing, as evidenced by reductions in this subdomain in patients receiving dialysis compared to those assigned to conservative management in the Mathew et al. BMC Psychology (2023) 11:151 Page 8 of 11 Table 5 Effect of baseline characteristics on KDQOL-SF36 mental health domains Mental composite β ± SE β p Vitality β ± SE β Patients living with advanced chronic kidney disease Social functioning Emotional well-being p β ± SE β p β ± SE β Black African ethnicity Source of income On dialysis programme Albumin Patients receiving dialysis R2 0.12, p < 0.001 0.18 ± 0.08 0.17 ± 0.08 – 0.27 ± 0.08 0.023 0.029 – < 0.001 R2 0.22, p < 0.001 – − 0.15 ± 0.07 − 0.44 ± 0.08 − 0.39 ± 0.08 Black African ethnicity R2adj 0.10, p < 0.001 – Source of income – Peritoneal dialysis Haemoglobin Albumin – − 0.33 ± 0.0 – R2adj 0.16, p < 0.001 – – − 0.40 ± 0.10 0.34 ± 0.10 – – R2adj 0.09, p = 0.003 0.22 ± 0.10 – – < 0.001 − 0.23 ± 0.10 < 0.001 – – R2adj 0.40, p < 0.001 – − 0.16 ± 0.08 0.51 ± 0.08 – − 0.23 ± 0.08 0.022 – 0.017 – – – – < 0.001 – p – 0.041 < 0.001 < 0.001 – 0.046 < 0.001 – 0.008 present study. Loss of body autonomy, dependency on KRT for survival, fear of future therapy-related complica- tions, and the constant threat of imminent mortality have been identified as significant contributors to emotional distress in patients receiving chronic dialysis [22]. Experience of these therapy-related stressors in the context of age-determined expectations of health may partially explain the inverse association between anxiety scores and age observed in this and other studies [23]. Younger patients may exhibit poorer emotion coping skills, in turn contributing to greater psychological dis- tress and poorer quality of life [24]. Patient participation in the KRT planning process may ameliorate fears around therapy and restore a degree of autonomy, in turn improving satisfaction with treat- ment and quality of life [4]. Proponents of PD further cite improved patient satisfaction with the modality arising from reduced impact of dialysis treatments on quality of life [22]. In South Africa, where resource constraints restrict patient choice of dialysis modality, this reported benefit of PD is often used to justify the institution of “PD-first” programmes. Actual evidence for any differ- ence in mental health and quality of life between dialy- sis modalities depends on few studies with contradictory findings [17, 25, 26]. Apparent differences in quality-of-life measurements between dialysis modalities may in part reflect con- founding effects of therapy-related factors. For exam- ple, selection for PD requires home circumstances of a higher socioeconomic status, a known factor in quality-of-life scores [27]. PD patients in this study were more likely to be employed than those on HD and were less frequent recipients of social grants than other treatment groups, evidencing better socio-economic circumstances in this group. Haemoglobin concentra- tion, which is known to affect both physical [11, 20] and mental [20, 21] quality-of-life scores, was lower in HD patients, reflecting a host of modality-related fac- tors including blood loss in the extracorporeal circuit and vascular access procedures. Patients prescribed PD may manifest lower albumin concentration due to pro- tein loss in dialysate effluent [28]; albumin has in some studies been associated with physical quality of life scores [29, 30]. Consistent with previous studies [11, 20, 21], higher haemoglobin concentration in the present study was independently associated with better physical quality of life, vitality, and reduced anxiety scores. Improved tis- sue oxygenation as haemoglobin levels rise is a probable contributor to physical quality of life and vitality meas- ures; amelioration of anaemia-related symptoms may similarly account for improvements in reported anxi- ety. Albumin has traditionally been viewed as a marker for nutritional status in patients living with advanced CKD, although factors such as inflammation and fluid overload may also affect serum concentrations [28]. Such clinical correlates likely underlie the association between albumin and physical composite and vital- ity quality-of-life scores observed in this study, which are broadly consistent with previous findings reported by other researchers [29–31]. Interestingly, in the pre- sent study albumin demonstrated an inverse correlation with emotional well-being. Improvements in physi- cal health and fatigue associated with increased albu- min concentration may exacerbate discontent with the limitations imposed on daily living by the diagnosis of Mathew et al. BMC Psychology (2023) 11:151 Page 9 of 11 advanced chronic kidney disease, leading to lower emo- tional well-being. Further study is, however, required to test this hypothesis. In this study, prescription of PD was independently associated with reduced anxiety and improved emo- tional well-being. Metanalysis has suggested some improvement in emotional distress and psychologi- cal well-being measures for patients on PD compared those on in-centre HD [32]. Additionally, PD recipients were more likely to maintain financial functionality, a finding which is consistent with other studies [33]. Despite these salutatory effects, overall composite men- tal score was lower in PD patients in the present study than in patients on HD, reflecting the contribution of lower-scoring domains such as social functioning, vitality, and the related subjective assessment of emo- tional role fulfilment (role–emotional) to the composite mental score. Reduced social functioning and vital- ity in patients receiving PD has been reported in other studies [33] and reflect the continuous nature of the modality with the need to perform frequent exchanges and associated disordered sleep [34]. Poorer physical composite score in the PD cohort of the present study is likely to have been influenced by the significantly poorer pain score reported by these patients. Abdom- inal discomfort engendered by frequent dialysate indwells [34] in patients receiving PD is likely to have resulted in a higher pain score in this group relative to those receiving intermittent HD. Limitations There are limitations to the present study. Treatment groups were not well matched for age, dialysis vintage, and comorbidities, and had significant differences in a number of measured laboratory parameters. Although not measured, there is also likely to have been significant variation in eGFR between treatment groups, as well as Kt/V urea clearance between the HD and PD cohorts. The systematics of provision of dialytic therapy in the local context, and the effect of different treatments on lab parameters, make these differences unavoidable, and their distribution in this study is reflective of clinical real- ity. Limitations imposed by the cross-sectional nature of this study should also be acknowledged. Although the exclusion of recently admitted patients and the admin- istration of the questionnaires on routine appointment days were employed to limit the effects of recent pertur- bations on responses, it may be that long-term pooled data would be better reflective of the burden of mental health disorders in these patients. Finally, the single-cen- tre nature of this study may limit the generalizability of the findings. Conclusions The present study confirms the significant prevalence of anxiety, depression, and reduced quality- of-life in patients living with advanced CKD, but also offers hope. In particular, the direct effect of PD on reducing anxiety and improving emotional well-being, and the potential indirect effect of PD on mental health through facilita- tion of employment, provides reassurance on the “PD- first” programmes adopted by many state units in South Africa and other LMIC. Deleterious effects of CKD on quality-of-life may be ameliorated through attention to haemoglobin targets, and through nutritional interven- tions to improve albumin. Younger patients receiving KRT may particularly require additional mental health support to cope with the demands of treatment. Abbreviations Alb ANOVA CAPD CKD CM eGFR HADS Hb HD HR-QOL IQR KDIGO KDOQI KDQOL-SF KF LTRRT MWUT PD Pmp PO4 PTH ROPD RRT SD SF-36 QOL WHO X2 Albumin Analysis of Variance Continuous Ambulatory Peritoneal Dialysis Chronic Kidney Disease Conservative Management Estimated Glomerular Filtration Rate Hospital Anxiety and Depression Scale Haemoglobin Haemodialysis Health Related Quality of Life Interquartile Range Kidney Disease Improving Global Outcome Kidney Disease Outcome Quality Initiative Kidney Disease Quality of Life-Short Form Kidney Failure Long Term Renal Replacement Therapy Mann Whitney U Test Peritoneal Dialysis Per Million Population Phosphate Parathyroid Hormone Renal Out Patient Department Renal Replacement Therapy Standard Deviation Short Form-36 Quality of Life World Health Organisation Chi Squared Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s40359- 023- 01196-1. Additional file 1. Multicollinearity analysis for included regression parameters Additional file 2. Hospital Anxiety and Depression Scale Additional file 3. KDQOL-36 Survey Acknowledgements We acknowledge all participants that were included in this study. Author contributions NM and MD contributed toward the design of the study. NM collected the data. MD, FK and ZC acted as Supervisors. MD provided statistical analysis. Mathew et al. BMC Psychology (2023) 11:151 Page 10 of 11 Data analysis was done by NM, MD, ZC. Manuscript write-up was done by NM, MD. MD, FK, ZC critically reviewed and edited manuscript. All authors contrib- uted significantly and have read and approved the final manuscript. 12. Tannor EK, Archer E, Kapembwa K, van Schalkwyk SC, Davids MR. Qual- ity of life in patients on chronic dialysis in South Africa: a comparative mixed methods study. BMC Nephrol. 2017;18:4. Funding No funding was required for this study. Availability of data and materials Data is available on request from the corresponding author. Declarations Ethical approval and consent to participate Approval for this study was obtained from the Human Research Ethics Committee of the University of the Witwatersrand, Johannesburg, South Africa (protocol number M200635). Institutional approval was obtained from Helen Joseph Hospital. Informed consent was obtained from all patients. All methods were carried out in accordance with the Declaration of Helsinki and relevant guidelines and regulations. Consent for publication All authors viewed the final draft and consented to its publication. Competing interests The authors declare that they have no competing interests. Author details 1 University of the Witwatersrand, Johannesburg, South Africa. 2 Division of Nephrology, Helen Joseph Hospital, Johannesburg, South Africa. 3 Depart- ment of Psychology, Helen Joseph Hospital, Johannesburg, South Africa. Received: 4 August 2022 Accepted: 2 May 2023 References 1. Naicker S. Burden of end-stage renal disease in sub-saharan Africa. Clin Nephrol. 2010;74(Suppl. 1):1–5. 2. George C, Mogueo A, Okpechi I, Echouffo-Tcheugui JB, Kengne AP. Chronic kidney disease in low-income to middle-income countries: the case for increased screening. BMJ Glob Health. 2017;2(2):1–10. 3. Davids MR, Jardine T, Marais N, Sebastian S, Davids T, Jacobs JC. South African Renal Registry report 2019. Afr J Nephrol. 2021;24(1):95–106. 4. Callahan MB. Using quality of life measurement to enhance interdisci- plinary collaboration. Adv Renal Replace Ther. 2001;8(2):148–51. 5. Moosa MR. The state of kidney transplantation in South Africa. S Afr J Med. 2019;109:235–40. 6. Mapes DL, Lopes AA, Satayathum S, Mccullough KP, Goodkin DA, Loca- telli F, Fukuhara S, Young EW, Kurokawa K, Saito A, Bommer J, Wolfe RA, Held PJ, Port FK. Health-related quality of life as a predictor of mortality and hospitalization: the Dialysis Outcomes and Practice Patterns Study (DOPPS). Kidney Int. 2003;64(1):339–49. 7. Oshus TBH, Preljevic VT, Sandvik L, Leivestad T, Nordhus IH, Dammen T, Os I. Mortality and health-related quality of life in prevalent dialysis patients: comparison between 12-items and 36-items short-form health survey. Health Qual Life Outcomes. 2012;10(46):1–9. 8. Nagasawa H, Tachi T, Sugita I, Esaki H, Yoshida A, Kanematsu Y, Noguchi Y, Kobayashi Y, Ichikawa E, Tsuchiya T, Teramachi H. The effect of quality of life on medication compliance among dialysis patients. Front Phar- macol. 2018;9:488. Lee MB, Bargman JM. Survival by dialysis modality—who cares? Clin J Am Soc Nephrol. 2016;11(6):1083–7. 9. 10. Seedat YK, MacIntosh CG, Subban JV. Quality of life for patients in an end-stage renal disease programme. S Afr Med J. 1987;71(8):500–4. 11. Okpechi IG, Nthithe T, Swanepoel CR. Health-related quality of life in patients on haemodialysis and peritoneal dialysis. Saudi J Kidney Dis. 2013;24(3):519–26. 13. Loosman WL, Siegert CEH, Korzec A, Honig A. Validity of the Hospi- tal Anxiety and Depression Scale and the Beck Depression Inven- tory for use in end-stage renal disease patients. Br J Clin Psychol. 2010;49(4):507–16. 14. Aiyegbusi OL, Kyte D, Cockwell P, Marshall T, Gheorghe A, Keeley T, Gheorghe A, Keeley T, Slade A, Calvert M. Measurement proper- ties of patient-reported outcome measures (PROMs) used in adult patients with chronic kidney disease : a systematic review. PLoS ONE. 2017;12(6): e0179733. 15. Herman AA, Stein DJ, Seedat S, Heeringa SG, Moomal H, Williams DR. The South African stress and health (SASH) study: 12-month and life- time prevalence of common mental disorders. S Afr Med J. 2009;99(5 Pt 2):339–44. 16. Murtagh FEM, Addington-hall J, Higginson IJ. The prevalence of symptoms in end-stage renal disease: a systematic review. Adv Chronic Kidney Dis. 2007;14(1):82–99. 17. Goh ZS, Griva K. Anxiety and depression in patients with end-stage renal disease : impact and management challenges—a narrative review. Int J Renovasc Dis. 2018;11:93–102. 18. Zalai D, Szeifert L, Novak M. Psychological distress and depression in patients with chronic kidney disease. Semin Dial. 2012;25(4):428–38. 19. Palmer S, Vecchio M, Craig JC, Tonelli M, Johnson DW, Nicolucci A, Pel- legrini F, Saglimbene V, Logroscino G, Fishbane S, Strippoli GFM. Preva- lence of depression in chronic kidney disease: systematic review and meta-analysis of observational studies. Kidney Int. 2013;84(1):179–91. 20. Mujais SK, Story K, Brouillette J, Takano T, Soroka S, Franek C, Mendelssohn D, Finkelstein FO. Health-related quality of life in CKD patients: correlates and evolution over time. Clin J Am Soc Nephrol. 2009;4(8):1293–301. 21. Pagels AA, Söderkvist BK, Medin C, Hylander B, Heiwe S. Health-related quality of life in different stages of chronic kidney disease and at initia- tion of dialysis treatment. Health Qual Life Outcomes. 2012;10:71. 22. Juergensen E, Wuerth D, Finkelstein SH, Juergensen PH, Bekui A, Finkel- stein FO. Hemodialysis and peritoneal dialysis: patients’ assessment of their satisfaction with therapy and the impact of the therapy on their lives. Clin J Am Soc Nephrol. 2006;1(6):1191–6. 23. Mok MMY, Liu CKM, Lam FM, Kwan LPY, Chan GCW, Ma MKM, Yap DYH, Chiu F, Choy CBY, Tang SCW, Chan TM. A longitudinal study on the prevalence and risk factors for depression and anxiety, quality of life and clinical outcomes in incident peritoneal dialysis patients. Perit Dial Int. 2019;39(1):74–82. 24. Ulusoy SI, Kal O. Relationship among coping strategies, quality of life, and anxiety and depressive disorders in haemodialysis patients. Ther Apheresis Dial. 2019;24(2):189–96. 25. Tian N, Chen N, Li PK-T. Depression in dialysis patients. Curr Opin Neph- rol Hypertens. 2016;21(8):639–46. 26. Donahue S, Quinn DK, Cukor D, Kimmel PL. Anxiety presentations and treatments in populations with kidney disease. Semin Nephrol. 2021;41(6):516–25. 27. Sesso R, Rodrigues-Neto JF, Ferraz MB. Impact of socioeconomic status on the quality of life in ESRD patients. Am J Kidney Dis. 2003;41(1):186–95. 28. Guest S. Hypoalbuminemia in peritoneal dialysis patients. Adv Perit Dial. 2013;29:55–60. 29. Mingardi G, Cornalba L, Cortinovis E, Ruggiata R, Mosconi P, Apolone G. Health-related quality of life in dialysis patients. A report from an Italian study using the SF-36 Health Survey. Nephrol Dial Transplant. 1999;14(6):1503–10. 30. Mittal S, Ahern L, Flaster E, Mittal VS, Maesaka J, Fishbane S. Self- assessed quality of life in peritoneal dialysis patients. Am J Nephrol. 2001;21:215–20. 31. Jhamb M, Pike F, Ramer S, Argyropoulos C, Steel J, Dew MA, Weisbord SD, Wessfield L, Unruh M. Impact of fatigue on outcomes in the hemo- dialysis (HEMO) study. Am J Nephrol. 2011;33:515–23. 32. Cameron J, Whiteside C, Katz J, Devins GM. Differences in quality of life across renal replacement therapies: a meta-analytic comparison. Am J Kidney Dis. 2000;35(4):629–37. Mathew et al. BMC Psychology (2023) 11:151 Page 11 of 11 33. Wu AW, Fink NE, Marsh-Manzi JVR, Meyer KB, Finkelstein FO, Chapman MM, Power NR. Changes in quality of life during hemodialysis and peritoneal dialysis treatment: generic and disease specific measures. J Am Soc Nephrol. 2004;15(3):743–53. 34. Aguiar R, Pei M, Qureshi AR, Lindholm B. Health-related quality of life in peritoneal dialysis patients: a narrative review. Semin Dial. 2018;32(5):453–62. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s12934-021-01548-9	Yuan et al. Microb Cell Fact (2021) 20:53 https://doi.org/10.1186/s12934-021-01548-9 Microbial Cell Factories RESEARCH Open Access The role of the gut microbiota on the metabolic status of obese children Xin Yuan1, Ruimin Chen1* , Kenneth L. McCormick2, Ying Zhang1, Xiangquan Lin1 and Xiaohong Yang1 Abstract Background: The term “metabolically healthy obese (MHO)” denotes a hale and salutary status, yet this connotation has not been validated in children, and may, in fact, be a misnomer. As pertains to obesity, the gut microbiota has garnered attention as conceivably a nosogenic or, on the other hand, protective participator. Objective: This study explored the characteristics of the fecal microbiota of obese Chinese children and adolescents of disparate metabolic statuses, and the associations between their gut microbiota and circulating proinflamma- tory factors, such as IL-6, TNF-α, lipopolysaccharide-binding protein (LBP), and a cytokine up-regulator and mediator, leptin. Results: Based on weight and metabolic status, the 86 Chinese children (ages 5–15 years) were divided into three groups: metabolically healthy obese (MHO, n 23), and healthy normal weight controls (Con, n 21). In the MUO subjects, the phylum Tenericutes, as well as the alpha and beta diversity, were significantly reduced compared with the controls. Furthermore, Phylum Synergistetes and genus Bacteroides were more prevalent in the MHO population compared with controls. For the MHO group, Spearman’s correlation analysis revealed that serum IL-6 positively correlated with genus Paraprevotella, LBP was positively correlated with genus Roseburia and Faecalibacterium, and negatively correlated with genus Lactobacillus, and leptin correlated posi- tively with genus Phascolarctobacterium and negatively with genus Dialister (all p < 0.05). 42), metabolic unhealthy obese (MUO, n = = = Conclusion: Although there are distinct differences in the characteristic gut microbiota of the MUO population versus MHO, dysbiosis of gut microsystem is already extant in the MHO cohort. The abundance of some metabolism- related bacteria associates with the degree of circulating inflammatory compounds, suggesting that dysbiosis of gut microbiota, present in the MHO children, conceivably serves as a compensatory or remedial response to a surfeit of nutrients. Keywords: Metabolically healthy obese, Children, 16s rRNA, Gut microbiota Introduction The global epidemic of childhood obesity, and the accom- panying rise in the prevalence of endocrine, metabolic, and cardiovascular comorbidities, is perhaps the most impactful and ubiquitous public health disorder of the modern world [1]. In the context of this pandemic, a Correspondence: chenrm321@163.com 1 Department of Endocrinology, Fuzhou Children’s Hospital of Fujian Medical University, NO. 145, 817 Middle Road, Fuzhou 350005, China Full list of author information is available at the end of the article distinct group of youth with obesity who are devoid of metabolic disturbances—so-called “metabolically healthy obese” (MHO)—have been identified. Obesity notwith- standing, by definition MHO children retain a favora- ble metabolic profile, with preserved insulin sensitivity along with normal blood pressure, glucose homeostasis, lipids, and liver enzymes. Moreover, their hormonal, inflammation, and immune profiles are seemingly imper- vious to obesity [2]. First described in obese adults, the MHO phenotype has also been extensively studied in young people with obesity [2]. Arguably, MHO may be © The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/ zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Yuan et al. Microb Cell Fact (2021) 20:53 Page 2 of 13 a transitional stage to the far more common, more high- risk, conventional cardio-metabolic obese phenotype. Regardless of the aforesaid normal biochemical charac- teristics of MHO, the risk for cardiovascular disease per- sists since the MHO phenotype may be unstable, thereby transitory [3, 4]. Among the non-genetic factors associated with obesity, the gut microbiota has garnered attention as an obesity regulator given the robust correlations in animal stud- ies between gut microbiota and body weight. Obese individuals, whether adults or children, have increased abundance in Firmicutes in concert with decreased in Bacteroidetes [5, 6]. The distinctive gut microbiota prevalent in obese subjects is recognized as promoting an unhealthy metabolic obese (MUO) phenotype with attendant comorbidities, such as increased endotoxemia, intestinal and systemic inflammation, as well as insulin resistance. An altered gut microbiota has been implicated in obesity and type 2 diabetes mellitus (T2DM) inso- far as a decrement in certain species and gene richness have been linked to adiposity, dyslipidemia, and insulin resistance [7]. Hence, the clinical repercussions aside, it is plausible that differences in the gut microbiota could dictate whether an obese child is metabolically fit (MHO) or not (MUO) [8, 9]. Obesity and related metabolic disorders are associated with gut microbiota dysbiosis, disrupted intestinal bar- rier and chronic inflammation [10]. For instance, obese Mexican children and adolescents had increased levels of leptin and C-reactive protein, which were associated with changes in the gut microbiota [11]. However, the asso- ciation between gut microbiota and proinflammatory cytokines, such as IL-6, TNF-α and lipopolysaccharide- binding protein (LBP), has not been fully investigated in children of varying metabolic statues. Firstly, this study examined the metabolic heterogeneity of obese children as it relates to the composition of the gut microbiota. And, as a secondary end point, identify metabolic-spe- cific bacteria which associate with serum inflammatory factors incriminated in obesity comorbidities. Results Study participants Based on weight status, the metabolically stable cohort subjects (n = 63) were subdivided as MHO (n = 42) or Con (n = 21). The age of the 86 participates ranged from 5.5 to 14.3 years, with a mean of 9.76 ± 1.93 years. There were 65 obese children, of whom 23 were MUO and 42 were MHO. The BMI of other 21 children were normal. Age, weight, BMI, BMI-Z, WHtR, SBP, TG and LDL-c in the MUO group were significantly higher than the Con and MHO children, and HDL-c in the in the MUO group were significantly lower than the Con and MHO children (all p < 0.05, Table 1). The weight, BMI, BMI-Z, WHR, WHtR, SBP, DBP, TG, LDL-c, IL-6, TNF-α, LBP and leptin were signifi- cantly higher in the MHO group than the Con children, and HDL-c in the MHO group were significantly lower than the Con group (all p < 0.05). There was no statisti- cal difference in age, gender, FPG and fasting TC between MHO and Con (all p > 0.05, Table 1). Microbiota profiles in different metabolic status subjects A total of 918,578 sequencing reads were obtained from 86 fecal samples, with an average value of 10,681 counts per sample. We identified an overall of 146 OTUs, among which 136 OTU with ≥ 2 counts, and they were grouped in 9 phylum and 38 families. Abundance profiling in different metabolic status subjects Grouping OTUs at phylum level, and applying the Mann–Whitney U test on the relative abundances of phyla for the two groups, the relative abundances of phy- lum Tenericutes was more prevalent in the metabolically healthy cohorts (Con and MHO children) compared to the MUO group (p = 0.006, Additional file 1: Table S1 and Fig. 1a). On OTUs at the genera level, by Mann–Whitney U-test, including all the genera (merging small taxa with counts < 10), we identified that genera Anaerostipes, Alis- tipes, Desulfovibrio, Fusobacterium, Gemmiger, Odori- bacter, Oscillospira and Parabacteroides were more prevalent in the metabolically healthy cohorts (Con and MHO children) versus MUO children, yet the genus Dorea was more prevalent in MUO (p < 0.05; Fig. 1b, Table 2). Alpha‑ and beta‑diversity in different metabolic status subjects To assess the overall differences of microbial community structures in metabolic healthy and MUO subjects, we measured ecological parameters based on alpha-diver- sity. The alpha-diversity analysis showed significantly higher diversity in metabolic healthy subjects (Con and MHO children) in comparison to MUO participants (p < 0.05, Fig. 2a, b, Additional file 1: Table S2). To determine the differences between microbial com- munity profiles in metabolic healthy and MUO subjects, we calculated beta-diversity. By Distance method Bray– Curtis dissimilarities PCoA analysis, the gut microbiota samples from Con and MHO children were clustered together and separated partly from the MUO group. Upon analysis, the first coordinate (Axis.1) explained the 18.6% of the inter sample variance the second coordinate (Axis.2) explained the 14.5% of the inter sample variance Yuan et al. Microb Cell Fact (2021) 20:53 Page 3 of 13 Table 1 Anthropometric profiles and laboratory measurements MUO (n = 23) Metabolic healthy subjects Total (n = 63) MHO (n = 42) Con (n = 21) Age (year) Male (%) Weight (kg) BMI (kg/m2) BMI-Z WHR WHtR SBP (mmHg) DBP (mmHg) FPG (mmol/L) TC (mmol/L) TG (mmol/L) LDL-c (mmol/L) HDL-c (mmol/L) Leptin (μg/mL) TNF-α (pg/mL) IL-6 (μg/mL) LBP (μg/mL) 10.96 1.69 ± 65.2 61.4 ± 27.02 2.81 11.5 2.75 ± 0.61 ± ± 0.89 0.05 0.04 0.55 ± 116.45 8.77 ± 5.72 65.09 5.09 4.54 1.62 ± 0.67 0.90 0.99 2.65 0.66 ± ± ± ± ± 0.24 1.48 1.24 2.70 ± 47.50 1.76 ± 25.63 ± 0.86 34.8 (29.55, 41.20) 0.86 0.06 1.84 14.6* 4.91* ± 1.53* 8.36* 0.06* ± 5.79 ± 0.39 0.62 0.30* 9.32 50.8 ± 43.0 ± 21.80 1.77 ± ± 0.50 ± 101.52 62.57 4.87 4.30 0.86 ± ± ± ± ± 2.31 0.53* 1.58 2.23 ± 48.48 1.65 ± 0.30* 1.83 18.77 ± 0.93 33.66 (27.01, 38.95) 1.68* 12.4* 3.14* ± 0.60 9.47 54.8 ± 49.6 ± 24.65 2.74 ± ± 0.88 0.05 0.04 0.53 ± 105.51 6.96* ± 6.45 ± 0.38* 0.57 0.33* 63.81 4.82 4.39 0.93 2.45 0.48 ± ± ± ± ± 1.51 3.10 ± 53.43 1.86 ± 0.30* 1.65 17.88 ± 1.04 33.28 (27.75, 41.22) 9.02 2.14 ± 42.9 8.5# 1.91# 0.79# ± 0.06# 0.03# 5.51# 3.56# ± ± 0.40 29.9 ± 16.11 ± 0.16 − 0.84 ± 0.43 ± 94.48 60.38 4.97 ± ± ± ± 2.03 1.71 4.14 0.72 0.69 0.19# 0.54# 0.26# 0.35# 16.81# ± 0.42# 1.23 27.18 (22.02, 36.61)# ± 38.59 0.51 ± ± MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con controls, BMI body mass index, BMI-Z BMI standard deviation Z score, WHR waist-to-hip ratios, TC total cholesterol, TG triglyceride, LDL-c low-density lipoprotein cholesterol, HDL-c high density lipoprotein cholesterol, LBP lipopolysaccharide-binding protein Compared with the MUO group, p < 0.05 # Compared with the MHO group. Data is expressed either as mean ± SD or median (25th–75th centiles) in metabolic healthy subjects (Con and MHO children) in comparison to MUO participants (P = 0.038, Fig. 2e, Additional file 1: Table S3). Bacterial taxa differences in different metabolic status subjects We next used LEfSe analysis to identify bacteria in which the relative abundance was significantly increased or decreased in each phenotypic category. The Con and MHO children had members of the phylum Tenericutes, class Deltaproteobacteria, Mollicutes, order Desulfovi- brionales, RF39, family Christensenellaceae, Odoribacte- raceae, Porphyromonadaceae, Ruminococcaceae, genera Anaerostipes, Oscillospira, Odoribacter, Gemmiger, Para- bacteroides, Alistipes, that were significantly higher than MUO subjects. Furthermore, the MUO subjects had members of the genus Fusobacterium that were sig- nificantly higher than the Con and MHO children (all p < 0.05, Fig. 3a, b). Microbiota profiles in obese children with different metabolic status Abundance profiling Grouping OTUs at phylum level, and applying the Mann–Whitney U test on the relative abundances of phyla for the MHO and MUO groups, the relative abun- dance of phylum Tenericutes was more prevalent in the MHO group compared to the MUO group (p = 0.027, Table 3 and Fig. 1c). On OTUs at the genera level, by Mann–Whitney U analysis, including all the genera (merging small taxa with counts < 10), we identified that genera Desulfovibrio, Parabacteroides and Gemmiger were more prevalent in MHO subjects compared to MUO subjects (p = 0.027, 0.040 and 0.047, respectively; Fig. 1d). Alpha‑ and beta‑diversity between MHO and MUO subjects Regarding alpha-diversity, in both the MHO and MUO group, the analysis exposed significantly higher diversity in MHO subjects versus MUO participants (all p < 0.05, Fig. 2c, d, Additional file 1: Table S2). Regarding beta-diversity, by an unweighted-UniFrac method, the MHO group was lower than the MUO group (p = 0.021, Additional file 1: Table S3). Bacterial taxa differences between MHO and MUO subjects LEfSe analysis showed MHO subjects had members of the phylum Tenericutes, class Deltaproteobacte- ria, Mollicutes, order Desulfovibrionales, RF39, family Christensenellaceae, Odoribacteraceae, Rikenellaceae, Yuan et al. Microb Cell Fact (2021) 20:53 Page 4 of 13 a MUO MHO & Con 0.00 0.25 0.50 Relative Abundance 0.75 b MUO MHO & Con 1.00 0.00 0.25 0.50 0.75 1.00 Phylum Firmicutes Bacteroidetes Proteobacteria Actinobacteria Fusobacteria Tenericutes Verrucomicrobia TM7 Synergistetes c MUO MHO Bacteroides Not_Assigned Prevotella Megamonas Faecalibacterium Roseburia Phascolarctobacterium Ruminococcus Dialister Blautia Sutterella Bifidobacterium Alistipes Lachnospira Parabacteroides Streptococcus Relative Abundance Acidaminococcus Bilophila Fusobacterium Enterococcus Desulfovibrio Citrobacter Paraprevotella Odoribacter Anaerostipes Eubacterium Akkermansia Lactobacillus Turicibacter Gemmiger Oscillospira Dorea Coprococcus Butyricicoccus Butyricimonas Lactococcus SMB53 Lachnobacterium Cetobacterium Rothia Mitsuokella Holdemania Catenibacterium Actinomyces Weissella Anaerotruncus Klebsiella Clostridium Megasphaera Oxalobacter Morganella Adlercreutzia Coprobacillus Pseudoramibacter_Eubacterium Comamonas Granulicatella Eggerthella Pyramidobacter Abiotrophia Actinobacillus Aggregatibacter Leuconostoc Veillonella Enterobacter Haemophilus Genus d MUO MHO 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 Relative Abundance Relative Abundance Phylum Bacteroidetes Firmicutes Proteobacteria Actinobacteria Fusobacteria Tenericutes Verrucomicrobia e MHO Con Bacteroides Not_Assigned Prevotella Megamonas Phascolarctobacterium Dialister Sutterella Faecalibacterium Bifidobacterium Alistipes Parabacteroides Roseburia Oxalobacter Comamonas Streptococcus Ruminococcus Klebsiella Blautia Enterobacter Veillonella Haemophilus Acidaminococcus Oscillospira Megasphaera Clostridium Lachnospira Granulicatella SMB53 Bilophila Dorea Gemmiger Fusobacterium Coprococcus Citrobacter Paraprevotella Eubacterium Desulfovibrio Odoribacter Butyricicoccus Turicibacter Coprobacillus Actinobacillus Akkermansia Lactobacillus Cetobacterium Lactococcus Butyricimonas Anaerostipes Holdemania Catenibacterium Actinomyces Morganella Rothia Lachnobacterium Aggregatibacter Leuconostoc Genus f MHO Con 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 Relative Abundance Relative Abundance Phylum Firmicutes Bacteroidetes Proteobacteria Actinobacteria Fusobacteria Tenericutes Verrucomicrobia TM7 Synergistetes Cyanobacteria Genus Bacteroides Not_Assigned Faecalibacterium Prevotella Megamonas Roseburia Ruminococcus Phascolarctobacterium Blautia Dialister Alistipes Parabacteroides Gemmiger Comamonas Eggerthella Pyramidobacter Bifidobacterium Oscillospira Sutterella Coprococcus Megasphaera Clostridium Klebsiella Lachnospira Veillonella Dorea Streptococcus Haemophilus Acidaminococcus Parvimonas Granulicatella Abiotrophia Bilophila Desulfovibrio Fusobacterium Odoribacter Paraprevotella Anaerostipes Lactobacillus Akkermansia Lachnobacterium Eubacterium Turicibacter Butyricimonas Mitsuokella Actinobacillus Christensenella Pseudoramibacter_Eubacterium Cetobacterium SMB53 Enterobacter Anaerotruncus Holdemania Oxalobacter Citrobacter Actinomyces Adlercreutzia Lactococcus Weissella Coprobacillus Rothia Aggregatibacter Leuconostoc Fig. 1 Bar chart representing Mann–Whitney U-test results on operational taxonomic units (OTUs) grouped in phyla (a, c, e) and in genus (b, d, f) of the different metabolic status groups. Each column in the plot represents a group, and each color in the column represents the percentage of relative abundance for each OTU. MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con controls Yuan et al. Microb Cell Fact (2021) 20:53 Page 5 of 13 Table 2 The mean relative abundance of gut microbiota with significantly differences in different metabolic status at genera level MUO MHO and Con Anaerostipes Odoribacter Desulfovibrio Alistipes Fusobacterium Dorea Gemmiger Oscillospira Parabacteroides 0.001 0.000 0.000 0.010 0.001 0.012 0.007 0.008 0.007 0.001 0.002 0.003 0.023 0.002 0.005 0.013 0.010 0.020 Z − − − − − − − − − 2.084 2.122 2.142 2.182 2.185 2.288 2.320 2.445 2.552 P value 0.037 0.034 0.032 0.029 0.029 0.022 0.020 0.014 0.011 MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con controls Desulfovibrionaceae, Porphyromonadaceae, Ruminococ- caceae, genus Gemmiger, Parabacteroides that were sig- nificantly higher than MUO subjects (all p < 0.05, Fig. 3c, d). Microbiota profiles in MHO and Con children with different weight status Abundance profiling Grouping OTUs at phylum level, the relative abundances of phylum Synergistetes was more prevalent in the MHO group compared to the Con group (p < 0.05, Fig. 1e, Table 4). On OTUs at the genera level, including all the genera (merging small taxa with counts < 10), genera Anaer- otruncus, Bacteroides, Adlercreutzia and Pyramidobacter were more prevalent in MHO subjects versus MUO sub- jects (p < 0.05; Fig. 1f ). Alpha‑ and beta‑diversity between different weight status Regarding alpha-diversity, the Shannon diversity index, Observed OTUs, Faith’s phylogenetic diversity and Pie- lou’s evenness based on OTU distribution did not reveal any significant difference between MHO and Con (all p > 0.05, Additional file 1: Table S2); also, beta-diversity did not differ significantly between these two groups. Importantly, none of the comparisons were significantly different (all p > 0.05) after correction for multiple testing (Additional file 1: Table S3). MHO &Con MUO Group MHO &Con MUO Group MHO MUO b 60 40 20 d 60 40 20 1 o a h C : x e d n I y t i s r e v i d - a h p l A 1 o a h C : x e d n I y t i s r e v i d - a h p l A MHO &Con MUO MHO MUO MHO MUO Group MHO &Con MUO Group MHO MUO group MHO &Con MUO a 3.0 2.5 2.0 1.5 1.0 n o n n a h S : x e d n I y t i s r e v i d - a h p l A c 3 2 1 n o n n a h S : x e d n I y t i s r e v i d - a h p l A e 0.50 0.25 0.00 -0.25 -0.50 ] % 5 4 1 [ . 2 . s i x A -0.6 -0.3 0.0 0.3 0.6 Axis.1 [18.6%] Fig. 2 Characterization of alpha- and beta-diversity of the gut microbiota in Con, MUO and MHO groups. The y-axes show the Shannon index (a, c) and Chao1 richness index (b, d). The x-axes show the phenotypic categories. Additional data are in Additional file 1: Table S2. Principal coordinates analysis (PCoA) plot of Con and MHO children and MUO subjects (e). The plots show the first two principal coordinates (axes) for PCoA using Bray–Curtis Distance method. MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con controls Bacterial taxa differences in MHO and Con children of different weight status LEfSe analysis showed MHO subjects had members of the phylum Synergistetes, class Synergistia, order Syn- ergistales, Erysipetotrichales, family Dethiosulfovibrion- aceae, genus Pyramidobacter were significantly higher than the Con-, however, the latter had members of the Yuan et al. Microb Cell Fact (2021) 20:53 Page 6 of 13 a MH MUO c MHO e Con MHO b MH MUO d f MHO MUO Con MHO Fig. 3 Differential biomarkers associated with different metabolic status. A linear discriminant effect size (LeFse) analysis have been performed (α value 2.0). MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con controls 0.05, logarithmic LDA score threshold = = Yuan et al. Microb Cell Fact (2021) 20:53 Page 7 of 13 Table 3 The mean relative abundance of gut microbiota obese subjects with different metabolic status at phylum level MHO MUO Actinobacteria Bacteroidetes Firmicutes Fusobacteria Proteobacteria Tenericutes Verrucomicrobia 0.012 0.453 0.393 0.006 0.132 0.003 0.001 0.025 0.371 0.321 0.016 0.267 0.000 0.000 z − − − − − − − 0.783 0.823 0.919 1.494 0.535 2.212 1.480 p value 0.434 0.410 0.358 0.135 0.593 0.027 0.139 MHO, metabolically healthy obese; MUO: metabolic unhealthy obese Italicized value P < 0.05 Table 4 The mean relative abundance of gut microbiota with significantly differences in obese subjects with different metabolic status at genera level MHO Con Actinobacteria Bacteroidetes Cyanobacteria Firmicutes Fusobacteria Proteobacteria Synergistetes Tenericutes TM7 Verrucomicrobia 0.012 0.319 0.000 0.572 0.006 0.088 0.000 0.002 0.000 0.001 0.018 0.377 0.000 0.531 0.014 0.057 0.000 0.002 0.000 0.001 MHO, metabolically healthy obese; Con, control Italicized value P < 0.05 Z − − − − − − − − − − 1.181 1.006 1.245 0.831 0.324 1.881 1.964 1.408 0.481 0.177 P value 0.238 0.314 0.213 0.406 0.746 0.060 0.050 0.159 0.630 0.859 family Bacteroidaceae, genus Anaerotruncus that were significantly higher (all p < 0.05, Fig. 3e, f ). Correlations between inflammatory factors and bacterial abundance To evaluate correlations between bacteria and serum inflammatory factors (IL-6, TNF-α and leptin), Spear- man’s rho cut-off values were assessed, taking into account r > 0.4, r < − 0.4 (p < 0.05, Additional file 1: Table S4). For MUO subjects, Spearman’s correlation analy- sis revealed that IL-6 positively correlated with genus Lactococcus, TNF-α positively correlated with phylum Bacteroidetes, negatively correlated with genus Citro- bacter. LBP positively correlated with genus Prevotella, Odoribacter, and negatively correlated with genus Bifi- dobacterium, Streptococcus, Roseburia, Clostridium and Veillonella. Leptin positively correlated with genus Eubacterium and negatively correlated with genus Fae- calibacterium and Lachnospira (all p < 0.05, Additional file 1: Table S4). For MHO subjects, Spearman’s correlation analy- sis revealed that serum IL-6 positively correlated with genus Paraprevotella. LBP positively correlated with genus Roseburia and Faecalibacterium, and negatively correlated with genus Lactobacillus. Leptin positively correlated with phylum Bacteroidetes, Firmicutes, genus Phascolarctobacterium and negatively correlated with genus Dialister (all p < 0.05). There was no association between the bacteria and TNF α at the genus level (all p > 0.05). Metabolic pathway predictions A total of 15 KEGG pathways were generated using the composition of the fecal microbiota based on PICRUSt2 in the metabolic healthy cohorts (MHO and Con sub- jects) versus MUO subjects (Fig. 4, Additional file 1: Table S5). Importantly, the glucose metabolism pathways, including GDP-mannose biosynthesis and superpathway of UDP-N-acetylglucosamine-derived O-antigen building blocks biosynthesis, were increased in metabolic healthy cohorts and, conversely, the superpathway of fucose and rhamnose degradation were alternated in the metabolic healthy cohorts (all p < 0.05). In the comparison between MHO and MUO subjects, we obtained 3 differential pathways including superpathway of fucose and rham- nose degradation, photorespiration, and sucrose degrada- tion III, which were also observed significantly different between the metabolic healthy cohorts (MHO and Con subjects) versus MUO subjects (Fig. 4, Additional file 1: Table S6). Moreover, 11 differential metabolic pat- terns differentially expressed resulted in the compari- son between MHO versus Con (Fig. 4, Additional file 1: Table S7). Discussion Recognized for decades, there is wide-ranging het- erogeneity among obese individuals as to their risk for developing metabolic dysfunction and its attendant com- plications [12]. Also well-established, and which may contribute to this metabolic heterogeneity, is the fact those with central obesity are more prone to develop- ing T2DM and cardiovascular disease than those with peripheral obesity [13]. In this study, to indirectly address the issue of fat distribution, we found there were no sig- nificant differences in WHR and WHtR between the two obese cohorts, MHO vs. MUO. A chronic low-grade inflammation, triggered by nutri- ent surplus, is a constituent of obesity. Adipose-origi- nated metabolic inflammation develops pari passu with insulin resistance and, as such, is a key element in the Yuan et al. Microb Cell Fact (2021) 20:53 Page 8 of 13 a MHO&Con b c Con -0.25 -0.20 -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 -0.25 -0.20 -0.15 -0.10 -0.05 0.00 0.05 Fig. 4 KEGGs biomarkers associated with the three metabolic statuses. MUO metabolic unhealthy obese, MHO metabolically healthy obese, Con controls Yuan et al. Microb Cell Fact (2021) 20:53 Page 9 of 13 metabolic syndrome [14]. In this study, we found there were no significant differences in serum IL-6, TNF-α, LBP and leptin between MHO and MUO subjects. It stands to reason that, besides these cytokines, other biochemical factors likely contribute to the metabolic diverseness in obese subjects. Or, perhaps, the concen- trations of circulating compounds—such as those above- mentioned—poorly reflect those found in extracellular or intracellular tissues. Evidence can be adduced that the gut microbiota is involved in the aetiology of obesity and obesity-related complications such as nonalcoholic fatty liver disease, insulin resistance and T2DM [15, 16]. These disorders are characterized by alterations in the diversity of the gut microbiota, and the relative abundance of certain genera. And bacteria-generated metabolites, translocated from the gut across a disrupted intestinal barrier, can affect several metabolic organs, such as the liver and adipose, thereby contributing to systemic metabolic inflammation [17]. Recently, several animal studies concluded that an optimal healthy-like gut microbiota may bestow a more propitious obese phenotype [18, 19]. For instance, the abundance of Bacteroidetes and Tenericutes were closely aligned with bile acid metabolism and obesity-related inflammation in a murine model of the metabolic syn- drome [20]. In our study, we corroborate this finding: reduced abundance of Tenericutes in the MUO group compared with the metabolically healthy groups (MHO and Con). Moreover, individuals with diminished insulin sensitivity had lower abundance of Tenericutes [21]. And, in animal experiments, administration of hydrogenated xanthohumol, which mitigates the metabolic syndrome by altering gut microbiota diversity and abundance, specifically, a reduction in Bacteroidetes and Teneri- cutes [20]. These results suggested an important role of Tenericutes in metabolism. We also observed greater abundance of Anaerostipes in the MHO and Con cohort, as well as the alpha and beta diversity. Using separate- sample Mendelian randomization to obtain estimates of the associations of 27 genera of gut microbiota with cardiovascular disease risks, Anaerostipes was identified as being nominally associated with T2DM [22], and this effect may be a result of butyrate production [23]. These results buttress the notion of dysbiosis in the gut micro- biota of MUO individuals. To characterize the gut microbiota in obese children of different metabolic status, we further analyze the MHO and the MUO groups. The abundance of Tenericutes was significantly reduced in the MUO group compared with the metabolic healthy children, indicating that Teneri- cutes is related to the metabolic state, and the bacterial imbalance is independent of weight. Previously reported, the abundance of Parabacteroides was significantly decreased in obese subjects with metabolic syndrome [6], and nonalcoholic fatty liver disease [24], and negatively correlated with weight gain and leptin plasma levels [25]. And germane to our findings, both genera Gemmiger [26] and Parabacteroides [27] are gut bacteria negatively associated with obesity and disturbed host metabolism. In accordance, we found that that the fecal abundance of these bacteria was significantly higher in the MHO group compared with MUO. The genera Parabacteroides are short-chain fatty acids (SCFAs)-producing bacteria. SCFAs are low molecular weight molecules produced from fermentation of dietary fiber or polysaccharides by gut microbiota. Absorbed by the intestinal epithelium into the blood, they can beget physiological disorders in the host, such as deranged lipid metabolism and intestinal environment imbalances [28, 29]. In our determination, alpha and beta diversity were significantly higher in Con and MHO children compared with the MUO group, again supporting the notion of dys- biosis in the unhealthy MUO population. Notwithstanding that the gut microbiota of obese individuals with metabolic syndrome may indeed be unhealthy, is the gut microbiota of the MHO popula- tion really healthy? We compared the characteristic of gut microbiota in the Con and MHO children of differ- ent weights. Even though there was no significant differ- ence in alpha and beta diversity, the relative abundances of phylum Synergistetes and genus Bacteroides were ele- vated in the MHO group compared to the Con children. Based on a metagenomic approach and bioinformatics analysis in obese adults, it is plausible that an abundance of the microbiota taxa Bacteroides could portent the evo- lution to T2DM [30]. Alterations in gut ecology can propel inflammatory pathways in several tissues, resulting in glucose intoler- ance and CVD [31, 32]. In rodents, a disturbance in the tripartite interactions between the microbiota, bile acids, and host metabolism, along with the bacterial production of lipopolysaccharides (LPS, i.e., endotoxemia), can beget derangements in glucose homeostasis [16, 26]. LBP is an acute inflammation phase protein that complexes with LPS and facilitates binding with CD14. In adolescents, serum LBP robustly correlates positively with indices of abnormal glucose and lipid metabolism. Herein, we found that, depending on the metabolic status, the serum levels of classic proinflammatory factors IL-6, TNF-α, LBP and leptin were related to the abundance of various fecal bac- teria. Notably, in MHO children, serum leptin correlated positively with genus Phascolarctobacterium and nega- tively with Dialister—the latter genera observed with low abundance in obese children [33]. And, relevant to our findings, it is noteworthy that Phascolarctobacterium is Yuan et al. Microb Cell Fact (2021) 20:53 Page 10 of 13 purportedly a biomarker for adult T2DM [30]. In high fat diet obese mice with insulin resistance, Prevotella was deemed as pro-inflammatory and, of note, its abundance in our study correlated with serum LBP [34]. As illus- trated in our MHO children and the above-cited studies in humans, the gut microbiota is a marquee player in pre- serving normal metabolism despite obesity or, perhaps, an ephemeral protective microbiota destined to change with transition to MUO. Compared to the metabolic healthy cohorts in the MUO children, several pathways associated with glu- cose and lipid metabolism pathways, such as fucose and rhamnose degradation and sucrose degradation III were increased. Conversely, mannan degradation was mark- edly decreased. Of interest, serum fucose levels are higher in the T2DM patients compared to healthy cohorts [35]. Mannan-oligosaccharide in the diet improves the meta- bolic syndrome in mice, alternatively insulin resistance and dyslipidemia [36, 37]. We found that bacterial fucose and rhamnose degradation and sucrose degradation III were increased in the MUO subjects compared with the MHO subjects, inferring that the change was independ- ent of weight. However, insofar as serum levels of fucose were undetectable, and the dietary intake of sucrose and mannan were not assessed in our study, future longitudi- nal studies could conceivably unravel the intricate, pos- sibly causual, relationships between the gut microbiota, obesity, and aberrant intermediary host metabolism. Conclusion In aggregate, the MUO population had lower alpha- and beta-diversity, and lower abundance of Tenericutes, inferring a robust intricate inter-relationship between gut bacterial ecology and host metabolic state. In the MHO population, phylum Synergistetes and genus Bacteroides and Phasco- larctobacterium were more prevalent, and the abundance of some metabolism-related bacteria correlated with circu- lating proinflammatory factors, suggesting that compared to healthy controls, dysbiosis of gut microbiota was already extant in the MHO children, and conceivably a compensa- tory or remedial response to a surfeit of nutrients. Methods Study population This study was approved by the Ethics Committee of the Fuzhou Children’s Hospital of Fujian Medical University and, in all cases, informed consent was obtained. The cross-sectional study consisted of participants managed by Fuzhou Children’s Hospital of Fujian Medi- cal University from September 2017 to March 2018. This study was limited to participants who met the following criteria: (a) ages between 5 to 15 years old, and (b) resi- dence of Fujian province. The exclusion criteria were as follows: any endo- crine disorder, history of antibiotic therapy in the past 3 months prior to the enrollment, chronic gastrointesti- nal illness or use of gastro-intestinal-related medication, or diarrheal disease (World Health Organization defini- tion) in the past 1 month. Clinical assessment Height and weight were measured by trained nurses. BMI-Z scores were calculated based on reference values of Li et al. [38]. At the end of normal expiration, waist and hip circumference were measured to the nearest 0.5 cm using standard technique with nonelastic tape. Waist circumference was measured at a point midway between the lower border of the ribs and the iliac crest, and hip circumference was measured at the widest part of the hip. A waist-to-hip ratio (WHR) was calculated by waist circumference (cm) divided by hip circumference (cm) and a waist-to-height ratio (WHtR) by waist cir- cumference (cm) divided by height (cm). Laboratory methods All participants maintained their usual dietary pattern at least 3 days before blood sampling. After 12 h of fast- ing, 10 mL venous blood was drawn by registered nurses. All blood samples were stored at − 80 ℃, and analyzed within two weeks of sampling. Serum IL-6 was meas- ured using a commercial ELISA kit (Abcam, UK), with an 4.4% inter-assay coefficient of variation (CV). Serum TNF-α levels was measured using a commercial ELISA kit (Abcam, UK), with inter-assay and intra-assay CVs of 3.3% and 9%, respectively, and serum leptin assayed using a commercial ELISA kit (Abcam, UK), with inter- assay and intra-assay CVs of 2.4% and 2.7%, respectively. The serum LBP levels were measured using a commercial ELISA kit (Abnova, Taiwan, China), with inter-assay and intra-assay CV 9.8–17.8% and 6.1%, respectively. Fasting plasma glucose (FPG) and plasma lipids, including total cholesterol (TC), triglyceride (TG), high-density lipopro- tein cholesterol (HDL-c) and low density lipoprotein cho- lesterol (LDL-c), were assayed by standard methods using specific reagents (Beckman Coulter AU5800, USA). Fast- ing insulin (INS) was determined by a chemiluminescent immunoassay (IMMULITE 2000, Siemens Healthcare Diagnostics Products Limited, Germany). Fecal samples were collected and processed as previously described [39]. Definition of metabolic unhealthy Metabolic syndrome parameters were applied accord- ing to 2019 Expert Committees [40], and MUO was Yuan et al. Microb Cell Fact (2021) 20:53 Page 11 of 13 defined by the presence of at least one of the following metabolic traits: (1) FPG ≥ 5.6 mmol/L; (2) systolic blood pressure ≥ 90th percentile for gender and age; (3) fasting HDL-C < 1.03 mmol/L; and (4) fasting TG ≥ 1.7 mmol/L. Genomic DNA extraction and library construction The microbial community DNA was extracted and quantified as previously described [39]. Variable regions V3–V4 of bacterial 16s rRNA gene were amplified with degenerate PCR primers [39]. Libraries were qualified by the Agilent 2100 bioanalyzer (Agilent, USA). The validated libraries were used for sequencing on Illumina MiSeq platform (BGI, Shenzhen, China) following the standard pipeline of Illumina, and generating 2 × 300 bp paired-end reads. Statistical analysis Statistical analyses of clinical data were performed using the Statistical Package for the Social Sciences software version 23.0 (SPSS Inc. Chicago, IL, USA). The normal- ity of the data was tested by Kolmogorov–Smirnov test. Data are expressed as mean ± SD or median (25th–75th percentiles). Comparisons of the results were assessed using independent samples t test, Mann–Whitney U test and Kruskal–Wallis test, depending on the type of data distribution (e.g., non parametric). Comparison of rates between two groups was by chi-square. A value of P < 0.05 was deemed statistically significant. Statistical analysis of 16s rRNA sequencing data were performed on alpha- and beta-diversity measurements, which was done by software QIIME2 (v2019.7) [41]. Kruskal–Wallis Test was adopted for two groups com- parison. Linear discriminant analysis Effect Size (LEfSe) Analysis was assessed by software LEFSE [42]. To pre- dict metagenome functional content from 16S rRNA gene surveys, Picrust2 [43] have been applied to obtain the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, and STAMP [44] was used to analyze the dif- ferential pathways. Supplementary Information The online version contains supplementary material available at https ://doi. org/10.1186/s1293 4-021-01548 -9. Additional file 1: Table S1. The mean relative abundance of gut micro- biota in different metabolic status at phylum level. Table S2. Comparison of alpha-diversity in obese subjects with different metabolic status. Table S3. Comparison of beta-diversity between different metabolic status. Table S4. Spearman’s correlation table on OTUs and inflamma- tory factors in MHO and MUO groups. Table S5. KEGGs biomarkers in MHO and Con subjects compared with MUO subjects. Table S6. KEGGs biomarkers in MHO and MUO subjects. Table S7. KEGGs biomarkers in MHO and Con subjects. Acknowledgements The authors are grateful to all the participants. Authors’ contributions XY drafted the initial manuscript; RMC conceptualized and designed the study, and reviewed and revised the manuscript; KLM assisted in data analysis and manuscript composition; YZ and XHY collected cases; XQL did the laboratory testing. All authors read and approved the final manuscript. Funding This study was supported by Technology Innovation Team Train Project of Fuzhou Health Committee in China (2016-S-wp1), and sponsored by key Clinical Specialty Discipline Construction Program of Fuzhou, Fujian, P.R.C. (201610191) and Fuzhou Children’s Medical Center (2018080310). Availability of data and materials The original contributions presented in the study are publicly available. The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2017) in National Genomics Data Center (Nucleic Acids Res 2020), Beijing Institute of Genomics (China National Center for Bioinformation), Chinese Academy of Sciences, under accession number CRA003010 that are publicly accessible at https ://bigd.big.ac.cn/gsa. Ethics approval and consent to participate This study was reviewed and approved by the Ethics Committee of Fuzhou Children’s Hospital of Fujian Medical University, and was conducted in agreement with the Declaration of Helsinki Principles. Informed consent was obtained from all individual participants included in the study. Consent for publication Informed consent for publication was obtained from all individual participants included in the study. Competing interests The authors declare that they have no competing interests. Author details 1 Department of Endocrinology, Fuzhou Children’s Hospital of Fujian Medical University, NO. 145, 817 Middle Road, Fuzhou 350005, China. 2 Division of Pedi- atric Endocrinology and Diabetes, University of Alabama at Birmingham, Birmingham, AL 35233, USA. Received: 8 December 2020 Accepted: 18 February 2021 References 1. The CR, Picture B. Outrunning child obesity trends. BMJ. 2018;363:k4362. https ://doi.org/10.1136/bmj.k4362 . 2. Damanhoury S, Newton AS, Rashid M, Hartling L, Byrne JLS, Ball GDC. Defining metabolically healthy obesity in children: a scoping review. Obes Rev. 2018;19(11):1476–91. https ://doi.org/10.1111/obr.12721 . 3. Greenhill C. Redefining metabolically healthy obesity. Nat Rev Endocrinol. 2018;14(8):440. https ://doi.org/10.1038/s4157 4-018-0046-9. 4. Garcia-Moll X. Obesity and prognosis: time to forget about metabolically healthy obesity. Eur Heart J. 2018;39(5):407–9. https ://doi.org/10.1093/ eurhe artj/ehx53 5. 5. Gérard P. Gut microbiota and obesity. Cell Mol Life Sci. 2016;73(1):147–62. https ://doi.org/10.1007/s0001 8-015-2061-5. 6. Gallardo-Becerra L, Cornejo-Granados F, García-López R, Valdez-Lara A, Bikel S, Canizales-Quinteros S, López-Contreras BE, Mendoza-Vargas A, Nielsen H, Ochoa-Leyva A. Metatranscriptomic analysis to define the Secrebiome, and 16S rRNA profiling of the gut microbiome in obesity and metabolic syndrome of Mexican children. Microb Cell Fact. 2020;19(1):61. https ://doi.org/10.1186/s1293 4-020-01319 -y. 7. Ma Q, Li Y, Li P, Wang M, Wang J, Tang Z, Wang T, Luo L, Wang C, Wang T, Zhao B. Research progress in the relationship between type 2 diabetes mellitus and intestinal flora. Biomed Pharmacother. 2019;117:109138. https ://doi.org/10.1016/j.bioph a.2019.10913 8. Yuan et al. Microb Cell Fact (2021) 20:53 Page 12 of 13 8. Rial SA, Karelis AD, Bergeron KF, Mounier C. Gut microbiota and metabolic health: the potential beneficial effects of a medium chain triglyceride diet in obese individuals. Nutrients. 2016;8(5):281. https ://doi.org/10.3390/ nu805 0281. Iacobini C, Pugliese G, Blasetti Fantauzzi C, Federici M, Menini S. Meta- bolically healthy versus metabolically unhealthy obesity. Metabolism. 2019;92:51–60. https ://doi.org/10.1016/j.metab ol.2018.11.009. 9. 10. Yuan X, Chen R, Ouyang Q, Lin X, Ai Z, Zhang Y, Yang X. Novel associations of serum adropin and lipopolysaccharide-binding protein versus lipid profiles in childhood obesity. J Pediatr Endocrinol Metab. 2020;33(2):265– 70. https ://doi.org/10.1515/jpem-2019-0329. 11. Nirmalkar K, Murugesan S, Pizano-Zárate ML, Villalobos-Flores LE, García- González C, Morales-Hernández RM, Nuñez-Hernández JA, Hernández- Quiroz F, Romero-Figueroa MDS, Hernández-Guerrero C, Hoyo-Vadillo C, García-Mena J. Gut microbiota and endothelial dysfunction markers in obese mexican children and adolescents. Nutrients. 2018;10(12):2009. https ://doi.org/10.3390/nu101 22009 . 12. Viveiros A, Oudit GY. The dual nature of obesity in metabolic pro- gramming: quantity versus quality of adipose tissue. Clin Sci. 2020;134(18):2447–51. https ://doi.org/10.1042/CS202 01028 . 13. Brandão I, Martins MJ, Monteiro R. Metabolically healthy obesity- heterogeneity in definitions and unconventional factors. Metabolites. 2020;10(2):E48. https ://doi.org/10.3390/metab o1002 0048. 14. Kacířová M, Zmeškalová A, Kořínková L, Železná B, Kuneš J, Maletínská L. Inflammation: major denominator of obesity, type 2 diabetes and Alzheimer’s disease-like pathology? Clin Sci. 2020;134(5):547–70. https :// doi.org/10.1042/CS201 91313 . 15. Canfora EE, Meex RCR, Venema K, Blaak EE. Gut microbial metabolites in obesity, NAFLD and T2DM. Nat Rev Endocrinol. 2019;15(5):261–73. https ://doi.org/10.1038/s4157 4-019-0156-z. 16. Pedersen HK, Gudmundsdottir V, Nielsen HB, Hyotylainen T, Nielsen T, Jensen BA, Forslund K, Hildebrand F, Prifti E, Falony G, Le Chatelier E, Levenez F, Doré J, Mattila I, Plichta DR, Pöhö P, Hellgren LI, Arumugam M, Sunagawa S, Vieira-Silva S, Jørgensen T, Holm JB, Trošt K, MetaHIT Consortium, Kristiansen K, Brix S, Raes J, Wang J, Hansen T, Bork P, Brunak S, Oresic M, Ehrlich SD, Pedersen O. Human gut microbes impact host serum metabolome and insulin sensitivity. Nature. 2016;535(7612):376– 81. https ://doi.org/10.1038/natur e1864 6. 17. Tilg H, Zmora N, Adolph TE, Elinav E. The intestinal microbiota fuelling metabolic inflammation. Nat Rev Immunol. 2020;20(1):40–54. https ://doi. org/10.1038/s4157 7-019-0198-4. 18. Serino M, Luche E, Gres S, Baylac A, Bergé M, Cenac C, Waget A, Klopp P, Iacovoni J, Klopp C, Mariette J, Bouchez O, Lluch J, Ouarné F, Monsan P, Valet P, Roques C, Amar J, Bouloumié A, Théodorou V, Burcelin R. Metabolic adaptation to a high-fat diet is associated with a change in the gut microbiota. Gut. 2012;61(4):543–53. https ://doi.org/10.1136/gutjn l-2011-30101 2. 19. Neyrinck AM, Van Hée VF, Piront N, De Backer F, Toussaint O, Cani PD, Delzenne NM. Wheat-derived arabinoxylan oligosaccharides with prebiotic effect increase satietogenic gut peptides and reduce metabolic endotoxemia in diet-induced obese mice. Nutr Diabetes. 2012;2(1):e28. https ://doi.org/10.1038/nutd.2011.24. 20. Zhang Y, Bobe G, Revel JS, Rodrigues RR, Sharpton TJ, Fantacone ML, Raslan K, Miranda CL, Lowry MB, Blakemore PR, Morgun A, Shulzhenko N, Maier CS, Stevens JF, Gombart AF. Improvements in metabolic syndrome by xanthohumol derivatives are linked to altered gut microbiota and bile acid metabolism. Mol Nutr Food Res. 2020;64(1):e1900789. https ://doi. org/10.1002/mnfr.20190 0789. 21. Benedict C, Vogel H, Jonas W, Woting A, Blaut M, Schürmann A, Ceder- naes J. Gut microbiota and glucometabolic alterations in response to recurrent partial sleep deprivation in normal-weight young individu- als. Mol Metab. 2016;5(12):1175–86. https ://doi.org/10.1016/j.molme t.2016.10.003. 22. Yang Q, Lin SL, Kwok MK, Leung GM, Schooling CM. The roles of 27 gen- era of human gut microbiota in ischemic heart disease, type 2 diabetes mellitus, and their risk factors: a mendelian randomization study. Am J Epidemiol. 2018;187(9):1916–22. https ://doi.org/10.1093/aje/kwy09 6. 23. Koh A, De Vadder F, Kovatcheva-Datchary P, Bäckhed F. From dietary fiber to host physiology: short-chain fatty acids as key bacterial metabolites. Cell. 2016;165(6):1332–45. https ://doi.org/10.1016/j.cell.2016.05.041. 24. Monga Kravetz A, Testerman T, Galuppo B, Graf J, Pierpont B, Siebel S, Feinn R, Santoro N. Effect of gut microbiota and PNPLA3 rs738409 vari- ant on nonalcoholic fatty liver disease (NAFLD) in obese youth. J Clin Endocrinol Metab. 2020;105(10):dgaa382. https ://doi.org/10.1210/cline m/dgaa3 82. 25. Carbajo-Pescador S, Porras D, García-Mediavilla MV, Martínez-Flórez S, Juarez-Fernández M, Cuevas MJ, Mauriz JL, González-Gallego J, Nistal E, Sánchez-Campos S. Beneficial effects of exercise on gut microbiota func- tionality and barrier integrity, and gut-liver crosstalk in an in vivo model of early obesity and non-alcoholic fatty liver disease. Dis Model Mech. 2019;12(5):dmm039206. https ://doi.org/10.1242/dmm.03920 6. 26. Brown JM, Hazen SL. Microbial modulation of cardiovascular disease. Nat Rev Microbiol. 2018;16:171–81. https ://doi.org/10.1038/nrmic ro.2017. 27. Patrone V, Vajana E, Minuti A, Callegari ML, Federico A, Loguercio C, Dallio M, Tolone S, Docimo L, Morelli L. Postoperative changes in fecal bacterial communities and fermentation products in obese patients undergo- ing bilio-intestinal bypass. Front Microbiol. 2016;7:200. https ://doi. org/10.3389/fmicb .2016.00200 . 28. Jing N, Liu X, Jin M, Yang X, Hu X, Li C, Zhao K. Fubrick tea attenuates high-fat diet induced fat deposition and metabolic disorder by regulating gut microbiota and caffeine metabolism. Food Funct. 2020;11(8):6971– 86. https ://doi.org/10.1039/d0fo0 1282c . 29. Hold GL. The gut microbiota, dietary extremes and exercise. Gut. 2014;63(12):1838–9. https ://doi.org/10.1136/gutjn l-2014-30730 5. 30. Wang TY, Zhang XQ, Chen AL, Zhang J, Lv BH, Ma MH, Lian J, Wu YX, Zhou YT, Ma CC, Dong RJ, Ge DY, Gao SH, Jiang GJ. A comparative study of microbial community and functions of type 2 diabetes mellitus patients with obesity and healthy people. Appl Microbiol Biotechnol. 2020;104(16):7143–53. https ://doi.org/10.1007/s0025 3-020-10689 -7. 31. Biassoni R, Di Marco E, Squillario M, Barla A, Piccolo G, Ugolotti E, Gatti C, Minuto N, Patti G, Maghnie M, d’Annunzio G. Gut microbiota in T1DM- onset pediatric patients: machine-learning algorithms to classify microor- ganisms as disease linked. J Clin Endocrinol Metab. 2020;105(9):dgaa407. https ://doi.org/10.1210/cline m/dgaa4 07. 32. van de Guchte M, Blottière HM, Doré J. Humans as holobionts: implica- tions for prevention and therapy. Microbiome. 2018;6:81. https ://doi. org/10.1186/s4016 8-018-0466-8. 33. Wahlström A, Sayin SI, Marschall HU, Bäckhed F. Intestinal crosstalk between bile acids and microbiota and its impact on host metabolism. Cell Metab. 2016;24(1):41–50. https ://doi.org/10.1016/j.cmet.2016.05.005. 34. Cai W, Xu J, Li G, Liu T, Guo X, Wang H, Luo L. Ethanol extract of propolis prevents high-fat diet-induced insulin resistance and obesity in association with modulation of gut microbiota in mice. Food Res Int. 2020;130:108939. https ://doi.org/10.1016/j.foodr es.2019.10893 9. 35. Al-Aama JY, Al Mahdi HB, Salama MA, Bakur KH, Alhozali A, Mosli HH, Bahijri SM, Bahieldin A, Willmitzer L, Edris S. Detection of secondary metabolites as biomarkers for the early diagnosis and prevention of type 2 diabetes. Diabetes Metab Syndr Obes. 2019;17(12):2675–84. https ://doi. org/10.2147/DMSO.S2155 28. 36. Wang H, Zhang X, Wang S, Li H, Lu Z, Shi J, Xu Z. Mannan-oligosaccharide modulates the obesity and gut microbiota in high-fat diet-fed mice. Food Funct. 2018;9(7):3916–29. https ://doi.org/10.1039/c8fo0 0209f . 37. Chen H, Nie Q, Hu J, Huang X, Zhang K, Pan S, Nie S. Hypoglycemic and hypolipidemic effects of glucomannan extracted from Konjac on type 2 diabetic rats. J Agric Food Chem. 2019;67(18):5278–88. https ://doi. org/10.1021/acs.jafc.9b011 92. 38. Li H, Zong XN, Ji CY, Mi J. Body mass index cut-offs for overweight and obesity in Chinese children and adolescents aged 2–18 years. Zhonghua Liu Xing Bing Xue Za Zhi. 2010;31(6):616–20. 39. Yuan X, Chen R, Zhang Y, Lin X, Yang X. Sexual dimorphism of gut micro- biota at different pubertal status. Microb Cell Fact. 2020;19(1):152. https :// doi.org/10.1186/s1293 4-020-01412 -2. 40. Liu C, Wang C, Guan S, Liu H, Wu X, Zhang Z, Gu X, Zhang Y, Zhao Y, Tse LA, Fang X. The prevalence of metabolically healthy and unhealthy obesity according to different criteria. Obes Facts. 2019;12(1):78–90. https ://doi.org/10.1159/00049 5852. 41. Bolyen E, Rideout JR, Dillon MR, Bokulich NA, Abnet CC, Al-Ghalith GA, Alexander H, Alm EJ, Arumugam M, Asnicar F, Bai Y, Bisanz JE, Bittinger K, Brejnrod A, Brislawn CJ, Brown CT, Callahan BJ, Caraballo-Rodríguez AM, Chase J, Cope EK, Da Silva R, Diener C, Dorrestein PC, Douglas GM, Durall DM, Duvallet C, Edwardson CF, Ernst M, Estaki M, Fouquier J, Gauglitz JM, Yuan et al. Microb Cell Fact (2021) 20:53 Page 13 of 13 Gibbons SM, Gibson DL, Gonzalez A, Gorlick K, Guo J, Hillmann B, Holmes S, Holste H, Huttenhower C, Huttley GA, Janssen S, Jarmusch AK, Jiang L, Kaehler BD, Kang KB, Keefe CR, Keim P, Kelley ST, Knights D, Koester I, Kosciolek T, Kreps J, Langille MGI, Lee J, Ley R, Liu YX, Loftfield E, Lozupone C, Maher M, Marotz C, Martin BD, McDonald D, McIver LJ, Melnik AV, Metcalf JL, Morgan SC, Morton JT, Naimey AT, Navas-Molina JA, Nothias LF, Orchanian SB, Pearson T, Peoples SL, Petras D, Preuss ML, Pruesse E, Rasmussen LB, Rivers A, Robeson MS 2nd, Rosenthal P, Segata N, Shaffer M, Shiffer A, Sinha R, Song SJ, Spear JR, Swafford AD, Thompson LR, Torres PJ, Trinh P, Tripathi A, Turnbaugh PJ, Ul-Hasan S, van der Hooft JJJ, Vargas F, Vázquez-Baeza Y, Vogtmann E, von Hippel M, Walters W, Wan Y, Wang M, Warren J, Weber KC, Williamson CHD, Willis AD, Xu ZZ, Zaneveld JR, Zhang Y, Zhu Q, Knight R, Caporaso JG. Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nat Biotechnol. 2019;37(8):852–7. https ://doi.org/10.1038/s4158 7-019-0209-9. 42. Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, Hutten- hower C. Metagenomic biomarker discovery and explanation. Genome Biol. 2011;12(6):R60. https ://doi.org/10.1186/gb-2011-12-6-r60. 43. Douglas GM, Maffei VJ, Zaneveld JR, Yurgel SN, Brown JR, Taylor CM, Huttenhower C, Langille MGI. PICRUSt2 for prediction of metagenome functions. Nat Biotechnol. 2020;38(6):685–8. https ://doi.org/10.1038/ s4158 7-020-0548-6. 44. Parks DH, Tyson GW, Hugenholtz P, Beiko RG. STAMP: statistical analysis of taxonomic and functional profiles. Bioinformatics. 2014;30(21):3123–4. https ://doi.org/10.1093/bioin forma tics/btu49 4. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1242_dev.201085	© 2023. Published by The Company of Biologists Ltd \| Development (2023) 150, dev201085. doi:10.1242/dev.201085 RESEARCH ARTICLE NHR-23 activity is necessary for C. elegans developmental progression and apical extracellular matrix structure and function Londen C. Johnson1, An A. Vo1, John C. Clancy1, Krista M. Myles1, Murugesan Pooranachithra2, Joseph Aguilera1, Max T. Levenson1, Chloe Wohlenberg1, Andreas Rechtsteiner1, James Matthew Ragle1, Andrew D. Chisholm2 and Jordan D. Ward1,* ABSTRACT Nematode molting is a remarkable process where animals must repeatedly build a new apical extracellular matrix (aECM) beneath a previously built aECM that is subsequently shed. The nuclear hormone receptor NHR-23 (also known as NR1F1) is an important regulator of C. elegans molting. NHR-23 expression oscillates in the epidermal epithelium, and soma-specific NHR-23 depletion causes severe developmental delay and death. Tissue-specific RNAi suggests that nhr-23 acts primarily in seam and hypodermal cells. NHR-23 coordinates the expression of factors involved in molting, lipid transport/metabolism and remodeling of the aECM. NHR-23 depletion causes dampened expression of a nas-37 promoter reporter and a loss of reporter oscillation. The cuticle collagen ROL- 6 and zona pellucida protein NOAH-1 display aberrant annular localization and severe disorganization over the seam cells after NHR-23 depletion, while the expression of the adult-specific cuticle collagen BLI-1 is diminished and frequently found in patches. Consistent with these localization defects, the cuticle barrier is severely compromised when NHR-23 is depleted. Together, this work provides insight into how NHR-23 acts in the seam and hypodermal cells to coordinate aECM regeneration during development. KEY WORDS: C. elegans, Molting, NHR-23, Nuclear hormone receptor, Apical extracellular matrix, Auxin-inducible degron INTRODUCTION Molting is a crucial developmental process required for the growth of all ecdysozoans: a clade comprising an estimated total of 4.5 million living species (Telford et al., 2008). Invertebrate molting involves conserved processes such as apical extracellular matrix (aECM) remodeling, intracellular trafficking and oscillatory gene expression (Lažetić and Fay, 2017). Molting is also a process of interest for developing new drugs against parasitic nematodes 1Department of Molecular, Cell, and Developmental Biology, University of California Santa Cruz, Santa Cruz, CA 95064, USA. 2Department of Cell and Developmental Biology, School of Biological Sciences, University of California San Diego, La Jolla, CA 92093, USA. Author for correspondence ( jward2@ucsc.edu) L.C.J., 0000-0002-8196-4957; A.A.V., 0000-0002-3130-6235; J.C.C., 0000- 0003-2867-9740; K.M.M., 0000-0002-1281-7793; M.P., 0000-0003-3506-7785; J.A., 0000-0003-1824-361X; M.T.L., 0000-0002-1470-3968; C.W., 0000-0002- 2848-4372; J.M.R., 0000-0002-6626-2615; A.D.C., 0000-0001-5091-0537; J.D.W., 0000-0001-9870-8936 This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed. Handling Editor: Swathi Arur Received 7 July 2022; Accepted 11 April 2023 (Ghedin et al., 2007). Parasitic nematodes cause over a billion human infections each year and loss of livestock and crops (Ward, 2015b). However, very few drugs exist, and resistance to those drugs is emerging rapidly (Ward, 2015b). Although many genes involved in nematode molting have been identified in C. elegans (Frand et al., 2005), little is known about how their gene products are coordinated to promote aECM remodeling and the generation and release of a new cuticle. Nematodes progress through four periodic larval stages (L1-L4) before becoming a reproductive adult (Brenner, 1974). The end of each larval stage is punctuated by a molt that involves trafficking and secretion of aECM components, assembly of a new aECM underneath the old cuticle, followed by separation of the cuticle from the underlying aECM (apolysis) and shedding of the old cuticle (ecdysis) (Lažetić and Fay, 2017). Apolysis coincides with a sleep- like behavior called lethargus (Singh and Sulston, 1978). The cuticle is a collagenous exoskeleton secreted by hypodermal and seam epithelial cells (Page and Johnstone, 2007). The outermost layer (glycocalyx) is rich in carbohydrates and mucins (Nelson et al., 1983; Singh and Sulston, 1978). Beneath the glycocalyx is the glycolipid- and lipid-rich epicuticle, which is postulated to function as a hydrophobic surface barrier (Blaxter, 1993; Blaxter et al., 1992). Underlying the epicuticle is a layer comprising mainly cuticlin proteins. Between the epidermal membrane and the epicuticle are collagen-rich layers, layer with collagen struts (Lažetić and Fay, 2017; Page and Johnstone, 2007). Sets of collagens oscillate in expression over the course of each molt and are classified into three groups based on temporal expression: early, intermediate and late collagens (Johnstone and Barry, 1996; Page and Johnstone, 2007). Although the specific localization of most collagens is unknown, collagens from the same group are thought to be in the same layer (Lažetić and Fay, 2017). A transient structure, the sheath or pre-cuticle, is formed during each larval stage and is thought to pattern the cuticle (Cohen and Sundaram, 2020). Many of the components of this pre-cuticle are related to mammalian matrix proteins. The C. elegans pre-cuticle contains zona pellucida proteins, proteins related to small leucine-rich proteoglycans and lipid transporters in the lipocalin family (Cohen and Sundaram, 2020; Cohen et al., 2020; Forman-Rubinsky et al., 2017; Kelley et al., 2015; Sapio et al., 2005; Vuong-Brender et al., 2017). including a fluid-filled medial Nuclear hormone receptor (NHR) transcription factors are key regulators of molting in insects and nematodes (King-Jones and Thummel, 2005; Taubert et al., 2011). NHRs are characterized by a ligand-binding domain (LBD) that has the potential to bind small molecules such as ligands and dietary-derived metabolites (Taubert et al., 2011). NHRs have a canonical zinc-finger DNA-binding domain (DBD) with an unstructured hinge region between the DBD and LBD that is subject to post-translational regulation (Antebi, 2015; Campbell et al., 2008; Ward et al., 2013). A single, conserved nuclear hormone 1 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 receptor, NHR-23 (also known as NR1F1; hereafter referred to as NHR-23), which is an ortholog of DHR3 in insects and of RORα in mammals, is a key regulator of C. elegans molting. NHR-23 is also necessary for spermatogenesis (Ragle et al., 2020, 2022). nhr-23 mutation or inactivation by RNAi leads to embryonic lethality, larval arrest, ecdysis defects and morphology defects (Frand et al., 2005; Gissendanner et al., 2004; Kostrouchova et al., 1998, 2001). nhr-23 mRNA expression oscillates over the course of each larval stage, peaking at mid-larval stage and falling at the molt stage (Gissendanner et al., 2004; Kostrouchova et al., 2001). nhr-23 is necessary for all four larval molts and it regulates microRNAs such as let-7 and lin-4 (Kostrouchova et al., 1998, 2001; Patel et al., 2022; Kinney et al., 2023 preprint). let-7 also regulates nhr-23, suggesting that a feedback loop might coordinate molting with developmental timing (Patel et al., 2022). The NHR-23 insect ortholog (DHR3) is part of the molting gene regulatory network (Lam et al., 1997; Ruaud et al., 2010), and the mammalian ortholog (RORα) regulates circadian rhythms, lipid metabolism and immunity (Jetten, 2009). However, how NHR-23 regulates molting and whether it is part of the core oscillator that promotes rhythmic gene expression over each larval stage is poorly understood (Tsiairis and Großhans, 2021). We show here that NHR-23 protein oscillates and rapid NHR-23 depletion via an auxin-inducible degron causes severe developmental delay and death. Analysis of NHR-23 target genes suggests a role in coordinating aECM assembly, remodeling and construction of specific cuticular structures. NHR-23 depletion causes aberrant localization of the collagens ROL-6 and BLI-1, and of the pre-cuticle factor NOAH-1. NHR-23 activity is necessary in seam and hypodermal cells to promote molting and timely development. Our work reveals when and where NHR-23 acts to promote molting. RESULTS NHR-23 protein oscillates during development Expression of nhr-23 mRNA oscillates throughout each C. elegans larval stage (Gissendanner et al., 2004; Hendriks et al., 2014; Kostrouchova et al., 2001; Meeuse et al., 2020). However, mRNA expression profiles do not always correlate with protein levels (de Sousa Abreu et al., 2009; Vogel and Marcotte, 2012). To determine whether NHR-23 protein oscillates, we monitored the expression of an endogenously tagged NHR-23::GFP, over 28 h (wrd8 allele; Ragle et al., 2020). During L1, oscillating expression of NHR-23:: GFP was observed in the nuclei of seam and hypodermal cells (Fig. S1). Expression of NHR-23 rises and falls before the completion of the 1st molt (Fig. 1A). As GFP tags can affect the expression or stability of proteins (Agbulut et al., 2006; Baens et al., 2006), we assayed the expression of NHR-23 through western blotting time- courses using an nhr-23::AID::3xFLAG strain (kry61 allele; Zhang et al., 2015). These experiments confirmed that NHR-23:: AID::3xFLAG also oscillates (Fig. 1B,C), similar to our NHR-23:: GFP imaging experiments (Fig. 1A). We detected three distinct NHR-23 bands; the lowest band is consistent with the size of NHR- 23b/f (Fig. 1B,C, Fig. S2). The upper two bands were larger than expected, which could reflect post-translational modification or that NHR-23 migrates aberrantly during SDS-PAGE. All observed NHR-23 bands oscillated, although the lowest band was expressed more strongly than the other bands (Fig. 1B,C). The peak in NHR- 23 expression was earlier in each larval stage than previous qRT-PCR approaches suggested (Gissendanner et al., 2004; Kostrouchova et al., 2001) and more in line with recent RNA-seq and imaging data (Meeuse et al., 2020; Kinney et al., 2023 preprint). To understand how NHR-23 promotes molting, it is important to clarify when it peaks in expression. We examined NHR-23::GFP expression over the course of the 4th larval stage, as vulva morphology allows for more precise sub-staging (Mok et al., 2015). NHR-23::GFP expression was first detectable in L4.1 larvae, peaked in expression from L4.2-L4.3, disappeared in the vulva by L4.5 and in the head by L4.6, and then remained undetectable (Fig. 1D,E, Fig. S1B,C), similar to what the observations of Kinney et al. (2023) of NHR-23::mScarlet. NHR-23 depletion causes severe developmental delay Given that NHR-23 oscillates, we tested when NHR-23 was necessary for molting during the L1 larval stage. In a pilot experiment, we consistently found weaker depletion phenotypes using an nhr-23::GFP::AID::3xFLAG strain (Ragle et al., 2020) compared with the nhr-23::AID::3xFLAG strain (Zhang et al., 2015) (Table S1). We therefore transitioned to using the nhr-23:: AID::3xFLAG strain for all experiments involving phenotypic analysis. We first tested NHR-23::AID::3xFLAG depletion kinetics and found robust depletion within 15 min of exposure to 4 mM auxin; levels remained low over the 16 h timecourse (Fig. 2A). We performed timed depletion experiments in the L1 larval stage (Fig. 2B) using an eft-3p::TIR1::mRuby2 control strain (henceforth referred to as TIR1) and a strain that permits somatic depletion of NHR-23 in the presence of auxin (eft-3p::TIR1:: mRuby2; nhr-23::AID::3xFLAG; henceforth referred to as nhr- 23::AID; TIR1). L1 larvae were synchronized by starvation arrest and release onto MYOB plates. Animals were shifted onto either 4 mM auxin or control plates every 2 h and scored for molting defects at 20 h post-release from starvation (Fig. 2B). TIR1 control animals shifted onto control or auxin and nhr-23::AID; TIR1 animals shifted onto control plates all reached the L2 stage (Fig. 2C,D). In contrast, nhr-23::AID; TIR1 animals shifted onto auxin displayed multiple phenotypes, including molting defects, internal vacuoles and developmental abnormalities such as disrupted tails and viable animals with a squashed morphology (Fig. 2C). We then quantified the molting defects, scoring animals unable to shed their cuticles or with morphological abnormalities (Fig. 2D). Animals shifted onto auxin within the first 4 h post-release tended to arrest as L1 larvae with few molting or morphological defects (Fig. 2D). To determine at what stage these animals arrested, we used an hlh-8p::GFP promoter reporter to monitor the M cell lineage (Harfe et al., 1998). Newly hatched L1 animals have one M cell that undergoes a stereotypical series of divisions to produce 16 M lineage cells by the L1 molt (Fig. S3A) (Sulston and Horvitz, 1977). nhr-23::AID; TIR1 animals shifted onto auxin at 3 h post-release were late reaching L1 based on M cell number (Fig. S3B,C). When NHR-23 animals were scored 72 h post-auxin shift, we observed molting defects and necrotic animals (Fig. 2E, Fig. S4A), suggestive of a severe developmental delay. We also observed arrested animals with a wild-type morphology that were likely L1 larvae by size (Fig. 2E). Shifts between 6 and 12 h post-release resulted in increased morphological and molting defects (Fig. 2D). Animals shifted onto auxin at 9 h reached the end of the L1 stage by M cell number (Fig. S3B,C). Shifts at 14 and 16 h post-release resulted in wild- type L2 animals, as judged by size (Fig. 2D). To test whether NHR-23 was similarly required in other larval stages, we performed depletion experiments later in development during which we could use vulval morphology to score progression through the L4 stage. We performed depletion experiments shifting early L3 animals onto auxin and monitored development by scoring vulva morphology 23 h later. The majority of TIR1 animals grown on control or auxin plates and nhr-23::AID; TIR1 animals grown on 2 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Fig. 1. NHR-23 protein oscillates during development. (A) Representative images from a timecourse monitoring the expression of endogenous NHR-23:: GFP protein. Two biological replicates were performed and the images are representative of the number of animals indicated at each timepoint. Scale bars: 20 µm. (B,C) Anti-FLAG immunoblot analysis of synchronized nhr-23::AID::3xFLAG animals monitoring the expression of NHR-23 3xFLAG across two timecourses: 4-28 h (B) and 22-34 h (C). Stain-free analysis, which visualizes total protein on the membrane, is provided as a loading control. The blots are representative of three experimental replicates. The times at which animals were observed to enter lethargus is indicated in A-C. (D,E) NHR-23::GFP protein expression in the vulva (D) and head (E) during the L4 larval stage. Animals were staged based on vulval morphology (Mok et al., 2015). Images are representative of 20 animals examined over four independent experiments. Scale bars: 5 µm in D; 20 µm in E. nhr-23::GFP::AID::3xFLAG (A,C,E) and nhr- 23::AID::3xFLAG (B,C) are previously described endogenous knock-ins that produce C-terminal translational fusions to all known nhr-23 isoforms (Ragle et al., 2020; Zhang et al., 2015). control plates reached late L4, with a fraction of the population in early or mid-L4 (Fig. 2F, Fig. S4B). In contrast, when nhr-23::AID; TIR1 animals were shifted to auxin, the vast majority of animals were early L4 larvae with a fraction of the population remaining in L3 (Fig. 2F, Fig. S4B). Repeating these experiments scoring later timepoints revealed that NHR-23-depleted animals were slowly progressing through development. After 72 h on auxin, we observed adult animals, as evidenced by the presence of oocytes (Fig. 2G, Fig. S4C), while all control animals were adults. A subset of NHR- 23-depleted L4 larvae could not be precisely staged due to aberrant vulval morphology (Fig. S4C). These animals had large vulval lumens reminiscent of a 4.3 stage vulva (Mok et al., 2015), but we also observed adults with a similar large vulval lumen (Fig. S4C). A fraction of nhr-23::AID; TIR1 animals grown for 72 h on auxin were dead and appeared necrotic, with large fluid filled spaces within the animal (Fig. 2G, Fig. S4C). These depletion experiments indicate that NHR-23 is required for timely developmental progression through multiple larval stages, consistent with previous reports of nhr-23 inactivation by RNAi (Macneil et al., 2013; Patel et al., 2022). T N E M P O L E V E D 3 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Fig. 2. Distinct phenotypes are observed when NHR-23 is depleted early or late in L1 larvae. (A) An anti-FLAG immunoblot to monitor NHR-23 depletion. Synchronized nhr-23::AID::3xFLAG; TIR1 and TIR1 animals were grown on MYOB plates for 12 h and then shifted onto control or auxin plates. Lysates were collected at the indicated time points. An anti-α-tubulin immunoblot is provided as a loading control. The blots are representative of three experimental replicates. (B) Schematic of the experimental set-up. Synchronized L1 larvae were grown on control media (light-gray bands) and transferred to 4 mM auxin (dark-gray bands) every 2 h. Animals remained on auxin until they were imaged at 20 h post-release. (C) Representative images of phenotypes observed when nhr-23::AID::3xFLAG; TIR1 animals were exposed to auxin in the first larval stage. (D) Percentage of animals of the indicated genotypes with unshed cuticles and morphological abnormalities after exposure to auxin. The number of animals scored (n) is provided. (E) Synchronized nhr-23::AID; TIR1 L1 larvae were released on auxin and scored for molting defects and death 72 h post-release. (F) Animals of the indicated genotype were synchronized and shifted onto auxin at 25 h post-release and imaged 23 h later (48 h post-release). Animals were staged based on vulval morphology (Mok et al., 2015). L4 larval stages were grouped as early L4 (4.0-4.2), mid L4 (4.3-4.5) and late L4 (4.6-4.9). (G) nhr-23::AID; TIR1 animals were treated as in F and scored after 72 h on auxin. The number of animals assayed for each genotype and condition is indicated at the top of the graphs in E-G. T N E M P O L E V E D 4 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 NHR-23 regulates oscillating genes involved in aECM biogenesis and regulation of protease activity To gain insight into how NHR-23 depletion causes developmental delay, we analyzed the expression of nhr-23-regulated genes by mining existing microarray (Kouns et al., 2011) and oscillatory gene expression (Meeuse et al., 2020) datasets (Table S2). Of the 265 nhr-23-regulated genes in the microarray dataset, 236 (89%) were oscillatory with the bulk of genes peaking in expression between 180° and 360° (Fig. 3A, Table S2); nhr-23 mRNA peaks at 178.11° (Meeuse et al., 2020). A similar trend was observed in a recent analysis of these datasets (Tsiairis and Großhans, 2021). In contrast, 10-20% of C. elegans genes oscillate in expression (Hendriks et al., 2014; Kim et al., 2013; Meeuse et al., 2020) (two-tailed χ2 test P<0.0001). nhr-23-regulated oscillating genes Fig. 3. NHR-23 regulates oscillating genes involved in aECM biogenesis. (A) Radar chart plotting amplitude over the phase of peak expression of nhr- 23-regulated genes from Kouns et al. (2011). The dotted circles indicate amplitude, with the innermost circle representing an amplitude of 1 and the outermost circle representing an amplitude of 4. Created with Biorender.com. (B) The data from A plotted as a scatter plot and functionally annotated (Table S2). We converted the 360° oscillation to developmental timing over a 9 h larval stage so that 1 hour=40°. The green shaded area represents NHR-23 expression based on RNA-seq ( peak phase=178.11°, amplitude=2.11; Meeuse et al., 2020), imaging (Fig. 1A,D,E) and western blotting data (Fig. 1B,C). (C) The genes in the aECM component class were subdivided into more specific classes and plotted as in B. (D) Average signal from NHR-23 ChIP-seq data (Gerstein et al., 2010) for nhr-23-regulated oscillating genes with peak phases at 135°-254.9° (early), 255-14.9° (middle) and 15°-134.9 (late). Expression of nhr-23 mRNA peaks at 178.11°. Signal is plotted relative to transcription start site (TSS) and transcription end site (TES). The average NHR-23 signal for all 21,600 C. elegans genes (all genes) and the 20% of genes with the lowest expression (no expression) are shown for reference. The mean signal is plotted with a line and the 95% confidence interval of the mean is indicated by the shaded area. (E) Number of NHR-23 peaks flanking and within nhr-23-regulated genes. The genes are binned by their peak phase relative to adjusted larval stage time. Non-oscillating (non-osc) nhr-23 genes are also depicted. T N E M P O L E V E D 5 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 regulation regulation, transcriptional were enriched in gene ontology classes such as ‘cuticle structure’, ‘regulation of endopeptidases’ and ‘metalloprotease activity’ (Table S3). These gene ontology classes are a specific subset of the functions enriched in all oscillating genes (Table S4). To provide biological context, we converted the 360° period to developmental time, assuming a 9 h larval stage and set the molt from the start of lethargus (45°) to the end of ecdysis (135°) as in Meeuse et al. (2023). Most nhr-23-regulated genes involved in aECM structure/function, cholesterol metabolism, and molting signal transduction had peak amplitudes within 3 h of the nhr-23 expression peak (Fig. 3B). Genes involved in the blocked protein unfolding response, as well as some aECM genes, peaked later in each larval stage (Fig. 3B). Although the C. elegans genome encodes 181 collagen genes (Teuscher et al., 2019), nhr-23 only including all of the furrow collagens regulated 13 collagens, implicated in epidermal damage sensing; these furrow collagens are in the ‘early’ collagen group (dpy-2, dpy-3, dpy-7 and dpy-10; Table S2, Fig. 3C) (Johnstone and Barry, 1996; Page and Johnstone, 2007; Dodd et al., 2018). Other nhr-23-regulated in body intermediate morphology (sqt-1, sqt-2 and rol-6; Table S2, Fig. 3C) (Kramer and Johnson, 1993; Johnstone and Barry, 1996; Page and Johnstone, 2007). There are also several sets of genes involved in aECM biogenesis and remodeling, such as lipocalins, proteases, protease inhibitors, fibrillin, PAN and zona pellucida (ZP) domain- containing proteins, and leucine-rich repeat proteins (Fig. 3C). Although not enriched as a gene ontology class, NHR-23 also regulates several transcription factors implicated in molting or energy metabolism ( peb-1, nhr-91 and dpy-20; Table S2, Fig. 3B) (Clark et al., 1995; Fernandez et al., 2004; Kasuga et al., 2013). collagens have well-described roles To explore direct targets of NHR-23, we analyzed an NHR-23 L3 ChIP-seq dataset (Gerstein et al., 2010). As expected, NHR-23 was enriched in the promoter region of genes (Fig. 3D, Fig. S5, Table S5). There was also notable enrichment downstream of the transcriptional end site (Fig. 3D, Fig. S5, Table S5). We then examined whether NHR-23 was enriched near nhr-23-regulated oscillating genes (Table S2). Genes that peaked in expression in the hour after the nhr-23 mRNA peak in expression (Fig. 3E; hour 2) had the highest average number of NHR-23 peaks flanking and within their gene body. The average number of NHR-23 peaks flanking and within gene bodies declined, and few peaks were detected flanking genes between larval stage hours 7-9 (Fig. 3E). Genes that peak in expression close to when nhr-23 mRNA levels peak tend to have higher NHR-23 levels upstream, downstream and within their gene bodies (Fig. 3D, early genes). As NHR-23-regulated oscillating genes were enriched in proteases and protease inhibitors (Table S2), we tested how NHR- 23 depletion affected expression of a nas-37 promoter reporter; nas- 37 is a protease implicated in C. elegans ecdysis (Davis et al., 2004). We released synchronized nhr-23::AID, TIR1 L1 larvae carrying a nas-37p::GFP::PEST reporter onto control and auxin plates, and monitored GFP expression over development. Animals growing on control media displayed oscillating reporter activity with peaks at 18 and 30 h (Fig. 4A,B). NHR-23 depleted animals exhibited a delayed onset of reporter expression and only a subset of animals expressed the reporter (Fig. 4A,B). There was only one pulse of nas- 37p::GFP::PEST expression, after which we detected no reporter activity (Fig. 4A,B). Together, these data confirm that NHR-23 regulates nas-37 expression and that, following NHR-23 depletion, there is a weaker pulse of target gene expression followed by a failure to express again. produces embryonic NHR-23 is necessary for NOAH-1 localization to the aECM Inactivation of the nhr-23-regulated predicted protease inhibitor disorganization with gene mlt-11 accompanying lethality reminiscent of noah-1 mutants (Ragle et al., 2022; Vuong-Brender et al., 2017). NOAH-1 is a zona pellucida domain protein that is part of the pre-cuticular aECM in embryos and larvae, a transient structure thought to play a role in patterning the cuticle and in embryonic elongation (Cohen and Sundaram, 2020). We introduced an mNeonGreen::3xFLAG (mNG::3xFLAG) tag after the ZP domain, which is similar to the insertion site for a previously generated mCherry knock-in (Fig. 5A; Vuong-Brender et al., 2017). noah-1::mNG::3xFLAG(int) animals had a wild-type brood size (Fig. S6A) and in noah-1:: mNG::3xFLAG(int), nhr-23::AID; TIR1 lysates there was a band of the expected full-length protein size (∼150 kDa) as well as a lower ∼50 kDa band (Fig. 5B). noah-1 is only predicted to have a single isoform so this might reflect a cleavage or degradation product. ZP proteins are frequently cleaved after their ZP domain (Gupta, 2021; Kiefer and Saling, 2002). Although there is a predicted furin RXXR cleavage site at the start of the CFCS domain, the smaller isoform is consistent with cleavage immediately after the ZP domain (Fig. 5A,B). It will be valuable in the future to create isoform-specific mNeonGreen knock-ins to determine where each localizes. NOAH-1::mNG was observed in the cuticle, and in punctate and tubular structures in the hypodermis (Fig. 5C). Hypodermal NOAH-1::mNG significantly overlapped with the lysosomal marker NUC-1::mCherry (Fig. 5C,D) (Guo et al., 2010; Clancy et al., 2023). NOAH-1::mNG was detected in lysosomes and weakly in the aECM from L4.1-L4.4, and then reached its maximum aECM intensity between L4.5 and L4.7, localizing to thick bands reminiscent of annuli (Fig. 5E). aECM expression decreased from L4.8 to young adulthood (Fig. 5E). NOAH-1::mNG also localized to alae from L4.6 onwards, similar to reports for NOAH-1::sfGFP (Katz et al., 2022). We next examined the effect of NHR-23-depletion on NOAH-1 expression and localization. In L4 larvae, we observed the expected vulval lumen localization and aECM expression (Cohen et al., 2020; Katz et al., 2022), as well as localization to the rectum and excretory (Fig. 5F). NHR-23 depletion did not affect NOAH-1 duct localization in the excretory duct or rectum (Fig. 5F). Although vulval morphology is aberrant in NHR-23-depleted animals, NOAH- 1 still localizes to cell surfaces and to lysosomes (Fig. 5F). In mid-L4, NOAH-1 is normally localized to the cuticle and lysosomes, and weakly localizes to alae (Fig. 5F). NHR-23 depletion caused a loss of NOAH-1::mNG localization to annuli, the formation of small irregular punctae and accumulation of NOAH-1::mNG in the aECM over the seam cells (Fig. 5F). Together, these data indicate that NHR-23 is dispensable for NOAH-1 expression, but necessary for correct localization in the cuticle. NHR-23 depletion causes defects in aECM structure and function Given the aberrant NOAH-1 localization (Fig. 5) and nhr-23- mediated regulation of early and intermediate collagens (Table S2), we next examined whether NHR-23 depletion affected aECM structure. We used CRISPR/Cas9-mediated genome editing in nhr- 23::AID; TIR1 animals to introduce mNG::3xFLAG cassettes into the intermediate collagen rol-6 (C-terminal translational fusion) and bli-1, a collagen that forms struts in the adult cuticle (internal translational fusion) (Fig. 6A). We also introduced a 3xFLAG:: mNG into bli-1 at the same location as the mNG::3xFLAG knock- in. rol-6 is regulated by nhr-23, whereas bli-1 is an adult-specific T N E M P O L E V E D 6 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Fig. 4. NHR-23 depletion causes reduced expression of a nas-37p::GFP::PEST promoter reporter. (A) nas-37p::GFP::PEST expression timecourse. Synchronized nhr-23::AID, TIR1; nas-37p::GFP::PEST L1 larvae were released on control or auxin plates and scored for head or hypodermal GFP expression every 2 h. Fifty animals per time point were scored in two independent experiments and the percentage of animals expressing GFP are presented. (B) Representative images of nhr-23::AID, TIR1; nas-37p::GFP::PEST animals grown on control or auxin plates at the indicated time points. Scale bars: 20 µm. collagen expressed in L4 larvae, so its expression would not be interrogated in the microarray experiment (Table S2) (Kouns et al., 2011). We first confirmed that each knock-in had a wild-type brood size (Fig. S6A) and then tested fusion protein size by western blotting. Anti-FLAG immunoblotting on nhr-23::AID; TIR1, rol- 6::mNG::3xFLAG lysates detected a single band of the expected size (∼70 kDa; Fig. 6B). In nhr-23::AID; TIR1, bli-1::3xFLAG:: mNG, there was a single band consistent with N-terminal processing of a monomer at an RXXR furin cleavage site that would remove 10 kDa (125 kDa predicted size). We could only detect protein by extracting soluble cuticle components. BLI-1::3xFLAG::mNG and BLI-1::mNG::3xFLAG had identical localization to one another (Fig. S6B), and were similar to an extrachromosomal transgenic BLI-1::GFP reporter (Tong et al., 2009); all subsequent experiments were performed using bli-1::mNG::3xFLAG(int). Further details of BLI-1::mNG localization will be described elsewhere (Adams et al., submitted J. Adams and A.D.C., unpublished). We first characterized expression of each translational fusion during the 4th larval stage, using vulval morphology to stage animals (Mok et al., 2015). ROL-6::mNG was first detected in thin bands reminiscent of furrows with a jagged pattern over seam cells (L4.1-L4.3; Fig. 6D). In L4.4-L4.5, some thicker aggregations began appearing; by L4.6-L4.9, ROL-6::mNG relocalized to thicker bands reminiscent of annuli (Fig. 6D). In L4.9, a fibrous pattern could also be observed (Fig. 6D). BLI-1::mNG expression was not detected until L4.5, when it was detected in hypodermal cells (Fig. 6E). In L4.4-L4.8, BLI-1::mNG was observed in a punctate localization in rows with some irregular brighter punctae (Fig. 6E). By L4.9, BLI-1::mNG was found in rows of regularly spaced punctae and was consistently excluded from the cuticle over seam cells throughout L4 and adulthood (Fig. 6E). To examine the impact of NHR-23 depletion on BLI-1 and ROL- 6 localization, we shifted nhr-23::AID; TIR1; rol-6::mNG or nhr- 23::AID; TIR1, bli-1::mNG animals onto control or auxin plates in early L3. We examined animals after 23 h on control plates when animals were mid-late L4s. We scored NHR-23-depleted animals after 47 h on auxin plates due to the developmental delay; this approach ensured that control- and auxin-treated animals were stage matched. ROL-6::mNG expression appeared unaffected by NHR- 23-depletion, but the furrow localization was irregular and thicker, and we observed gaps in the annular ROL-6::mNG over the seam cells (Fig. 6F). Many animals had a corset of unshed cuticle to T N E M P O L E V E D 7 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Fig. 5. NHR-23 depletion causes defective NOAH-1 localization. (A) Domain prediction for NOAH-1. ZP, zona pellucida; CFCS, consensus furin cleavage site. Predicted size of NOAH-1 and NOAH-1 with an internal mNeonGreen::3xFLAG is provided in kDa. (B) Anti-FLAG immunoblots on lysates from noah-1:: mNG::3xFLAG(int) mid-L4 animals. An age-matched wild-type control is included in each experiment. An anti-α-tubulin immunoblot was used as a loading control. (C) Representative images from a noah-1::mNG::3xFLAG(int); mlt-11p::NUC-1::mCherry mid-L4 animal. Arrow indicates cuticle localization. Scale bars: 5 µm. (D) Manders’ co-localization analysis of NOAH-1::mNG::3xFLAG(int) and the lysosome marker NUC-1::mCherry. Three biological replicates were performed analyzing a total of 29 animals. The horizontal line indicates the median. (E) NOAH-1::mNG expression timecourse through L4; animals were staged by vulva morphology (Mok et al., 2015). 100 animals were examined over three independent replicates. Scale bars: 10 µm. (F) Representative NOAH- 1::mNG and DIC images of the indicated tissues of noah-1::mNG::3xFLAG(int) and nhr-23::AID; TIR1 animals grown on control or auxin plates. Arrows indicate aberrant NOAH-1::mNG localization in the aECM above the seam cells, and wild-type localization to excretory duct and rectum. Two biological replicates were performed; the images represent 100% of animals scored (n=51 for control and n=44 for auxin). Scale bars: 20 µm for all images except the noah-1::mNG excretory duct, rectal epithelium and vulva. T N E M P O L E V E D 8 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Fig. 6. NHR-23-depletion causes ROL-6 and NOAH-1 localization defect, and reduced expression and mislocalization of BLI-1. (A) Cartoon of ROL-6 and BLI-1 domains with knock-in position of mNG::3xFLAG or 3xFLAG::mNG tags. Predicted protein size is provided in kDa. (B,C) Immunoblots on lysates from rol-6::mNG::3xFLAG mid-L4 animals (B) and bli-1::3xFLAG::mNG(int) mid-L4 animals (C). An age-matched wild-type control is included in each experiment. An anti-α-tubulin (B) or anti-actin (C) immunoblot was used as a loading control. For the blots in C, the anti-mNeonGreen blot was performed on a soluble cuticle fraction and the anti-actin control was performed on a soluble intracellular fraction (see Materials and Methods). An arrow indicates the position of the BLI-1::3xFLAG::mNG band. (D) ROL-6::mNG and (E) BLI-1::mNG expression timecourse through L4; animals were staged by vulva morphology (Mok et al., 2015). One hundred animals were examined over three independent replicates. Scale bars: 10 µm. Arrow indicates fibrous localization pattern in D. (F) Representative images of nhr-23::AID; TIR1, rol-6::mNeonGreen::3xFLAG (rol-6::mNG) animals grown on control or auxin plates. Arrow indicates mislocalized ROL-6::mNG in the aECM above the seam cells. Images from control and auxin-treated animals are representative of 40/40 animals observed in three independent experiments. (G) Representative images of nhr-23::AID; TIR1, rol-6::mNG animals grown on auxin with a corset phenotype. Corset is indicated by an arrow. 24/40 animals displayed this phenotype. (H) Representative images of nhr-23::AID; TIR1, bli-1::mNeonGreen::3xFLAG (bli-1::mNG) animals grown on control or auxin plates. Three experimental replicates were performed. Control images are representative of 42/42 animals scored. For auxin- treated animals, 38/38 exhibited dim expression of BLI-1::mNG and 30/38 had patchy BLI-1::mNG localization (arrow). Scale bars: 20 µm in F-H. which ROL-6::mNG localized (Fig. 6G). We next addressed whether NHR-23 depletion affected the formation of struts in the cuticle medial layer. NHR-23 depletion caused reduced expression of BLI-1::mNG and a loss of the punctate pattern (Fig. 6H). BLI-1:: mNG weakly localized to annuli and animals displayed bright disorganized patches of BLI-1::mNG in the cuticle (Fig. 6H). BLI- 1::mNG expression appeared dimmed in the nhr-23::AID; TIR1 background compared with a wild-type background, which would T N E M P O L E V E D 9 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 be consistent with low level auxin-independent NHR-23 depletion (Fig. S6B). Given the defects in aECM structure after NHR-23 depletion, we tested whether the epidermal barrier was intact. First, we incubated control and auxin-treated nhr-23::AID, TIR1 animals with the cuticle-impermeable, cell membrane-permeable Hoechst 33258 dye, which stains nuclei. N2 and TIR1 animals grown on control or auxin plates and nhr-23::AID, TIR1 animals grown on control plates exhibited no staining, whereas almost all animals on auxin plates displayed Hoechst staining (Fig. 7A). Animals shifted onto auxin early or late in L1 robustly expressed an nlp-29p::GFP reporter (Fig. 7B, Fig. S7), which can be activated by infection, acute stress and physical damage to the cuticle (Pujol et al., 2008; Zugasti and Ewbank, 2009). We further tested the cuticle integrity in a hypo- osmotic shock assay in which animals with defective cuticle barriers rapidly die after exposure to water (Kage-Nakadai et al., 2010). TIR1 animals grown on control media or auxin were completely viable in water (Fig. 7C). nhr-23::AID, TIR1 control animals had a weakly penetrant sensitivity to hypo-osmotic shock, while the same strain grown on auxin rapidly died after hypo-osmotic shock (Fig. 7C). These data suggest that NHR-23 is required for cuticle barrier establishment or maintenance. animals, 100% of nhr-23(RNAi) nhr-23 is necessary in seam and hypodermal cells for larval development Finally, we used a tissue-specific RNAi approach to determine in which cells nhr-23 was necessary for molting. In RNAi-proficient N2 exhibited developmental delay and molting defects (Fig. 8A). nhr-23 knockdown in body-wall muscle and intestinal cells produced no molting defects and we observed only mild developmental delay (Fig. 8A). Vulval precursor cell-specific nhr-23(RNAi) caused moderate molting defects and developmental delay (Fig. 8A). We observed highly penetrant developmental delay and molting defects when we performed nhr-23 RNAi in a strain using a wrt-2 promoter for tissue-specific RNAi (Fig. 8A). wrt-2 is expressed in seam cells, animals in this SCMp RNAi rectal cells and hypodermal cells (Fig. S8A,B; Aspöck et al., 1999). Given this result and the defective ROL-6::mNG and NOAH-1:: mNG (Figs 5 and 6) localization over seam cells, we generated more tissue-restricted RNAi strains. We used a minimal seam cell- specific enhancer with a pes-10 minimal promoter (Ashley et al., 2021), which is robustly expressed in seam cells with some weak hypodermal expression (Fig. S8C,D). To test the specificity of this strain, we introduced a his-72::mNG knock-in. In animals treated with control RNAi, we observed nuclear expression in seam, hypodermal, intestinal, vulval and germline cells, as expected for a histone H3 fusion (Fig. S9). mNeonGreen RNAi caused reduced HIS-72::mNG expression in seam, hypodermal syncytium and intestinal nuclei. nhr-23(RNAi) strain phenocopied nhr-23(RNAi) in N2 animals with almost all animals exhibiting developmental delay and molting defects (Fig. 8A). Notably, intestine-specific nhr-23 RNAi did not cause phenotypes (Fig. 8A). We then constructed a hypodermis-specific RNAi strain using the semo-1 promoter (Kaletsky et al., 2018; Köhnlein et al., 2020). nhr-23 RNAi in this strain produced molting defects and developmental delay, although with less penetrance compared with our SCMp-specific RNAi strain (Fig. 8A). To test the effect of SCMp-specific nhr-23 knockdown on the aECM, we introduced a BLI-1::mNG knock-ins in our seam cell-specific RNAi strain. nhr- 23 depletion in this strain caused reduced BLI-1 levels relative to (Fig. 8B). This BLI-1::mNG reduction was control RNAi comparable with both nhr-23 RNAi in a wild-type control (Fig. 8B) and with NHR-23 protein depletion in the soma (Fig. 6H). Together, these data indicate that nhr-23 is necessary in the seam and hypodermal cells timely developmental progression, completion of molting and aECM formation. for DISCUSSION C. elegans molting is a powerful system for understanding developmentally programmed aECM regeneration. In this study, we determine when and where NHR-23 acts to promote molting. NHR-23 oscillates and its depletion causes severe developmental Fig. 7. Depletion of NHR-23 leads to a defective cuticle barrier. (A) Permeability to Hoechst 33258 dye. Synchronized animals of the indicated genotype were shifted onto control or auxin plates at 25 h post-release and grown for 23 h. Animals were washed off plates and incubated with the cuticle impermeable/membrane permeable Hoechst 33258 dye and scored for nuclear staining in the body. The data are from two biological replicates. P<0.000001 (two-tailed Student’s t-test). (B) Representative images of GFP expression in animals of the indicated genotype after an early and late shift to control or auxin plates. The images are representative of 50 animals observed during each of two experimental replicates. (C) nhr-23::AID; TIR1 and TIR1 animals were subjected to hypo-osmotic shock and scored every 5 min. Ten animals were assayed for each genotype and condition, with three biological replicates. Data are averages of the three biological replicates±s.d. 10 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Fig. 8. nhr-23 is necessary in seam cells for developmental progression, molting and BLI-1 localization. (A) Tissue-specific RNAi. A timed egg lay of animals of the indicated strains was performed on control or nhr-23 RNAi plates and plates were scored 3 days later. The promoters used to reconstitute RNAi, their tissue specificity and oscillation phase (when applicable) are provided. The number of progeny scored from two biological replicates is indicated. Developmental delay was scored as a failure to reach adulthood after 72 h of growth. Molting defects included animals trapped in cuticles, animals dragging cuticles and animals with cuticle corsets. (B) L2 bli-1::mNG (RNAi competent) and bli-1::mNG; SCMp::rde-1; rde-1(ne300) (SCMp-specific RNAi) animals were shifted onto control or nhr-23 RNAi plates at L2 and grown until stage L4.8. Equal exposure times were used for all images. Scale bars: 10 µm. Images are representative of 20 animals over two independent experimental replicates. delay. nhr-23-regulated genes are enriched for cuticle components and protease inhibitors, and NHR-23 is enriched at the transcription start and end sites of target genes. NHR-23 depletion causes aberrant localization of the early collagen ROL-6 and the pre-cuticle component NOAH-1. These cuticle defects are correlated with a loss of the cuticle permeability barrier function. Loss of NHR-23 function in L4 stages also causes severely reduced levels of the adult collagen BLI-1, which is also mislocalized. These cuticle defects are correlated with a loss of the cuticle barrier. Tissue-specific RNAi suggests that nhr-23 is necessary in seam and hypodermal cells. NHR-23 is necessary for cuticle structure and function NHR-23 binds more robustly in the promoter and transcription end site of genes that peak in expression closer to the nhr-23 mRNA peak in expression (Fig. 3). One intriguing possibility is that the earlier NHR-23-regulated genes may be more sensitive to NHR-23 levels. Consistent with this model, there appear to be more NHR-23 peaks flanking and within early nhr-23-regulated genes (Fig. 3, Tables S2, S5). This model would align with E. coli amino acid biosynthesis gene regulation where enzymes earlier in the pathway have more responsive promoters with higher activity (Zaslaver et al., 2004). Interestingly, the non-oscillating nhr-23-regulated genes had a range in number of NHR-23 peaks. It is unclear why NHR-23 oscillation is failing to drive oscillation of these target genes. Some possible explanations include more stable mRNA transcripts in comparison with the transcripts of the oscillating genes or non-oscillating combinatorial transcription factor. regulation involving gene a NHR-23 regulates the expression of nas-37, a protease implicated in molting (Fig. 4; Frand et al., 2005). Curiously, two other targets (noah-1 and rol-6) seemed to be expressed at near wild-type levels but displayed aberrant localization (Figs 5 and 6). These data highlight a potential limitation of a single timepoint gene expression study. A shift in phase of an oscillating gene without change in amplitude could create the appearance of up- or downregulation, depending on the timepoint sampled (Tsiairis and Großhans, 2021). 11 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 The nhr-23(RNAi) microarray is a useful starting point to understand how NHR-23 promotes molting, but an RNA-seq timecourse on NHR-23 depleted animals may be necessary to identify regulated genes. Such an approach was necessary to determine promotes developmental timing (Hauser et al., 2022 preprint). factor, BLMP-1, how the pioneer NHR-23 depletion caused aberrant localization of NOAH-1 and ROL-6, lower levels of BLI-1 and a severe barrier defect (Figs 5-7). There are numerous nhr-23-regulated genes that are implicated in the epithelial barrier. The cuticle furrow formed by six nhr-23- regulated collagens (dpy-2, dpy-3, dpy-7, dpy-8, dpy-9 and dpy-10) is thought to be monitored by a sensor that coordinates several stress responses (Dodd et al., 2018). An RNAi screen of 91 collagens found that inactivation of only these six collagens caused a barrier defect (Sandhu et al., 2021). Inactivation of a subset of the furrow collagens causes permeability to Hoechst 33458 dye and elevated nlp-29p::GFP reporter activity (Dodd et al., 2018), consistent with our NHR-23-depletion data (Fig. 7). bus-8 is a predicted glycosyltransferase that plays a role in the epithelial barrier and is also regulated by nhr-23; its peak expression follows that of the furrow collagens (Table S2). Screening nhr-23-regulated genes may reveal other genes implicated in the epithelial barrier and the peak phase could provide insight into how this barrier is constructed. NHR-23-depletion causes developmental delay and failed molting What is driving NHR-23-depleted animals to eventually attempt to molt? One model is that NHR-23 depletion is incomplete and the remaining NHR-23 is sufficient to drive low levels of target gene expression. Consistent with this idea, NHR-23 depletion or knockdown results in developmental delay but not an arrest (Fig. 2; MacNeil et al., 2013; Patel et al., 2022). Low sustained NHR-23-levels could eventually allow the accumulation of factors that initiate apolysis but with insufficient expression for execution. Consistent with this idea, an NHR-23-regulated promoter reporter (nas-37p::GFP::PEST) peaked at a similar time in control and NHR-23-depleted animals (Fig. 4), but was expressed at lower levels. nas-37 is necessary for C. elegans ecdysis and recombinant NAS-37 promotes the formation of retractile rings, a structure associated with molting, in the parasitic nematode H. contortus (Davis et al., 2004). Alternatively, other factors could eventually promote molting. Molting can be uncoupled from the stage-specific developmental events controlled by the heterochronic pathway, which leads to death when animals attempt molting before completion of cell division and differentiation (Ruaud and Bessereau, 2006). Molting is thought to be controlled by a recently discovered oscillator, although the mechanism remains to be fully elucidated (Tsiairis and Großhans, 2021). It is possible that other candidate components of this oscillator, such as BLMP-1, GRH-1, NHR-25, MYRF-1 or BED-1, could eventually drive animals to molt in the absence of NHR-23 (Hauser et al., 2022 preprint; Meeuse et al., 2023; Stec et al., 2021; Stojanovski et al., 2022). nhr-23 is necessary in seam and hypodermal cells to promote molting Seam cell-enriched nhr-23 RNAi caused severe developmental delay and molting defects (Fig. 8). Hypodermal-enriched nhr-23 knockdown produced less penetrant phenotypes (Fig. 8). The most severe defects in ROL-6 and NOAH-1 localization occur above the seam cells (Figs 5 and 6). Determining how NHR-23 regulates BLI- 1 expression and formation of struts will provide insight into aECM assembly. nhr-23 is expressed in other epidermal cells that produce the pre-cuticle and cuticle, such as vulval precursor cells, rectal cells and excretory duct cells. NHR-23 depletion did not produce obvious defects in localization of the pre-cuticle component NOAH-1 in these cells (Fig. 5). We did not see phenotypes such as excretory duct or pore lumen dilation, which would be indicative of excretory system defects (Gill et al., 2016). We observed vulval morphology defects after NHR-23 depletion, but it is not clear whether that is due to NHR-23 regulation of this specialized aECM or whether it is the consequence of developmental delay and failure to molt. Tissue-specific RNAi or protein depletion will be required to determine whether NHR-23 activity is necessary in these epithelial cells or whether it predominantly functions in seam and hypodermal cells. Our data indicate the importance of tissue- specific RNAi strain validation. SCMp drove reporter expression robustly in seam cells (Fig. S8). However, when we made a tissue- specific RNAi strain using SCMp, we observed depletion in seam, hypodermal and intestinal cells (Fig. S9). RNAi has an inherent amplification of dsRNA triggers by RNA-dependent RNA polymerases (Zhang and Ruvkun, 2012). Weak promoter activity in undesired tissues could reconstitute RNAi in these tissues. A systematic analysis of tissue-specific RNAi strains is an important future direction for interpreting studies using these strains. Future perspectives This work highlights the power of timed protein depletion for dissecting the role of oscillating developmental regulators in development. As transcription factors coordinate the expression of batteries of genes in a given biological process, future work will reveal how NHR-23 coordinates apical ECM remodeling, apolysis and ecdysis. MATERIALS AND METHODS C. elegans strains and culture C. elegans strains (see Table 1) were cultured as originally described (Brenner, 1974), except worms were grown on MYOB instead of NGM. MYOB was made as previously described (Church et al., 1995). Animals were cultured at 20°C for all assays, unless otherwise indicated. For general strain propagation, animals were grown at 15°C according to standard protocols. Brood sizes were controlled by picking L4 larvae to individual wells of a six-well plate seeded with OP50 and incubating the plate at 20°C. Animals were transferred to new plates daily over 4 days. Two days post- transfer, the number of hatched progeny and unfertilized eggs were scored. Genome editing and transgenesis mNeonGreen::3xFLAG knock-ins into bli-1, noah-1 and rol-6, and mNeonGreen knock-ins into his-72 were generated by injection of Cas9 ribonucleoprotein complexes [700 ng/µl Alt-R S.p. Cas9 Nuclease V3 (IDT), 115 ng/µl crRNA and 250 ng/µl IDT tracrRNA] and a dsDNA repair template (25-50 ng/μl) created by PCR amplification of a plasmid template (Paix et al., 2014, 2015). The bli-1 and noah-1 knock-ins were internal and the mNG::3xFLAG cassette was flanked by flexible glycine- and serine-rich linker sequences. Generation of the BLI-1::3xFLAG::mNG(int) used in Fig. 6C will be described elsewhere (J. Adams and A.D.C., unpublished). The C-terminal rol-6::mNG::3xFLAG fusion encoded a flexible linker between the end of rol-6 and the start of mNG. The noah-1::mNG::3xFLAG(int) knock-in used a pJW2332 repair template, which will be described elsewhere (J.M.R. and J.D.W., unpublished). The PCR products were melted to boost editing efficiency, as previously described (Ghanta and Mello, 2020). crRNAs used are provided in Table S6. Oligonucleotides used for repair template generation from template pJW2172 (Ashley et al., 2021) and for genotyping are provided in Table S7. Plasmids used are provided in Table S8. To generate JDW371 and JDW510 (seam cell and hypodermal-specific tissue-specific RNAi), we crossed a jsTi1493 landing pad for recombination-mediated cassette exchange T N E M P O L E V E D 12 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Table 1. Strains used in this study Strain Construction Genotype CZ27591 Microinjection JDW29 JDW180 JDW258 Ragle et al. (2020) Microinjection Ashley et al. (2021) bli-1( ju1789[bli-1::3xFLAG::mNG(int)]) II nhr-23(wrd8[nhr-23::GFP^AID::3xFLAG]) I wrdEx12[SCMp::NLS-mScarlet (dpi)-tbb-2 3’UTR+pCFJ90 (myo-2p::mCherry)] wrdSi22 [eft-3p::TIR1::F2A::mTagBFP2::AID::NLS::tbb-2 3’UTR+SEC, I:-5.32] , nhr-23(kry61(nhr-23::AID-TEV- JDW259 Ashley et al. (2021) JDW322 JDW348 JDW349 Microinjection Crossing Crossing JDW330 JDW354 Crossing Crossing JDW355 Crossing JDW356 Crossing JDW357 Crossing JDW370 Microinjection JDW371 SEC excision of JDW370 JDW376 Clancy et al. (2023) 3xFLAG)) I wrdSi55[eft-3p::TIR1::F2A::mTagBFP2::AID::NLS::tbb-2 3’UTR, I:-5.32] , nhr-23(kry61(nhr-23::AID-TEV- 3xFLAG)) I wrdEx29[wrt-2p::NLS-mScarlet (dpi)-tbb-2 3’UTR] ieSi57 [Peft-3::TIR1::mRuby::unc-54 3′UTR, cb-unc-119(+)] II; frIs7 [nlp-29p::GFP+col-12p::DsRed] IV nhr-23(kry61(nhr-23::AID-TEV-3xFLAG))I; ieSi57 [Peft-3::TIR1::mRuby::unc-54 3′UTR, cb-unc-119(+)] II; frIs7 [nlp- 29p::GFP+col-12p::DsRed] IV.x jsTi1493 [mosL loxP mex-5p FLP sl2 mNeonGreen rpl-28p FRT GFP-HIS-58 FRT3 mosR] IV; rde-1 (ne300) V wrdSi60[eft-3p::TIR1::F2A::mTagBFP2::AID::NLS::tbb-2:tbb-2 3’UTR, I:-5.32] I; ayIs6 [hlh-8::GFP fusion+dpy- 20(+)] X wrdSi61[eft-3p::TIR1::F2A::mTagBFP2::AID::NLS::tbb-2:tbb-2 3’UTR, I:-5.32] , nhr-23(kry61(nhr-23::AID-TEV- 3xFLAG)) I; ayIs6 [hlh-8::GFP fusion+dpy-20(+)] X wrdSi59[eft-3p::TIR1::F2A::mTagBFP2::AID::NLS::tbb-2 3’UTR, I:-5.32], daf-16(mu86) I; ayIs6 [hlh-8::GFP fusion+dpy-20(+)] X wrdSi59[eft-3p::TIR1::F2A::mTagBFP2::AID::NLS::tbb-2 3’UTR, I:-5.32], nhr-23(kry61(nhr-23::AID-TEV- 3xFLAG)), daf-16(mu86) I; ayIs6 [hlh-8::GFP fusion+dpy-20(+)] X jsTi1493 {mosL::loxP::mex-5p::FLP::sl2::mNeonGreen::rpl-28p::FRT [wrdSi64(SEC::loxP::SCMp::rde-1 CDS+3’UTR)] FRT3::mosR} IV ; rde-1 (ne300) V jsTi1493 {mosL loxP [wrdSi65(SCMp::rde-1 CDS+3’UTR)] FRT3::mosR} IV ; rde-1 (ne300) V jsTi1493 {mosL::loxP::mex-5p::FLP::sl2::mNeonGreen::rpl-28p::FRT [wrdSi67(SEC::loxP::mlt-11p (−2.8 kb)::nuc- 1::mCherry-tbb-2 3’UTR)] FRT3::mosR} IV JDW389 JDW395 Microinjection Crossing bli-1(wrd84[bli-1::mNG::3xFLAG(int)]) II wrdSi73[eft-3p::TIR1::F2A::mTagBFP2::AID::NLS::tbb-2 3’UTR] , nhr-23(kry61(nhr-23::AID-TEV-3xFLAG)) I; JDW459 JDW460 JDW463 Microinjection Microinjection Microinjection JDW472 Microinjection JDW475 Microinjection JDW509 Microinjection JDW510 SEC excision of JDW525 JDW509 Microinjection JDW537 Cross of JDW376xJDW460 JDW538 SEC excision of KRY87 KRY88 NM5179 JDW537 Zhang et al. (2015) Zhang et al. (2015) Nonet (2020) oxIs134[Pnas-37::GFP:rODC(pest)(pWD95@90 ng/µl),lin-15(+)] nhr-23(kry61(nhr-23::AID-TEV-3xFLAG))I; ieSi57 [Peft-3::TIR1::mRuby::unc-54 3′UTR, cb-unc-119(+)] II noah-1(wrd119[noah-1::internal mNeonGreen (dpi)::3xFLAG::linker]) I nhr-23(kry61(nhr-23::AID-TEV-3xFLAG)) I ; ieSi57 [Peft-3::TIR1::mRuby::unc-54 3′UTR, cb-unc-119(+)], bli- 1(wrd122[bli-1::internal linker::mNeonGreen (dpi)::3xFLAG::linker]) II rol-6(wrd123[rol-6::C-term mNeonGreen (dpi)::3xFLAG::linker]) I ; jsTi1493 wrdSi65[SCMmin::pes-10delta::rde-1 CDS+3’UTR in pLF3FShC] IV; rde-1(ne300) V noah-1(wrd126[noah-1::internal linker::mNeonGreen (dpi)::3xFLAG::linker]), nhr-23(kry61(nhr-23::AID-TEV- 3xFLAG)) I ; ieSi57 [Peft-3::TIR1::mRuby::unc-54 3′UTR, cb-unc-119(+)] II jsTi1493 {mosL::loxP::mex-5p::FLP::sl2::mNeonGreen::rpl-28p::FRT [wrdSi96(SEC::loxP::semo-1 (Y37A1B.5)p:: rde-1 CDS+3’UTR)] FRT3::mosR} IV ; rde-1 (ne300) V jsTi1493 {mosL loxP [wrdSi97(semo-1p::rde-1 CDS+3’UTR)] FRT3::mosR} IV ; rde-1 (ne300) V his-72(wrd142[mNeonGreen::his-72]) III ; jsTi1493 {mosL loxP [wrdSi65(SCMp::rde-1 CDS+3’UTR)] FRT3::mosR} IV ; rde-1 (ne300) V noah-1(wrd119[noah-1::internal mNeonGreen (dpi)::3xFLAG::linker]) I; jsTi1493 wrdSi67{mosL::loxP::mex-5p:: FLP::sl2::mNeonGreen::rpl-28p::FRT [wrdSi67(SEC::loxP::mlt-11p (−2.8 kb)::nuc-1::mCherry-tbb-2 3’UTR)] FRT3::mosR} IV noah-1(wrd119[noah-1::internal mNeonGreen (dpi)::3xFLAG::linker]) I; jsTi1493 wrdSi104{mosL::loxP::mlt-11p (−2.8 kb)::nuc-1::mCherry-tbb-2 3’UTR)] FRT3::mosR} IV nhr-23(kry61(nhr-23::AID-TEV-3xFLAG)) I nhr-23(kry61(nhr-23::AID-TEV-3xFLAG)) I; ieSi57 [Peft-3::TIR1::mRuby::unc-54 3′UTR, cb-unc-119(+)] II jsTi1493 [mosL loxP mex-5p FLP sl2 mNeonGreen rpl-28p FRT GFP-HIS-58 FRT3 mosR] IV Strains provided by the Caenorhabditis Genetics Center Strain Genotype N2 CA1200 CF1038 PD4666 QK52 VP303 JU2039 WM118 MGH171 NM5179 WM45 IG274 Wild type ieSi57 [Peft-3::TIR1::mRuby::unc-54 3′UTR, cb-unc-119 (+)] II (Zhang et al., 2015) daf-16(mu86) I (Lin et al., 1997) ayIs6 [hlh-8::GFP fusion+dpy-20(+)] X (Harfe et al., 1998) rde-1(ne219) V; xkIs99(wrt-2p::rde-1::unc-54 3’UTR) (Melo and Ruvkun, 2012) rde-1(ne219) V; kbIs7[nhx-2p::rde-1+rol-6(su1006)]) (Espelt et al., 2005) mfIs70 [lin-31p::rde-1+myo2p::GFP] IV ; rde-1(ne219) V (Barkoulas et al., 2013) rde-1(ne300) V ; neIs9(neIs9 [myo-3::HA::RDE-1+rol-6(su1006)]) X (Watts et al., 2020) alxIs9 [vha-6p::sid-1::SL2::GFP] sid-1(qt9) V (Melo and Ruvkun, 2012) jsTi1493 [mosL loxP mex-5p FLP sl2 mNeonGreen rpl-28p FRT GFP-HIS-58 FRT3 mosR] IV (Nonet, 2020) rde-1(ne300) V (Tabara et al., 1999) frIs7 [nlp-29p::GFP+col-12p::DsRed] IV (Pujol et al., 2008) Other strains Strain Genotype EG3200 oxIs134[Pnas-37::GFP:rODC (PEST) (pWD95@90 ng/µl),lin-15(+)]; lin-15(n765ts) X (Davis et al., 2004; provided by D. Fay, University of Wyoming, Laramie, WY, USA) T N E M P O L E V E D 13 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 (RMCE) (Nonet, 2020) into an rde-1(ne300) null mutant, generating JDW330. We used Gibson cloning to introduce a promoterless rde-1 genomic coding sequence+3′UTR fragment into the RMCE integration vector pLF3FShC (Nonet, 2020), creating pJW2247. This vector can be linearized with AvrII+BsiWI double-digestion and promoters can be introduced by Gibson cloning (Gibson et al., 2009). We generated a SCMp (seam cell-specific) and semo-1p (hypodermis specific) by Gibson cloning of promoters into linearized pJW2247, constructing pJW2236 and pJW2264, respectively. These vectors were integrated into jsTi1493 landing pad in JDW330 using RMCE, as described previously (Nonet, 2020). wrt-2p and SCMp promoter reporters were constructed in pJW1836 (NLS::mScarlet::tbb-2 3′ UTR) (Ashley et al., 2021) by Gibson cloning. Plasmids were injected into N2 animals at 50 ng/µl with a pCFJ90 co-injection marker at 10 ng/µl for the SCMp promoter reporter and a pRF4 co-injection marker at 50 ng/µl for the wrt-2 promoter reporter (Frøkjaer- Jensen et al., 2008; Mello et al., 1991). Transgenic lines propagating extrachromosomal arrays were generated as previously described (Mello et al., 1991). Genomic and knock-in sequences are provided as GenBank-compatible .ape files on figshare.com (File S1, https://doi.org/10.6084/m9.figshare. 22270525.v1). Sequence files for plasmids are provided as Genbank- compatible .ape files on figshare.com (File S2, https://doi.org/10.6084/m9. figshare.22270522.v1). Auxin treatment Control and auxin media/plates were made as described by Ragle et al. (2020). Control media consisted of MYOB agar+0.25% ethanol. Auxin media was made by dissolving indole 3-acetic acid (IAA; Alfa Aesar, AAA1055622) in 100% ethanol to 1.6 M and then mixing it into melted MYOB agar at 55°C to a final concentration of 4 mM before pouring plates. Control plates contained 0.25% ethanol. Temperature of the media was monitored with a Lasergrip 1080 infrared thermometer gun (Etekcity). Plates were seeded with E. coli OP50 and incubated overnight at room temperature. Plates were stored for up to 1 month at 4°C before use. For most auxin treatment experiments, animals were synchronized by alkaline bleaching (www.protocols.io/view/ward-lab-alkaline-bleaching-protocol- cbq9smz61). The incubated in M9 buffer supplemented with 5 mg/ml cholesterol at 20°C for 24 h and arrested L1 larvae were released onto the indicated type of MYOB plate. For the experiment in Table S1, a timed egg lay was performed. Twenty adult hermaphrodites animals of the indicated genotype were picked onto control (0.25% ethanol) or auxin (4 mM IAA) plates and allowed to lay eggs for 2 h. The adult animals were removed and plates were incubated for 48 h at 25°C. collected eggs were Microscopy Synchronized animals were collected from MYOB, control or auxin plates by either picking or washing off plates. For washing, 1000 µl of M9 and 0.05% gelatin was added to the plate or well, agitated to suspend animals in M9 and gelatin, and then transferred to a 1.5 ml tube. Animals were spun at 700 g for 1 min. The medium was then aspirated off and animals were resuspended in 500 µl M9 and 0.05% gelatin with 5 mM levamisole. 12 µl of animals in M9 and 0.05% levamisole solution were placed on slides with a 2% agarose pad and secured with a coverslip. For picking, animals were transferred to a 10 µl drop of M9+5 mM levamisole on a 2% agarose pad on a slide and secured with a coverslip. Images were acquired using a Plan- Apochromat 40×/1.3 Oil DIC lens, a Plan-Apochromat 63×/1.4 Oil DIC lens or a Plan-Apochromat 100×/1.4 OIL DIC lens on an AxioImager M2 microscope (Carl Zeiss Microscopy) equipped with a Colibri 7 LED light source and an Axiocam 506 mono camera. Acquired images were processed using Fiji software (version: 2.0.0- rc-69/1.52p) (Schindelin et al., 2012). For direct comparisons within a figure, we set the exposure conditions to avoid pixel saturation of the brightest sample and kept equivalent exposure for imaging of the other samples. Colocalization using a Manders’ co-efficient (Manders et al., 1992, 1993) was performed as described previously (Clancy et al., 2023). Phenotypic analysis For the phenotypic analysis in Fig. 2, synchronized CA1200 or KRY88 larvae were released onto MYOB plates and then shifted onto control or auxin plates every 2 h up to 16 h. Animals were collected as described in the Microscopy section and imaged by DIC microscopy to score for morphology and shedding of the L1 cuticles. For the M cell lineage experiments in Fig. S3, synchronized animals of the indicated genotypes were released onto MYOB plates and shifted onto control or auxin plates at 3 h (early shift) or 9 h (late shift) post-release and then imaged at 24 h post- release, as described in the Microscopy section. M cells were counted and recorded. For the analyses in Fig. 2E, synchronized nhr-23::AID, TIR1 larvae were released on auxin plates and scored for viability and molting defects 72 h later. For the L3 shift experiments (Fig. 2F,G) synchronized animals of the indicated genotype were grown on MYOB. At 25 h post- release, they were shifted onto six-well control or auxin MYOB plates seeded with OP50. Animals were collected 23 or 72 h later, as described in the Microscopy section and imaged by DIC microscopy using a 63× lens to stage animals according to vulval morphology (Mok et al., 2015). For the reporter timecourse, synchronized nhr-23::AID, TIR1, nas-37p::GFP:: PEST animals were released on control or auxin plates. They were then scored for GFP expression using a PlanApo 5.0×/0.5 objective on a M165 FC stereomicroscope (Leica) equipped with an X-cite FIRE LED lightsource (Excelitas) and long-pass GFP filter set (Leica, 10447407). We scored GFP expression in the head and hypodermis and did not score rectal GFP expression, as expression perdured in this tissue after head and hypodermal GFP expression ceased. Barrier assays Hoechst 33258 staining was performed as described previously (Moribe et al., 2004), except that we used 10 µg/ml of Hoechst 33258 as in Ward et al. (2014). Two biological replicates were performed examining 50 animals per experiment. The number of animals with either head or hypodermal nuclei staining with Hoechst 33258 were scored under a 10× DIC objective. Representative images were then taken with equivalent exposures using a 63× Oil DIC lens, as described in the imaging section. Hypo-osmotic stress sensitivity assays were performed on L4 stage larvae as described previously (Ward et al., 2014), except we used 20 µl of deionized H20. Each strain was assayed in triplicate. Western blots Animals were synchronized by alkaline bleaching, as described in the Auxin treatment section. For the blots in Figs 5B and 6B, 30 animals of the indicated genotype and stage were picked into 30 µl of M9+0.05% gelatin and 10 µl of 4×Laemmli sample buffer was added and then samples were heated at 95°C for 5 min and stored at −80°C until they were resolved by SDS-PAGE. For the blots in Fig. 6C, synchronized animals of the indicated genotype were collected at the L4 stage. Extracts of soluble cuticle proteins were prepared as described previously (Cox et al., 1981) with minor modifications. Briefly, worms were collected using M9J buffer in a 15 ml Falcon tube and washed three times with M9J buffer. Worm pellets were resuspended in 5 ml of sonication buffer and incubated on ice for 10 min and sonicated (Misonix XL-2000 Series Ultrasonic Liquid Processor, 10×10 s pulse with 10 s break) with 30 μl of 0.1 M PMSF. After sonication, samples were centrifuged at 6010 g for 2 min at 4°C and supernatants stored as fraction 1 (F1; intracellular proteins) at −20°C for the purpose of using it as loading control. Pellets were washed three times with sonication buffer and the resultant pellets were resuspended in 100 μl sonication buffer, boiled at 95°C for 2 min with 1 ml of ST buffer and incubated overnight with rotation at room temperature. After incubation, samples were centrifuged at 6010 g for 3 min and supernatants were stored as fraction 2 (F2) at −20°C. Subsequently, cuticle pellets were washed three times with 0.5% Triton X-100 and boiled at 95°C for 2 min with 300 μl of ST buffer with 5% β-mercaptoethanol and incubated with rotation at room temperature for 16 h. Finally, the samples were centrifuged at 6010 g for 3 min and supernatants collected and stored as fraction 3 (F3; soluble cuticle proteins). F3 was precipitated with methanol/ chloroform and resuspended in 100 μl of rehydration buffer. For the remaining blots, animals of the indicated genotype were washed out of wells on a six-well plate at the indicated timepoints with M9+0.05% gelatin (VWR, 97062-620), transferred to a 1.5 ml tube and washed twice more with M9+0.05% gelatin, as previously described (Vo et al., 2021). Animals were pelleted, transferred into a 100 µl volume to a new 1.5 ml tube, and flash frozen in liquid nitrogen, then stored at −80°C. Before 14 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 SDS-PAGE, samples were freeze-cracked twice in liquid nitrogen or on dry ice, Laemmli sample buffer was added to 1× and samples were heated to 95°C for 5 min. For the western blots in Fig. 1C,D, 6000 KRY88 animals per well were transferred onto a six-well MYOB plate seeded with OP50. For the western blot in Fig. 2B, 6000 synchronized KRY88 animals were plated onto control or auxin plates. Control animals were harvested at peak NHR- 23 expression (12 h of growth on control plates) and auxin-treated animals were shifted onto auxin at this time; samples were taken at the indicated time points, as described above. For the western blots in Fig. S7, 500 animals of the indicated genotype were transferred into wells of a six-well control or auxin plate seeded with OP50. Animals were incubated for 24 h at 20°C and collected as described above. For the blots in Fig. 1C,D, 10 µl of lysate and 7.5 µl of a 1:1 mix of Amersham ECL Rainbow Molecular Weight Markers (95040-114) and Precision Plus Protein Unstained Standards (1610363) was resolved by SDS-PAGE using precast 4-20% MiniProtean TGX Stain Free Gels (Bio- Rad). The blot in Fig. S7, the lysates had much more variable levels of total protein, particularly the NHR-23-depleted samples that produced an early larval arrest. Therefore, we quantified the signal of the most intense band by Stain Free imaging (Posch et al., 2013) using a Bio-Rad ChemiDoc imaging system, and ran a new gel with normalized loading. The lysates in Figs 2A, 5B, 6B and S7 were resolved as described above except we used a Spectra Multicolor Broad Range Protein Ladder (ThermoFisher Scientific, 26623). Proteins were transferred to a polyvinylidene difluoride membrane by semi- dry transfer with a TransBlot Turbo (Bio-Rad). In Figs 1C and 7C, total protein, pre- and post-transfer, was monitored using the stain-free fluorophore as described (Posch et al., 2013; Ward, 2015a). Membranes were washed in TBST and blocked in TBST+5% milk (TBST-M; Nestle Carnation Instant Nonfat Dry Milk) for 1 h at room temperature. Blots were rocked in primary antibodies in TBST-M overnight at 4°C and then washed four times for 5 min each with TBST. For primary antibodies conjugated with horseradish peroxidase (HRP), the blots were developed after the last TBST wash. Otherwise, blots were incubated with HRP-conjugated secondary antibodies in TBST-M for 1 h at room temperature followed by four 5 min TBST washes and then developed as described below. The primary antibody used for Figs 1C,D, 2A, 5B and 6B was horseradish peroxidase (HRP)-conjugated anti-FLAG M2 (Sigma-Aldrich, A8592- 5x1MG, Lot #SLCB9703) at a 1:2000 dilution. Precision Protein StrepTactin-HRP Conjugate (Bio-Rad, 1610381, Lot #64426657) was included with the primary antibody at a 1:10,000 dilution to visualize the protein size standard during blot imaging for Figs 1C,D. For the blot in Figs 2A, 5B and 6B, we used mouse anti-alpha-tubulin 12G10 concentrated [Developmental Studies Hybridoma Bank, supernatant at 1:2000 (Fig. 2A, Fig. S7) or 1:4000 (Figs 5B, 6B)]. The secondary antibodies were Digital anti-mouse (Kindle Biosciences LLC, R1005) diluted 1:10,000 (Fig. 2A, Fig. S7) or 1:20,000 (Fig. S6B,C). Blots were incubated for 5 min with 1 ml of Supersignal West Femto Maximum Sensitivity Substrate (ThermoFisher Scientific, 34095) and imaged using the ‘chemi high-resolution’ setting on a Bio-Rad ChemiDoc MP System. For the blots in Fig. 6C, 20 μl was loaded on to NuPAGE 4 to 12% protein gels with 7 μl of 4× NuPAGE LDS sample buffer as described by the manufacturer (Invitrogen, NP0322BOX) and run for 2 h at 100 V, transferred to 0.45 μm nitrocellulose membrane (Bio-Rad) at 4°C for 90 min using Bio- Rad WET transfer system. After transfer, membranes were blocked in Intercept blocking buffer (LI-COR: 927-60001) for 1 h. Later, the membrane was incubated overnight in anti-mNeonGreen primary antibody (ChromoTek, 32F6) at 1:5000 dilution at 4°C on rocker. Membranes were washed three times with TBST and incubated with secondary antibody (Sigma 12-349) at 1:100,000 dilution at room temperature on rocker for 1 h and washed three times with TBST. Blots were visualized with Supersignal West Femto detection kit (Thermo Fisher) using a LI-COR Odyssey imager. Similarly, to test the loading control, 10 μl of F1 samples were loaded and, after transfer to nitrocellulose membrane, probed with monoclonal mouse anti-actin clone C4 (ThermoFisher Scientific: ICN691001). ‘-c’ the x=amplitude×cos( phase) and y=amplitude×sin( phase). For Figs 3D and S5, L3 NHR-23 ChIP-Seq data was downloaded from http://www.modencode. org (accession ID modEncode_3837). Wig files were converted to bigwig format and loaded into the Bioconductor package SeqPlots (Huber et al., 2015; Stempor and Ahringer, 2016). A bed file with the start and end coordinates of the CDS of all genes was generated (Wormbase WS220) and loaded into SeqPlots. SeqPlots aligned all genes by the start and end position, and scaled the coding sequence of each gene to 2 kb. The average signal over all aligned genes was calculated in 25 bp windows from 1 kb upstream of the start site to 1 kb downstream of the end site. Peak calls for NHR-23 in L3 (Fig. 4C) were obtained from GEO (accession numbers: GSM1183659, GSM1183660, GSM1183661 and GSM118366). Only peak calls validated by both replicates were considered. Peaks were assigned to the following genome features: gene body (WormBase WS220); 1 kb upstream of the start site of coding genes; between 1 kb and 3 kb upstream of the gene start site; and between 3 kb and 5 kb upstream of the gene start site. The same intervals were chosen downstream of the gene end sites. An NHR-23 peak was assigned to a feature if it overlapped with it by at least 100 bp. The peak analysis data are in Table S5. The total number of peaks in all bins was tallied and is presented in Table S2 and in Fig. 3D. For Figs 3D and 4E, the phase for the oscillating genes in Table S1 was converted to time during the molting cycle by assuming a 9 h larval stage, which makes each hour=40°. We used the phasing from Meeuse et al. (2023) with lethargus starting at 45° and ecdysis ending at 135°. The phase in hours was then plotted along the x-axis and the amplitude of the oscillating genes was plotted on the y-axis. The gene annotation was based on the Concise Description and Automated Description attributes downloaded from Wormbase. RNAi RNAi feeding plates were made by melting a solidified bottle of MYOB and adding carbenicillin (25 μg/ml) and IPTG (8 mM) once cooled to 55°C. dsRNA-expressing E. coli bacteria were streaked on LB plates with ampicillin (100 µg/ml) and grown overnight at 37°C. A single colony was picked into 25 ml LB with 100 µg/ml ampicillin and 12.5 µg/ml tetracycline and shaken overnight at 37°C at 220 RPM. The next day, the liquid culture was pelleted and resuspended in 1.25 ml LB with 100 µg/ml ampicillin, resulting in a 20× concentration from the original overnight liquid culture. 90 µl of the liquid culture was spread per small RNAi plate and allowed to dry. The plates were incubated at room temperature in the dark for 3 days. Ten adult worms from each strain were allowed to lay eggs on plates spread with E. coli bacteria containing either the control plasmid L4440 or the nhr- 23 RNAi knockdown plasmid for 1-2 h. The adults were removed and the eggs allowed to develop at 20°C for 2 or 3 days, depending on the experiment. For tissue-specific RNAi experiments, we used RNAi defective rde-1 mutant animals carrying rde-1 transgenes to rescue RNAi in specific tissues (Qadota et al., 2007) The one exception was strain MGH171, which used an intestinal-specific rescue of sid-1 in an RNAi defective sid-1 mutant (Melo and Ruvkun, 2012). Statistical analysis Statistical tests and numbers of animals analyzed are detailed in figure legends. Acknowledgements We thank Chris Hammell, David Matus and Julian Ceron Madrigal for their critical reading of the manuscript. We thank David Fay, Miles Pufall, Helge Grosshans and Ali Shariati for helpful discussions. We thank Jennifer Adams for generation of initial rol-6 and bli-1 mNeonGreen knock-ins. Some strains were provided by the Caenorhabditis Genetics Center, which is funded by the NIH Office of Research Infrastructure Programs [P40 OD010440]. L.C.J. was funded by an NIGMS training grant [T32 GM133391]. The anti-alpha tubulin 12G10 monoclonal antibody developed by J. Frankel and E.M. Nelson of the University of Iowa was obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242. Bioinformatics Fig. 3A was generated in R (R Core Team). Phase and amplitude in Table S2 were converted to x and y coordinates for each gene by calculating Competing interests The authors declare no competing or financial interests. 15 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Author contributions Conceptualization: L.C.J., J.M.R., J.D.W.; Methodology: L.C.J., M.P., J.M.R., J.D.W.; Validation: L.C.J., K.M.M., M.P., J.M.R., J.D.W.; Formal analysis: L.C.J., A.A.V., J.C.C., K.M.M., M.P., J.A., M.T.L., A.R., J.M.R., A.D.C., J.D.W.; Investigation: L.C.J., A.A.V., J.C.C., K.M.M., M.P., J.A., M.T.L., C.W., A.R., J.M.R., A.D.C., J.D.W.; Resources: L.C.J., A.R., J.M.R., J.D.W.; Data curation: L.C.J., J.D.W.; Writing - original draft: L.C.J., J.D.W.; Writing - review & editing: A.A.V., J.C.C., K.M.M., M.P., J.A., M.T.L., C.W., J.M.R., A.D.C.; Visualization: L.C.J., M.P., J.A., A.R., J.M.R., J.D.W.; Supervision: J.M.R., A.D.C., J.D.W.; Project administration: J.D.W.; Funding acquisition: A.D.C., J.D.W. Funding This work was funded by the National Institutes of Health (NIH) National Institute of General Medical Sciences (NIGMS) (R00GM107345 and R01GM138701) and the NIH Office of the Director (R21OD033663) to J.D.W. M.P. and A.D.C. were supported by the National Institute of General Medical Sciences (R35GM134970). Open Access funding provided by the University of California. Deposited in PMC for immediate release. Data availability All relevant data can be found within the article and its supplementary information. Peer review history The peer review history is available online at https://journals.biologists.com/dev/ lookup/doi/10.1242/dev.201085.reviewer-comments.pdf References Agbulut, O., Coirault, C., Niederlä nder, N., Huet, A., Vicart, P., Hagège, A., Puceat, M. and Menasché, P. (2006). GFP expression in muscle cells impairs actin-myosin interactions: implications for cell therapy. Nat. Methods 3, 331. doi:10.1038/nmeth0506-331 Antebi, A. (2015). Nuclear receptor signal transduction in C. elegans. WormBook, 1-49. doi:10.1895/wormbook.1.64.2 Ashley, G. E., Duong, T., Levenson, M. T., Martinez, M. A. Q., Johnson, L. C., Hibshman, J. D., Saeger, H. N., Palmisano, N. J., Doonan, R., Martinez- for Mendez, R. et al. (2021). An expanded auxin-inducible degron toolkit Caenorhabditis elegans. Genetics 217, iyab006. doi:10.1093/genetics/iyab006 Aspö ck, G., Kagoshima, H., Niklaus, G. and Bü rglin, T. R. (1999). Caenorhabditis elegans has scores of hedgehog related genes: sequence and expression analysis. Genome Res. 9, 909-923. doi:10.1101/gr.9.10.909 Baens, M., Noels, H., Broeckx, V., Hagens, S., Fevery, S., Billiau, A. D., Vankelecom, H. and Marynen, P. (2006). The dark side of EGFP: defective polyubiquitination. PLoS ONE 1, e54. doi:10.1371/journal.pone.0000054 Barkoulas, M., van Zon, J. S., Milloz, J., van Oudenaarden, A. and Félix, M.-A. (2013). Robustness and Epistasis in the C. elegans vulval signaling network revealed by pathway dosage modulation. Dev. Cell 24, 64-75. doi:10.1016/j. devcel.2012.12.001 Blaxter, M. L. (1993). Cuticle surface proteins of wild type and mutant Caenorhabditis elegans. J. Biol. Chem. 268, 6600-6609. doi:10.1016/S0021- 9258(18)53293-2 Blaxter, M. L., Page, A. P., Rudin, W. and Maizels, R. M. (1992). Nematode surface coats: actively evading immunity. Parasitol. Today 8, 243-247. doi:10.1016/0169- 4758(92)90126-M Brenner, S. (1974). The genetics of Caenorhabditis elegans. Genetics 77, 71-94. doi:10.1093/genetics/77.1.71 Campbell, L. A., Faivre, E. J., Show, M. D., Ingraham, J. G., Flinders, J., Gross, J. D. and Ingraham, H. A. (2008). Decreased recognition of SUMO-sensitive target genes following modification of SF-1 (NR5A1). Mol. Cell. Biol. 28, 7476-7486. doi:10.1128/MCB.00103-08 Church, D. L., Guan, K. L. and Lambie, E. J. (1995). Three genes of the MAP kinase cascade, mek-2, mpk-1/sur-1 and let-60 ras, are required for meiotic cell cycle progression in Caenorhabditis elegans. Development 121, 2525-2535. doi:10.1242/dev.121.8.2525 Clancy, J. C., Vo, A. A., Myles, K. M., Levenson, M. T., Ragle, J. M. and Ward, J. D. (2023). Experimental considerations for study of C. elegans lysosomal proteins. G3 13, jkad032. doi:10.1093/g3journal/jkad032 Clark, D. V., Suleman, D. S., Beckenbach, K. A., Gilchrist, E. J. and Baillie, D. L. the dpy-20 gene of (1995). Molecular cloning and characterization of Caenorhabditis elegans. Mol. Gen. Genet. 247, 367-378. doi:10.1007/ BF00293205 Cohen, J. D. and Sundaram, M. V. (2020). C. elegans apical extracellular matrices shape epithelia. J. Dev. Biol. 8, 23. doi:10.3390/jdb8040023 Cohen, J. D., Sparacio, A. P., Belfi, A. C., Forman-Rubinsky, R., Hall, D. H., Maul-Newby, H., Frand, A. R. and Sundaram, M. V. (2020). A multi-layered and dynamic apical extracellular matrix shapes the vulva lumen in Caenorhabditis elegans. eLife 9, e57874. doi:10.7554/eLife.57874 Cox, G. N., Kusch, M. and Edgar, R. S. (1981). Cuticle of Caenorhabditis elegans: its isolation and partial characterization. J. Cell Biol. 90, 7-17. doi:10.1083/jcb.90. 1.7 Davis, M. W., Birnie, A. J., Chan, A. C., Page, A. P. and Jorgensen, E. M. (2004). A conserved metalloprotease mediates ecdysis in Caenorhabditis elegans. Development 131, 6001-6008. doi:10.1242/dev.01454 de Sousa Abreu, R., Penalva, L. O., Marcotte, E. M. and Vogel, C. (2009). Global signatures of protein and mRNA expression levels. Mol. Biosyst. 5, 1512-1526. doi:10.1039/b908315d Dodd, W., Tang, L., Lone, J.-C., Wimberly, K., Wu, C.-W., Consalvo, C., Wright, J. E., Pujol, N. and Choe, K. P. (2018). A damage sensor associated with the cuticle coordinates three core environmental stress responses in Caenorhabditis elegans. Genetics 208, 1467-1482. doi:10.1534/genetics.118.300827 Espelt, M. V., Estevez, A. Y., Yin, X. and Strange, K. (2005). Oscillatory Ca2+ signaling in the isolated Caenorhabditis elegans intestine: role of the inositol- 1,4,5-trisphosphate receptor and phospholipases C beta and gamma. J. Gen. Physiol. 126, 379-392. doi:10.1085/jgp.200509355 Fernandez, A. P., Gibbons, J. and Okkema, P. G. (2004). C. elegans peb-1 mutants exhibit pleiotropic defects in molting, feeding, and morphology. Dev. Biol. 276, 352-366. doi:10.1016/j.ydbio.2004.08.040 Forman-Rubinsky, R., Cohen, J. D. and Sundaram, M. V. (2017). Lipocalins are required for apical extracellular matrix organization and remodeling in Caenorhabditis elegans. Genetics 207, 625-642. doi:10.1534/genetics.117. 300207 Frand, A. R., Russel, S. and Ruvkun, G. (2005). Functional Genomic Analysis of C. elegans Molting. PLoS Biol. 3, e312. doi:10.1371/journal.pbio.0030312 Frøkjaer-Jensen, C., Davis, M. W., Hopkins, C. E., Newman, B. J., Thummel, J. M., Olesen, S.-P., Grunnet, M. and Jorgensen, E. M. (2008). Single-copy insertion of transgenes in Caenorhabditis elegans. Nat. Genet. 40, 1375-1383. doi:10.1038/ng.248 Gerstein, M. B., Lu, Z. J., Van Nostrand, E. L., Cheng, C., Arshinoff, B. I., Liu, T., Yip, K. Y., Robilotto, R., Rechtsteiner, A., Ikegami, K. et al. (2010). Integrative analysis of the Caenorhabditis elegans genome by the modENCODE Project. Science 330, 1775-1787. doi:10.1126/science.1196914 Ghanta, K. S. and Mello, C. C. (2020). Melting dsDNA donor molecules greatly improves precision genome editing in Caenorhabditis elegans. Genetics 216, 643-650. doi:10.1534/genetics.120.303564 Ghedin, E., Wang, S., Spiro, D., Caler, E., Zhao, Q., Crabtree, J., Allen, J. E., Delcher, A. L., Guiliano, D. B., Miranda-Saavedra, D. et al. (2007). Draft genome of the filarial nematode parasite Brugia malayi. Science 317, 1756-1760. doi:10.1126/science.1145406 Gibson, D. G., Young, L., Chuang, R.-Y., Venter, J. C., Hutchison, C. A. and Smith, H. O. (2009). Enzymatic assembly of DNA molecules up to several hundred kilobases. Nat. Methods 6, 343-345. doi:10.1038/nmeth.1318 Gill, H. K., Cohen, J. D., Ayala-Figueroa, J., Forman-Rubinsky, R., Poggioli, C., Bickard, K., Parry, J. M., Pu, P., Hall, D. H. and Sundaram, M. V. (2016). Integrity of narrow epithelial tubes in the C. elegans excretory system requires a transient luminal matrix. PLoS Genet. 12, e1006205. doi:10.1371/journal.pgen. 1006205 Gissendanner, C. R., Crossgrove, K., Kraus, K. A., Maina, C. V. and Sluder, A. E. receptor genes in (2004). Expression and function of conserved nuclear Caenorhabditis elegans. Dev. Biol. 266, 399-416. doi:10.1016/j.ydbio.2003.10. 014 Guo, P., Hu, T., Zhang, J., Jiang, S. and Wang, X. (2010). Sequential action of Caenorhabditis elegans Rab GTPases regulates phagolysosome formation during apoptotic cell degradation. Proc. Natl. Acad. Sci. USA 107, 18016-18021. doi:10.1073/pnas.1008946107 Gupta, S. K. (2021). Human Zona Pellucida glycoproteins: binding characteristics with human spermatozoa and induction of acrosome reaction. Front. Cell Dev. Biol. 9, 619868. doi:10.3389/fcell.2021.619868 Harfe, B. D., Gomes, A. V., Kenyon, C., Liu, J., Krause, M. and Fire, A. (1998). Analysis of a Caenorhabditis elegans Twist homolog identifies conserved and divergent aspects of mesodermal patterning. Genes Dev. 12, 2623-2635. doi:10. 1101/gad.12.16.2623 Hauser, Y. P., Meeuse, M. W. M., Gaidatzis, D. and Großhans, H. (2022). The BLMP-1 transcription factor promotes oscillatory gene expression to achieve timely molting. bioRxiv 2021.07.05.450828. Hendriks, G.-J., Gaidatzis, D., Aeschimann, F. and Großhans, H. (2014). Extensive oscillatory gene expression during C. elegans larval development. Mol. Cell 53, 380-392. doi:10.1016/j.molcel.2013.12.013 Huber, W., Carey, V. J., Gentleman, R., Anders, S., Carlson, M., Carvalho, B. S., Bravo, H. C., Davis, S., Gatto, L., Girke, T. et al. (2015). Orchestrating high- throughput genomic analysis with Bioconductor. Nat. Methods 12, 115-121. doi:10.1038/nmeth.3252 Jetten, A. M. (2009). Retinoid-related orphan receptors (RORs): critical roles in development, immunity, circadian rhythm, and cellular metabolism. Nucl. Recept. Signal. 7, e003. doi:10.1621/nrs.07003 Johnstone, I. L. and Barry, J. D. (1996). Temporal reiteration of a precise gene expression pattern during nematode development. EMBO J. 15, 3633-3639. doi:10.1002/j.1460-2075.1996.tb00732.x 16 T N E M P O L E V E D RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Kage-Nakadai, E., Kobuna, H., Kimura, M., Gengyo-Ando, K., Inoue, T., Arai, H. and Mitani, S. (2010). Two very long chain fatty acid acyl-CoA synthetase genes, acs-20 and acs-22, have roles in the cuticle surface barrier in Caenorhabditis elegans. PLoS ONE 5, e8857. doi:10.1371/journal.pone.0008857 Kaletsky, R., Yao, V., Williams, A., Runnels, A. M., Tadych, A., Zhou, S., Troyanskaya, O. G. and Murphy, C. T. (2018). Transcriptome analysis of adult Caenorhabditis elegans cells reveals tissue-specific gene and isoform expression. PLoS Genet. 14, e1007559. doi:10.1371/journal.pgen.1007559 Kasuga, H., Fukuyama, M., Kitazawa, A., Kontani, K. and Katada, T. (2013). The microRNA miR-235 couples blast-cell quiescence to the nutritional state. Nature 497, 503-506. doi:10.1038/nature12117 Katz, S. S., Barker, T. J., Maul-Newby, H. M., Sparacio, A. P., Nguyen, K. C. Q., Maybrun, C. L., Belfi, A., Cohen, J. D., Hall, D. H., Sundaram, M. V. et al. (2022). A transient apical extracellular matrix relays cytoskeletal patterns to shape permanent acellular ridges on the surface of adult C. elegans. PLoS Genet. 18, e1010348. doi:10.1371/journal.pgen.1010348 Kelley, M., Yochem, J., Krieg, M., Calixto, A., Heiman, M. G., Kuzmanov, A., Meli, V., Chalfie, M., Goodman, M. B., Shaham, S. et al. (2015). FBN-1, a fibrillin-related protein, is required for resistance of the epidermis to mechanical deformation during C. elegans embryogenesis. eLife 4, 1629. doi:10.7554/eLife. 06565 Kiefer, S. M. L. and Saling, P. (2002). Proteolytic processing of human zona pellucida proteins. Biol. Reprod. 66, 407-414. doi:10.1095/biolreprod66.2.407 Kim, D., Grü n, D. and van Oudenaarden, A. (2013). Dampening of expression oscillations by synchronous regulation of a microRNA and its target. Nat. Genet. 45, 1337-1344. doi:10.1038/ng.2763 King-Jones, K. and Thummel, C. S. (2005). Nuclear receptors-a perspective from Drosophila. Nat. Rev. Genet. 6, 311-323. doi:10.1038/nrg1581 Kinney, B., Sahu, S., Stec, N., Hills-Muckey, K., Adams, D. W., Wang, J., Jaremko, M., Joshua-Tor, L., Keil, W. and Hammell, C. M. (2023). A circadian- like gene network regulates heterochronic miRNA transcription in C. elegans. bioRxiv 2022.09.26.509508. Kö hnlein, K., Urban, N., Guerrero-Gó mez, D., Steinbrenner, H., Urbánek, P., Priebs, J., Koch, P., Kaether, C., Miranda-Vizuete, A. and Klotz, L.-O. (2020). A Caenorhabditis elegans ortholog of human selenium-binding protein 1 is a pro- aging factor protecting against selenite toxicity. Redox Biol. 28, 101323. doi:10. 1016/j.redox.2019.101323 Kostrouchova, M., Krause, M., Kostrouch, Z. and Rall, J. E. (1998). CHR3: a Caenorhabditis elegans orphan nuclear hormone receptor required for proper epidermal development and molting. Development 125, 1617-1626. doi:10.1242/ dev.125.9.1617 Kostrouchova, M., Krause, M., Kostrouch, Z. and Rall, J. E. (2001). Nuclear hormone receptor CHR3 is a critical regulator of all four larval molts of the nematode Caenorhabditis elegans. Proc. Natl. Acad. Sci. USA 98, 7360-7365. doi:10.1073/pnas.131171898 Kouns, N. A., Nakielna, J., Behensky, F., Krause, M. W., Kostrouch, Z. and Kostrouchova, M. (2011). NHR-23 dependent collagen and hedgehog-related genes required for molting. Biochem. Biophys. Res. Commun. 413, 515-520. doi:10.1016/j.bbrc.2011.08.124 Kramer, J. M. and Johnson, J. J. (1993). Analysis of mutations in the sqt-1 and rol- 6 collagen genes of Caenorhabditis elegans. Genetics 135, 1035-1045. doi:10. 1093/genetics/135.4.1035 Lam, G. T., Jiang, C. and Thummel, C. S. (1997). Coordination of larval and prepupal gene expression by the DHR3 orphan receptor during Drosophila metamorphosis. Development 124, 1757-1769. doi:10.1242/dev.124.9.1757 Laž etić, V. and Fay, D. S. (2017). Molting in C. elegans. Worm 6, e1330246. doi:10. 1080/21624054.2017.1330246 Lin, K., Dorman, J. B., Rodan, A. and Kenyon, C. (1997). daf-16: an HNF-3/ forkhead family member that can function to double the life-span of Caenorhabditis elegans. Science 278, 1319-1322. doi:10.1126/science.278. 5341.1319 Macneil, L. T., Watson, E., Arda, H. E., Zhu, L. J. and Walhout, A. J. M. (2013). Diet-induced developmental acceleration independent of TOR and insulin in C. elegans. Cell 153, 240-252. doi:10.1016/j.cell.2013.02.049 Manders, E. M., Stap, J., Brakenhoff, G. J., van Driel, R. and Aten, J. A. (1992). Dynamics of three-dimensional replication patterns during the S-phase, analysed by double labelling of DNA and confocal microscopy. J. Cell Sci. 103, 857-862. doi:10.1242/jcs.103.3.857 Manders, E. M. M., Verbeek, F. J. and Aten, J. A. (1993). Measurement of co- localization of objects in dual-colour confocal images. J. Microsc. 169, 375-382. doi:10.1111/j.1365-2818.1993.tb03313.x Meeuse, M. W. M., Hauser, Y. P., Morales Moya, L. J., Hendriks, G. J., Eglinger, J., Bogaarts, G., Tsiairis, C. and Großhans, H. (2020). Developmental function and state transitions of a gene expression oscillator in Caenorhabditis elegans. Mol. Syst. Biol. 16, e9975. doi:10.15252/msb.20209498 Meeuse, M. W. M., Hauser, Y. P., Nahar, S., Smith, A. A. T., Braun, K., Azzi, C., Rempfler, M. and Großhans, H. (2023). C. elegans molting requires rhythmic accumulation of the Grainyhead/LSF transcription factor GRH-1. EMBO J. 42, e111895. doi:10.15252/embj.2022111895 Mello, C. C., Kramer, J. M., Stinchcomb, D. and Ambros, V. (1991). Efficient gene in C. elegans: extrachromosomal maintenance and integration of transfer transforming sequences. EMBO J. 10, 3959-3970. doi:10.1002/j.1460-2075. 1991.tb04966.x Melo, J. A. and Ruvkun, G. (2012). Inactivation of conserved C. elegans genes engages pathogen- and xenobiotic-associated defenses. Cell 149, 452-466. doi:10.1016/j.cell.2012.02.050 Mok, D. Z. L., Sternberg, P. W. and Inoue, T. (2015). Morphologically defined sub- stages of C. elegans vulval development in the fourth larval stage. BMC Dev. Biol. 15, 26. doi:10.1186/s12861-015-0076-7 Moribe, H., Yochem, J., Yamada, H., Tabuse, Y., Fujimoto, T. and Mekada, E. integrity in (2004). Tetraspanin protein (TSP-15) is required for epidermal Caenorhabditis elegans. J. Cell Sci. 117, 5209-5220. doi:10.1242/jcs.01403 Nelson, F. K., Albert, P. S. and Riddle, D. L. (1983). Fine structure of the Caenorhabditis elegans secretory-excretory system. J. Ultrastruct. Res. 82, 156-171. doi:10.1016/S0022-5320(83)90050-3 Nonet, M. L. (2020). Efficient transgenesis in Caenorhabditis elegans using Flp recombinase-mediated cassette exchange. Genetics 215, 903-921. doi:10.1534/ genetics.120.303388 Page, A. P. and Johnstone, I. L. (2007). The cuticle. WormBook, 1-15. doi:10.1895/ wormbook.1.138 Paix, A., Wang, Y., Smith, H. E., Lee, C.-Y. S., Calidas, D., Lu, T., Smith, J., Schmidt, H., Krause, M. W. and Seydoux, G. (2014). Scalable and versatile genome editing using linear DNAs with microhomology to Cas9 Sites in Caenorhabditis elegans. Genetics 198, 1347-1356. doi:10.1534/genetics.114. 170423 Paix, A., Folkmann, A., Rasoloson, D. and Seydoux, G. (2015). High efficiency, homology-directed genome editing in Caenorhabditis elegans using CRISPR- Cas9 Ribonucleoprotein complexes. Genetics 201, 47-54. doi:10.1534/genetics. 115.179382 Patel, R., Galagali, H., Kim, J. K. and Frand, A. R. (2022). Feedback between a retinoid-related nuclear receptor and the let-7 microRNAs controls the pace and number of molting cycles in C. elegans. eLife 11, e80010. doi:10.7554/eLife. 80010 Posch, A., Kohn, J., Oh, K., Hammond, M. and Liu, N. (2013). V3 stain-free workflow for a practical, convenient, and reliable total protein loading control in western blotting. J. Vis. Exp. 82, 50948. doi:10.3791/50948 Pujol, N., Cypowyj, S., Ziegler, K., Millet, A., Astrain, A., Goncharov, A., Jin, Y., Chisholm, A. D. and Ewbank, J. J. (2008). Distinct innate immune responses to infection and wounding in the C. elegans epidermis. Curr. Biol. 18, 481-489. doi:10.1016/j.cub.2008.02.079 Qadota, H., Inoue, M., Hikita, T., Kö ppen, M., Hardin, J. D., Amano, M., Moerman, D. G. and Kaibuchi, K. (2007). Establishment of a tissue-specific RNAi system in C. elegans. Gene 400, 166-173. doi:10.1016/j.gene.2007.06.020 R Core Team (2007). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ Ragle, J. M., Aita, A. L., Morrison, K. N., Martinez-Mendez, R., Saeger, H. N., Ashley, G. A., Johnson, L. C., Schubert, K. A., Shakes, D. C. and Ward, J. D. (2020). The conserved molting/circadian rhythm regulator NHR-23/NR1F1 serves as an essential co-regulator of C. elegans spermatogenesis. Development 147, dev193862. doi:10.1242/dev.193862 Ragle, J. M., Morrison, K. N., Vo, A. A., Johnson, Z. E., Hernandez Lopez, J., Rechtsteiner, A., Shakes, D. C. and Ward, J. D. (2022). NHR-23 and SPE-44 regulate distinct sets of genes during C. elegans spermatogenesis. G3 12, jkac256. doi:10.1093/g3journal/jkac256 Ruaud, A.-F. and Bessereau, J.-L. (2006). Activation of nicotinic receptors in C. elegans. from the molting timer uncouples a developmental Development 133, 2211-2222. doi:10.1242/dev.02392 timer Ruaud, A.-F., Lam, G. and Thummel, C. S. (2010). The Drosophila nuclear receptors DHR3 and βFTZ-F1 control overlapping developmental responses in late embryos. Development 137, 123-131. doi:10.1242/dev.042036 Sandhu, A., Badal, D., Sheokand, R., Tyagi, S. and Singh, V. (2021). Specific collagens maintain the cuticle permeability barrier in Caenorhabditis elegans. Genetics 217, iyaa047. doi:10.1093/genetics/iyaa047 Sapio, M. R., Hilliard, M. A., Cermola, M., Favre, R. and Bazzicalupo, P. (2005). The Zona Pellucida domain containing proteins, CUT-1, CUT-3 and CUT-5, play essential roles in the development of the larval alae in Caenorhabditis elegans. Dev. Biol. 282, 231-245. doi:10.1016/j.ydbio.2005.03.011 Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B. et al. (2012). Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, 676-682. doi:10.1038/nmeth.2019 Singh, R. N. and Sulston, J. E. (1978). Some observations on molting in C. elegans. Nematologica 24, 63-71. doi:10.1163/187529278X00074 Stec, N., Doerfel, K., Hills-Muckey, K., Ettorre, V. M., Ercan, S., Keil, W. and Hammell, C. M. (2021). An epigenetic priming mechanism mediated by nutrient sensing regulates transcriptional output during C. elegans development. Curr. Biol. 31, 809-826.e6. doi:10.1016/j.cub.2020.11.060 T N E M P O L E V E D 17 RESEARCH ARTICLE Development (2023) 150, dev201085. doi:10.1242/dev.201085 Stempor, P. and Ahringer, J. (2016). SeqPlots - Interactive software for exploratory data analyses, pattern discovery and visualization in genomics. Wellcome Open Res. 1, 14. doi:10.12688/wellcomeopenres.10004.1 Stojanovski, K., Großhans, H. and Towbin, B. D. (2022). Coupling of growth rate and developmental tempo reduces body size heterogeneity in C. elegans. Nat. Commun. 13, 3132. doi:10.1038/s41467-022-29720-8 Sulston, J. E. and Horvitz, H. R. (1977). Post-embryonic cell lineages of the nematode, Caenorhabditis elegans. Dev. Biol. 56, 110-156. doi:10.1016/0012- 1606(77)90158-0 Tabara, H., Sarkissian, M., Kelly, W. G., Fleenor, J., Grishok, A., Timmons, L., Fire, A. and Mello, C. C. (1999). The rde-1 gene, RNA interference, and transposon silencing in C. elegans. Cell 99, 123-132. doi:10.1016/S0092- 8674(00)81644-X Taubert, S., Ward, J. D. and Yamamoto, K. R. (2011). Nuclear hormone receptors in nematodes: Evolution and function. Mol. Cell. Endocrinol. 334, 49-55. doi:10. 1016/j.mce.2010.04.021 Telford, M. J., Bourlat, S. J., Economou, A., Papillon, D. and Rota-Stabelli, O. (2008). The evolution of the Ecdysozoa. Philos. Trans. R. Soc. B Biol. Sci. 363, 1529-1537. doi:10.1098/rstb.2007.2243 Teuscher, A. C., Jongsma, E., Davis, M. N., Statzer, C., Gebauer, J. M., Naba, A. and Ewald, C. Y. (2019). The in-silico characterization of the Caenorhabditis elegans matrisome and proposal of a novel collagen classification. Matrix Biol. Plus 1, 100001. doi:10.1016/j.mbplus.2018.11.001 Tong, A., Lynn, G., Ngo, V., Wong, D., Moseley, S. L., Ewbank, J. J., Goncharov, A., Wu, Y.-C., Pujol, N. and Chisholm, A. D. (2009). Negative regulation of Caenorhabditis elegans epidermal damage responses by death-associated protein kinase. Proc. Natl. Acad. Sci. USA 106, 1457-1461. doi:10.1073/pnas. 0809339106 Tsiairis, C. and Großhans, H. (2021). Chapter Two - Gene expression oscillations in C. elegans underlie a new developmental clock. In Current Topics in Developmental Biology (ed. S. Jarriault and B. Podbilewicz), pp. 19-43. Academic Press. Vo, A. A., Levenson, M. T., Ragle, J. M. and Ward, J. D. (2021). Efficient generation of a single-copy eft-3p::TIR1::F2A:: BFP::AID::NLS allele in the C. elegans ttTi5605 insertion site through recombination-mediated cassette exchange. MicroPublication Biol. 2021, 10.17912/micropub.biology.000425. Vogel, C. and Marcotte, E. M. (2012). Insights into the regulation of protein abundance from proteomic and transcriptomic analyses. Nat. Rev. Genet. 13, 227-232. doi:10.1038/nrg3185 Vuong-Brender, T. T. K., Suman, S. K. and Labouesse, M. (2017). The apical ECM preserves embryonic integrity and distributes mechanical stress during morphogenesis. Development 144, 4336-4349. doi:10.1242/dev.150383 Ward, J. D. (2015a). Rapid and precise engineering of the Caenorhabditis elegans genome with lethal mutation co-conversion and inactivation of NHEJ repair. Genetics 199, 363-377. doi:10.1534/genetics.114.172361 Ward, J. D. tools from (2015b). Rendering the intractable more tractable: Caenorhabditis elegans ripe for import into parasitic nematodes. Genetics 201, 1279-1294. doi:10.1534/genetics.115.182717 Ward, J. D., Bojanala, N., Bernal, T., Ashrafi, K., Asahina, M. and Yamamoto, K. R. (2013). Sumoylated NHR-25/NR5A regulates cell fate during C. elegans vulval development. PLoS Genet. 9, e1003992. doi:10.1371/journal.pgen. 1003992 Ward, J. D., Mullaney, B., Schiller, B. J., He, L. D., Petnic, S. E., Couillault, C., Pujol, N., Bernal, T. U., Van Gilst, M. R., Ashrafi, K. et al. (2014). Defects in the C. elegans acyl-CoA synthase, acs-3, and nuclear hormone receptor, nhr-25, cause sensitivity to distinct, but overlapping stresses. PLoS ONE 9, e92552. doi:10.1371/journal.pone.0092552 Watts, J. S., Harrison, H. F., Omi, S., Guenthers, Q., Dalelio, J., Pujol, N. and Watts, J. L. (2020). New strains for tissue-specific RNAi studies in Caenorhabditis elegans. G3 10, 4167-4176. doi:10.1534/g3.120.401749 Zaslaver, A., Mayo, A. E., Rosenberg, R., Bashkin, P., Sberro, H., Tsalyuk, M., Surette, M. G. and Alon, U. (2004). Just-in-time transcription program in metabolic pathways. Nat. Genet. 36, 486-491. doi:10.1038/ng1348 Zhang, C. and Ruvkun, G. (2012). New insights into siRNA amplification and RNAi. RNA Biol. 9, 1045-1049. doi:10.4161/rna.21246 Zhang, L., Ward, J. D., Cheng, Z. and Dernburg, A. F. (2015). The auxin-inducible degradation (AID) system enables versatile conditional protein depletion in C. elegans. Development 142, 4374-4384. doi:10.1242/dev.129635 Zugasti, O. and Ewbank, J. J. (2009). Neuroimmune regulation of antimicrobial peptide expression by a noncanonical TGF-β signaling pathway in Caenorhabditis elegans epidermis. Nat. Immunol. 10, 249-256. doi:10.1038/ni.1700 T N E M P O L E V E D 18
10.1186_s12964-023-01131-2	Laakmann et al. Cell Communication and Signaling (2023) 21:111 https://doi.org/10.1186/s12964-023-01131-2 RESEARCH Cell Communication and Signaling Open Access Bacterial extracellular vesicles repress the vascular protective factor RNase1 in human lung endothelial cells Katrin Laakmann1, Jorina Mona Eckersberg1, Moritz Hapke1, Marie Wiegand1, Jeff Bierwagen1, Isabell Beinborn1, Christian Preußer2, Elke Pogge von Strandmann2, Thomas Heimerl3, Bernd Schmeck1,3,4,5,6 and Anna Lena Jung1,4* Abstract Background Sepsis is one of the leading causes of death worldwide and characterized by blood stream infections associated with a dysregulated host response and endothelial cell (EC) dysfunction. Ribonuclease 1 (RNase1) acts as a protective factor of vascular homeostasis and is known to be repressed by massive and persistent inflammation, associated to the development of vascular pathologies. Bacterial extracellular vesicles (bEVs) are released upon infec- tion and may interact with ECs to mediate EC barrier dysfunction. Here, we investigated the impact of bEVs of sepsis- related pathogens on human EC RNase1 regulation. Methods bEVs from sepsis-associated bacteria were isolated via ultrafiltration and size exclusion chromatography and used for stimulation of human lung microvascular ECs combined with and without signaling pathway inhibitor treatments. Results bEVs from Escherichia coli, Klebsiella pneumoniae and Salmonella enterica serovar Typhimurium significantly reduced RNase1 mRNA and protein expression and activated ECs, while TLR2-inducing bEVs from Streptococcus pneu- moniae did not. These effects were mediated via LPS-dependent TLR4 signaling cascades as they could be blocked by Polymyxin B. Additionally, LPS-free ClearColi™ had no impact on RNase1. Further characterization of TLR4 downstream pathways involving NF-кB and p38, as well as JAK1/STAT1 signaling, revealed that RNase1 mRNA regulation is medi- ated via a p38-dependent mechanism. Conclusion Blood stream bEVs from gram-negative, sepsis-associated bacteria reduce the vascular protective factor RNase1, opening new avenues for therapeutical intervention of EC dysfunction via promotion of RNase1 integrity. Keywords Sepsis, OMV, Ribonuclease 1, Endothelium, Inflammation, TLR4, Polymyxin B, p38 Correspondence: Anna Lena Jung anna.jung@uni-marburg.de Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 2 of 16 Introduction Sepsis is one of the leading causes of death worldwide with approximately 50 million cases per year [1]. Devel- opment and progression of sepsis is mainly caused by bacterial infections of the lung, urinary tract, skin and intestine [1–5]. Accordingly, frequent pathogens like gram-positive Streptococcus pneumoniae (Sp) as well as gram-negative, increasingly antibiotic-resistant Entero- bacteriaceae like Escherichia coli (Ec), Klebsiella pneu- moniae (Kp) or Salmonella enterica (Sal) are of major interest [6, 7]. Exemplarily, severe lung infections cause a breakdown of the alveolar membrane and enable the invasion of bacterial pathogens into the bloodstream [3, 7]. These systemic infections provoke an imbalanced immune response of the host that finally ends up in vas- cular barrier breakdown, multi-organ failure and death [8, 9]. In this regard, sepsis progression is characterized by excessive production of pro-inflammatory cytokines and a pro-coagulant state of the vasculature that further promotes endothelial cell (EC) dysfunction [2, 3, 5]. Besides classical inflammatory factors like cytokines or chemokines, extracellular vesicles (EVs) play an essential role during infection and inflammation as novel media- tors of intercellular communication [10]. EVs are small, nano- to micrometer sized, spherical membrane enclosed structures that can be released by eukaryotic host cells as well as bacteria (bEVs; [11]). bEVs from bacteria can further be separated into two classes, outer membrane vesicles (OMVs) that are secreted by gram-negative bac- teria and membrane vesicles (MVs) that are secreted by gram-positive bacteria [11–13]. In respect to pneumonia, various studies describe a wide range of bEV functions in host–pathogen interactions (reviewed in: [11]) and an additional impact of bEVs in blood stream infection is currently under consideration. Thereby, bEVs can either be released by circulating pathogens, but are also capable of reaching distant organs independently of the secreting bacteria [14], where they fulfill an essential role in pro- gression of sepsis-associated dysfunction of the endothe- lium [15–20]. The pathogen-EC interaction is primarily regulated by bacterial immune agonists like lipopolysac- charide (LPS) or lipoprotein that are also associated to the bEV surface [21]. Those known pattern-recognition receptor ligands can further induce inflammation and intracellular signaling in the endothelium, such as activa- tion of Toll-like receptors (TLRs) that promote disrup- tion of the EC barrier and immune function [16, 22]. Cell membrane-bound TLRs like TLR2 or TLR4 can sense bEVs and induce downstream signaling via the MyD88/ IRAK-1 axis that is mediated by IRAK-1 degradation to further promote NF-κB translocation to the nucleus as well as p38 phosphorylation and translocation to acti- vate associated gene expression. Additionally, TLR4 can also induce TRIF-dependent signaling upon endocytosis that promotes IRF3/7-mediated type I interferon (IFN) production and subsequent activation o the f type I IFN receptor (IFNAR) and the JAK/STAT pathway (Fig. S1) [23–25]. ECs function as a protective barrier to separate the blood from the surrounding tissue and act as regulators of vascular homeostasis [26–28]. Upon inflammation, ECs get activated, which is characterized for instance by secretion of pro- and antithrombotic factors as well as upregulation of pro-inflammatory mediators like adhe- sion molecules and cytokines (e.g., intercellular adhesion molecule 1 (ICAM-1), C-X-C motif chemokine ligand 8 (CXCL8), C-X-C motif chemokine ligand 10 (CXCL10)) [15, 26–29]. Thereby, massive and persistent inflamma- tion can tremendously harm the homeostatic function of the endothelium and is involved in development of vas- cular dysfunctions such as sepsis-associated formation of micro-thrombosis or intravascular coagulation [15, 27]. Endothelial Ribonuclease 1 (RNase1) is known as an important vessel- and tissue-protective factor [30–33], countering the damage associated molecular pattern extracellular RNA (eRNA) to prevent excessive EC inflammation [34]. However, massive inflammation results in eRNA-induced EC inflammation and RNase1 repression, effected by massive amounts of pro-inflam- matory mediators like tumor necrosis factor alpha (TNF-α) [34–36], that promotes EC dysfunction and development of vascular pathologies including throm- bosis [30, 34, 37–41]. Besides these data, recent studies also investigated the role of the RNase1-eRNA system in sepsis suggesting a potential RNase1 repression dur- ing disease progression as well as a protective function of RNase1 during sepsis-associated tissue- and organ dam- age [42, 43]. Altogether, the current literature suggests a causal interaction between sepsis-associated EC dys- function and RNase1 repression that is still insufficiently studied. Here, we investigated the impact of bEVs from differ- ent sepsis-associated bacterial pathogens on the regu- lation of the vessel-protective factor RNase1 and the underlying signaling cascades in human lung ECs. In this study, we found that OMVs from the gram-negative bacteria Escherichia coli, Klebsiella pneumoniae and Sal- monella enterica serovar Typhimurium repress endothe- lial RNase1 via an LPS-induced TLR4-IRAK-1 and p38-mediated mechanism to promote EC inflammation that favors development of endothelial dysfunction inde- pendent of the investigated donor bacteria. Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 3 of 16 Material and methods Bacterial vesicle generation and isolation Escherichia coli (#25922, Ec), Klebsiella pneumoniae (#700721/MGH78578, Kp), and Salmonella enterica serovar Typhimurium (#14028, Sal) were purchased from American Type Culture Collection (ATCC; Manassas, USA). ClearColi™ BL21 (cC) were obtained from BioCat GmbH (Heidelberg, Germany). Streptococcus pneumo- niae (D39, Sp) were kindly provided by Sven Hammer- schmidt (University Greifswald, Germany). Ec, Kp and Sal were cultivated on MacConkey agar plates, cC in LB with 1% NaCl and Sp on blood agar plates overnight at 37°C and 5% CO2. Bacteria were then used for inocula- tion of liquid culture media (LB: Ec, Kp, Sal; LB + 1% NaCl: cC; THY: Sp) and grown until they reached the late exponential phase with shaking at 160 rpm and 37°C (MaxQ 6000, Thermo Fisher Scientific, Waltham, USA) excluding Sp, which were cultivated without shaking. To obtain a medium control for following stimulation experiments, LB medium and THY medium was han- dled in parallel to bacterial liquid cultures during vesicle preparation serving as medium controls without bacte- rial growth. Next, bacterial cultures, LB and THY media were centrifuged three times at 4,800 xg for 20 min at 4 °C (Multifuge X3R, Thermo Fisher Scientific) and residual bacteria were removed from the supernatant by using a 0.22 µm pore sterile filter unit. Bacterial vesicles and medium were further processed by ultrafiltration and size exclusion chromatography (UF-SEC) as follows: Bacterial/medium supernatants were concentrated using a 100 kDa molecular weight cut-off filter (Merck Milli- pore, Burlington, Sigma Aldrich, St. Lois, USA) to a final volume of 500 µl, which was further processed by size exclusion chromatography using the qEVoriginal/ 70 nm Gen 2 columns (IZON Science LTD, Lyon, France). Vesi- cle elution was proceeded using 0.1 µm filtered PBS and 24 SEC-fractions were collected (500 µl/fraction). Vesicle enriched fractions (7–12) were concentrated to a final volume between 200 and 400 µl using molecular weight cut-off filters (Merck Millipore) and particle concentra- tion was determined by nano-flow cytometry (nanoFCM; NanoFCM Co., Ltd, Nottingham, UK). Equal amounts of vesicles per cell (multiplicity of vesicles of 1000; MOV1000) were used for stimulation experiments. Each bacterial vesicle preparation was checked for contami- nating bacteria by plating on blood agar plates. Vesicles were aliquoted and stored at -20°C as working stocks. nanoFCM nanoFCM measurements of bEV preparations to deter- mine bEV concentration and size distribution were performed as previously described by Bierwagen et al. 2023 [44]. Transmission Electron Microscopy (TEM) TEM was performed to validate intact vesicle structures as previously described by Bierwagen et al. 2023 [44]. Cell culture Cells used in this study were cultivated in a humidi- fied incubator at 37°C with 5% CO2. Human microvas- cular lung endothelial cells (HULEC-5a) (CRL-3244™, ATCC) were cultured in human microvascular endothelial cell medium MCDB 131 (Gibco™, Thermo Fisher Scientific) supplemented with 10% fetal calf serum (Capricorn Scientific GmbH, Ebsdorfergrund, Germany), 1% penicillin and streptomycin (Gibco™, Thermo Fisher Scientific), 10 mM GlutaMax™ (Gibco®, Thermo Fisher Scientific), 10 ng/ml EGF (Merck Millipore, Sigma Aldrich) and 1 µg/ml hydro- cortisone (Th. Geyer Ingredients GmbH & Co. KG, Höxter, Germany). Cells were cultured up to passage 20 for all experiments. Stimulation of endothelial cells Cells were seeded with 3.8104 cells/cm2 overnight fol- lowed by stimulation for 16 or 24 h (mRNA expression and ELISA) or 0–180 min (Western Blot) as indi- cated: TNF-α (10 ng/ml) (R&D Systems, Inc., Min- neapolis, USA), LPS from Escherichia coli O111:B4 (100 ng/ml) (cell culture grade, Sigma Aldrich), LPS from Salmonella minnesota R595 (100 ng/ml) (cell culture grade, Enzo Life Sciences, Lausen, Switzer- land), IFN-ɣ (250 ng/ml) (Promo Cell, Heidelberg, Germany) or with vesicles from gram-negative bacte- ria (OMVs) from Escherichia coli (EcOMV), Klebsiella pneumoniae (KpOMV), Salmonella enterica sero- var Typhimurium (SalOMV), Clear coli (cCOMV) or MVs from gram-positive Streptococcus pneumoniae (SpMV) with MOV1000, respectively. For Polymyxin B treatment (PB; Merck Millipore), OMVs were preincu- bated for 1 h with the LPS neutralizing antibiotic PB (20 µg/ml) (Merck Millipore) followed by stimulation as indicated. For inhibitor experiments, HULEC-5a were pretreated for 1 h with the JAK1 inhibitor Ruxoli- tinib (5 µM; JAKi) (Biozol Diagnostics Vertrieb GmbH, Eching, Germany), the NF-κB inhibitor BAY11- 7082 (5 µM; NF-κBi) or the p38 inhibitor SB202190 (10 µM; p38i) (Merck Millipore) prior to indicated stimulation. Dimethyl sulfoxide (DMSO, Carl Roth GmbH + Co. KG, Karlsruhe, Germany) served as sol- vent control. Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 4 of 16 Table 1 Primer sequences 5′➔3’ Primer Forward Reverse CXCL8 ACT GAG AGT GAT TGA GAG TGGAC AAC CCT CTG CAC CCA GTT TTC CXCL10 CTG CCA TTC TGA TTT GCT GCC GAT GCA GGT ACA GCG TAC AG ICAM-1 CTC AGT CAG TGT GAC CGC CCT TCT GAG ACC TCT GGC RNase1 GCT GCA GAT CCA GGC TTT TCT AAT TTC TTG GCC CGG GAT TC GGG RPS18 GCG GCG GAA AAT AGC CTT TG GAT CAC ACG TTC CAC CTC ATC TLR2 TLR4 GGC CAG CAA ATT ACC TGT GTG AGG CGG ACA TCC TGA ACC T TGA GAC CAG AAA GCT GGG AG ACT CTG GAT GGG GTT TCC TG RNA isolation and quantitative reverse transcription PCR Total RNA was isolated from HULEC-5a and cDNA was generated as described previously [36]. mRNA tran- script expression of RNase1, CXCL8, ICAM-1, CXCL10, TLR2 and TLR4 was analyzed by quantitative reverse transcription PCR (qPCR) compared to RPS18 that served as internal control. Respective primer pairs are listed in Table 1 (Metabion international AG, Planegg/ Steinkirchen, Germany). qPCR was performed using LUNA® Universal qPCR Master Mix (New England Bio- labs, Ipswich, USA) and the QuantStudioTM System and QuantStudioTM Design & Analysis Software v1.3.1 (both Thermo Fisher Scientific) according to the manufacturer’s instructions. The 2-ΔΔct method was used for calculation of the fold-induction and qPCR results were normalized to the corresponding control cells [45]. LDH‑assay Cytotoxicity of HULEC-5a upon stimulation was deter- mined via lactate dehydrogenase (LDH) release into supernatants compared to a total lysis (TL) represent- ing 100% cell death using the Pierce™ LDH Cytotoxicity Assay Kit (Roche, Mannheim, Germany) according to the manufacturer’s protocol. The absorbance was meas- ured using the infinite F200Pro microplate reader (Tecan, Männedorf, Switzerland). Protein isolation for Western Blot HULEC-5a were seeded and stimulated as described before for 0–180 min as indicated. For isolation of total protein, cells were washed once with PBS and stored dry at -20°C until further processing. Cells were scraped and lysed in RIPA buffer (containing cOmplete™ mini pro- tease inhibitor cocktail and PhosSTOP™, Merck Mil- lipore) followed by sonication via ultrasound for 5 min (30 s on/off at 4°C). Samples were centrifuged for 20 min at 13,000 xg at 4 °C to remove cellular debris. For frac- tionation of cytosolic and nuclear proteins, HULEC- 5a were seeded and stimulated as described above for 30 min. After stimulation, cells were scraped in 1 ml PBS and centrifuged at 240 xg at 4°C for 2 min to pellet cells. The supernatant was discarded, and the pellet was resuspended in fractionation buffer A (10 mM HEPES, pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, sup- plemented with protease inhibitor as well as 0.5 mM DTT) and incubated for 15 min on ice. Cell disruption was additionally promoted by multiple aspirating of the suspension using a 26G needle and syringe, followed by centrifugation for 2 min at 4,400 xg at 4°C. For cytosolic protein isolation, the supernatant was centrifuged for additional 20 min at 20,000 xg at 4°C and stored at -80°C until further use. For nuclear protein isolation, the pel- let was washed twice with fractionation buffer A and the nuclei containing pellet was resuspended in fractionation buffer B (20 mM HEPES, pH 7.9, 400 mM NaCl, 1 mM EDTA, 1 mM EGTA) and incubated for 30–60 min at 4 °C with shaking at 1,000 rpm. Samples were centri- fuged for 20 min at 20,000 xg at 4°C and supernatants were stored at -80°C. Protein concentration of all sam- ples was measured using the Pierce BCA protein assay kit according to the manufacturer’s instructions (Thermo Fisher Scientific). Twenty-five microgram protein per sample and fifteen microgram for fractionation samples were loaded and separated on a 10% or 12.5% SDS PAGE gel (30 min at 80 V and subsequently ~ 120 min for total protein samples or ~ 180 min for fractionation samples at 120 V), respectively, followed by protein transfer and immobilization via wet blot procedure using a Perfect- Blue™ Tank Electro Blotter (VWR International, Radnor, USA) and Towbin buffer on 0.2 µm nitrocellulose mem- brane (GE Healthcare, Chicago, USA) for 1 h with 100 V at 4°C. Membranes were blocked and exposed to anti- bodies targeting IRAK-1 (4359S), phospho-p38 (Thr180/ Tyr182) (9211S), p38 (9212S), STAT1 (D1K9Y, 14994S), phospho-STAT1 (Tyr701) (58D6, 9167S) (Cell Signaling, Cambridge, UK), p65 (F-6; sc-8008), Lamin A/C (H-110; sc-20681), α1c-Tubulin (MH-87, sc-134239) and β-actin (I-19, sc-1616) (Santa Cruz Biotechnology, Heidelberg, Germany) followed by incubation with respective sec- ondary, HRP-conjugated antibodies: mouse anti-rabbit IgG-HRP (L27A9, 5127S; Cell Signaling) or anti-mouse m-IgGк BP-HRP (sc-516102; Santa Cruz Biotechnol- ogy). Chemiluminescence was detected using Amer- sham™ ECL™ Prime Western Blotting Detection Reagent (RPN2236, cytiva, Merck Millipore) and visualized by the ADVANCED Fluorescence and ECL Imager (Intas Sci- ence Imaging Instruments, Germany). ELISA Supernatants of stimulated HULEC-5a were used for protein detection of CXCL8 and RNase1 via ELISA. CXCL8 ELISA was performed as recommended by Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 5 of 16 Fig. 1 Characterization of bacterial outer membrane vesicles. A) Mean size of SEC-purified OMVs from Escherichia coli (EcOMV), Klebsiella pneumoniae (KpOMVs) and Salmonella enterica serovar Typhimurium (SalOMV) and their size distribution profile was determined by nanoFCM. n = 3, line at mean, One-way ANOVA with Tukey‘s multiple comparison test; ns: not significant. B) Vesicle shape and structure was validated by transmission electron microscopy (TEM). Scale bar: 50 nm the manufacturer’s protocol using the BD OptEIATM Human IL-8 ELISA Set (555244, BD Biosciences, Franklin Lakes, USA). RNase1 ELISA was performed using the human Ribonuclease A Matched ELISA Anti- body Pair Set (SEK13468, Sino Biological, Inc., Beijing, China) as recommended by the manufacturer’s instruc- tions. Detection of HRP-mediated signal was per- formed using BD OptEIATM TMB Substrate Reagent Set (555214, BD Biosciences) and the absorbance was measured using the microplate reader infinite F200Pro (Tecan). Statistical analyses Statistical analyses were performed using GraphPad Prism Version 9.5.0 (730) (GraphPad Software, La Jolla, CA, USA). qPCR results are expressed as log2 transformed data with line at mean. ELISA results are expressed as linear data, line at mean. One-way or two-way ANOVA were performed as indicated with subsequent multiple comparison using recom- mended post-tests as indicated in the figure legend. Results were considered significant at p ≤ 0.05, which is labelled with * or # in the figures. Availability of data and materials All data generated or analyzed during this study are included in this article and its supplementary file. Results Characterization of bacterial vesicles To investigate the impact of bacterial vesicles from the sepsis-associated gram-negative pathogens Escherichia coli (EcOMVs), Klebsiella pneumoniae (KpOMVs) and Salmonella enterica serovar Typhimurium (SalOMVs) as well as the gram-positive pathogen Streptococcus pneu- moniae (SpMVs) on endothelial RNase1, OMVs and MVs were isolated from liquid bacterial culture via UF-SEC (Fig. S2A). Particle concentration and size distribution of isolated vesicles was investigated by nanoFCM. No significant differences were observed between the differ- ent vesicle types showing mean sizes of approximately 60–70 nm as well as their size distribution profile, peak- ing at ~ 50 nm (Fig. 1A and Fig. S2B-D). Further charac- terization of isolated bacterial vesicles was performed by TEM imaging showing intact, spherical, membrane enclosed structures for all tested vesicle types (Fig. 1B). OMVs from Ec, Kp, Sal repress RNase1 and activate the endothelium To analyze the regulatory potential of OMVs and MVs on endothelial RNase1 and their impact on proinflammatory EC activation, HULEC-5a were exposed to TNF-α (10 ng/ ml) as positive control [36], EcOMVs, KpOMVs, SalOMVs and SpMVs with MOV1000 or left untreated as control (Ctrl) for 16 and 24 h (Fig. 2, Figs. S3A and S4). UF-SEC Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 6 of 16 Fig. 2 OMVs repress endothelial RNase1 and induce proinflammatory EC activation. Human lung microvascular ECs (HULEC-5a) were stimulated for 16 h and 24 h with SEC-purified OMVs (MOV1000) from Ec, Kp and Sal or TNF-α (10 ng/ml) or left untreated as control (Ctrl). mRNA expression of A) RNase1 C) CXCL8, E) ICAM-1 and F) CXCL10 was determined by qPCR, normalized to RPS18 and untreated cells (Ctrl). B) RNase1 and D) CXCL8 protein release in supernatants of 16 h stimulated HULEC-5a was measured by ELISA, depicted as x-fold protein relative to control (Ctrl). n = 3–4, line at mean, One-way ANOVA with Dunnett‘s multiple comparison test compared to respective Ctrl. p < 0.05, p < 0.01, p < 0.001, *p < 0.0001 processed LB- or THY-medium served as medium con- trol to assess potential effects of medium components on EC activation (Fig. S3B-C). Interestingly, RNase1 mRNA expression was significantly repressed by EcOMVs, KpOMVs, SalOMVs treatment after 16 h, similar to stimulation with the RNase1-repressive cytokine TNF-α. This effect was less prominent at 24 h, although RNase1 mRNA remained suppressed (Fig. 2A). In contrast, no effect on RNase1 mRNA could be observed by stimula- tion with the gram-positive SpMVs (Fig. S4A) as well as the LB and THY-medium controls (Fig. S3C). Similar to TNF-α, OMVs from Ec, Kp, Sal also repressed secretion of RNase1 protein measured by ELISA after 16 h stimulation (Fig. 2B), while SpMV did not regulate RNase1 protein levels (Fig. S4B). Besides RNase1 regulation, all OMVs proinflammatory activated HULEC-5a after 16 and 24 h stimulation, as indicated by mRNA expression of CXCL8, ICAM-1 and CXCL10 (Fig. 2C, E–F), while this was not the case for SpMV (Fig. S4C-E). Additionally, protein secretion of CXCL8 was also increased by OMV treat- ment (Fig. 2D). To ensure cell viability upon OMV and MV treatment, cytotoxicity of respective stimulants was obtained by LDH assay showing no significant increase in cytotoxicity upon exposure above a threshold of 30% cytotoxicity (Fig. S3A-B). Accordingly, OMVs from sepsis- associated gram-negative bacteria Ec, Kp and Sal specifi- cally repressed endothelial RNase1 and activated human lung ECs, in contrast to gram-positive MVs from Sp. Based on these findings, further analysis focused on 16 h OMV treatment due to the observed significant RNase1 regulation in conjunction with a strong proinflammatory activation of the cells. OMVs induce a TLR4‑dependent signaling cascade in human ECs to repress RNase1 To understand the underlying mechanisms, we investi- gated the basal mRNA expression of specific TLRs that are needed to sense those bEV types on the cell sur- face: TLR2 for SpMVs and TLR4 for all tested OMV types. Basal mRNA expression of TLR2 and TLR4 was investigated by qPCR. These data revealed substan- tial differences in TLR2 and TLR4 mRNA expression in HULEC-5a with low abundance for TLR2 compared to high abundance for TLR4 (Fig. S5). Thus, unrespon- siveness of HULEC-5a to SpMV might be associated to the low availability of TLR2, while abundant TLR4 Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 7 of 16 Fig. 3 OMVs repress endothelial RNase1 and induce proinflammatory EC activation in a TLR4-dependent manner. SEC-purified OMVs from Ec, Kp and Sal were pretreated for 1 h with 20 µg/ml Polymyxin B (PB) and further used for stimulation of HULEC-5a (MOV1000) for 16 h, as well as TNF-α (10 ng/ml), LPS from Ec (LPSEc) or Sal (LPSSal) (100 ng/ml) or left untreated as control (Ctrl). mRNA expression of A) RNase1, C) CXCL8, E) ICAM-1 and F) CXCL10 was determined by qPCR, normalized to RPS18 and Ctrl. n = 3–5, line at mean, Two-way ANOVA with Tukey ‘s multiple comparison test. B) RNase1 protein release and D) CXCL8 protein release from supernatants of 16 h stimulated HULEC-5a was measured by ELISA, depicted as x-fold protein relative to Ctrl. n = 4, line at mean, One-way ANOVA with Tukey ‘s multiple comparison test. compared to corresponding Ctrl, # as indicated. p < 0.05, p < 0.01, p < 0.001, *p < 0.0001. #p < 0.05, ##p < 0.01, ####p < 0.0001 expression is a prerequisite for the responsiveness of these cells to Ec, Kp and Sal OMVs. To further unravel the intracellular signaling cascade that transmits OMV- mediated RNase1 repression, we addressed LPS that is exposed on the OMV surface as potent TLR4 agonist to induce proinflammatory signaling cascades via either the MyD88/IRAK-1 or the TRIF axis (Fig. S1) [23–25]. To block the potential LPS-mediated TLR4 induction by OMVs, the LPS neutralizing antibiotic Polymyxin B (PB) was used for further analysis (Fig. S1, Fig. 3) [46, 47]. EcOMVs, KpOMVs and SalOMVs (MOV1000) were pre- incubated with PB (20 µg/ml) for 1 h, prior to 16 h stim- ulation of HULEC-5a with untreated or PB-pretreated OMVs. For comparison purposes, cells were stimulated with LPS from E. coli (LPSEc) or Salmonella (LPSSal) (100 ng/ml), TNF-α (10 ng/ml) or left unstimulated (Ctrl). As an additional control, OMVs from endotoxin free Clear coli (cCOMVs), that carry a modified lipid A and have therefore a non-functional version of LPS, served as control as these bacteria are not able to elicit TLR4-mediated endotoxic responses [48]. cCOMVs were comparable in size to all other used OMVs and did not show any difference in their size distribution profile (Fig. S2B and D). In accordance with previous results, TNF-α and OMVs significantly reduced RNase1 mRNA expression, while LPS from Ec and Sal only showed tendencies of RNase1 repression. Compared to that, stimulation with cCOMVs and the LPS-blocking agent PB did not influence RNase1 mRNA expression. Remarkably, stimulation of HULEC- 5a with PB-pretreated OMVs (grey bars) significantly recovered RNase1 mRNA compared to the respective untreated OMVs (white bars) for Ec, Kp and SalOMVs (Fig. 3A). Similar results were observed on protein level, where TNF-α and untreated OMVs, except cCOMVs, repressed RNase1 protein release which could be recov- ered by OMV-pretreatment with PB (Fig. 3B). In addi- tion to RNase1 regulation, TNF-α, LPSEc, LPSSal and untreated OMVs, except cCOMVs, significantly induced proinflammatory EC activation as demonstrated by CXCL8 mRNA induction and protein release (Fig. 3C and D), ICAM-1 mRNA (Fig. 3E) and CXCL10 mRNA (Fig. 3F) compared to PB and Ctrl treatment. In addition, PB-pretreated OMVs were also not capable to induce a Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 8 of 16 Fig. 4 OMVs induce proinflammatory signaling cascades in human lung ECs. A) Human lung microvascular ECs (HULEC-5a) were stimulated with SEC-purified KpOMVs (MOV1000) or left untreated as control (Ctrl). Expression and phosphorylation of IRAK-1, p38 and STAT1 after OMV exposure for up to 180 min were determined by Western Blot. β-actin served as a loading control. B) HULEC-5a were stimulated for 180 min with Kp or cCOMVs (MOV1000) alone or in combination with PB (20 µg/ml) or left untreated for Ctrl. Expression and phosphorylation of IRAK-1 and p38 was determined by Western Blot. β-actin served as a loading control. C) HULEC-5a were incubated with KpOMV for up to 45 min or D) incubated with KpOMV for 30 min with or without 1 h pretreatment with PB (20 µg/ml). Samples were fractionated into cytosolic and nuclear proteins and analyzed by Western Blot for p65 and p38. Tubulin served as a cytoplasmic loading control, while Lamin A/C served as a nuclear loading control. One representative result of three biological independent replicates is shown strong proinflammatory response. Additionally, none of the stimulations induced significant changes in cytotox- icity (Fig. S3D). To further validate the previous observations on pro- tein level, activation of the TLR4-associated intracellular signaling molecules (Fig. S1) in KpOMV-stimulated ECs (MOV1000 for 0–180 min) was analyzed. Degradation of IRAK-1, phosphorylation of p38 (at threonine 180 and tyrosine 182) and phosphorylation of STAT1 (at tyrosine 701) were investigated by Western Blot along with the loading control β-actin (Fig. 4A). IRAK-1 degradation started at approximately 60 min after OMV exposure, while phosphorylation of p38 already increased 30 min after OMV exposure. Phosphorylation of STAT1 was prominent after 180 min of stimulation. To investigate the involvement of TLR4 in the observed activation pat- tern in ECs, cells were further stimulated with KpOMVs alone or in combination with PB (Fig. 4B). While cCOMVs served as negative control. Kp and cCOMVs were added to HULEC-5a cells at MOV1000 for 180 min with or without PB pretreatment for 1 h. Combination of KpOMVs with PB blocked degradation of IRAK-1 and phosphorylation of p38. In contrast to KpOMVs, cCOMVs lacked the capacity to induce degradation of IRAK-1 along with phosphorylation of p38 (Fig. 4B). Additionally, nuclear translocation of p65 and p38 were investigated as mean of NF-κB and p38 signaling activation and their nuclear translocation both occurred 30 min after KpOMV addition (Fig. 4C). Interestingly, PB pretreatment of KpOMVs reduced p65 and p38 translo- cation (Fig. 4D). Altogether, these results indicate that OMVs induce LPS-mediated TLR4 signaling and associ- ated downstream activation via the MyD88/IRAK-1 axis that further results in activation of NF-κB and p38 signal- ing to block RNase1 in human ECs. OMVs repress endothelial RNase1 mRNA via p38 signaling To further identify the responsible downstream signal- ing pathway for OMV-mediated RNase1 repression, we performed OMV-stimulation experiments combined with signaling molecule inhibitors targeting the type I interferon- or IL-6-mediated JAK/STAT pathway (JAKi: Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 9 of 16 Fig. 5 OMV-mediated RNase1 mRNA repression is independent of JAK/STAT and NF-κB signaling. HULEC-5a were pretreated for 1 h with 5 µM JAKi (Ruxolitinib) or NF-κBi (BAY-11–7082) followed by 16 h stimulation with SEC-purified OMVs from Ec, Kp and Sal (MOV1000) as well as IFN-γ (250 ng/ml) or TNF-α (10 ng/ml) as indicated, left untreated as control (Ctrl) or treated with DMSO as solvent control (Ctrl-DMSO). mRNA expression of RNase1 (A and C), B) CXCL10 or D) ICAM-1 was determined by qPCR, normalized to RPS18 and the respective Ctrl. n = 3–6, line at mean, Two-way ANOVA with Tukey‘s multiple comparison test. compared to corresponding Ctrl, # as indicated. p < 0.05, p < 0.01, p < 0.001, *p < 0.0001. #p < 0.05, ##p < 0.01, ####p < 0.0001 JAK1 inhibitor; Ruxolitinib; 5 µM), the MyD88/IRAK-1/ NF-κB pathway (NF-κBi: NF-κB inhibitor; BAY11-7082; 5 µM) and the MyD88/IRAK-1/p38 pathway (p38i: p38 inhibitor; SB202190; 10 µM). Therefore, HULEC-5a were pretreated for 1 h with JAKi, NF-κBi (Fig. 5) or p38i (Fig. 6; grey bars) prior to 16 h OMV (MOV1000), IFN-γ (250 ng/ml) or TNF-α (10 ng/ml) stimulation (white bars) as indicated. Untreated and DMSO treated cells served as controls (Ctrl, Ctrl-DMSO). Investigation of JAK/STAT signaling using JAKi showed a slight upregu- lation of RNase1 mRNA by the inhibitor itself com- pared to the respective controls. Additionally, IFN-γ as an initiator of JAK/STAT signaling only slightly affected RNase1 mRNA levels without reaching significance, that was not affected by JAKi treatment (Fig. 5A). In line with our previous data for STAT1 phosphorylation (Fig. 4A), RNase1 mRNA expression was repressed upon EcOMV, KpOMV and SalOMV stimulation independent of JAKi treatment (Fig. 5A). In addition, mRNA expression of the JAK/STAT-dependent mRNA CXCL10 [49] was sig- nificantly increased upon IFN-γ and OMV treatment that was blocked to basal level by JAKi pretreatment, indicat- ing successful pathway inhibition by the applied inhibitor (Fig. 5B). As JAK/STAT signaling seems not to be respon- sible for OMV-mediated RNase1 regulation, we further investigated the MyD88/IRAK-1/NF-κB pathway using NF-κBi (Fig. 5C-D). Interestingly, the inhibitor itself slightly repressed RNase1 mRNA expression and could not recover TNF-α- or OMV-mediated RNase1 repres- sion on mRNA level. On the contrary, NF-кBi further intensified the repressive effect of TNF-α and OMVs on RNase1 expression (Fig. 5C). Functionality of the inhibi- tor was demonstrated by mRNA expression of NF-κB- dependent ICAM-1 [50, 51]. Here, TNF-α or OMV induced ICAM-1 mRNA expression was successfully reduced upon NF-κBi treatment compared to samples without inhibitor pretreatment (Fig. 5D). Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 10 of 16 Fig. 6 OMV-mediated RNase1 mRNA repression depends on p38 signaling. HULEC-5a were pretreated for 1 h with 10 µM p38i (SB202190) followed by 16 h stimulation with SEC-purified OMVs from Ec, Kp and Sal (MOV1000) as well as TNF-α (10 ng/ml), treated with DMSO as solvent control (Ctrl-DMSO) or left untreated as control (Ctrl). mRNA expression of A) RNase1 and C) CXCL10 was determined by qPCR, normalized to RPS18 and the respective Ctrl. n = 3–5, line at mean, Two-way ANOVA with Tukey‘s multiple comparison test. B) RNase1 protein release from 16 h stimulated HULEC-5a was measured by ELISA, depicted as x-fold protein relative to Ctrl. D) HULEC-5a were pretreated with p38i (10 µM for 1 h) followed by 180 min KpOMV stimulation (MOV1000). Expression and phosphorylation of p38 and STAT1 were determined by Western Blot. β-actin served as a loading control. One representative result is shown. A, C) n = 4, line at mean; Two-way ANOVA with Tukey‘s multiple comparison test. compared to corresponding Ctrl, # as indicated. p < 0.05, p < 0.01, p < 0.001, *p < 0.0001. #p < 0.05, ##p < 0.01. B) n = 3, line at mean; One-way ANOVA with Tukey ‘s multiple comparison test. p < 0.05 As a third potential pathway regulating OMV-medi- ated RNase1 repression, the MyD88/IRAK-1/p38 axis was analyzed using p38i (Fig. 6). Similar to previous data obtained with PB-pretreated OMVs, p38i prein- cubation (grey bars) of HULEC-5a markedly recovered RNase1 mRNA expression after treatment with TNF- α, EcOMV, KpOMV or SalOMV (white bars) to almost basal levels compared to p38i alone or the respective controls. Furthermore, this effect even reached signifi- cance in the case of KpOMVs (Fig. 6A). Interestingly, RNase1 protein secretion was significantly repressed by TNF-α and OMV exposure independent of p38i treatment, which did not rescue the protein release (Fig. 6B). In addition, TNF-α, EcOMV, KpOMV or SalOMV induced CXCL10 mRNA expression was significantly reduced by p38i, verifying a functional Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 11 of 16 inhibition (Fig. 6C) [49, 52], which is further promoted by successful blocking of p38 phosphorylation upon KpOMV treatment and reduced phosphorylation of STAT1 (Fig. 6D). Additionally, stimulation of HULEC- 5a with all aforementioned stimuli and inhibitors was not cytotoxic to the cells (Fig. S3E-G). Altogether our data provides evidence that OMV- mediated RNase1 mRNA repression is regulated via the TLR4/MyD88/IRAK-1/p38 cascade in human lung endothelial cells. Discussion Sepsis is a life-threatening condition caused by a dereg- ulated immune response to infection that leads to organ dysfunction and accounts for almost 20% of all global mortalities [1, 53]. In this context, vascular dysfunc- tion is a key feature of sepsis, which can lead to bleed- ing, multi-organ failure and death [8, 9, 54]. Besides the classical inflammatory mediators like cytokines or thrombotic factors, EVs were found to be key players in intercellular communication during infection [11]. Bac- teria-derived EVs can be secreted into the blood stream during systemic infection, but are also able to reach the blood prior to the bacteria and disseminate further and faster [14, 55]. Moreover, they are capable of causing- sepsis like systemic inflammation [56] and are associ- ated with complications of sepsis such as disseminated intravascular coagulation and cardiac malfunction [18, 57, 58]. To understand how bEVs influence sepsis progression and endothelial dysfunction, we investigated the impact of bEVs from the sepsis-associated pathogens Escheri- chia coli, Klebsiella pneumoniae, Salmonella enterica serovar Typhimurium and Streptococcus pneumoniae on RNase1, a vessel protective factor, in human lung ECs. Gram-negative OMVs from Ec, Kp and Sal significantly reduced RNase1 expression compared to the gram-pos- itive MVs from Sp. Furthermore, we demonstrated that the OMV-mediated RNase1 repression is regulated via LPS-induced activation of the TLR4/MyD88/IRAK-1 axis, with p38 playing a crucial role in inflamed human lung ECs. RNase1 is a vessel-protective factor that counteracts the effects of eRNA, whose downregulation is linked to various vascular pathologies, like thrombosis, myocardial infarction, atherosclerosis and stroke [30, 41]. Although insufficiently studied in bacterial infections and sep- sis, RNase1 administration has been shown to block the eRNA-mediated mechanism of alveolar epithelial cell infection by Streptococcus pneumoniae [59]. Addition- ally, RNase1 levels are elevated in serum in the early dis- ease stage of sepsis and can act as a prognostic factor for the development of multi-organ failure [43]. However, sepsis progression leads to an increase in serum levels of RNase1 antagonists, eRNA and RNase-Inhibitor, poten- tially repressing RNase1. Besides that, studies in mice suggest that RNase1 administration can prevent sepsis- associated tissue and organ damage [42], highlighting its potential as a therapeutic intervention. These find- ings, combined with the profound influence of RNase1 repression in thrombotic diseases, suggest that the RNase1-eRNA system plays a crucial role in infectious and systemic diseases like sepsis, potentially promoting disease progression and a fatal outcome. Future investigation into RNase1 in the context of (pneumogenic) sepsis is crucial, as pneumonia is the leading cause of sepsis. To this end, we investigated the impact of bEVs from sepsis-associated pathogens on RNase1 regulation in human lung microvascular ECs (HULEC-5a) as a model system. ECs in the pulmonary microvasculature play a critical role in gas exchange [60, 61] and also act as first line of defense by initiating the immune response against invading pathogens [62]. We found that only OMVs from the gram-negative bacteria Ec, Kp and Sal activated the endothelium and repressed RNase1, while gram-positive MVs from Sp could not. The difference may be due to surface-exposed bacterial toxins, with lipoprotein being the major toxin exposed on SpMVs and LPS being present on OMVs [63, 64]. Previous studies demonstrated that SpMVs activated proinflammatory signaling cascades in primary human macrophages via exposed lipoprotein and TLR2 [44, 63, 65]. HULEC-5a may be insensitive to SpMVs due to low TLR2 expression, whereas TLR4 is expressed and cells were more responsive to OMV-associated LPS from Ec, Kp and Sal. This is in line with the literature, showing low expression of TLR2 and high expression of TLR4 in der- mal microvascular ECs [66]. As consequence, OMVs from Ec, Kp and Sal robustly activated the endothelium, that is consistent with data from macrophages [44, 67], as indicated by increased expression of the key inflammatory cytokines CXCL8 and CXCL10 and the cell adhesion molecule ICAM-1 [68, 69] and repression of RNase1 mRNA and protein. This effect might be attributed to the OMV-exposed LPS, as confirmed by treatment with Polymyxin B (PB), an LPS neutralizing peptide antibiotic [46], that could block the OMV-mediated RNase1 repression and EC activation. In comparison, OMVs from cC carrying a non-functional LPS did not influence RNase1 or the proinflammatory EC activation. These data are in line with literature that demonstrates a potent function of PB in blocking OMV- exposed LPS-mediated inflammation [70, 71]. The RNase1 recovering effect of PB in HULEC-5a indi- cated an LPS/TLR4-dependent mechanism responsi- ble for inducing RNase1 repression. Investigation of the Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 12 of 16 underlying intracellular signaling induced by OMV stim- ulation of human lung ECs revealed activation of both the MyD88/IRAK-1 and TRIF axes [23–25, 68]. These findings align with literature investigating OMV-medi- ated signal transduction, where Legionella pneumophila OMVs were shown to proinflammatory activate mac- rophages via TLR2, IRAK-1 and NF-кB [72], while OMVs from Porphyromonas gingivalis mediate endothelial nitric oxide synthase suppression in human ECs via ERK1/2 and p38 signaling [73]. Interestingly, PB treatment of OMVs only blocked the activation of the MyD88/IRAK-1 axis as indicated by reduced IRAK-1 degradation, p38 phosphorylation and translocation and p65 translocation. These results point towards a MyD88/IRAK-1-dependent signaling cascade for RNase1 repression via NF-кB and p38. Similar regulations of LPS-induced TLR4 signaling and PB were obtained by Cheng et al., who demonstrated an impact of PB on LPS-induced endotoxemia in mice via the TLR4/MyD88 axis [74]. Signaling pathway inhibitors targeting specific downstream molecules of TLR4 were used for verification. The JAK1 inhibitor Ruxolitinib was used to investigate the TLR4/IRF axis that provokes acti- vation of IFN signaling via JAK1/STAT1 [23–25]. How- ever, Ruxolitinib treatment of HULEC-5a prior to OMV stimulation did not prevent RNase1 repression and Rux- olitinib treatment alone increased RNase1 expression. Little is known about the underlying molecular pathways of RNase1 regulation in ECs, but referred data primar- ily focused on p38 and histone deacetylase 2 (HDAC2)- mediated mechanisms [35, 36]. Besides the possibility of off-target effects of Ruxolitinib, JAK1 can also be asso- ciated to IL-6 signaling [75]. Unpublished data by our group suggest that HULEC-5a produce IL-6 in response to OMV stimulation, and IL-6/IL-6R treatment represses RNase1. Thus, it might be possible that JAK1 can be associated to RNase1 signaling via an autocrine feedback loop in which OMVs trigger the release of IL-6 via TLR4/ NF-кB/p38 signaling [68, 76, 77]. This in turn can reduce RNase1 expression via JAK1 as well as self-perpetuates IL-6 to promote persistent inflammation. As IL-6 is also known as a key inflammatory mediator during cytokine storm in sepsis [2], further investigation of a possible impact of Ruxolitinib, JAK1 and IL-6 in RNase1-associ- ated EC dysfunction might be of future interest. Thereby, Ruxolitinib could be suitable to prevent sepsis-associated IL-6 release to impede EC dysfunction via promoting RNase1 expression. As the downstream signaling of TLR4/MyD88/IRAK-1 can either be mediated via NF-κB or p38, we further investigated the impact of these on OMV-mediated RNase1 repression using BAY11-7082 as NF-κB inhibitor and SB202190 as p38 inhibitor. BAY11-7082 itself slightly downregulated RNase1 and elevated the repressive effect of TNF-α and OMVs. However, previous studies by Gan- sler et al. and our group did not observe any impact of NF-κB inhibition on RNase1 mRNA expression in pri- mary human umbilical vein ECs [34, 35]. Furthermore, JNK signaling was found to be important for physi- ological RNase1 expression, as JNK inhibition repressed RNase1 [35]. These findings suggest that RNase1 regula- tion in ECs may be organ-specific and vary depending on the physiological demands of the vascular bed [78–81]. ECs from different organs can activate organ-specific pathways and express organ-specific transporters and surface markers [60], and even within a specific vascular bed, there can be differences in EC function [82]. This heterogeneity may explain differences in reactivity of the microvascular lung endothelium used in this study com- pared to ECs from the large umbilical vein used in previ- ous studies addressing RNase1 regulations. Compared to these findings, blocking the p38 signal- ing cascade with SB202190 recovered RNase1 mRNA, indicating that OMV-mediated TLR4 activation signals via MAPK p38 to repress RNase1 expression which is in line with previous studies on TNF-α-mediated RNase1 mRNA repression [35]. TNF-α is a major regulator of RNase1 [39, 41, 83, 84] and can repress it through acti- vation of an HDAC2-containing NuRD repressor com- plex, which deacetylates the RNASE1 promoter region and prevents RNase1 transcription [36]. This pathway is mediated via p38 signaling as its inhibition impedes repressor complex recruitment and RNASE1 chromatin modulation upon TNF-α stimulation [35, 36]. There- fore, p38 MAPK negatively regulates gene expression and is associated with known RNase1 repressive stimuli like TNF-α or IL-1β as well as LPS [35, 85–87]. Although RNase1 mRNA expression was recovered by p38 inhibi- tion, its protein release was still affected by OMVs as well as TNF-α treatment. As p38 is known as a major regu- lator of intracellular signaling, inhibition of this pathway can also influence a variety of cellular factors that might trigger RNase1 repression [85, 86, 88, 89]. Based on our data we conclude that an unknown factor that acts downstream of TLR4 affects RNase1 protein release, as PB treatment efficiently recovered RNase1 protein upon OMV stimulation. However, further investigation needs to be done to maintain RNase1 integrity in the inflamed endothelium. The presented results provide strong evi- dence for a p38-mediated RNase1 mRNA regulation upon OMV treatment and identify p38 as a key repressor of RNase1, consistent with our previously published data on TNF-α mediated RNase1 regulation [35, 36]. Furthermore, our study suggests potential therapeu- tic strategies for sepsis-associated vascular dysfunction by restoring RNase1 function and vascular homeosta- sis. While p38 inhibitors have shown promise in clinical Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 13 of 16 trials [90–93] by decreasing serum levels of proinflam- matory cytokines [94], our data indicates that p38 inhi- bition alone might not be sufficient to restore RNase1 function and vascular integrity. Thus, a combination of p38 inhibitors with other complementary therapeutic options should be considered. One particularly interest- ing option is PB treatment, as it represents an approved drug and has already been investigated in several clini- cal trials for sepsis intervention, where it improved vari- ous clinical outcomes, such as mean arterial pressure, ventilator-free days, and mortality [95–99]. PB treatment fully recovered RNase1 mRNA and protein and has been shown to reduce the proapoptotic function of plasma of septic patients in sepsis-induced acute renal failure [100]. Additionally, new innovations such as PB-releasing nano- particles have been developed to target bacterial accumu- lation and vascular inflammation [101] and thereby could be a valuable tool for sepsis intervention through RNase1 recovery in human ECs. Conclusions In summary, we observed that EVs released by different gram-negative, (pneumogenic) sepsis-associated bacteria repress RNase1, a key modulator of vascular integrity, in human lung microvascular ECs. This effect is mediated via signaling through TLR4/MyD88/IRAK-1 and p38 and can be prevented by LPS neutralization via PB and in part p38 inhibition. Thereby, our data provides new insights in the regulation of sepsis-induced vascular dysfunction and offers novel treatment options to prevent sepsis- associated RNase1 repression and consecutive vascular breakdown. Abbreviations bEVs cC CXCL8 CXCL10 DMSO Ec eRNA EVs HDAC2 ICAM-1 IFN IFNAR JAKi Kp LDH LPS MAPK MVs NF-кBi OMVs p38i PB qPCR RNase1 Bacterial extracellular vesicles Clear Coli C-X-C motif chemokine ligand 8 C-X-C motif chemokine ligand 10 Dimethyl sulfoxide Escherichia coli Extracellular RNA Extracellular vesicles Histone deacetylase 2 Intercellular adhesion molecule 1 Interferon Interferon alpha/beta receptor Janus activated kinase inhibitor (Ruxolitinib) Klebsiella pneumoniae Lactate dehydrogenase Lipopolysaccharide Mitogen activated protein kinase Membrane vesicles Nuclear factor kappa B inhibitor (BAY11-7081) Outer membrane vesicles P38 inhibitor (SB202190) Polymyxin B Quantitative reverse transcription PCR Ribonuclease 1 Sal Sp TLR TNF-α UF-SEC Salmonella enterica serovar Typhimurium Streptococcus pneumoniae Toll-like receptor Tumor necrosis factor alpha Ultrafiltration and size exclusion chromatography Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12964- 023- 01131-2. Additional file 1: Figure S1. OMV-mediated intracellular signaling cas- cades. Figure S2. Isolation procedure and characterization of bEVs. Figure S3. Impact of OMVs and inhibitor treatments on HULEC-5a. Figure S4. SpMVs do not activate human lung ECs. Figure S5. Basal TLR2 and TLR4 mRNA expression in human lung ECs. Acknowledgements We thank Sabrina von Einem for her help in editing and proof reading the manuscript and all other lab members for support and discussion. All graphi- cal illustrations were created with BioRender.com. Authors’ contributions KL, BS, ALJ contributed to conception and design of the study; KL, JME, MH, MW, JB, IB, performed experiments; CP performed vesicle quantification; TH performed TEM; KL, JME, MH, IB, ALJ analyzed the data; KL, ALJ created the figures; KL, ALJ wrote the manuscript; all authors contributed to manuscript revision, read and approved the submitted version. Funding Open Access funding enabled and organized by Projekt DEAL. Parts of this work were funded by the Von-Behring Röntgen-Stiftung (69–0012) to KL, the Hessisches Ministerium für Wissenschaft und Kunst (LOEWE Diffusible Signals) to BS and ALJ, the German Ministry for Education and Research (BMBF) (ERACo-SysMed2 SysMed-COPD-FKZ 031L0140, PerMedCOPD-FKZ 01EK2203A) to BS, the Von-Behring Röntgen-Stiftung (66-LV07) to BS and the Deutsche Forschungsgemeinschaft (SFB/TR-84 TP C01) to BS. Declarations Competing interests All authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Author details 1 Institute for Lung Research, Universities of Giessen and Marburg Lung Center, Philipps-University Marburg, German Center for Lung Research (DZL), Marburg, Germany. 2 Institute for Tumor Immunology and Core Facility – Extracellular Vesicles, Philipps-University Marburg, Marburg, Germany. 3 Center for Synthetic Microbiology (SYNMIKRO), Philipps-University Marburg, Marburg, Germany. 4 Core Facility Flow Cytometry – Bacterial Vesicles, Philipps-Univer- sity Marburg, Marburg, Germany. 5 Department of Pulmonary and Critical Care Medicine, Philipps-University Marburg, Marburg, Germany. 6 Member of the German Center for Infectious Disease Research (DZIF), Marburg, Germany. Received: 8 March 2023 Accepted: 17 April 2023 References 1. Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, et al. Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the Global Burden of Disease Study. Lancet. 2020;395:200–11. https:// doi. org/ 10. 1016/ S0140- 6736(19) 32989-7. Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 14 of 16 2. Chousterman BG, Swirski FK, Weber GF. Cytokine storm and sepsis disease pathogenesis. Semin Immunopathol. 2017;39:517–28. https:// doi. org/ 10. 1007/ s00281- 017- 0639-8. 3. Feldman C, Anderson R. Pneumonia as a systemic illness. Curr Opin Pulm Med. 2018;24:237–43. https:// doi. org/ 10. 1097/ MCP. 00000 00000 000466. 4. Minasyan H. Sepsis: mechanisms of bacterial injury to the patient. Scand J Trauma Resusc Emerg Med. 2019;27:19. https:// doi. org/ 10. 1186/ s13049- 019- 0596-4. Iba T, Levi M, Levy JH. Sepsis-induced coagulopathy and disseminated intravascular coagulation. Semin Thromb Hemost. 2020;46:89–95. https:// doi. org/ 10. 1055/s- 0039- 16949 95. 5. from interferons for cytokine responses. Front Immunol. 2017;8:29. https:// doi. org/ 10. 3389/ fimmu. 2017. 00029. 25. Rathinam VAK, Zhao Y, Shao F. Innate immunity to intracellular LPS. Nat Immunol. 2019;20:527–33. https:// doi. org/ 10. 1038/ s41590- 019- 0368-3. 26. Bochenek ML, Schafer K. Role of endothelial cells in acute and chronic thrombosis. Hamostaseologie. 2019;39:128–39. https:// doi. org/ 10. 1055/s- 0038- 16756 14. 27. Rajendran P, Rengarajan T, Thangavel J, Nishigaki Y, Sakthisekaran D, Sethi G, Nishigaki I. The vascular endothelium and human diseases. Int J Biol Sci. 2013;9:1057–69. https:// doi. org/ 10. 7150/ ijbs. 7502. 28. Pober JS, Sessa WC. Evolving functions of endothelial cells in inflamma- tion. Nat Rev Immunol. 2007;7:803–15. https:// doi. org/ 10. 1038/ nri21 71. 6. Luyt CE, Hekimian G, Koulenti D, Chastre J. Microbial cause of ICU- 29. Wang M, Hao HF, Leeper NJ, Zhu LY, Comm EC. Thrombotic regula- acquired pneumonia: hospital-acquired pneumonia versus ventilator- associated pneumonia. Curr Opin Crit Care. 2018;24:332–8. https:// doi. org/ 10. 1097/ MCC. 00000 00000 000526. 7. Prina E, Ranzani OT, Torres A. Community-acquired pneumonia. Lancet. 2015;386:1097–108. https:// doi. org/ 10. 1016/ S0140- 6736(15) 60733-4. 8. Cecconi M, Evans L, Levy M, Rhodes A. Sepsis and septic shock. Lancet. 2018;392:75–87. https:// doi. org/ 10. 1016/ S0140- 6736(18) 30696-2. 9. Hotchkiss RS, Karl IE. The pathophysiology and treatment of sepsis. N Engl J Med. 2003;348:138–50. https:// doi. org/ 10. 1056/ NEJMr a0213 33. tion from the endothelial cell perspectives. Arterioscl Throm Vas. 2018;38:E90–5. https:// doi. org/ 10. 1161/ Atvba ha. 118. 310367. 30. Bedenbender K, Schmeck BT. Endothelial ribonuclease 1 in cardiovas- cular and systemic inflammation. Front Cell Dev Biol. 2020;8:576491. https:// doi. org/ 10. 3389/ fcell. 2020. 576491. 31. Fischer S, Nishio M, Dadkhahi S, Gansler J, Saffarzadeh M, Shibamiy- ama A, et al. Expression and localisation of vascular ribonucleases in endothelial cells. Thromb Haemost. 2011;105:345–55. https:// doi. org/ 10. 1160/ TH10- 06- 0345. 10. Terrasini N, Lionetti V. Exosomes in critical illness. Crit Care Med. 32. Futami J, Tsushima Y, Murato Y, Tada H, Sasaki J, Seno M, Yamada H. 2017;45:1054–60. https:// doi. org/ 10. 1097/ CCM. 00000 00000 002328. 11. Jung AL, Schmeck B, Wiegand M, Bedenbender K, Benedikter BJ. The clinical role of host and bacterial-derived extracellular vesicles in pneu- monia. Adv Drug Deliv Rev. 2021. https:// doi. org/ 10. 1016/j. addr. 2021. 05. 021. 12. Doyle LM, Wang MZ. Overview of extracellular vesicles, their origin, composition, purpose, and methods for exosome isolation and analysis. Cells. 2019. https:// doi. org/ 10. 3390/ cells 80707 27. 13. Maas SLN, Breakefield XO, Weaver AM. Extracellular vesicles: unique intercellular delivery vehicles. Trends Cell Biol. 2017;27:172–88. https:// doi. org/ 10. 1016/j. tcb. 2016. 11. 003. 14. Bittel M, Reichert P, Sarfati I, Dressel A, Leikam S, Uderhardt S, et al. Visualizing transfer of microbial biomolecules by outer membrane vesicles in microbe-host-communication in vivo. J Extracell Vesicles. 2021;10:e12159. https:// doi. org/ 10. 1002/ jev2. 12159. 15. Joffre J, Hellman J, Ince C, Ait-Oufella H. Endothelial responses in sepsis. Am J Respir Crit Care Med. 2020;202:361–70. https:// doi. org/ 10. 1164/ rccm. 201910- 1911TR. 16. Lubkin A, Torres VJ. Bacteria and endothelial cells: a toxic relationship. Curr Opin Microbiol. 2017;35:58–63. https:// doi. org/ 10. 1016/j. mib. 2016. 11. 008. 17. Shah B, Sullivan CJ, Lonergan NE, Stanley S, Soult MC, Britt LD. Circulating bacterial membrane vesicles cause sepsis in rats. Shock. 2012;37:621–8. https:// doi. org/ 10. 1097/ SHK. 0b013 e3182 50de5d. 18. Svennerholm K, Park K-S, Wikström J, Lässer C, Crescitelli R, Shelke GV, et al. Escherichia coli outer membrane vesicles can contribute to sepsis induced cardiac dysfunction. Sci Rep-Uk. 2017;7:17434. https:// doi. org/ 10. 1038/ s41598- 017- 16363-9. 19. Yu YJ, Wang XH, Fan GC. Versatile effects of bacterium-released membrane vesicles on mammalian cells and infectious/inflammatory diseases. Acta Pharmacol Sin. 2018;39:514–33. https:// doi. org/ 10. 1038/ aps. 2017. 82. 20. Zhang Y, Meng H, Ma RS, He ZX, Wu XM, Cao MH, et al. Circulating microparticles, blood cells, and endothelium induce procoagulant activity in sepsis through phosphatidylserine exposure. Shock. 2016;45:299–307. https:// doi. org/ 10. 1097/ Shk. 00000 00000 000509. 21. Toyofuku M, Nomura N, Eberl L. Types and origins of bacterial mem- brane vesicles. Nat Rev Microbiol. 2019;17:13–24. https:// doi. org/ 10. 1038/ s41579- 018- 0112-2. 22. Soult MC, Lonergan NE, Shah B, Kim WK, Britt LD, Sullivan CJ. Outer membrane vesicles from pathogenic bacteria initiate an inflamma- tory response in human endothelial cells. J Surg Res. 2013;184:458–66. https:// doi. org/ 10. 1016/j. jss. 2013. 05. 035. 23. Gay NJ, Symmons MF, Gangloff M, Bryant CE. Assembly and localiza- tion of Toll-like receptor signalling complexes. Nat Rev Immunol. 2014;14:546–58. https:// doi. org/ 10. 1038/ nri37 13. 24. Majoros A, Platanitis E, Kernbauer-Hölzl E, Rosebrock F, Müller M, Decker T. Canonical and non-canonical aspects of JAK-STAT signaling: lessons Tissue-specific expression of pancreatic-type RNases and RNase inhibi- tor in humans. DNA Cell Biol. 1997;16:413–9. https:// doi. org/ 10. 1089/ dna. 1997. 16. 413. 33. Landre JB, Hewett PW, Olivot JM, Friedl P, Ko Y, Sachinidis A, Moen- ner M. Human endothelial cells selectively express large amounts of pancreatic-type ribonuclease (RNase 1). J Cell Biochem. 2002;86:540– 52. https:// doi. org/ 10. 1002/ jcb. 10234. 34. Gansler J, Preissner KT, Fischer S. Influence of proinflammatory stimuli on the expression of vascular ribonuclease 1 in endothelial cells. FASEB J. 2014;28:752–60. https:// doi. org/ 10. 1096/ fj. 13- 238600. 35. Bedenbender K, Beinborn I, Vollmeister E, Schmeck B. p38 and Casein Kinase 2 Mediate Ribonuclease 1 Repression in Inflamed Human Endothelial Cells via Promoter Remodeling Through Nucleosome Remodeling and Deacetylase Complex. Front Cell Dev Biol. 2020;8:604. https:// doi. org/ 10. 3389/ fcell. 2020. 563604. 36. Bedenbender K, Scheller N, Fischer S, Leiting S, Preissner KT, Schmeck BT, Vollmeister E. Inflammation-mediated deacetylation of the ribonu- clease 1 promoter via histone deacetylase 2 in endothelial cells. FASEB J. 2019;33:9017–29. https:// doi. org/ 10. 1096/ fj. 20190 0451R. 37. Fischer S, Preissner KT. Extracellular nucleic acids as novel alarm signals in the vascular system Mediators of defence and disease. Hamostase- ologie. 2013;33:37–42. https:// doi. org/ 10. 5482/ Hamo- 13- 01- 0001. 38. Garnett ER, Lomax JE, Mohammed BM, Gailani D, Sheehan JP, Raines RT. Phenotype of ribonuclease 1 deficiency in mice. RNA. 2019;25:921–34. https:// doi. org/ 10. 1261/ rna. 070433. 119. 39. Kannemeier C, Shibamiya A, Nakazawa F, Trusheim H, Ruppert C, Markart P, et al. Extracellular RNA constitutes a natural procoagulant cofactor in blood coagulation. Proc Natl Acad Sci U S A. 2007;104:6388– 93. https:// doi. org/ 10. 1073/ pnas. 06086 47104. 40. Simsekyilmaz S, Cabrera-Fuentes HA, Meiler S, Kostin S, Baumer Y, Liehn EA, et al. Role of extracellular RNA in atherosclerotic plaque formation in mice. Circulation. 2014;129:598–606. https:// doi. org/ 10. 1161/ CIRCU LATIO NAHA. 113. 002562. 41. Zernecke A, Preissner KT. Extracellular Ribonucleic Acids (RNA) Enter the Stage in Cardiovascular Disease. Circ Res. 2016;118:469–79. https:// doi. org/ 10. 1161/ CIRCR ESAHA. 115. 307961. 42. Zechendorf E, O’Riordan CE, Stiehler L, Wischmeyer N, Chiazza F, Col- lotta D, et al. Ribonuclease 1 attenuates septic cardiomyopathy and cardiac apoptosis in a murine model of polymicrobial sepsis. JCI Insight. 2020. https:// doi. org/ 10. 1172/ jci. insig ht. 131571. 43. Martin L, Koczera P, Simons N, Zechendorf E, Hoeger J, Marx G, Schuer- holz T. The Human Host Defense Ribonucleases 1, 3 and 7 Are Elevated in Patients with Sepsis after Major Surgery–A Pilot Study. Int J Mol Sci. 2016;17:294. https:// doi. org/ 10. 3390/ ijms1 70302 94. 44. Bierwagen J, Wiegand M, Laakmann K, Danov O, Limburg H, Herbel SM, Heimerl T, Dorna J, Jonigk D, Preußer C, Bertrams W, Braun A, Sewald K, Schulte LN, Bauer S, Pogge von Strandmann E, Böttcher-Friebertshäuser E, Schmeck B, Jung AL. Bacterial vesicles block viral replication in Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 15 of 16 macrophages via TLR4-TRIF-axis. Cell Commun Signal. 2023;21(1):65. https:// doi. org/ 10. 1186/ s12964- 023- 01086-4. 45. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2(T)(-Delta Delta C) method. Meth- ods. 2001;25:402–8. https:// doi. org/ 10. 1006/ meth. 2001. 1262. 46. Goode A, Yeh V, Bonev BB. Interactions of polymyxin B with lipopoly- saccharide-containing membranes. Faraday Discuss. 2021;232:317–29. https:// doi. org/ 10. 1039/ d1fd0 0036e. 64. Ellis TN, Kuehn MJ. Virulence and immunomodulatory roles of bacte- rial outer membrane vesicles. Microbiol Mol Biol Rev. 2010;74:81–94. https:// doi. org/ 10. 1128/ MMBR. 00031- 09. 65. Volgers C, Benedikter BJ, Grauls GE, Savelkoul PHM, Stassen FRM. Immunomodulatory role for membrane vesicles released by THP-1 macrophages and respiratory pathogens during macrophage infection. BMC Microbiol. 2017;17:216. https:// doi. org/ 10. 1186/ s12866- 017- 1122-3. 47. Cardoso LS, Araujo MI, Góes AM, Pacífico LG, Oliveira RR, Oliveira SC. 66. Faure E, Equils O, Sieling PA, Thomas L, Zhang FX, Kirschning CJ, et al. Polymyxin B as inhibitor of LPS contamination of Schistosoma mansoni recombinant proteins in human cytokine analysis. Microb Cell Fact. 2007;6:1. https:// doi. org/ 10. 1186/ 1475- 2859-6-1. 48. Mamat U, Woodard RW, Wilke K, Souvignier C, Mead D, Steinmetz E, et al. Endotoxin-free protein production—ClearColi™ technology. Nat Methods. 2013;10:916. https:// doi. org/ 10. 1038/ nmeth.f. 367. 49. Liu M, Guo S, Hibbert JM, Jain V, Singh N, Wilson NO, Stiles JK. CXCL10/ IP-10 in infectious diseases pathogenesis and potential therapeutic implications. Cytokine Growth Factor Rev. 2011;22:121–30. https:// doi. org/ 10. 1016/j. cytog fr. 2011. 06. 001. 50. Xue J, Thippegowda PB, Hu G, Bachmaier K, Christman JW, Malik AB, Tiruppathi C. NF-kappaB regulates thrombin-induced ICAM-1 gene expression in cooperation with NFAT by binding to the intronic NF- kappaB site in the ICAM-1 gene. Physiol Genomics. 2009;38:42–53. https:// doi. org/ 10. 1152/ physi olgen omics. 00012. 2009. 51 Holden NS, Catley MC, Cambridge LM, Barnes PJ, Newton R. ICAM-1 expression is highly NF-kappaB-dependent in A549 cells. No role for ERK and p38 MAPK. Eur J Biochem. 2004;271:785–91. https:// doi. org/ 10. 1111/j. 1432- 1033. 2004. 03982.x. 52. Xia J-B, Mao C-Z, Chen Z-Y, Liu G-H, Wu H-Y, Zhou D-C, et al. The CXCL10/CXCR3 axis promotes cardiac microvascular endothelial cell migration via the p38/FAK pathway in a proliferation-independent manner. Exp Mol Pathol. 2016;100:257–65. https:// doi. org/ 10. 1016/j. yexmp. 2016. 01. 010. 53. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315:801–10. https:// doi. org/ 10. 1001/ jama. 2016. 0287. 54. Greco E, Lupia E, Bosco O, Vizio B, Montrucchio G. Platelets and Multi- Organ Failure in Sepsis. Int J Mol Sci. 2017. https:// doi. org/ 10. 3390/ ijms1 81022 00. Bacterial lipopolysaccharide activates NF-kappaB through toll-like receptor 4 (TLR-4) in cultured human dermal endothelial cells. Dif- ferential expression of TLR-4 and TLR-2 in endothelial cells. J Biol Chem. 2000;275:11058–63. https:// doi. org/ 10. 1074/ jbc. 275. 15. 11058. 67. Watanabe S, Kumazawa Y, Inoue J. Liposomal lipopolysaccharide initi- ates TRIF-dependent signaling pathway independent of CD14. PLoS ONE. 2013;8:e60078. https:// doi. org/ 10. 1371/ journ al. pone. 00600 78. 68. Ciesielska A, Matyjek M, Kwiatkowska K. TLR4 and CD14 trafficking and its influence on LPS-induced pro-inflammatory signaling. Cell Mol Life Sci. 2021;78:1233–61. https:// doi. org/ 10. 1007/ s00018- 020- 03656-y. 69. Dauphinee SM, Karsan A. Lipopolysaccharide signaling in endothelial cells. Lab Invest. 2006;86:9–22. https:// doi. org/ 10. 1038/ labin vest. 37003 66. 70. Pfalzgraff A, Correa W, Heinbockel L, Schromm AB, Lübow C, Gisch N, et al. LPS-neutralizing peptides reduce outer membrane vesicle- induced inflammatory responses. Biochim Biophys Acta Mol Cell Biol Lipids. 2019;1864:1503–13. https:// doi. org/ 10. 1016/j. bbalip. 2019. 05. 018. 71. Kim Y-S, Choi E-J, Lee W-H, Choi S-J, Roh T-Y, Park J, et al. Extracellular vesicles, especially derived from Gram-negative bacteria, in indoor dust induce neutrophilic pulmonary inflammation associated with both Th1 and Th17 cell responses. Clin Exp Allergy. 2013;43:443–54. https:// doi. org/ 10. 1111/ cea. 12085. 72. Herkt CE, Caffrey BE, Surmann K, Blankenburg S, Gesell Salazar M, Jung AL, et al. A MicroRNA Network Controls Legionella pneumophila Replication in Human Macrophages via LGALS8 and MX1. MBio. 2020;11:e03155-e3219. https:// doi. org/ 10. 1128/ mBio. 03155- 19. 73. Jia Y, Guo B, Yang W, Zhao Q, Jia W, Wu Y. Rho kinase mediates Porphy- romonas gingivalis outer membrane vesicle-induced suppression of endothelial nitric oxide synthase through ERK1/2 and p38 MAPK. Arch Oral Biol. 2015;60:488–95. https:// doi. org/ 10. 1016/j. archo ralbio. 2014. 12. 009. 55. Michel LV, Gaborski T. Outer membrane vesicles as molecular biomark- 74. Cheng Y, Du J, Han J, Sun W, Gao F, Zhang P, et al. Polymyxin B Attenu- ers for Gram-negative sepsis: taking advantage of nature’s perfect packages. J Biol Chem. 2022;298:102483. https:// doi. org/ 10. 1016/j. jbc. 2022. 102483. ates LPS-Induced Death but Aggravates Radiation-Induced Death via TLR4-Myd88-IL-6 Pathway. Cell Physiol Biochem. 2017;42:1120–6. https:// doi. org/ 10. 1159/ 00047 8767. 56. Park K-S, Lee J, Lee C, Park HT, Kim J-W, Kim OY, et al. Sepsis-like systemic 75. Huarte E, Peel MT, Verbist K, Fay BL, Bassett R, Albeituni S, et al. Rux- inflammation induced by nano-sized extracellular vesicles from feces. Front Microbiol. 2018;9:1735. https:// doi. org/ 10. 3389/ fmicb. 2018. 01735. 57. Wang E, Liu Y, Qiu X, Tang Y, Wang H, Xiao X, et al. Bacteria-released olitinib, a JAK1/2 Inhibitor, Ameliorates Cytokine Storm in Experi- mental Models of Hyperinflammation Syndrome. Front Pharmacol. 2021;12:650295. https:// doi. org/ 10. 3389/ fphar. 2021. 650295. outer membrane vesicles promote disseminated intravascular coagula- tion. Thromb Res. 2019;178:26–33. https:// doi. org/ 10. 1016/j. throm res. 2019. 03. 019. 76. Kawai T, Akira S. Toll-like receptors and their crosstalk with other innate receptors in infection and immunity. Immunity. 2011;34:637–50. https:// doi. org/ 10. 1016/j. immuni. 2011. 05. 006. 58. Peng Y, Gao M, Liu Y, Qiu X, Cheng X, Yang X, et al. Bacterial outer mem- brane vesicles induce disseminated intravascular coagulation through the caspase-11-gasdermin D pathway. Thromb Res. 2020;196:159–66. https:// doi. org/ 10. 1016/j. throm res. 2020. 08. 013. 59. Zakrzewicz D, Bergmann S, Didiasova M, Giaimo BD, Borggrefe T, Mieth M, et al. Host-derived extracellular RNA promotes adhesion of Streptococcus pneumoniae to endothelial and epithelial cells. Sci Rep. 2016;6:37758. https:// doi. org/ 10. 1038/ srep3 7758. 60. Jambusaria A, Hong Z, Zhang L, Srivastava S, Jana A, Toth PT, et al. Endothelial heterogeneity across distinct vascular beds during homeo- stasis and inflammation. Elife. 2020. https:// doi. org/ 10. 7554/ eLife. 51413. 61. Tsuchiya T, Doi R, Obata T, Hatachi G, Nagayasu T. Lung Microvascular Niche, Repair, and Engineering. Front Bioeng Biotechnol. 2020;8:105. https:// doi. org/ 10. 3389/ fbioe. 2020. 00105. 62. Amersfoort J, Eelen G, Carmeliet P. Immunomodulation by endothelial cells - partnering up with the immune system? Nat Rev Immunol. 2022;22:576–88. https:// doi. org/ 10. 1038/ s41577- 022- 00694-4. 63. Briaud P, Carroll RK. Extracellular vesicle biogenesis and functions in 77. Medzhitov R, Horng T. Transcriptional control of the inflammatory response. Nat Rev Immunol. 2009;9:692–703. https:// doi. org/ 10. 1038/ nri26 34. 78. Hennigs JK, Matuszcak C, Trepel M, Körbelin J. Vascular endothelial cells: heterogeneity and targeting approaches. Cells. 2021. https:// doi. org/ 10. 3390/ cells 10102 712. 79. Kvietys PR, Granger DN. Endothelial cell monolayers as a tool for study- ing microvascular pathophysiology. Am J Physiol. 1997;273:G1189–99. https:// doi. org/ 10. 1152/ ajpgi. 1997. 273.6. G1189. 80. Marziano C, Genet G, Hirschi KK. Vascular endothelial cell specification in health and disease. Angiogenesis. 2021;24:213–36. https:// doi. org/ 10. 1007/ s10456- 021- 09785-7. 81. Trimm E, Red-Horse K. Vascular endothelial cell development and diversity. Nat Rev Cardiol. 2023;20:197–210. https:// doi. org/ 10. 1038/ s41569- 022- 00770-1. 82. Gillich A, Zhang F, Farmer CG, Travaglini KJ, Tan SY, Gu M, et al. Capillary cell-type specialization in the alveolus. Nature. 2020;586:785–9. https:// doi. org/ 10. 1038/ s41586- 020- 2822-7. gram-positive bacteria. Infect Immun. 2020. https:// doi. org/ 10. 1128/ IAI. 00433- 20. 83. Cabrera-Fuentes HA, Ruiz-Meana M, Simsekyilmaz S, Kostin S, Inserte J, Saffarzadeh M, et al. RNase1 prevents the damaging interplay between Laakmann et al. Cell Communication and Signaling (2023) 21:111 Page 16 of 16 extracellular RNA and tumour necrosis factor-alpha in cardiac ischae- mia/reperfusion injury. Thromb Haemost. 2014;112:1110–9. https:// doi. org/ 10. 1160/ TH14- 08- 0703. EUPHRATES Randomized Clinical Trial. JAMA. 2018;320:1455–63. https:// doi. org/ 10. 1001/ jama. 2018. 14618. 99. Klein DJ, Foster D, Walker PM, Bagshaw SM, Mekonnen H, Antonelli 84. Cabrera-Fuentes HA, Niemann B, Grieshaber P, Wollbrueck M, Gehron J, Preissner KT, Boning A. RNase1 as a potential mediator of remote ischaemic preconditioning for cardioprotectiondagger. Eur J Cardio- thorac Surg. 2015;48:732–7. https:// doi. org/ 10. 1093/ ejcts/ ezu519. M. Polymyxin B hemoperfusion in endotoxemic septic shock patients without extreme endotoxemia: a post hoc analysis of the EUPHRATES trial. Intensive Care Med. 2018;44:2205–12. https:// doi. org/ 10. 1007/ s00134- 018- 5463-7. 85. Ashwell JD. The many paths to p38 mitogen-activated protein kinase 100. Cantaluppi V, Assenzio B, Pasero D, Romanazzi GM, Pacitti A, Lanfranco G, et al. Polymyxin-B hemoperfusion inactivates circulating proapop- totic factors. Intensive Care Med. 2008;34:1638–45. https:// doi. org/ 10. 1007/ s00134- 008- 1124-6. 101. Zhang P, Ouyang Q, Zhai T, Sun J, Wu J, Qin F, et al. An inflammation- targeted nanoparticle with bacteria forced release of polymyxin B for pneumonia therapy. Nanoscale. 2022;14:15291–304. https:// doi. org/ 10. 1039/ D2NR0 2026B. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. activation in the immune system. Nat Rev Immunol. 2006;6:532–40. https:// doi. org/ 10. 1038/ nri18 65. 86. Ono K, Han J. The p38 signal transduction pathway: activation and function. Cell Signal. 2000;12:1–13. https:// doi. org/ 10. 1016/ S0898- 6568(99) 00071-6. 87. Raingeaud J, Gupta S, Rogers JS, Dickens M, Han J, Ulevitch RJ, Davis RJ. Pro-inflammatory cytokines and environmental stress cause p38 mitogen-activated protein kinase activation by dual phosphorylation on tyrosine and threonine. J Biol Chem. 1995;270:7420–6. https:// doi. org/ 10. 1074/ jbc. 270. 13. 7420. 88. Miyazawa K, Mori A, Miyata H, Akahane M, Ajisawa Y, Okudaira H. Regu- lation of interleukin-1beta-induced interleukin-6 gene expression in human fibroblast-like synoviocytes by p38 mitogen-activated protein kinase. J Biol Chem. 1998;273:24832–8. https:// doi. org/ 10. 1074/ jbc. 273. 38. 24832. 89. Marie C, Roman-Roman S, Rawadi G. Involvement of mitogen-activated protein kinase pathways in interleukin-8 production by human mono- cytes and polymorphonuclear cells stimulated with lipopolysaccharide or Mycoplasma fermentans membrane lipoproteins. Infect Immun. 1999;67:688–93. https:// doi. org/ 10. 1128/ IAI. 67.2. 688- 693. 1999. 90. Christie JD, Vaslef S, Chang PK, May AK, Gunn SR, Yang S, et al. A Randomized dose-escalation study of the safety and anti-inflammatory activity of the p38 mitogen-activated protein kinase inhibitor dilmapi- mod in severe trauma subjects at risk for acute respiratory distress syndrome. Crit Care Med. 2015;43:1859–69. https:// doi. org/ 10. 1097/ CCM. 00000 00000 001132. 91. Kumar S, Boehm J, Lee JC. p38 map kinases: Key signalling molecules as therapeutic targets for inflammatory diseases. Nat Rev Drug Discov. 2003;2:717–26. https:// doi. org/ 10. 1038/ nrd11 77. 92. Patnaik A, Haluska P, Tolcher AW, Erlichman C, Papadopoulos KP, Lensing JL, et al. A First-in-Human Phase I Study of the Oral p38 MAPK Inhibitor, Ralimetinib (LY2228820 Dimesylate), in Patients with Advanced Cancer. Clin Cancer Res. 2016;22:1095–102. https:// doi. org/ 10. 1158/ 1078- 0432. CCR- 15- 1718. 93. Cheriyan J, Webb AJ, Sarov-Blat L, Elkhawad M, Wallace SML, Mäki- Petäjä KM, et al. Inhibition of p38 mitogen-activated protein kinase improves nitric oxide-mediated vasodilatation and reduces inflamma- tion in hypercholesterolemia. Circulation. 2011;123:515–23. https:// doi. org/ 10. 1161/ CIRCU LATIO NAHA. 110. 971986. 94. Fijen JW, Zijlstra JG, de Boer P, Spanjersberg R, Tervaert JW, van der Werf TS, et al. Suppression of the clinical and cytokine response to endotoxin by RWJ-67657, a p38 mitogen-activated protein-kinase inhibitor, in healthy human volunteers. Clin Exp Immunol. 2001;124:16–20. https:// doi. org/ 10. 1046/j. 1365- 2249. 2001. 01485.x. 95. Srisawat N, Tungsanga S, Lumlertgul N, Komaenthammasophon C, Peerapornratana S, Thamrongsat N, et al. The effect of polymyxin B hemoperfusion on modulation of human leukocyte antigen DR in severe sepsis patients. Crit Care. 2018;22:279. https:// doi. org/ 10. 1186/ s13054- 018- 2077-y. 96. Yaroustovsky M, Abramyan M, Komardina E, Nazarova H, Popov D, Ply- ushch M, et al. Selective LPS Adsorption Using Polymyxin B-Immo- bilized Fiber Cartridges in Sepsis Patients Following Cardiac Surgery. Shock. 2018;49:658–66. https:// doi. org/ 10. 1097/ SHK. 00000 00000 001016. 97. Klein DJ, Foster D, Schorr CA, Kazempour K, Walker PM, Dellinger RP. The EUPHRATES trial (Evaluating the Use of Polymyxin B Hemoperfusion in a Randomized controlled trial of Adults Treated for Endotoxemia and Septic shock): study protocol for a randomized controlled trial. Trials. 2014;15:218. https:// doi. org/ 10. 1186/ 1745- 6215- 15- 218. 98. Dellinger RP, Bagshaw SM, Antonelli M, Foster DM, Klein DJ, Marshall JC, et al. Effect of Targeted Polymyxin B Hemoperfusion on 28-Day Mortal- ity in Patients With Septic Shock and Elevated Endotoxin Level: The • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s13071-019-3601-x	Rinaldi et al. Parasites Vectors (2019) 12:353 https://doi.org/10.1186/s13071-019-3601-x Parasites & Vectors RESEARCH Open Access Rapid assessment of faecal egg count and faecal egg count reduction through composite sampling in cattle Laura Rinaldi1, Alessandra Amadesi1, Elaudy Dufourd2, Antonio Bosco1, Marion Gadanho2, Anne Lehebel2, Maria Paola Maurelli1, Alain Chauvin2, Johannes Charlier3, Giuseppe Cringoli1, Nadine Ravinet2 and Christophe Chartier2 Abstract Background: Faecal egg counts (FEC) and the FEC reduction test (FECRT) for assessing gastrointestinal nematode (GIN) infection and efficacy of anthelmintics are rarely carried out on ruminant farms because of the cost of individual analyses. The use of pooled faecal samples is a promising method to reduce time and costs, but few studies are avail- able for cattle, especially on the evaluation of different pool sizes and FECRT application. Methods: A study was conducted to assess FEC strategies based on pooled faecal samples using different pool sizes and to evaluate the pen-side use of a portable FEC-kit for the assessment of FEC on cattle farms. A total of 19 farms representing 29 groups of cattle were investigated in Italy and France. On each farm, individual faecal samples from heifers were collected before (D0) and two weeks after (D14) anthelmintic treatment with ivermectin or benzimida- zoles. FEC were determined individually and as pooled samples using the Mini-FLOTAC technique. Four different pool sizes were used: 5 individual samples, 10 individual samples, global and global on-farm. Correlations and agreements between individual and pooled results were estimated with Spearman’s correlation coefficient and Lin’s concordance correlation coefficients, respectively. Results: High correlation and agreement coefficients were found between the mean of individual FEC and the mean of FEC of the different pool sizes when considering all FEC obtained at D0 and D14. However, these parameters were lower for FECR calculation due to a poorer estimate of FEC at D14 from the faecal pools. When using FEC from pooled samples only at D0, higher correlation and agreement coefficients were found between FECR data, the better results being obtained with pools of 5 samples. Interestingly, FEC obtained on pooled samples by the portable FEC-kit on- farm showed high correlation and agreement with FEC obtained on individual samples in the laboratory. This field approach has to be validated on a larger scale to assess its feasibility and reliability. Conclusions: The present study highlights that the pooling strategy and the use of portable FEC-kits on-farm are rapid and cost-effective procedures for the assessment of GIN egg excretion and can be used cautiously for FECR calculation following the administration of anthelmintics in cattle. Keywords: Gastrointestinal strongyles, Mini-FLOTAC , FECRT , Pooled faecal samples, Cattle Correspondence: lrinaldi@unina.it 1 Department of Veterinary Medicine and Animal Production, University of Napoli Federico II, CREMOPAR, Napoli, Italy Full list of author information is available at the end of the article © The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/ publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Rinaldi et al. Parasites Vectors (2019) 12:353 Page 2 of 8 Background Gastrointestinal nematode (GIN) parasites, also known as gastrointestinal strongyles (Strongylida, Trichostron- gyloidea), are amongst the most important production- limiting pathogens of grazing ruminants in Europe and globally (http://www.disco ntool s.eu) [1]. The negative impact of GIN on livestock farms is further exacerbated by the escalating spread of anthelmintic resistance (AR) [2], a phenomenon under the attention of the scientific community and stakeholders as demonstrated by several European initiatives including the COST Action COM- BAR (COMBatting Anthelmintic Resistance in Rumi- nants; https ://www.comba r-ca.eu/; CA16230) recently launched to coordinate research on the control of AR in helminth parasites of ruminants. One of the options to make GIN control practices more sustainable is to lower drug application frequency by targeting treatment (TT) to the whole group of ani- mals when infection is high while preserving a pool of unexposed parasites in refugia as free-living stages [3]. Diagnosis of gastrointestinal helminth infection is mainly based on the detection of worm eggs through faecal egg counts (FEC) [1]. The TT approach requires a relevant method (e.g. FEC) that indicates the worm burden of a given group despite the over-dispersed distribution of parasites within a group of animals [4]. Furthermore, there is an urgent need to obtain better information on the AR status in Europe and FEC are required to estimate anthelmintic efficacy/resistance by the faecal egg count reduction test (FECRT) [5]. To perform this test, the ideal group size is around 10 to 15 animals [6]. However, the cost of individual FEC is too high for ruminant farmers and makes veterinarians reluctant to increase FEC-based investigations [7]. As a result, on most ruminant farms, faecal diagnosis is rarely carried out, if at all [2]. A more regular employment of copromicroscopic monitoring of worm egg excretion could be facilitated by reducing the number of individ- ual FEC analyses through the use of composite (pooled) faecal samples in which equal amounts of faeces from several animals are mixed together and a single FEC is determined from the mixture as a proxy of the group mean FEC. Several studies have been performed in sheep compar- ing mean individual counts to pooled counts using dif- ferent pool sizes, ranging from three to ten samples, and different analytic sensitivities of the FEC technique, rang- ing from 10 to 50 eggs per gram (EPG) of faeces [8–12]. These studies indicated that pooling ovine faecal samples was a reliable procedure for assessing GIN FEC taking into account the level of FEC, the pool size and the ana- lytical sensitivity of the method [11]. Less is known about faecal pooling in cattle. Ward et al. [13] in Australia showed a good agreement between mean individual counts (n = 10) and mean composite counts (two pools of five), and George et al. [14] in the USA successfully tested the single pooling from a group of animals ranging from 9 to 19 individuals (mean num- ber of 15.7). However, these two studies were based either on a composite sample made from two pools of five individual faecal samples or on a single pool of all the individual samples and did not investigate the effect of different pool sizes on the FEC estimation. Besides pooling, field-applicable kits allowing on- farm implementation of FEC with easy-to-use devices to quickly analyze pooled samples are needed by the new generation of veterinarians and farmers to quantify hel- minth infection, anthelmintic efficacy and AR. Recently, portable FEC kits combined with a mobile phone applica- tion have been developed for image capture and specific worm egg quantification in horses and humans [15–17]. In order to further improve and evaluate the rapid and cost-effective evaluation of FEC and related FECR in cat- tle, field studies were conducted in order to: (i) further evaluate strategies to assess FEC based on pooled faecal samples (using different pool sizes); and (ii) develop and evaluate a portable FEC-kit in order to perform pooled FEC on-farm. Methods Study design and sampling Between June and October 2017, field trials were con- ducted on a total of 19 cattle farms located in Italy and France. Specifically, in Italy 10 beef cattle farms were included and selected in the Campania and Basilicata regions (southern Italy); cattle were crossbreeds (Lim- ousine, Podolica, Marchigiana). In France, 9 farms were included and selected in Normandy and Brittany regions (north-western France); they were Holstein or Normande breed dairy farms. In both countries, the farms were ini- tially randomly chosen within the selected regions and then the selection was mainly driven by the availability of the farmer and the presence of GIN positive cattle. Overall on each farm, individual faecal samples (20 g at least) from first or second grazing season heifers (aged from 6 to 20 months) were collected before (D0) and two weeks after (D14) anthelmintic treatment, i.e. ivermectin (IVM, injectable solution, 0.2 mg/kg of body weight) or albendazole/fenbendazole (ABZ/FBZ, oral suspension, 7.5 mg/kg of body weight). When the number of heif- ers on a given farm was much higher than 20 and thus exceeded the average value met on most farms, ani- mals were split into similar groups of 10/20 animals and assigned a different treatment. Rinaldi et al. Parasites Vectors (2019) 12:353 Page 3 of 8 In Italy, the animals were divided into 2 groups of 10 animals (one group treated with IVM and one with ABZ) on 3 farms; on 3 other farms, 20 cattle were treated with IVM and on 4 farms 20 cattle were treated with ABZ. Similarly, in France, on 2 farms the animals were divided into 2 groups of 18–20 animals, with on each farm one group treated with IVM and the other with FBZ; on 6 other farms, animals were divided into 5 and 6 groups of 11 to 18 animals, respectively; within each farm the groups were assigned to a treatment with either IVM or FBZ. On one farm, a single group of 9 animals was treated with FBZ. Therefore, a total of 29 groups of cat- tle were available for evaluating the relationship between mean FEC of the individuals and the composite samples, 13 groups (6 treated with IVM and 7 with ABZ) in Italy and 16 groups (7 treated with IVM and 9 with FBZ) in France. The total number of cattle farms, individual fae- cal samples and pools used for the study are provided in Fig. 1. Preparation of pooled samples and parasitological analysis At D0 and D14, bovine faecal samples were analyzed both individually and as pooled samples using the Mini- FLOTAC technique with a detection limit of 5 eggs per gram (EPG) of faeces, using a sodium chloride flotation solution (FS2, specific gravity = 1.200) [18]. Three dif- ferent pool sizes were used when possible (5 or 10 indi- vidual samples, global pooling) according to the protocol described in Rinaldi et al. [12] and Kenyon et al. [11]. Briefly, each sample was labelled, thoroughly homog- enized, individually examined and then composite (pooled) samples were prepared taking approximately 5 g of each sample with the collector of the Fill-FLOTAC [18]. It should be noted that the predefined pool sizes of 5 and 10 could not always be met at both D0 and D14 due to some practical constraints such as the exact number of animals in the group and an insufficient amount of fae- ces to perform the analysis of each pool. The actual pool Fig. 1 The number of Italian and French cattle farms, individual faecal samples and pools used for the study. Abbreviations: ABZ, albendazole; FBZ, fenbendazole; IVM, ivermectin Rinaldi et al. Parasites Vectors (2019) 12:353 Page 4 of 8 sizes (number of animals from which an individual fae- cal sample was included) ranged from 3 to 6 for pools of 5 and from 6 to 10 for pools of 10. The global pool was made from all the individuals whatever the group size (ranging from 9 to 20). At D0 and D14, the same animals were sampled and the same pools were prepared. When one sample was missing in a given pool, the correspond- ing sample was withdrawn before individual FECs were averaged. FECR on‑farm A portable FEC-kit was developed in order to perform pooled FEC on-farm. The kit consisted of 2 Fill-FLOTAC (for sample collection and weighing, homogenization, filtration and filling) and 2 Mini-FLOTAC devices [18], the flotation solution (FS2) and a portable (hand-held) microscope with batteries (Celestron, Torrance, CA, USA) for use on-farm. This portable FEC-kit was used on 10 farms to assess a global pool FEC at D0 and/or D14. Briefly, a single pooled sample was prepared taking 5 g of faeces from all individual samples using Fill-FLOTAC and then thoroughly mixed with a spatula in a large beaker. From this pool (90–100 g), a single sample of 5 g was taken by the Fill-FLOTAC and analyzed using the Mini-FLOTAC technique [18] combined with the reading by a senior researcher under the hand-held microscope. Coprocultures For each of the 29 groups of cattle, a pooled faecal cul- ture was performed at D0 and D14, following the pro- tocol described in MAFF [19]. Developed third-stage larvae (L3) were identified using the morphological keys proposed by van Wyk & Mayhew [20]. Identification and percentages of each nematode genera were conducted on 100 L3; if a sample had 100 or less L3 present, all larvae were identified. So, on the total number of larvae identi- fied, it was possible to give the percentage of each genus. Statistical analysis The mean FEC of individual and pooled samples were calculated as the arithmetic mean. Correlations between the different measures of FEC were assessed by Spear- man’s rho correlation coefficient (rs), the associated 95% confidence interval (CI) and P-value. Moreover, Lin’s concordance correlation coefficients (CCC) and the cor- responding 95% CI were calculated to quantify the agree- ment between the analysis from individual samples and each pool size (including those performed on-farm). Like a correlation, CCC ranges from − 1 to 1, with perfect agreement at 1. The strength of agreement was classified as poor, moderate, substantial or almost perfect for CCC values < 0.9, 0.90–0.95, 0.95–0.99 or > 0.99, respectively [21]. When examining individual samples, the FECR (%) was calculated according to the formula: FECR (%) = [1 − (arithmetic mean of post treatment indi- vidual FECs/ arithmetic mean of pre-treatment indi- vidual FECs)] × 100. For each size of pooled samples (5, 10, global), the FECR (%) was calculated as the percent reduction in pooled FEC at D14 compared to corre- sponding pooled FEC at D0: FECR (%) = [1 − (arithmetic mean of post treatment pooled FECs/ arithmetic mean of pre-treatment pooled FECs)] × 100, the number of pools ranging from 1 to 4. Spearman’s rs and Lin’s CCC were calculated as above between FECR (%) from indi- vidual and pooled samples. In addition, a further correla- tion analysis (rs and CCC) was done for the calculation of FECR (%) using a “mixed approach”, i.e. using FEC on D0 based on pooled samples and FEC on D14 based on individual samples. The following criterion was used for defining reduced efficacy: FECR < 95% and lower limit of 95% confidence interval < 90% [6]. The level of significance was set at a P-value < 0.05 for all tests. All statistical analyses were performed using GraphPad Prism v.5 (Graph Pad Software, San Diego, CA, USA) and SPSS Statistics v.23 (IBM, Armonk, NY, USA). Results FEC in individual and composite samples A total of 200 individual samples were analyzed in Italy and 252 in France. When calculated from individual samples, the mean GIN FEC at D0 and FECR (%) varied between 9.2–359 EPG and 73.3–100%, respectively, pro- viding reasonable variation in FEC and FECR (%) values to be tested in the pooling strategy. The correlation and the agreement between FEC results from individual means and pool means are reported in Table 1 and Fig. 2. Overall, the FEC results of pooled samples strongly correlated with those of individual sam- ples regardless of the pool sizes. When focusing on FEC values at D0 or D14, i.e. FEC ranging between 5–400 EPG and 0–69 EPG, respectively, Spearman’s rs values were notably lower for D14 FEC values. The overall level of agreement between the FEC from individual and pool means was substantial for pool of 5 (CCC = 0.99, P < 0.001), pool of 10 (CCC = 0.97, P < 0.001) or global pool (CCC = 0.97, P < 0.001). When considering results separately for D0 or D14, the agree- ment was substantial for pool of 5 (CCC = 0.98, P < 0.001 and CCC = 0.96, P < 0.001, respectively) and moderate for pool of 10 (CCC = 0.94, P < 0.001 and CCC = 0.95, P < 0.001, respectively) or global pool (CCC = 0.95, P < 0.001 and CCC = 0.94, P < 0.001, respectively). Rinaldi et al. Parasites Vectors (2019) 12:353 Page 5 of 8 Table 1 Spearman’s rho correlation coefficient (rs) and Lin’s concordance correlation coefficients (CCC) between FEC from individual and pooled samples according the pool size and the FEC values (whole, D0 or D14) and between FECR(%) from individual samples and FECR(%) from individual samples at D14 and pooled samples at D0 according the pool size Pool size No. of pools rs 95% CI CCC 95% CI Faecal egg count Pool of 5 samples Pool of 10 samples Global pool Global pool on-farm 58 42 58 26 0.98 0.95–0.99 0.99 0.98–0.99 0.97 0.92–0.99 0.97 0.94–0.98 0.95 0.91–0.97 0.97 0.95–0.98 0.94 0.82–0.98 0.93 0.88–0.96 Pool of 5 samples (D0) 29 0.98 0.93–0.99 0.98 0.96–0.99 Pool of 10 samples (D0) Global pool (D0) Pool of 5 samples (D14) Pool of 10 samples (D14) Global pool (D14) 21 29 29 21 29 Faecal egg count reduction Pool of 5 samples Pool of 10 samples Global pool Global pool on-farm 29 21 29 13 0.98 0.90–1.00 0.94 0.86–0.98 0.91 0.75–0.97 0.95 0.90–0.98 0.84 0.64–0.94 0.96 0.93–0.97 0.79 0.51–0.92 0.95 0.90–0.98 0.69 0.39–0.86 0.94 0.89–0.97 1.00 0.99–1.00 0.97 0.95–0.98 0.88 0.72–0.95 0.82 0.65–0.91 0.80 0.62–0.91 0.82 0.70–0.90 0.68 0.20–0.90 0.89 0.85–0.91 Regarding the diagnosis directly on-farm including D0 and D14 values, results showed a high correlation (rs = 0.94, P < 0.001) and a moderate level of agreement (CCC = 0.93, P < 0.001). The correlation between FECRs resulting from indi- vidual and composite samples showed rs values sig- nificant but moderate for pools of 5 samples (rs = 0.80, P < 0.001), 10 samples (rs = 0.77, P < 0.001) and global pools (rs = 0.67, P < 0.001). Similarly, CCC values indi- cated a poor and decreasing level of agreement for pool of 5 samples (CCC = 0.74; P < 0.001) and global pool (CCC = 0.49, P < 0.001). When considering a mixed determination of FECR using FEC at D0 based on pooled samples and FEC at D14 based on individual samples (Table 1), higher values were obtained both for Spearman’s correlation coefficients and for CCC val- ues. Specifically, the better results were obtained with pools of 5 samples (rs = 1.00, P < 0.001; CCC = 0.97, P < 0.001) and the worst with the global pool (rs = 0.80, P < 0.001; CCC = 0.82, P < 0.001). Data were less avail- able for global pool on-farm and indicated low cor- relation value (rs = 0.68, P < 0.001) and a poor level of agreement (CCC = 0.89, P < 0.001). Coprocultures In Italy, the following GIN genera were detected at D0 (pre-treatment): Cooperia (41%), Trichostrongylus (20%), Oesophagostomum (18%), Ostertagia (11%) and Haemon- chus (10%); at D14 (post-treatment) all samples were neg- ative for GIN larvae. In France, the following GIN genera were detected at D0 (pre-treatment): Cooperia (88%) and Ostertagia (12%). At D14 (post-treatment), the following GIN genera were detected: Cooperia (99%) and Osterta- gia (1%) on the farms treated with IVM, whilst very few numbers of Cooperia and Ostertagia were found at D14 on farms treated with FBZ. Discussion Diagnosis of GIN infections by the examination of indi- vidual faecal samples, although simple and effective, remains expensive and time-consuming which hampers widespread adoption by farmers. Over the last decade, thanks to the development of new diagnostic approaches and the improvement of the existing ones, considerable progress has been made to improve the performance (e.g. increasing the analytic sensitivity, accuracy and preci- sion) of FEC and FECR in livestock. However, to increase user-friendliness and uptake of the FEC and FECR by veterinarians and farmers, portable kits are required to make rapid decisions on the need to treat or to determine whether anthelmintics are effective [1]. In addition, promising results have been obtained in pilot studies using pooled faecal samples to decrease the workload and cost of conducting FEC in sheep and cat- tle [11, 12, 14]. Moreover, in all these studies, as well as in a recent study on a comparison between different FEC methods (McMaster, Wisconsin and Mini-FLOTAC) in four different livestock hosts (cattle, sheep, llamas and horses) [22], the good performance of Mini-FLOTAC was emphasized especially when high accuracy is impor- tant, such as when measuring FECR. In the light of these findings, in the present study a practical approach was developed for a rapid and accu- rate assessment of GIN infection intensity before and after anthelmintic treatment in cattle in Italy and France. The experiment was conducted in parallel in two coun- tries where the susceptibility of GIN could vary as it has been previously mentioned for small ruminants [12] but also encompassing potential variation in the laboratory settings where the tests were performed. The present study provides new insights into standardi- zation of FEC and FECRT on pooled faecal samples by comparing different pool sizes (five samples, ten samples and global) in cattle and the evaluation of a portable kit to perform pen-side FEC. Rinaldi et al. Parasites Vectors (2019) 12:353 Page 6 of 8 Fig. 2 The correlation in FEC (pre-treatment and post-treatment) based on the examination of individuals and pools of 5 (a), 10 (b), global pool (c) and global pool analysed directly on-farm (d) in Italy and France High correlation and agreement coefficients (Spear- man and Lin) were found between the mean of individual FECs and the mean of FECs of three different pool sizes (five samples, ten samples and global) when considering all FEC obtained at D0 and D14. Values were in the same range for the different pools (0.95 to 0.98 for rs and 0.97 to 0.99 for CCC) and indicated that any pooling strat- egy was efficient. However, when focusing on the lowest FECs, i.e. those obtained 14 days after anthelmintic treat- ment, correlations were noticeably lower suggesting a poorer estimation of FEC through pooling, due to a lot of zero data. These poor estimates of FEC at D14 were responsible for a poor FECR calculation. In contrast, when FEC determination at D14 was based on individual faecal samples, noticeably higher cor- relation/agreement values were found for FECR, par- ticularly for a pool of five samples. Our results globally confirm the previous data on pooled FEC/FECR obtained in sheep by Kenyon et al. [11] and Rinaldi et al. [12] and in cattle by Ward et al. [13] and George et al. [14] with different pooling strategies (pools of 5, 10 or 20; global pool of 9–19 animals). In the study of George et al. [14] involving 14 groups of cattle, the mean individual FEC ranged from 82 to 671 and from 0 to 210 EPG for pre- treatment and post-treatment sampling, respectively whereas the FECR (%) ranged from 53.1 to 100. The authors found very high correlation (rs = 0.92) and agree- ment (CCC = 0.95) of FECR (%) between individual and global pooling sampling (9–19 animals per pool). Such distributions in mean individual FEC and in FECR (%) have not been found in the context of the French and Ital- ian cattle production. Kenyon et al. [11] pointed out the importance of the EPG level and the EPG aggregation at D0 for the use of pooled faeces for FECR. Interestingly, FECs obtained on pooled samples by the portable FEC-kit on-farm showed high correlation and agreement with FECs obtained on individual samples in the laboratory. This field approach has to be validated on a larger scale to assess the feasibility and reliability of FECR calculation on-farm. Rinaldi et al. Parasites Vectors (2019) 12:353 Page 7 of 8 The present study also confirmed the findings by Geur- den et al. [23] with the full efficacy of ivermectin on cattle farms in Italy and the lack of efficacy on some farms in France. Conclusions The present study highlighted that the pooling strategy and the use of a portable FEC-kit on-farm are rapid and cost-effective procedures for the assessment of GIN egg excretion and can be used cautiously for FECR calcula- tion following administration of anthelmintics in cattle. The use of improved FEC and FECR together with har- monization of study design and interpretation [14, 24] would allow field surveys to be conducted on a larger scale than today. It would also promote uptake of diag- nostic procedures by veterinary practitioners in order to fill knowledge gaps in the burden of GIN infection and the efficacy of anthelmintics at both the European and global scale. For these reasons, the development of an automated system for reading and counting eggs based on the Mini-FLOTAC technique in the veteri- nary field is in progress. It uses remote support tools to assist veterinarians and farmers to optimize control strategies so that evidence-based parasite control strat- egies for livestock can be effectively implemented in the future. Abbreviations GIN: gastrointestinal nematode; GI strongyles: gastrointestinal strongyles; D0: pre-treatment; D14: post-treatment; IVM: ivermectin; ABZ: albendazole; FBZ: fenbendazole; FEC: faecal egg count; FECR: faecal egg count reduction; FECRT : faecal egg count reduction test; AR: anthelmintic resistance; TT: targeted treat- ment; rs: Spearman’s rho correlation coefficient; CI: confidence interval; CCC : Lin’s concordance correlation coefficients. Acknowledgements The authors would like to express sincere appreciation to Mirella Santaniello and Maria Elena Morgoglione for their technical collaboration. Authors’ contributions Conceived, designed and coordinated the study: GC, LR, CC and NR. Per- formed sampling and laboratory analyses: AA, AB, ED, MG, AL, MPM, AC. All authors contributed to data analysis and interpretation, and preparation of the manuscript. All authors read and approved the final manuscript. Funding This study was performed within the Italy/France Galileo Project (2016). This article is based upon work from COST Action COMBAR CA16230, supported by COST (European Cooperation in Science and Technology). Availability of data and materials All data generated or analysed during this study are included in this published article. The datasets used and/or analysed during the present study available from the corresponding author upon reasonable request. Ethics approval and consent to participate We obtained verbal informed consent from the owners of animals to collect the faecal samples. Consent for publication Not applicable. Competing interests The Mini-FLOTAC technique was developed and is patented by GC, but the patent has been handed over to the University of Naples ‘Federico II’. The fact that GC is the current patent holder of the Mini-FLOTAC and Fill-FLOTAC had no role in the preparation and submission of the protocols reported or the design and implementation of ongoing and future studies. To obtain Mini- FLOTAC or Fill-FLOTAC devices, a contribution is required that is used only to cover costs of production and packaging, and to contribute to the ongoing FLOTAC research. The remaining authors declare that they have no competing interests. Author details 1 Department of Veterinary Medicine and Animal Production, University of Napoli Federico II, CREMOPAR, Napoli, Italy. 2 BIOEPAR, INRA, Oniris, 44307 Nantes, France. 3 Kreavet, Hendrik Mertensstraat 17, 9150 Kruibeke, Belgium. Received: 27 February 2019 Accepted: 6 July 2019 References 1. Charlier J, Thamsborg SM, Bartley DJ, Skuce PJ, Kenyon F, Geurden T, et al. Mind the gaps in research on the control of gastrointestinal nematodes of farmed ruminants and pigs. Transbound Emerg Dis. 2018;65(Suppl. 1):217–34. 2. Vercruysse J, Charlier J, Van Dijk J, Morgan ER, Geary T, von Samson- Himmelstjerna G, et al. Control of helminth ruminant infections by 2030. Parasitology. 2018;145:1655–64. 3. Charlier J, Morgan ER, Rinaldi L, van Dijk J, Demeler J, Höglund J, et al. Practices to optimize gastrointestinal nematode control on sheep, goat and cattle farms in Europe using targeted (selective) treatments. Vet Rec. 2014;175:250–5. 4. Kenyon F, Hutchings F, Morgan-Davies C, van Dijk J, Bartley DJ. Worm control in livestock: bringing science to the field. Trends Parasitol. 2017;33:669–77. 5. Rose H, Rinaldi L, Bosco A, Mavrot F, de Waal T, Skuce P, et al. Widespread anthelmintic resistance in European farmed ruminants: a systematic review. Vet Record. 2015;176:546. 6. Coles GC, Bauer C, Borgsteede FH, Geerts S, Klei TR, Taylor MA, et al. World Association for the Advancement of Veterinary Parasitology (W.A.A.V.P.) methods for the detection of anthelmintic resistance in nematodes of veterinary importance. Vet Parasitol. 1992;44:35–44. 7. Cabaret J. Pro and cons of targeted selective treatment against digestive- tract strongyles of ruminants. Parasite. 2008;15:506–9. 8. Baldock FC, Lyndal-Murphy M, Pearse B. An assessment of a composite sampling method for counting strongyle eggs in sheep faeces. Aust Vet J. 1990;67:165–7. 9. Nicholls J, Obendorf DL. Application of a composite faecal egg count procedure in diagnostic parasitology. Vet Parasitol. 1994;52:337–42. 10. Morgan ER, Cavill L, Curry GE, Wood RM, Mitchell ES. Effects of aggrega- tion and sample size on composite faecal egg counts in sheep. Vet Parasitol. 2005;131:79–87. 11. Kenyon F, Rinaldi L, McBean D, Pepe P, Bosco A, Melville L, et al. Pooling sheep faecal samples for the assessment of anthelmintic drug efficacy using McMaster and Mini-FLOTAC in gastrointestinal strongyle and Nematodirus infection. Vet Parasitol. 2016;225:53–60. 12. Rinaldi L, Levecke B, Bosco A, Ianniello D, Pepe P, Charlier J, et al. Compari- son of individual and pooled faecal samples in sheep for the assessment of gastrointestinal strongyle infection intensity and anthelmintic drug efficacy using McMaster and Mini-FLOTAC. Vet Parasitol. 2014;205:216–23. 13. Ward MP, Lyndal-Murphy M, Baldock FC. Evaluation of a composite method for counting helminth eggs in cattle faeces. Vet Parasitol. 1997;73:181–7. 14. George MM, Paras KL, Howell SB, Kaplan RM. Utilization of composite fecal samples for detection of anthelmintic resistance in gastrointestinal nematodes of cattle. Vet Parasitol. 2017;240:24–9. 15. Scare JA, Slusarewicz P, Noel ML, Wielgus KM, Nielsen MK. Evaluation of accuracy and precision of a smartphone based automated parasite Rinaldi et al. Parasites Vectors (2019) 12:353 Page 8 of 8 egg counting system in comparison to the McMaster and Mini-FLOTAC methods. Vet Parasitol. 2017;247:85–92. 16. Slusarewicz P, Pagano S, Mills C, Popa G, Chow KM, Mendenhall M, et al. 21. McBride GB. A proposal for strength-of-agreement criteria for Lin’s concordance correlation coefficient. Hamilton: National Institute of Water and Atmospheric Research; 2005. p. 10. Automated parasite faecal egg counting using fluorescence labelling, smartphone image capture and computational image analysis. Int J Parasitol. 2016;46:485–93. 22. Paras KL, George MM, Vidyashankar AN, Kaplan RM. Comparison of fecal egg counting methods in four livestock species. Vet Parasitol. 2018;257:21–7. 17. Bogoch II, Andrews JR, Speich B, Utzinger J, Ame SM, Ali SM, et al. Mobile 23. Geurden T, Chartier C, Fanke J, di Regalbono AF, Traversa D, von Samson- phone microscopy for the diagnosis of soil-transmitted helminth infec- tions: a proof-of-concept study. Am J Trop Med Hyg. 2013;88:626–9. 18. Cringoli G, Maurelli MP, Levecke B, Bosco A, Vercruysse J, Utzinger J, et al. The Mini-FLOTAC technique for the diagnosis of helminth and protozoan infections in humans and animals. Nat Protoc. 2017;12:1723–32. 19. Ministry of Agriculture, Fisheries and Food (MAFF). Manual of veterinary parasitological techniques. 3rd ed. London: Her Majesty’s Stationary Office; 1986. p. 160. 20. van Wyk JA, Mayhew E. Morphological identification of parasitic nema- tode infective larvae of small ruminants and cattle: a practical lab guide. Onderstepoort J Vet Res. 2013;80:539. Himmelstjerna G, et al. Anthelmintic resistance to ivermectin and mox- idectin in gastrointestinal nematodes of cattle in Europe. Int J Parasitol Drugs Drug Resist. 2015;5:163–71. 24. Levecke B, Kaplan BM, Thamsborg SM, Torgerson PR, Vercruysse J, Dobson RJ. How to improve the standardization and the diagnostic performance of the fecal egg count reduction test? Vet Parasitol. 2018;253:71–8. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your research ? Choose BMC and benefit from:
10.1186_s40168-023-01493-2	Hauer et al. Microbiome (2023) 11:106 https://doi.org/10.1186/s40168-023-01493-2 RESEARCH Microbiome Open Access Geography, not lifestyle, explains the population structure of free-living and host-associated deep-sea hydrothermal vent snail symbionts Michelle A. Hauer1, Corinna Breusing1, Elizabeth Trembath‑Reichert2, Julie A. Huber3 and Roxanne A. Beinart1* Abstract Background Marine symbioses are predominantly established through horizontal acquisition of microbial symbionts from the environment. However, genetic and functional comparisons of free‑living populations of symbionts to their host‑associated counterparts are sparse. Here, we assembled the first genomes of the chemoautotrophic gammapro‑ teobacterial symbionts affiliated with the deep‑sea snail Alviniconcha hessleri from two separate hydrothermal vent fields of the Mariana Back‑Arc Basin. We used phylogenomic and population genomic methods to assess sequence and gene content variation between free‑living and host‑associated symbionts. Results Our phylogenomic analyses show that the free‑living and host‑associated symbionts of A. hessleri from both vent fields are populations of monophyletic strains from a single species. Furthermore, genetic structure and gene content analyses indicate that these symbiont populations are differentiated by vent field rather than by lifestyle. Conclusion Together, this work suggests that, despite the potential influence of host‑mediated acquisition and release processes on horizontally transmitted symbionts, geographic isolation and/or adaptation to local habitat con‑ ditions are important determinants of symbiont population structure and intra‑host composition. Keywords Symbiosis, Hydrothermal vents, Population genomics, Alviniconcha, Mariana Back‑Arc, Microbial biogeography Introduction Mutualistic animal-microbe associations are globally sig- nificant phenomena, shaping the ecology and evolution of both host animals and microbial symbionts [1]. These *Correspondence: Roxanne A. Beinart rbeinart@uri.edu 1 Graduate School of Oceanography, University of Rhode Island, Narragansett, RI, USA 2 School of Earth and Space Exploration, Arizona State University, Tempe, AZ, USA 3 Department of Marine Chemistry and Geochemistry, Woods Hole Oceanographic Institution, Woods Hole, Falmouth, MA, USA symbiotic associations are maintained by transmission of symbionts from host parent to progeny either [1] directly, for example via the germline (vertical transmission), [2] indirectly, for example through an environmental popu- lation of symbionts (hereafter referred to as “free-living” symbionts) (horizontal transmission), or [3] via a com- bination of both vertical and horizontal transmission (mixed mode transmission) [2]. Horizontal transmission is more commonly found in aquatic than terrestrial habitats, likely due to the ease with which microbes can be transported in water com- pared to air or soil [3]. However, even for marine symbi- oses where horizontally transmitted microbial symbionts are observed in the environment [4], it is not yet clear © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Hauer et al. Microbiome (2023) 11:106 Page 2 of 12 whether free-living, environmental populations of symbi- onts represent host-associated populations at the strain level, or whether their diversity and composition differs. Free-living symbiont populations may be shaped by local environmental conditions as well as the dynamic inter- actions with their host—for example, host animals may “seed” the environment by the release of their symbi- onts into the water column only upon host death [5] or via continuous release from live adults [6, 7]. In addition, ecological and evolutionary processes, such as dispersal barriers, natural selection, and genetic drift, can contrib- ute to the diversity and biogeography of environmental symbionts [8, 9]. Deep-sea hydrothermal vents are discontinuous, island-like habitats dominated by vent-endemic inver- tebrates that host primarily horizontally transmitted chemoautotrophic bacterial symbionts, making them opportune natural systems for understanding the bio- geography of free-living microbial symbionts. In these mutualisms, the symbiotic bacteria are either obtained during a narrow competence window in early develop- mental stages or throughout the lifetime of the host [10, 11] and are, in most cases, housed intracellularly within the host’s tissues, e.g., gill or trophosome. The symbionts oxidize chemical reductants (e.g., H2S, H2, CH4) in vent- ing fluids to generate energy for the production of organic matter, thereby providing the primary food source for the host in an otherwise oligotrophic deep ocean [12] and accounting for the high ecosystem productivity charac- teristic of hydrothermal vents [13–15]. Despite reliance on horizontal transmission, the major- ity of host species from hydrothermal vents affiliate with only one or two specific endosymbiont phylotypes (i.e., species or genera based on 16S rRNA gene sequence similarity) [12], possibly as a means to reduce the acqui- sition of cheaters [16]. While a significant number of studies have focused on the diversity, composition and structure of the host-associated symbiont populations (e.g., [10, 17–22]), their free-living, environmental stages remain poorly investigated [4], partly due to the difficulty of detecting low abundance free-living symbionts in envi- ronmental samples. As a consequence, few free-living symbiont studies exist. Most of these studies have so far relied on investigations of particular marker genes [4, 23]; only one used an -omics approach but was limited to a single metagenome [24]. A recent shotgun metagenomic study found putative free-living symbiont populations of the provannid snail Alviniconcha hessleri in low-temperature diffuse venting fluids at two distinct vent fields of the Mariana Back-Arc, Northwest Pacific (15.5–18° N), [25], providing a unique opportunity to compare free-living and host-associated stages of chemosynthetic symbionts at hydrothermal vents. Alviniconcha hessleri belongs to the dominant fauna at hydrothermal vents in the Mariana Back-Arc Basin, where it lives in nutritional endosymbiosis with one species of sulfur-oxidizing, environmentally acquired Gammaproteobacteria [26, 27]. Although patterns of host-symbiont phylogenetic discordance strongly sup- port a mode of horizontal transmission for the A. hessleri symbiont [26, 27], the exact dynamics of symbiont uptake and release are unknown. As an endemic species to the Mariana region, A. hessleri is currently listed as “Vul- nerable” on the International Union for Conservation of Nature Red List of Threatened Species (https:// www. iucnr edlist. org), highlighting the need to identify the fac- tors that contribute to its limited biogeographic range, including the population structure of its obligate micro- bial symbiont. In this study, we applied phylogenomic and popula- tion genomic methods to evaluate the evolutionary rela- tionships as well as the genetic and functional variation of Alviniconcha hessleri symbionts based on lifestyle by comparing free-living and host-associated symbiont pop- ulations collected from the same habitats. In addition, we addressed the effect of geography by comparing popula- tions of both host-associated and free-living symbionts between vent fields of the northern and central Mariana Back-Arc Basin that are approximately 300 km apart and differ notably in their geochemistry: the central vent sites are known to support both low-temperature diffuse flow and black smokers that emit high-temperature fluids, with high amounts of hydrogen sulfide (H2S), whereas the northern sites only harbor diffuse flow habitats with lower concentrations of H2S [25]. Methods Host‑associated symbiont collection, sequencing, and genome assemblies Three A. hessleri specimens each were collected from snail beds at the Illium vent field (3582 m) and the Voo- doo Crater-2 (VC2) location within the Hafa Adai vent field (3277 m) in the Mariana Back-Arc Basin (Fig. 1) using the remotely operated vehicle SuBastian on board the R/V Falkor in 2016. Symbiont-bearing snail gill tis- sues were dissected and stored in RNALater™ (Thermo Fisher Scientific, Inc.) at − 80 °C until DNA extraction with the Zymo Quick DNA 96 Plus and ZR-96 Clean-up kits (Zymo Research, Inc.). High-throughput Illumina 150-bp paired-end libraries for all six samples were pre- pared and sequenced by Novogene Corporation, Inc. (Beijing, China) with an average yield of 58 million total Illumina reads (Supplementary Table 1). In addition, one gill sample from each vent field—Hafa Adai 172 (VC2) and Illium 13—was selected for long-read Nanopore sequencing on 2–3 MinION flow cells (Oxford Nanopore Hauer et al. Microbiome (2023) 11:106 Page 3 of 12 Fig. 1 Map of the Mariana Back‑Arc Basin adapted from ref. [25], indicating the locations of the two main hydrothermal vent fields sampled in this study, Illium and Hafa Adai. Free‑living symbiont samples were obtained at low coverage from two additional vent sites investigated in ref. [25], Alice Springs and Burke. Colors associated with vent fields are used consistently across samples. A light cyan and teal are used for Hafa Adai, representing Voodoo Crater 1 and 2, respectively Technologies, Oxford, UK) using the SQK-LSK109 liga- tion kit (Supplementary Table 1). Financial constraints prevented the ability to perform long-read sequencing on all host-associated samples. Raw Illumina reads were trimmed with Trimmo- matic v0.36 [28], and common sequence contaminants were removed by mapping against the PhiX and human (GRCh38) reference genomes. Nanopore reads were base-called with Albacore (Oxford Nanopore Technolo- gies, Oxford, UK) and adapter-clipped with Porechop (https:// github. com/ rrwick/ Porec hop). For Illium, symbi- ont Illumina and Nanopore reads were extracted through mapping against a draft co-assembly constructed with MEGAHIT [29] and reassembled with SPAdes v3.13.1 [30] using kmers between 21 and 91 in increments of 10. The Hafa Adai 172 MEGAHIT assembly was low quality; Hauer et al. Microbiome (2023) 11:106 Page 4 of 12 therefore, for Hafa Adai 172, all trimmed Illumina and Nanopore reads were assembled with metaSPAdes v3.13.1 [31] using the same parameters. Manual bin- ning of the assemblies was conducted with GBTools [32], and contigs < 200 bp and < 500 bp were excluded from the Illium and Hafa Adai-VC2 assemblies, respectively. Assemblies were then scaffolded with SSPACE-Standard v3.0 [33] and SLR [34] and gapfilled with GapFiller v1-10 [35] and LR_GAPCLOSER [36]. Final assemblies were polished with Pilon [37]. Trimmed Illumina reads for the remaining samples—Illium 11, Illium 17, Hafa Adai 60, and Hafa Adai 64—were assembled with metaSPades v3.13.1 as described above and binned with MaxBin [38]. All metagenome-assembled genomes (MAGs) were qual- ity checked with Quast v5.0.2 [39] and CheckM v1.0.18 [40] and taxonomically assigned through the Genome Taxonomy Database toolkit [41]. Free‑living symbiont collection, sequencing, and genome assembly All sequences from free-living samples used here were retrieved from a previous study [25], including two high- quality MAGs of environmental A. hessleri symbionts from the Illium (GCA_003972985.1) and Hafa Adai-VC2 (GCA_003973075.1) vent sites. The fluid samples from which these MAGs were assembled were collected in direct vicinity of snail beds where the A. hessleri speci- mens for host-associated analyses were obtained [25]. We further included a third previously assembled free- living symbiont MAG from diffuse venting fluids at the Voodoo Crater-1 (VC1) (GCA_003973045.1) location within the Hafa Adai vent field, ~ 5 m from Hafa Adai- VC2. Though symbiont MAGs were not previously able to be assembled from the other vent sites sampled in ref. [25]—Burke, Alice Springs, Perseverance, and Hafa Adai- Alba—we attempted again to retrieve symbiont MAGs from these samples by assembling and binning the raw reads from these sites with our methods described above, but did not produce usable symbiont MAGs. All details of the hydrothermal fluid collection, sam- ple storage, sample processing, sequencing, assembly, and binning of metagenome-assembled genomes can be found in ref. [25]. Information about raw sequencing reads is provided in Supplementary Table 1. Genome similarity and phylogenomic analyses To confirm that all symbiont MAGs belong to the same bacterial species, we calculated average nucleotide iden- tities (ANIs) via FastANI [42]. A phylogenomic tree that included the six host-associated and the three free-living symbiont MAGs as well as reference genomes of other chemosynthetic Gammaproteobacteria (Supplementary Table 2) was then constructed with IQ-TREE2 [43] based on 70 single-copy core genes in the Bacteria_71 collec- tion [44]. Parameter choice for phylogenomic recon- structions followed ref. [18]. Population structure and gene content analysis To determine symbiont population structure according to geography and lifestyle, we inferred DNA sequence polymorphisms in the free-living and host-associated samples by mapping metagenomic reads to a pangenome created with Panaroo [45] from all nine symbiont MAGs. Variants were called and filtered following the pipeline in ref. [18]. All samples from Illium and Hafa Adai met our minimum 10 × coverage threshold. Free-living sam- ples from Burke and Alice Springs mapped at 5.9 × and 3.9 × coverage, respectively. The population structure and gene content analyses (see below) were repeated for these lower-coverage samples. All other free-living metagen- omic samples from the remaining vent sites collected in ref. [25] (i.e., Perseverance, Hafa Adai-Alba) had an insuf- ficient number of reads mapped for further analyses. Principal coordinate analysis (PCoA) plots were created based on nucleotide counts converted to Bray–Curtis dissimilarities with the ggplot2 [46] and vegan [47] pack- ages in Rstudio [48]. To quantify the qualitative variant calling results depicted in the PCoAs, fixation indices (FST) between individual metagenomic samples (wherein each individual gill metagenome was treated as a popu- lation) were calculated following ref. [49] and plotted with pheatmap [50]. The method from ref. [49] as well as scikit-allel (https:// github. com/ cggh/ scikit- allel) were further used to calculate pairwise FST values between samples pooled by lifestyle or vent field. Gene content variation among symbiont populations was determined via Pangenome-based Phylogenomic Analysis (PanPhlAn) [51] and visualized through a PCoA plot based on the Jaccard Similarity Coefficient. Genes that were uniquely associated with lifestyle and vent field, respectively, were extracted from the PanPhlAn gene presence/absence matrix. Functional predictions for these genes were either obtained from the Prokka [52] annotations created during pangenome construction or inferred by blasting the respective protein sequences against the NR database. Hypothetical and unknown pro- teins were further annotated via KEGG [53] and Alpha- fold [54]. Differences in gene content between symbiont populations were visualized through Likert plots with the HH package [55] in RStudio. Validation of free‑living symbiont populations To gain confidence that the symbionts detected in our environmental samples represented truly “free-living” symbiont stages as opposed to symbionts associated with host larvae or shed gill cells, we calculated the ratio of Hauer et al. Microbiome (2023) 11:106 Page 5 of 12 symbiont 16S rRNA genes to host mitochondrial CO1 genes in all nine samples by mapping raw metagen- omic reads from the snail gills to custom-generated Alviniconcha symbiont 16S rRNA and host mtCO1 gene databases. To account for false positive mappings, we created additional background databases consisting of select bacterial (SUP05 clade bacteria, Thiomicrospira, and Marinomonas) and mollusk gene sequences. Bacte- rial 16S rRNA genes were downloaded from SILVA [56], while all Alviniconcha and mollusk mtCO1 genes were downloaded from BOLD [57]. BBSplit (https:// sourc eforge. net/ proje cts/ bbmap/) was then used to sepa- rate Alviniconcha symbiont and host reads based on the taxon-specific and background 16S rRNA and CO1 gene databases. Results Free‑living and host‑associated symbionts belong to the same bacterial species Our analysis included nine A. hessleri symbiont MAGs from the Illium and Hafa Adai vent fields: six host-asso- ciated symbiont genomes assembled in this study, and three previously published, free-living symbiont genomes from the diffuse venting fluids around A. hessleri beds [25] (Supplementary Table 3). All host-associated and two of the three free-living MAGs were of very high qual- ity, with > 90% completeness and < 3% contamination. The third free-living MAG, Hafa Adai-VC1, had a medium quality (~ 67% completeness). ANI values between all MAGs were > 97.7% (Supplementary Table 4), suggesting that the nine A. hessleri symbiont genomes belong to the same bacterial species [42] within the genus Thiolapillus based on the Genome Taxonomy Database, and confirm- ing that the previously assembled free-living symbiont genomes were indeed A. hessleri symbionts (Supplemen- tary Table 3). Corroborating the ANI results, the nine A. hessleri MAGs were monophyletic in our phylogenomic analysis relative to the gammaproteobacterial symbionts of other vent invertebrates (Fig. 2). In agreement with phylogenetic analyses of the 16S rRNA gene [27], the nearest neighbors of the A. hessleri symbionts were the Ifremeria nautilei SOX symbiont, as well as Thiolapillus brandeum, a microbe not known to be symbiotic [58]. Environmental samples contain free‑living symbiont populations To investigate whether the symbionts observed in the dif- fuse flow samples were true free-living symbionts rather than symbionts associated with A. hessleri larvae or shed gill tissue, we calculated the ratio of symbiont 16S rRNA gene to host mitochondrial CO1 gene reads in all nine environmental and host-associated samples (Table 1). If the environmental symbiont samples were associated with larvae or host tissue debris/cells, we expect the ratio in the environmental and host-associated samples to be similar to one another. However, the 16S rRNA:mtCO1 ratio was consistently orders of magnitude higher in environmental samples than in host-associated samples, indicating the presence of a population of symbiont cells independent from host tissue. This finding provides evi- dence that our environmental samples include truly free- living A. hessleri symbiont populations. A. hessleri symbiont populations are structured primarily by vent field, not lifestyle Our genome assemblies from both host tissue and dif- fuse vent fluids likely represent the dominant symbiont strain in each sample, but do not reveal the full extent of strain-level population variation between samples. To determine whether A. hessleri symbionts form subpopu- lations consistent with geography or lifestyle, we created a pangenome out of the individual symbiont MAGs from the Illium and Hafa Adai vent fields that we used as ref- erence for variant calling (Supplementary Tables 5, 6). Our variant detection method resulted in 2177 sequence polymorphisms for investigation of population genomic structure based on FST and ordination analyses (Fig. 3). FST values were calculated pairwise between all nine populations (Fig. 3a), as well as between samples pooled by lifestyle and vent field. FST values range from 0 to 1, where an FST value of 1 indicates that samples form genetically isolated subpopulations, while an FST value of 0 indicates that the samples form a single, well-mixed population. Between individual samples, pairwise FSTs showed a moderate (0.2–0.5) to strong (> 0.5) differen- tiation, indicating that all samples represent distinct sub- populations with limited genetic exchange among each other (Fig. 3a, Supplementary Table 7). Genetic isolation among individual samples was typically stronger between (0.54–0.76) than within (0.21–0.46) vent fields (i.e., Illium vs Hafa Adai-VC2). Within vent sites, the degree of dif- ferentiation was comparable among samples independ- ent of lifestyle at Illium, while host-associated samples were more similar to one another than to free-living samples at Hafa Adai-VC2. When samples were pooled, overall pairwise FST values were markedly higher by vent field (0.47 ± 0.03 s.d.) than by lifestyle (0.05 ± 0.01 s.d.). The dominant effect of geography on symbiont popu- lation structure was supported by PCoAs where both free-living and host-associated samples from Illium clus- tered distinctly from Hafa Adai (VC1 and VC2) (Fig. 3b). Despite the fact that Hafa Adai-VC1 and -VC2 differ spa- tially by only ~ 5 m, the free-living VC1 sample formed its own distinct subpopulation from both host-associ- ated and free-living populations at VC2 (FST 0.58–0.67), Hauer et al. Microbiome (2023) 11:106 Page 6 of 12 Fig. 2 Cladogram branch transformed phylogenomic tree of chemosynthetic Gammaproteobacteria based on 70 single‑copy core genes. Genome accession numbers for all genomes included in the phylogenomic analysis can be found in Supplementary Table 1. A. hessleri symbionts are colored by vent field (Hafa Adai VC2: teal, Hafa Adai VC1: light cyan, Illium: yellow) with dots and triangles indicating free‑living or host‑associated lifestyle. Candidatus Pseudothioglobus singularis was used as outgroup for tree rooting suggesting very fine-scale geographic or environmental structuring. locations without evidence for (Supplementary Table 10). isolation-by-distance These patterns were consistent in analyses based on 1271 and 793 variant sites that included the free-living, low-coverage symbiont samples from Burke and Alice Springs, respectively (Supplementary Fig. 1, 4; Supple- mentary Tables 8, 9). Burke represented the most diver- gent population, reaching FST values > 0.8 in all pairwise comparisons. Although Alice Springs clustered closely with free-living and host-associated symbionts from Illium in the PCoAs, FST values indicated a high degree of genetic isolation for this population (FST > 0.7). Analy- ses with samples pooled by vent field confirmed patterns of strong genetic differentiation between geographic A. hessleri symbiont gene content differs by vent field, not lifestyle Gene content variation between symbiont populations was assessed based on lifestyle and geography. Similar to the population structure analyses, PCoA plots based on gene content variation across all nine host-associated and free- living populations revealed clustering by vent field but not by lifestyle: symbiont populations from Hafa Adai-VC1 and Hafa Adai-VC2 were more similar to one another than to Illium (Fig. 4), although Hafa Adai-VC1 clustered as an independent population from all other samples. Hauer et al. Microbiome (2023) 11:106 Page 7 of 12 Table 1 Percentage of unambiguous reads in environmental samples mapping to Alviniconcha symbiont 16S rRNA and mitochondrial CO1 genes, as well as the ratio of 16S rRNA to mtCO1 Sample Alviniconcha 16S rRNA Alviniconcha mtCO1 16S rRNA:CO1 Origin Host‑associated Hafa Adai 172 Host‑associated Hafa Adai 60 Host‑associated Hafa Adai 64 Host‑associated Illium 11 Host‑associated Illium 13 Host‑associated Illium 17 Free‑living Illium Free‑living Hafa Adai‑ VC2 Free‑living Hafa Adai‑ VC1 0.019% 0.016% 0.017% 0.011% 0.020% 0.014% 0.043% 0.084% 0.065% 0.0032% 0.0111% 0.0102% 0.0084% 0.0080% 0.0072% 0.0022% 0.0019% 0.0003% 5.83 Gill tissue 1.46 Gill tissue 1.67 Gill tissue 1.32 Gill tissue 2.50 Gill tissue 1.99 Gill tissue 19.73 Diffuse fluids 43.54 Diffuse fluids 241.70 Diffuse fluids Fig. 3 Symbiont population structure based on single nucleotide polymorphisms. A Heatmap of genome‑wide, pairwise fixation indices (FST) created using the pheatmap package in RStudio. FST values range from 0 to 1, where a value of 0 indicates no genetic differentiation, while a value of 1 indicates complete isolation among populations. “FL” and “HA” indicate free‑living and host‑associated symbionts, respectively. B PCoA plot based on Bray–Curtis distances illustrating the population structure of A. hessleri symbionts of different lifestyles and vent fields Gene content differed more substantially by geography than by lifestyle: the Illium symbionts had 44 unique gene clusters, and the Hafa Adai (VC1 and VC2) symbionts had 26 (Fig. 5a, Supplementary Table 11), while only three total gene clusters were unique by lifestyle (group_681 for host-associated; group_2104 and group_2131 for free- living). However, these genes could not be character- ized by any database we used for functional annotations. For all unique gene clusters across both biogeography and lifestyle, hypothetical and unknown proteins based on Prokka and the NR database were also assessed via KEGG and Alphafold, but yielded low-confidence results. Of the successfully annotated genes unique to the Illium symbionts, most were predicted to be involved in the mobilome and DNA metabolism, followed by membrane transport, virulence, disease, defense; RNA metabo- lism; sulfur metabolism; cell signaling and regulation; conjugation; iron metabolism; glycolysis and gluconeo- genesis; and detoxification and stress response. Genes unique to the Hafa Adai (VC1 & VC2) symbionts were predominantly associated with the mobilome, followed by membrane transport; RNA metabolism; motility and chemotaxis; DNA metabolism; virulence, disease and defense; and glycolysis and gluconeogenesis. Given the small-scale geographic structuring found between VC1 and VC2 at Hafa Adai, and given that VC2 has a larger sample size to represent its subpopulation, we also compared the unique genes between Illium and Hauer et al. Microbiome (2023) 11:106 Page 8 of 12 Fig. 4 PCoA plot based on Jaccard distances illustrating the difference in gene content between A. hessleri symbionts based on both lifestyle and vent field Hafa Adai VC2 symbionts alone (i.e., without VC1) (Sup- plementary Table 12, Fig. 5b). In this case, there were 62 unique gene clusters for symbionts from Illium and 28 unique gene clusters for symbionts from Hafa Adai- VC2 (Fig. 5b), i.e., two additional as compared to VC-1 and VC-2 combined. Only one of the genes unique to the Hafa Adai symbionts could be annotated and fell under the larger subcategory of “Virulence, Disease and Defense,” whereas unique genes of the Illium symbionts spanned a variety of metabolic functions. Analyses that included symbiont reads from Alice Springs and Burke (Supplementary Figs. 2, 3, 5; Supple- mentary Tables 13, 14) further supported the effect of geography over lifestyle on gene content variation in the A. hessleri symbionts. The population at Burke harbored a single unique, uncharacterized gene (Supplementary Table 13). When pooled with Illium as a “northern site,” additional genes unique to DNA metabolism and mem- brane transport were found, followed by genes involved in the mobilome, RNA metabolism, virulence, glycoly- sis and gluconeogenesis, cell signaling, conjugation, and stress response (Supplementary Fig. 3; Supplementary Table 13). Alice Springs harbored three uncharacterized or hypothetical genes. When all three northern sites (Alice Springs, Illium, and Burke) were pooled together, seven unique genes were found. Four of these were related to DNA metabolism, virulence, conjugation, and transposition (Supplementary Table 14). Since Alice Springs and Illium are more geochemically similar to one another than either vent is to Burke [25], we also inves- tigated the unique genes shared by these two vent fields alone: four unique genes were found, one of which fell under the functional category of virulence, disease, and defense. Discussion Here, we compared free-living and host-associated sym- biont populations of Alviniconcha hessleri from two vent fields in the Mariana Back-Arc. Based on ANI and taxo- nomic assignments, our nine representative, medium- to high-quality MAGs can be considered to represent a sin- gle species within the genus Thiolapillus [58]. Our results provide strong evidence that diffuse fluid flow microbial communities include populations of free-living symbi- onts, further supporting an expected model of horizontal transmission in Alviniconcha species [18, 59]. Both population structure and gene content analy- ses suggest that A. hessleri symbionts form subpopula- tions that segregate by geography more strongly than by lifestyle. These patterns agree with previous studies of non-symbiotic hydrothermal vent microbial communi- ties, which show that microbes are shaped by their local environment [60], as well as of host-associated A. hes- sleri symbiont biogeography at the 16S rRNA gene level [27] and other horizontally transmitted associations from Hauer et al. Microbiome (2023) 11:106 Page 9 of 12 Fig. 5 Likert plots showing A the number of unique genes between Illium and Hafa Adai (including both VC1 and VC2), and B the number of unique genes between Illium and Hafa Adai (VC2 only) hydrothermal vents, such as bathymodiolin mussels [49, 61, 62] and provannid snails [18], that have been shown to partner with habitat-specific symbiont strains. These results, therefore, provide further evidence for horizon- tal transmission in the A. hessleri symbiont system. Such uptake of environmental symbiont strains bears a risk of infection to the host by cheaters [16], but also enhances an animal’s ability to flexibly associate with locally availa- ble symbiont strains and, therefore, to maximize the hab- itat range in which they can settle [2, 18]. Furthermore, since hydrothermal vents are ephemeral and geochemi- cally dynamic habitats that harbor microbial communi- ties shaped by local environmental conditions [60], it may be ecologically and evolutionarily advantageous for vent animals to acquire symbiont strains that are likely locally adapted [63]. The dynamics of microbial interaction with the host during acquisition and release processes can have sig- nificant impacts on the population structure and com- position of horizontally transmitted symbionts. It is not known whether A. hessleri can replenish or recycle its symbionts, or if symbiont acquisition occurs only once upon settlement. For example, hydrothermal vent tube- worms seed the environment with their symbionts only upon death [5], Bathymodiolus mussels can acquire and release their symbionts throughout their lifetime [11, 64], and Vibrio fischeri symbionts are expelled every morning by their sepiolid squid host [65]. In V. fischeri, it is well established that evolution in the free-living stage—for example, via horizontal gene transfer—impacts the evo- lution of host-microbe interactions, though the role of novel mutations remains unclear [65]. Although A. hes- sleri symbionts were overall more strongly partitioned by geography than by lifestyle, all symbiont samples were genetically distinct from each other and formed separate free-living or host-associated subpopulations. These find- ings suggest that symbiont exchanges between host and environment throughout the lifetime of the host are lim- ited but might occur occasionally via symbiont uptake or release [49], thereby leading to mixing of host-associated and free-living symbiont pools. Periodic switching of symbiont strains could increase shared genetic variation among intra- and extra-host-symbiont populations, while maintaining geographic differentiation in the presence of dispersal barriers and/or environmental selection. All samples from Illium showed a comparably small degree of differentiation from each other, while samples from Hafa Adai were notably divergent between free-living and host-associated lifestyles. These patterns could arise from differences in the sampling locations of the free- living symbiont populations (e.g., distance from the snail beds) and/or the age of the Alviniconcha host individu- als. Although we do not have size-related data for the col- lected specimens, it is possible that the snail individuals from Hafa Adai were older than those from Illium, giv- ing host-associated symbiont populations more time to diverge from their free-living counterparts. Strong genetic differentiation between host-associated and free- living symbiont populations can be expected if hosts take up similar symbiont strains that have limited exchange with the environment post-infection, while the free-living symbiont population experiences more turnover. The high genetic isolation of symbiont populations observed between vent fields may reflect the influence of both neutral (e.g., dispersal barriers and isolation-by-dis- tance) and selective processes (e.g., adaptation to habitat differences between vent fields) on symbiont biogeog- raphy. Illium, Burke, and Alice Springs are all northern vent fields within the Mariana Back-Arc Basin that are characterized by sites of low-temperature diffuse fluid flow, while Hafa Adai is located further south and con- tains high-temperature black smokers [25]. Illium and Alice Springs are similar geochemically, notably in that they are both low in H2S concentrations, while Burke and Hafa Adai exhibit elevated H2S concentrations [25]. The close proximity (~ 360 m) and overlap in geochemi- cal characteristics between Alice Springs and Illium may explain why these vent fields clustered together in our population structure analyses. By contrast, Burke’s dis- tinct geochemical signature might contribute to the high Hauer et al. Microbiome (2023) 11:106 Page 10 of 12 genetic isolation seen for this vent field, despite its rela- tive proximity to Alice Springs and Illium (~ 4 km). Over- all, however, no clear pattern of isolation-by-distance was observed, indicating that ecological factors might play a more important role than dispersal barriers in shaping symbiont population structure, in agreement with the oceanographic connectivity between the northern and central Mariana Back-Arc Basin [66]. Interestingly, Hafa Adai-VC1—while more similar to Hafa Adai-VC2 than any other vent site—represented its own symbiont subpopulation, suggesting small-scale population structuring of symbionts within vent fields. Local patchiness of symbionts, as observed in our study, mirrors patterns found for host-associated symbionts of cold-seep vestimentiferan tubeworms [67] and Acropora corals [68]. Although Hafa Adai-VC1 and -VC2 were only ~ 5 m apart, it is possible that Alviniconcha hes- sleri symbionts have extremely low dispersal potential that could be further reduced by small-scale circulation within vent sites due to physical structuring in the sub- seafloor [69, 70]. Alternatively, micro-niche adaptation driven by locally fluctuating environmental conditions might contribute to these patterns. Among the identified differences in gene content, sym- bionts from Illium uniquely harbored genes related to iron and sulfur metabolism. As iron and sulfur concen- trations appear to be reduced at northern Mariana Back- Arc vents [25, 71] and are typically lower in diffuse flow than black smoker fluids such as those found at Hafa Adai, it is possible that symbionts at Illium harbor high affinity sulfur and Fe2+ transporters to efficiently obtain this essential element for their metabolism. All symbiont populations, including the low-coverage samples from Alice Springs and Burke, showed differences in the pres- ence of genes related to the mobilome and virulence, disease, and defense. This suggests that each vent field supports distinct viral communities that may uniquely infect and interact with the symbionts, as hydrothermal vent viruses have restricted bacterial and archaeal host ranges, and viral communities are typically endemic to a given vent site due to limited dispersal or environmen- tal selection [72, 73]. The high number of unique genes related to the mobilome may be a consequence of inte- grated phage-derived genetic material that reflect the local, free-living viral communities. Conclusions Our research demonstrates that Alviniconcha hes- sleri symbiont populations are primarily structured by geography rather than by their host-associated or free- living lifestyle. Future work using population genomic approaches should help clarify the predominant force(s) shaping the geographic population structure, as recent analyses of the symbionts associated with other Alvini- concha species suggest that both genetic drift and local adaptation play a role in symbiont biogeography [18]. Although our analyses indicate a weak effect of lifestyle on symbiont genetic structure, it is possible that free- living and host-associated populations are characterized by differences in gene expression. A comparison of gene expression between lifestyles may provide additional clarity on the extent to which these symbiont subpopu- lations differ functionally. Our work also strengthens previous evidence for horizontal symbiont transmission in Alviniconcha species [18, 59], despite the fact that almost nothing is currently known about the dynam- ics of symbiont acquisition and release in these species. Given that A. hessleri has been classified as “Vulnerable” on the IUCN Red List (https:// www. iucnr edlist. org) and is a dominant species at vents that are part of the Mari- anas Trench Marine National Monument, it is critical for future conservation and management that we understand the genetic connectivity of the symbiotic microbes that support this foundation species. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s40168‑ 023‑ 01493‑2. Additional file 1. Additional file 2: Supplementary Figure 1. Population structure analyses based on 1271 single nucleotide polymorphisms. Includes all nine samples from Illium and Hafa Adai plus symbiont reads from Burke. A) PCoA plot based on Bray‑Curtis distances exhibiting population genetic structuring by geography and lifestyle. B) Heatmap of genome‑wide, pairwise fixation indices (FST) created using the pheatmap package in RStudio. FST values range from 0 to 1, where a value of 0 indicates no genetic differentiation, while a value of 1 indicates complete isolation among populations. “FL” and “HA” indicate free‑living and host‑associated symbionts, respectively. Additional file 3: Supplementary Figure 2. PCoA plot based on Jaccard distances illustrating the difference in gene content between A. hessleri symbionts based on both lifestyle and vent field, including all nine samples from Illium and Hafa Adai in addition to the free‑living symbiont population from the Burke vent field. Additional file 4: Supplementary Figure 3. Barplot showing the num‑ ber of unique gene clusters per functional category between vent fields, including a “northern” site category which combines the Illium and Burke vent fields. Created using ggplot in RStudio. Additional file 5: Supplementary Figure 4. Population structure analy‑ ses based on 793 single nucleotide polymorphisms. Includes all nine sam‑ ples from Illium and Hafa Adai, plus symbiont reads from both the Burke and Alice Springs vent fields. A) PCoA plot based on Bray‑Curtis distances exhibiting population genetic structuring by geography and lifestyle. B) Heatmap of genome‑wide, pairwise fixation indices (FST) created using the pheatmap package in RStudio. FST values range from 0 to 1, where a value of 0 indicates no genetic differentiation, while a value of 1 indicates complete isolation among populations. “FL” and “HA” indicate free‑living and host‑associated symbionts, respectively. Additional file 6: Supplementary Figure 5. PCoA plot based on Jaccard distances illustrating the difference in gene content between A. hessleri Hauer et al. Microbiome (2023) 11:106 Page 11 of 12 symbionts based on both lifestyle and vent field, including all nine sam‑ ples from Illium and Hafa Adai in addition to free‑living symbiont samples from the Burke and Alice Springs vent fields. Acknowledgements We thank the Schmidt Ocean Institute, the captain, crew and pilots of the R/V Falkor and ROV SuBastian, as well as Bill Chadwick, David Butterfield, Verena Tunnicliffe, and Amanda Bates for their support in the sample collections. The data collected in this study includes work supported by the Schmidt Ocean Institute during the R/V Falkor cruise FK161129. Authors’ contributions M.H., C.B., R.A.B., and J.H. designed the study. J.H. collected the samples. E.T.R. and R.A.B. did laboratory work for the free‑living and host‑associated metagenomic samples, respectively. M.H., C.B., and E.T.R. performed bioinfor‑ matic analyses. M.H. drafted the manuscript. All authors edited, reviewed, and approved the text. Funding This work was funded by the National Science Foundation (grant number OCE‑1736932 to RAB and Graduate Research Fellowship award number 1747454 to MAH). JAH was funded by the NOAA Ocean Exploration and Research (OER) Program and the NSF Center for Dark Energy Biosphere Investi‑ gations (C‑DEBI) (OCE‑0939564). ETR was supported by the NASA Postdoctoral Fellowship with the NASA Astrobiology Institute and the L’Oréal USA For Women in Science Fellowship. This is C‑DEBI Contribution no. 614. Availability of data and materials The datasets supporting the conclusions of this article are available in the National Center for Biotechnology Information repository, under BioProject number PRJNA763533. The previously published, free‑living raw sequenc‑ ing reads and corresponding MAGs are available at the National Center for Biotechnology Information under BioProject PRJNA454888. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Received: 20 August 2022 Accepted: 11 February 2023 References 1. McFall‑Ngai M, Hadfield MG, Bosch TCG, Carey HV, Domazet‑Lošo T, Douglas AE, et al. Animals in a bacterial world, a new imperative for the life sciences. Proc Natl Acad Sci U S A. 2013;110(9):3229–36. 2. Vrijenhoek RC. Genetics and evolution of deep‑sea chemosynthetic bacteria and their invertebrate hosts. In: Kiel S, editor. The Vent and Seep Biota: Aspects from Microbes to Ecosystems. Dordrecht, Netherlands: Springer; 2010. p. 15–49. 3. Russell SL. Transmission mode is associated with environment type and taxa across bacteria‑eukaryote symbioses: a systematic review and meta‑ analysis. FEMS Microbiol Lett. 2019;366(3):1–13. 4. Harmer TL, Rotjan RD, Nussbaumer AD, Bright M, Ng AW, DeChaine EG, et al. Free‑living tubeworm endosymbionts found at deep‑sea vents. Appl Environ Microbiol. 2008;74(12):3895–8. 5. Klose J, Polz MF, Wagner M, Schimak MP, Gollner S, Bright M. Endosymbi‑ onts escape dead hydrothermal vent tubeworms to enrich the free‑living population. Proc Natl Acad Sci. 2015;112(36):11300–5. 6. Doino JA, McFall‑Ngai MJ. A transient exposure to symbiosis‑competent bacteria induces light organ morphogenesis in the host squid. Biol Bull. 1995;189(3):347–55. 7. Nyholm SV, McFall‑Ngai MJ. Sampling the light‑organ microenviron‑ ment of Euprymna scolopes: description of a population of host cells in association with the bacterial symbiont Vibrio fischeri. Biol Bull. 1998;195(2):89–97. 8. Martiny JBH, Bohannan BJM, Brown JH, Colwell RK, Fuhrman JA, Green JL, et al. Microbial biogeography: Putting microorganisms on the map. Nat Rev Microbiol. 2006;4(2):102–12. 9. Dickey JR, Swenie RA, Turner SC, Winfrey CC, Yaffar D, Padukone A, et al. The utility of macroecological rules for microbial biogeography. Front Ecol Evol. 2021;9:633155. 10. Nussbaumer AD, Fisher CR, Bright M. Horizontal endosymbiont transmis‑ sion in hydrothermal vent tubeworms. Nature. 2006;441(7091):345–8. 11. Wentrup C, Wendeberg A, Schimak M, Borowski C, Dubilier N. For‑ ever competent: Deep‑sea bivalves are colonized by their chem‑ osynthetic symbionts throughout their lifetime. Environ Microbiol. 2014;16(12):3699–713. 12. Dubilier N, Bergin C, Lott C. Symbiotic diversity in marine animals: the art of harnessing chemosynthesis. Nat Rev Microbiol. 2008;6:725–40. 13. Olins HC, Rogers DR, Frank KL, Vidoudez C, Girguis PR. Assessing the influence of physical, geochemical and biological factors on anaerobic microbial primary productivity within hydrothermal vent chimneys. Geobiology. 2013;11(3):279–93. 14. McNichol J, Stryhanyuk H, Sylva SP, Thomas F, Musat N, Seewald JS, et al. Primary productivity below the seafloor at deep‑sea hot springs. Proc Natl Acad Sci U S A. 2018;115(26):6756–61. 15. Beinart RA, Gartman A, Sanders JG, Luther GW, Girguis PR. The uptake and excretion of partially oxidized sulfur expands the repertoire of energy resources metabolized by hydrothermal vent symbioses. Proc R Soc B Biol Sci. 2015;282:20142811. 16. Douglas AE. Conflict, cheats and the persistence of symbioses. New Phytol. 2008;177(4):849–58. 17. Lan Y, Sun J, Chen C, Sun Y, Zhou Y, Yang Y, et al. Hologenome analysis reveals dual symbiosis in the deep‑sea hydrothermal vent snail Gigan- topelta aegis. Nat Commun. 2021;12(1):1–15. 18. Breusing C, Genetti M, Russell SL, Corbett‑Detig RB, Beinart RA. Horizontal transmission enables flexible associations with locally adapted symbiont strains in deep‑sea hydrothermal vent symbioses. Proc Natl Acad Sci USA. 2022;119(14):1–11. 19. Beinart RA, Sanders JG, Faure B, Sylva SP, Lee RW, Becker EL, et al. Evidence for the role of endosymbionts in regional‑scale habitat partitioning by hydrothermal vent symbioses. Proc Natl Acad Sci U S A. 2012;109(47):1–10. 20. Beinart RA, Luo C, Konstantinidis KT, Stewart FJ, Girguis PR. The bacte‑ rial symbionts of closely related hydrothermal vent snails with distinct geochemical habitats show broad similarity in chemoautotrophic gene content. Front Microbiol. 2019;10:1818. 21. Breusing C, Johnson SB, Tunnicliffe V, Vrijenhoek RC. Population struc‑ ture and connectivity in Indo‑Pacific deep‑sea mussels of the Bathymo- diolus septemdierum complex. Conserv Genet. 2015;16(6):1415–30. 22. Hinzke T, Kleiner M, Meister M, Schlüter R, Hentschker C, Pané‑Farré J, et al. Bacterial symbiont subpopulations have different roles in a deep‑ sea symbiosis. Elife. 2021;10:1–45. 23. Ücker M. Metagenomic analyses of a deep‑sea mussel symbiosis. Dis‑ sertation, University of Bremen. 2021. https:// doi. org/ 10. 26092/ elib/ 672. 24. Polzin J, Arevalo P, Nussbaumer T, Polz MF, Bright M. Polyclonal symbi‑ ont populations in hydrothermal vent tubeworms and the environ‑ ment. Proc R Soc B Biol Sci. 2019;286:20181281. 25. Trembath‑Reichert E, Butterfield DA, Huber JA. Active subseafloor microbial communities from Mariana back‑arc venting fluids share metabolic strategies across different thermal niches and taxa. ISME J. 2019;13(9):2264–79. 26. Endow K, Ohta S. The symbiotic relationship between bacteria and a mesogastropod snail, Alviniconcha hessleri, collected from hydrother‑ mal vents of the Mariana Back‑Arc Basin. Bull Japanese Soc Microb Ecol. 1989;3(2):73–82. 27. Breusing C, Johnson SB, Tunnicliffe V, Clague DA, Vrijenhoek RC, Beinart RA. Allopatric and sympatric drivers of speciation in Alviniconcha hydrothermal vent snails. Mol Biol Evol. 2020;37(12):3469–84. Hauer et al. Microbiome (2023) 11:106 Page 12 of 12 28. Bolger AM, Lohse M, Usadel B. Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics. 2014;30(15):2114–20. 29. Li D, Liu CM, Luo R, Sadakane K, Lam TW. MEGAHIT: an ultra‑fast single‑ node solution for large and complex metagenomics assembly via succinct de Bruijn graph. Bioinformatics. 2015;31(10):1674–6. 54. Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583–9. 55. Heiberger RM, Holland B. Statistical Analysis and Data Display: An Inter‑ mediate Course with Examples in R. 2nd ed. New York: Springer; 2015. 30. Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, 56. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA et al. SPAdes: A new genome assembly algorithm and its applications to single‑cell sequencing. J Comput Biol. 2012;19(5):455–77. ribosomal RNA gene database project: improved data processing and web‑based tools. Nucleic Acids Res. 2013;41(D1):590–6. 31. Nurk S, Meleshko D, Korobeynikov A, Pevzner PA. MetaSPAdes: A new 57. Ratnasingham S, Hebert PDN. BOLD: The Barcode of Life Data System. versatile metagenomic assembler. Genome Res. 2017;27(5):824–34. Mol Ecol Notes. 2007;7(3):355–64. 32. Seah BKB, Gruber‑Vodicka HR. gbtools: Interactive visualization of metagenome bins in R. Front Microbiol. 2015;6:1451. 33. Boetzer M, Henkel CV, Jansen HJ, Butler D, Pirovano W. Scaffolding pre‑ assembled contigs using SSPACE. Bioinformatics. 2011;27(4):578–9. 34. Luo J, Lyu M, Chen R, Zhang X, Luo H, Yan C. SLR: a scaffolding algo‑ rithm based on long reads and contig classification. BMC Bioinformat‑ ics. 2019;20(1):1–11. 35 Boetzer M, Pirovano W. Toward almost closed genomes with GapFiller. Genome Biol. 2012;13(6):R56. 58. Nunoura T, Takaki Y, Kazama H, Kakuta J, Shimamura S, Makita H. Physi‑ ological and genomic features of a novel sulfur‑oxidizing Gammapro‑ teobacterium belonging to a previously uncultivated symbiotic lineage isolated from a hydrothermal vent. PLoS ONE. 2014;9(8):e104959. 59. Suzuki Y, Kojima S, Sasaki T, Suzuki M, Utsumi T, Watanabe H, et al. Host‑symbiont relationships in hydrothermal vent gastropods of the genus Alviniconcha from the Southwest Pacific. Appl Environ Microbiol. 2006;72(2):1388–93. 60. Dick GJ. The microbiomes of deep‑sea hydrothermal vents: distributed 36. Xu GC, Xu TJ, Zhu R, Zhang Y, Li SQ, Wang HW, et al. LR‑Gapcloser: A globally, shaped locally. Nat Rev Microbiol. 2019;17(5):271–83. tiling path‑based gap closer that uses long reads to complete genome assembly. Gigascience. 2018;8(1):1–14. 37. Walker BJ, Abeel T, Shea T, Priest M, Abouelliel A, Sakthikumar S, et al. Pilon: An integrated tool for comprehensive microbial variant detection and genome assembly improvement. PLoS ONE. 2014;9(11):e112963. 38. Wu Y‑W, Tang Y‑H, Tringe SG, Simmons BA, Singer SW. MaxBin: an automated binning method to recover individual genomes from metagenomes using. Microbiome. 2014;2(26):4904–9. 39. Gurevich A, Saveliev V, Vyahhi N, Tesler G. QUAST: quality assessment tool for genome assemblies. Bioinformatics. 2013;29(8):1072–5. 40. Parks DH, Imelfort M, Skennerton CT, Hugenholtz P, Tyson GW. CheckM: assessing the quality of microbial genomes recovered from isolates, single cells, and metagenomes. Genome Res. 2015;25(7):1043–55. 41. Chaumeil PA, Mussig AJ, Hugenholtz P, Parks DH. GTDB‑Tk: A toolkit to classify genomes with the genome taxonomy database. Bioinformatics. 2020;36(6):1925–7. 42. Jain C, Rodriguez‑R LM, Phillippy AM, Konstantinidis KT, Aluru S. High throughput ANI analysis of 90K prokaryotic genomes reveals clear spe‑ cies boundaries. Nat Commun. 2018;9(1):5114. 43. Minh BQ, Schmidt HA, Chernomor O, Schrempf D, Woodhams MD, Von Haeseler A, et al. IQ‑TREE 2: new models and efficient meth‑ ods for phylogenetic inference in the genomic era. Mol Biol Evol. 2020;37(5):1530–4. 44. Eren AM, Esen OC, Quince C, Vineis JH, Morrison HG, Sogin ML, et al. Anvi’o: An advanced analysis and visualization platformfor ’omics data. PeerJ. 2015;2015(10):1–29. 45. Tonkin‑Hill G, MacAlasdair N, Ruis C, Weimann A, Horesh G, Lees JA, et al. Producing polished prokaryotic pangenomes with the Panaroo pipeline. Genome Biol. 2020;21(1):1–21. 46. Villanueva RAM, Chen ZJ. ggplot2: Elegant Graphics for Data Analysis (2nd ed.). Meas Interdiscip Res Perspect. 2019;17(3):160–7. 47. Oksanen AJ, Blanchet FG, Kindt R, Legen‑ P, Minchin PR, Hara RBO, et al. vegan: Community Ecology Package. 2018. https:// CRAN.R‑ proje ct. org/ packa ge= vegan. 48. RStudio Team. RStudio: Integrated Development Environment for R. Boston, MA: RStudio, PBC; 2021. 61. Ücker M, Ansorge R, Sato Y, Sayavedra L, Breusing C, Dubilier N. Deep‑sea mussels from a hybrid zone on the Mid‑Atlantic Ridge host genetically indistinguishable symbionts. ISME J. 2021;15(10):3076–83. 62. Brzechffa C, Goffredi SK. Contrasting influences on bacterial symbiont specificity by co‑occurring deep‑sea mussels and tubeworms. Environ Microbiol Rep. 2021;13(2):104–11. 63 Vrijenhoek RC. Genetic diversity and connectivity of deep‑sea hydrother‑ mal vent metapopulations. Mol Ecol. 2010;19:4391–411. 64. Kádár E, Bettencourt R, Costa V, Santos RS, Lobo‑da‑Cunha A, Dando P. Experimentally induced endosymbiont loss and re‑acquirement in the hydrothermal vent bivalve Bathymodiolus azoricus. J Exp Mar Bio Ecol. 2005;318(1):99–110. 65. Soto W, Travisano M, Tolleson AR, Nishiguchi MK. Symbiont evolution during the free‑living phase can improve host colonization. Microbiology (United Kingdom). 2019;165(2):174–87. 66. Breusing C, Johnson SB, Mitarai S, Beinart RA, Tunnicliffe V. Differential patterns of connectivity in Western Pacific hydrothermal vent metapopu‑ lations: a comparison of biophysical and genetic models. Evol Appl. 2021. https:// doi. org/ 10. 1111/ eva. 13326. 67. Vrijenhoek RC, Duhaime M, Jones WJ. Subtype variation among bacterial endosymbionts of tubeworms (Annelida: Siboglinidae) from the Gulf of California. Biol Bull. 2007;212(3):180–4. 68. Davies SW, Moreland KN, Wham DC, Kanke MR, Matz MV. Cladocopium community divergence in two Acropora coral hosts across multiple spatial scales. Mol Ecol. 2020;29(23):4559–72. 69. Stewart LC, Algar CK, Fortunato CS, Larson BI, Vallino JJ, Huber JA, et al. Fluid geochemistry, local hydrology, and metabolic activity define methanogen community size and composition in deep‑sea hydrother‑ mal vents. ISME J. 2019;13(7):1711–21. 70. Fortunato CS, Larson B, Butterfield DA, Huber JA. Spatially distinct, 71. temporally stable microbial populations mediate biogeochemical cycling at and below the seafloor in hydrothermal vent fluids. Environ Microbiol. 2018;20(2):769–84. Ishibashi J ichiro, Tsunogai U, Toki T, Ebina N, Gamo T, Sano Y, et al. Chemical composition of hydrothermal fluids in the central and southern Mariana Trough backarc basin. Deep Res Part II Top Stud Oceanogr. 2015;121:126–136. 49. Romero Picazo D, Dagan T, Ansorge R, Petersen JM, Dubilier N, Kupczok A. Horizontally transmitted symbiont populations in deep‑sea mussels are genetically isolated. ISME J. 2019;13(12):2954–68. 72. Paez‑Espino D, Eloe‑Fadrosh EA, Pavlopoulos GA, Thomas AD, Hunte‑ mann M, Mikhailova N, et al. Uncovering Earth’s virome. Nature. 2016;536(7617):425–30. 50. Kolde R. pheatmap: Pretty Heatmaps. 2019. https:// CRAN.R‑ proje ct. org/ 73. Thomas E, Anderson RE, Li V, Rogan LJ, Huber JA. Diverse viruses in packa ge= pheat map. 51. Beghini F, McIver LJ, Blanco‑Míguez A, Dubois L, Asnicar F, Maharjan S, et al. Integrating taxonomic, functional, and strain‑level profiling of diverse microbial communities with biobakery 3. Elife. 2021;10:1–42. 52. Seemann T. Prokka: rapid prokaryotic genome annotation. Bioinformatics. 2014;30(14):2068–9. 53. Kanehisa M, Sato Y, Kawashima M, Furumichi M, Tanabe M. KEGG as a reference resource for gene and protein annotation. Nucleic Acids Res. 2016;44(D1):D457‑462. deep‑sea hydrothermal vent fluids have restricted dispersal across ocean basins. mSystems. 2021;6(3):1–18. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub‑ lished maps and institutional affiliations.
10.1242_bio.059967	© 2023. Published by The Company of Biologists Ltd \| Biology Open (2023) 12, bio059967. doi:10.1242/bio.059967 RESEARCH ARTICLE Context-dependent thermolability of sex determination in a lacertid lizard with heteromorphic sex chromosomes Alexander Hansson1,2,, Erik Wapstra2, Geoffrey M. While2, Willow R. Lindsay1 and Mats Olsson1 ABSTRACT In rarer, but the individual. reaction norms, permanent Developmental conditions can profoundly impact key life history traits In cases where offspring sex is driven by of developmental changes to the phenotype can fundamentally alter life history trajectories. Sex determination mechanisms in reptiles are remarkably diverse, including well-characterised genetic and temperature-dependent sex determination. increasingly more commonly documented cases, sex can also be determined by a combination of the two, with temperature overriding the genetically determined sex. Thus, sex-by-temperature interactions is a mechanism that can be contextually labile, where reaction norms of sex against developmental environment might only be observable under certain conditions. We examine the effects of incubation temperature on hatchling sex in an oviparous lizard with clearly defined heteromorphic sex chromosomes presumed to determine sex solely on a genetic basis. We also test the repeatability of our results by replicating incubation experiments across 3 years. We show that warmer temperatures may override chromosomal sex and cause an overproduction of daughters. However, this effect was inconsistent among years, with high temperature only resulting in a daughter-significant bias in one year. Warm-incubated daughters were more efficient at converting yolk into tissue, which would allow for greater resource allocation to other fitness-related processes, such as growth. This suggests that thermolabile sex determination could be a trait under selection. More energy-efficient embryos also produced faster-growing offspring, suggesting that energy utilization patterns of the embryo were maintained into the juvenile stage, which could have important implications for the ontogenetic development and evolution of life histories. KEY WORDS: Sex determination, Development, Incubation temperature, Reptile INTRODUCTION The determination of phenotypic sex is fundamentally important for development and life history of an individual with downstream consequences for sex ratio structure of populations (West et al., 2002). Unsurprisingly, much scientific effort has been allocated to 1Department of Biological and Environmental Sciences, University of Gothenburg, Box 463, 405 30, Gothenburg, Sweden. 2School of Natural Sciences, University of Tasmania, Private Bag 55, Hobart, TAS, 7001 Australia. Author for correspondence (alexander.hansson@bioenv.gu.se) A.H., 0000-0002-7179-9361; E.W., 0000-0002-2050-8026; G.M.W., 0000-0001- 8122-9322; W.R.L., 0000-0001-8581-2558; M.O., 0000-0002-4130-1323 This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed. Received 19 April 2023; Accepted 20 April 2023 in mammals understanding the diversity of ways in which sex is determined across the animal kingdom. Sex determination mechanisms of mammals and birds are highly conserved, where sex is determined genetically at conception by genes contained in sex chromosomes, and female with male heterogamety (XY) heterogamety (ZW) in birds. In contrast, reptiles show the full diversity of sex determination mechanisms (Janzen and Phillips, 2006; Bachtrog et al., 2014) and have proven vital to the study of vertebrate sex determination evolution (Shine, 1999; Shine et al., 2002; Warner and Shine, 2008b; Pen et al., 2010). In vertebrates, sex is generally determined either by genes at conception (genetic sex determination, GSD), or environmental cues (ESD), most commonly temperature during embryogenesis (temperature- dependent sex determination, TSD). Under some conditions, and in some taxa, the defining mechanism may also be a combination of genes and temperature, when environment can override genotypic sex (Shine et al., 2002; Valenzuela et al., 2003; Sarre et al., 2004; Hill et al., 2022). These types of systems have been hypothesised to be widespread in reptiles (Holleley et al., 2016), with documented cases becoming increasingly more common (see reviews Schwanz et al., 2020; Schwanz and Georges, 2021). In ESD species, Frequency-dependent selection will typically act on a system where allocation towards one sex yields greater fitness returns, driven by sex ratio disequilibrium and differential reproductive costs associated with sons and daughters (Frank, 1990). In species with sex genes, sex is determined at conception and facultative alterations to mendelian parity are commonly caused by differential and Willard, 1973; fertilization and/or mortality (Trivers Krackow, 1995). the environment provides additional mechanisms for the adjustment of sex ratios, by parents adaptively altering nesting site and phenology, which can overproduce the sex giving the greatest return on investment as a function of sex-specific environmental optima (Doody et al., 2006; Warner and Shine, 2008a). Such sex-specific fitness differences are the most accepted evolutionary explanation of ESD (Charnov and Bull, 1977; Shine et al., 1995; Shine, 1999; Warner and Shine, stimuli 2008b). However, when directional consistently maintain sex ratio bias in a population, frequency- dependent selection will favour a transition to GSD or random sex determination (Conover et al., 1992; Schwanz and Georges, 2021). sex determination system can be difficult, especially in long-lived species where direct measurements of life-time fitness are impractical. We therefore often must accept early-life traits as predictors of survival and lifetime reproductive output. This makes predictors of growth rate prime candidates for fitness related traits, as it likely influences the time to, and size at, maturity (Angilletta, 2004). Such changes to growth rate could be caused by thermal effects on physiological efficiency by altering basal metabolic rate (Nord and Nilsson, 2011; Spencer and Janzen, 2014). This would allow a greater surplus of energy to be allocated to fitness-related Assigning an adaptive value thermosensitive environmental to a 1 n e p O y g o o B l i RESEARCH ARTICLE Biology Open (2023) 12, bio059967. doi:10.1242/bio.059967 processes, such as growth (Steyermark, 2002). If such a relationship between temperature and metabolic rate were to differentially that a affect male and female offspring, thermosensitivity of sex determination is a trait under selection (Ligon et al., 2009). it would suggest Recently, environmental involvement in the sex determination of GSD species have received growing scientific attention and is proposed to be an ancestrally conserved trait in reptiles (Holleley et al., 2016; Schwanz and Georges, 2021). We test this hypothesis by examining the influence of incubation temperature on offspring sex in a lacertid lizard with clearly defined heteromorphic sex chromosomes, with female heterogamety (ZZ/ZW). We also duplicate the experiment across 3 years to determine whether the results are repeatable, or context-dependent. Our previous work on this system suggests the potential for labile sex allocation, where females overproduce daughters when mated to males of perceived higher quality (Olsson et al., 2005a,b). Furthermore, we examine whether the physiological efficiency of embryos is sex-specific by determining the conversion of yolk-to-tissue mass within each treatment, and if such effects are maintained post-hatching by influencing juvenile growth rates. RESULTS During the 3-year study, a total of 81 clutches were produced including 574 fertilised eggs combined (with an average clutch size of 7.09 eggs) of which 561 hatched. Incubation duration was inversely related to incubation temperature (linear mixed model regression; t=–67.87, d.f.=501.1, P<0.0001), with eggs from 23°C hatching after an average of 50.6±0.2 days (40–56 days), 25°C eggs after 38.6±0.2 days (33–44 days), and 27°C eggs after 33.8 ±0.1 days (31–39 days). The eggs producing male or female hatchlings did not differ in size (linear mixed model regression; t=–0.693, d.f.=497.2, P=0.49), and egg mass was the strongest determinant of hatchling mass, regardless of incubation treatment (Tables 1 and 3). Male hatchlings from the 27°C were heavier compared their female counterparts when correcting for egg mass and egg mass-by-sex interaction, while an opposite trend, approaching significance, was found in the 23°C treatment (Table 3). However, the average weight differences between the sexes were miniscule (<0.5% difference in the 27°C treatment) and thus unlikely to be ecologically important. Fig. 1. Percent males produced by the three temperature treatments (23, 25, 27°C) in order of increasing colour saturation, and among the 3 years of the study. Incubation temperature affected hatchling sex in a year- dependent manner. We observed a significant effect of incubation temperature on hatchling sex in 2018, when more female hatchlings were produced from eggs incubated at 27°C compared to those in 23°C, with an overall production of 70% and 50% females, respectively (Fig. 1, Table 2). Similar female bias, albeit non- significant, was observed in 2020, with a production of 69% females in 27°C against 53% females in 23°C. It is worth noting that the sample sizes in 2020 were about half of that in 2018 and 2019, which could have influenced the statistical power of the model and the results in this year. Sex ratios were unaffected by incubation treatments in 2019. Female hatchlings were significantly better than males at converting egg mass into hatchling mass when incubated at the temperature that overproduced female hatchlings (27°C), whereas the opposite was true for eggs incubated at 23°C, with males being the more efficient sex at converting egg mass into hatchling mass (Table 3, Fig. 2). This variation in energy efficiency extended beyond embryogenesis, with more energy-efficient eggs also producing (linear mixed model growing regression; t=2.23, d.f.=222.0, P=0.027; Fig. 3). juveniles faster Table 1. Descriptive statistics of average (± s.e.) egg mass, hatchling body mass and SVL, separated by year, sex, and incubation treatment Trait Egg mass (g) Hatchling mass (g) Hatchling SVL (mm) f, female; m, male. Year 2018 2019 2020 2018 2019 2020 2018 2019 2020 Sex m f m f m f m f m f m f m f m f m f n 41 41 32 39 15 17 41 41 31 39 15 17 41 41 32 39 15 17 23°C 0.477±0.010 0.497±0.010 0.516±0.012 0.491±0.007 0.465±0.008 0.486±0.011 0.501±0.012 0.497±0.009 0.524±0.013 0.507±0.009 0.496±0.010 0.505±0.012 27.66±0.21 28.41±0.18 29.00±0.19 29.59±0.15 28.60±0.21 29.88±0.19 n 37 50 33 39 18 14 37 50 33 39 18 14 37 50 33 39 18 14 25°C 0.477±0.009 0.493±0.011 0.504±0.009 0.496±0.006 0.468±0.011 0.475±0.018 0.491±0.010 0.497±0.011 0.514±0.010 0.495±0.007 0.500±0.009 0.496±0.016 27.57±0.18 28.28±0.19 29.06±0.17 29.69±0.17 28.67±0.21 29.43±0.29 n 25 57 32 39 10 22 25 57 32 39 10 22 24 57 32 39 10 22 27°C 0.487±0.014 0.485±0.008 0.506±0.008 0.500±0.009 0.444±0.012 0.480±0.011 0.483±0.011 0.485±0.009 0.503±0.008 0.502±0.009 0.469±0.019 0.469±0.011 27.75±0.21 28.09±0.17 29.13±0.19 29.77±0.16 28.30±0.15 29.59±0.20 2 n e p O y g o o B i l RESEARCH ARTICLE Biology Open (2023) 12, bio059967. doi:10.1242/bio.059967 Table 2. Numbers of hatchlings of each sex produced in the three incubation temperature treatments, separated by the 3 years of the study, and the summary statistics from the analyses of by-year effects of incubation treatment on hatchling sex Year 2018 2019 2020 Sex n(23°C) n(25°C) n(27°C) m f m f m f 41 41 32 39 15 17 37 50 33 39 18 14 25 57 32 39 10 22 f, female; m, male. χ2 6.47 P 0.039 0.01 0.99 4.03 0.13 chromosomal DISCUSSION We continue to uncover remarkable complexities of squamate sex determination. Recent research has highlighted the importance of this group for our understanding of sex determination mechanisms and their evolution (Shine et al., 2002; Quinn et al., 2007; Capel, 2017; Pennell et al., 2018; Schwanz and Georges, 2021). Our findings add to this research by providing the first robust evidence of thermolability in the sex determination of a lacertid lizard. Specifically, we show that high incubation temperature may override sex causing an overproduction of daughters. However, this effect was inconsistent among years, with temperature only resulting in a significant sex-bias in one year of the study (and a trend in the same direction in another year). Such inconsistencies are not uncommon in sex-determination experiments in GSD species. Studies on birds are notable for showing context-dependent sex-ratio adjustments within and among studies, effects often varying between study year and population (Korsten et al., 2006; Henderson et al., 2014), potentially calling into question conclusions about the generality of sex ratio adjustments. The research of sex allocation in reptiles also show clear instances of a context-dependent reaction norm between environment and sex determination in GSD species. One such dependency has been observed as seasonal sex ratio variation within a population. The turtle with TSD, Chrysemys picta, experienced seasonal changes to the direction of sex ratio shifts where the same incubation temperature produced 72% males early in the season, and 76% females late in the season (Bowden et al., 2000). Such thermolability of sex determination has also been observed between altitudinally and latitudinally separated populations, as shown in both a viviparous and an oviparous lizard (Wapstra et al., 2004; Pen et al., 2010; Li et al., 2022). Thus, reaction norms of offspring sex against developmental temperature may themselves be labile, perhaps overlooked because replicated studies among contexts (in this case years) are rare. This can for limit interpretations and predictions of reaction norms sex-determination and sex-differentiation to the immediate conditions at which they are evaluated, conditions that might be predictive of the evolution of sex determination and its norm of reaction. This would be especially true in species that determine sex based on an interaction of genes and environment, where effects on offspring sex might only be observable as subtle biases under extreme environmental conditions. involvement In species with heteromorphic sex chromosomes where sex is determined at conception, biased sex ratios can be achieved by processes that cause differential fertilization by X and Y sperm ( primary sex ratio bias), or by means of differential mortality of male and female embryos (secondary sex ratio bias). Because Lacerta agilis has clearly defined sex chromosomes with female heterogamety (Srikulnath et al., 2014; Lisachov et al., 2020), and we used a split-brood design with a very high hatching success, the overproduction of females cannot be explained by parental effects, differential fertilization, or sex-specific embryo mortality. A candidate proximate mechanism explaining the observed thermal in the sex determination of L. agilis is an acquired thermosensitivity during sex differentiation causing a mismatch between genetic and phenotypic sex (sex reversal). The phenomenon of sex reversal has only been adequately demonstrated with a mismatch between phenotypic and genetic sex in three reptiles – all lizards – thus far, two oviparous [the bearded dragon, Pogona vitticeps (Quinn et al., 2007; Holleley et al., 2015; Castelli et al., 2020), the Eastern three-lined skink, Bassiana duperreyi (Shine et al., 2002; Radder et al., 2008; Quinn et al., 2009)], and one viviparous [the spotted snow skink, Niveoscincus ocellatus (Wapstra et al., 2004; Pen et al., 2010; Cunningham et al., 2017; Hill et al., 2018, 2022)]. Although there are only a few cases confirmed of sex reversal, phylogenetic inference has established TSD as an ancestral reptilian trait (Janzen and Phillips, 2006) and evidence of thermosensitive sex determining genes in GSD species to be ancestrally conserved (Valenzuela, 2007). This suggests that the capacity for sex reversals might be widespread in reptiles (see Holleley et al., 2016; Schwanz and Georges, 2021 for reviews). Recent discoveries of GSD and environmental effects (GSD+EE) from several species lend further weight to this hypothesis, including data from the Jacky dragon (Amphibolurus muricatus), the multi-ocellated racerunner (Eremias multiocellata), the common collared lizard (Crotaphytus collaris), the Japanese gecko (Gekko japonicus), and the yellow-bellied water skink (Eulamprus heatwolei) (Wang et al., 2015; Cornejo-Páramo et al., 2020; Wiggins et al., 2020; Whiteley et al., 2021; Li et al., 2022). Moreover, local variation in sex determination systems has recently been found in P. vitticeps and proposed to be the result of genetic adaptations to local environmental conditions (Castelli et al., 2020). The sand lizard population used in this study was experimentally Table 3. By-treatment effects of egg mass and sex, and their interaction on hatchling mass Incubation temperature 23°C 25°C 27°C n 184 191 185 Egg mass t(171.7)=9.24 P<0.0001 β=0.690±0.075 t(167.1)=12.54 P<0.0001 β=0.747±0.060 t(137.6)=13.39 P<0.0001 β=0.866±0.065 Effects Sex t(168.0)= −1.87 P=0.063 β=-0.087±0.047 t(143.0)= 0.279 P=0.781 β=0.010±0.035 t(167.8)= 3.45 P<0.0008 β=0.149±0.043 Interaction t(169.4)=2.064 P=0.041 β=0.197±0.095 t(143.2)=-0.039 P=0.989 β=-0.003±0.072 t(169.7)=-3.42 P<0.0008 β=-0.302±0.088 3 n e p O y g o o B i l RESEARCH ARTICLE Biology Open (2023) 12, bio059967. doi:10.1242/bio.059967 Fig. 2. Interaction plot of within-treatment relationships between egg mass and hatchling mass of male (blue) and female (red) hatchlings. Slopes are based on linear regressions and shaded area denotes standard error range. founded on an island about 20 years ago. We have recently found evidence for increased plasticity in the reproductive biology of island-females compared to their mainland counterparts (Olsson et al., 2018). Future work should be dedicated to determining whether the observed sex bias is in fact due to sex reversal and whether a similar in mainland populations, or whether it is a locally acquired trait unique to this island population. If the latter is found to be true, this population will offer exciting opportunities to study the evolution of thermosensitive sex determination in reptiles. thermosensitive sex determination exist The evolution of thermosensitivity in GSD species is likely to be under similar selective pressures as those in a TSD system. Sex genes may ensure balanced sex ratios under normal temperatures but can be overridden by extreme temperatures to overproduce the sex with greater potential fitness in the more extreme environment. One proposed mechanism of sex reversal is the acquisition of a Fig. 3. Relationship between residual conversion efficiency of eggs and juvenile growth rate (mm/day). Data points are separated by treatment (23, 25, 27°C) in order of increasing colour saturation. thermosensitive gene dosage in the homogametic sex, in which extreme temperatures may inactivate the sex-determining gene (Quinn et al., 2007, 2011; Ezaz et al., 2009). In agreement with our results, all confirmed cases of reptile sex-reversal are unidirectional, the homogametic sex is the reversing one, producing XX males and ZZ females and causing a sex ratio bias towards the heterogametic sex. This is predicted by evolutionary theory where sex reversal is suboptimal or in the heterogametic sex from second generation YY and WW chromosomal combinations (Bull, 1981; Schwanz et al., 2013). lethal the inconsistent thermosensitivity of Our split-brood design is arguably a robust approach for analysing among- and within-year thermal effects on sex and sex phenotype. Thus, determination in L. agilis can only be explained by some additional factor outside the scope of our experimental design. One candidate hypothesis is natural yearly temperature variation, with 2018 being a notable thermal outlier, and indeed the year in which we observed a significant female bias in the warmest incubation treatment. The average daily temperature in 2018 during the month of May – when females emerge from hibernation and most ovulations occur (Olsson and Madsen, 1996; Olsson et al., 2011a) – was over 5°C warmer compared to the two subsequent years. There is little experimental research of pre-ovipositional environmental effects on sex determination, but a sensitivity to sex hormone inhibitors has shown to be present at oviposition in some (Shine et al., 2007). squamates and perhaps even earlier to, and during egg Furthermore, maternal production, has been observed in a skink to affect the thermolability of sex determination independent of incubation temperature (Schwanz, 2016). For such effects to explain our influence must alter the inconsistent results, pre-ovipositional thermosensitivity of sexual differentiation post-oviposition due to our split-brood design. The importance of hormonal or epigenetic effects should not be underestimated, especially in squamates where most embryos complete a substantial portion, about 25–40%, of their development prior to oviposition (Shine, 1983; DeMarco, temperature prior 4 n e p O y g o o B i l RESEARCH ARTICLE Biology Open (2023) 12, bio059967. doi:10.1242/bio.059967 in result to that cause(s) of what Regardless resource allocation to other 1993). This warrants future investigation of pre-ovipositional influence on downstream thermosensitivity of sex determination. proximate an the selective pressure acting on overproduction of one sex, GSD+EE species is likely similar in ESD species. Evolutionary explanations of ESD mainly rest upon the presence of sex-specific differential fitness linked to development conditions (Shine et al., 1995; Schwanz and Georges, 2021). We show that daughters incubated in the warmest, female-biased, treatment were more efficient at converting yolk into tissue mass. The opposite was true in the coolest incubation treatment, in which male embryos from large eggs were more efficient converters. If this is caused by a sex-specific differential energy efficiency, it would support the observed thermolability of sex determination as an adaptive trait. Energy efficiency, in this context, refers to the portion of the total energy budget utilised for basal metabolism and if reduced could allow for greater fitness-related processes, such as growth (Steyermark, 2002; Ligon et al., 2009). Sex-specific responses to overwintering temperature, shown as differential metabolic expenditure and growth, have previously been observed in the painted turtle (Chrysemys picta), in which male and female hatchlings had different optimal overwintering temperatures (Spencer and Janzen, 2014). This was argued to be a proximate mechanism for the adaptive maintenance of TSD in this turtle by to the current producing the sex best suited, metabolically, environmental conditions. Similar effects have recently been shown in zebra in which incubation temperature differentially affect the basal metabolic rate of male and female hatchlings (Wada et al., 2015; Gurley et al., 2018). If such changes to basal metabolic rate were to extend beyond embryogenesis, it could have major implications for life history trade-offs throughout life, such as time to, and size at, maturity, and the capacity for future reproductive investments (Burton et al., 2011; Ben-Ezra and Burness, 2017). We observed exactly this; individuals produced from more energy-efficient embryos also grew at a faster rate after hatching, suggesting that the proposed effects on energy utilization could be extended into the juvenile – and perhaps later – life stages. This would further support the hypothesis that the observed thermolabile sex determination in L. agilis is a trait under selection. (Taeniopygia guttata), finches In conclusion, we provide the first evidence of a context- dependent thermolabile aspect to the sex determination of a lacertid lizard. We also show that this reaction norm might be maintained through selection by differential energy efficiency of male and female embryos. This hypothesis is further strengthened by our observations of more energy-efficient embryos producing faster- growing juveniles, suggesting prolonged benefits. Future studies will inevitably uncover the prevalence of reaction norms involved in sex determination of the increasing evidence of such reaction norms clearly highlight the necessity to consider these effects going forward. We also stress the importance of considering context when studying modes of sex determination, where snapshot studies might not be sufficient to draw meaningful conclusions of negative results, and the biological importance of positive results can easily be exaggerated. reptiles with sex genes. However, the most northern-occurring oviparous lizard in Europe. Our study population was experimentally founded about 20 years ago on a small island (St. Keholmen) located on the southwestern coast of Sweden (57°29′ N 11°56′ E) (see Lindsay et al., 2020), which is near the northern limit of the species’ range. Females produce a single clutch of 4–15 eggs each year, but can, when exposed to optimal conditions, produce a second clutch (Olsson and Shine, 1997a,b). We collected adult female sand lizards by noose or by hand after Spring emergence in early May of 2018–2020 and palpated for eggs. Egg-carrying females were brought back to the laboratory and housed in cages (500×400×350 mm) with a flat basking rock on a moistened sand substrate to act as a favourable oviposition site. The ambient temperature was set to fluctuate daily between 15 and 20°C, simulating natural conditions, and a 40W spotlight positioned over the basking rock allowed behavioural thermoregulation to 40°C body temperature. Females were fed live meal worms dusted with calcium and multivitamins and provided with water ad libitum. Egg collection and incubation Gravid females (n(2018)=33, n(2019)=29, n(2020)=16) were closely monitored and eggs were removed from the cage within hours of oviposition. In the 3 years combined, females laid a total of 574 fertilised eggs (n(2018)=258, n(2019)=218, n(2020)=98). Three females produced a second clutch in 2019, while no second clutches were produced in the two other years. Due to the nature of this longitudinally studied population, some females were included in more than 1 year, with a total of ten females included in more than 1 year (nine in 2018, ten in 2019, two in 2020). At oviposition, eggs were removed from the female cage and brushed clean from sand and moisture, weighed to the nearest 0.01 g and placed individually in plastic cups half-filled with moist vermiculite (1:8 water to vermiculite by volume). The cups were sealed with plastic cling wrap and rubber bands to prevent moisture loss. To test the influence of incubation temperature on the developing embryo, we divided sibling eggs among three constant incubation temperature treatments in accordance with a split-brood design which minimises the risk of confounding parental and treatment effects (Via, 1993). The temperatures used were 23, 25 and 27°C, based around the optimal incubation temperature of L. agilis; 25°C that minimise developmental abnormalities and asymmetries (Zakharov, 1989). Although natural nests are unlikely to experience such warm conditions consistently throughout incubation, this species practice uterine retention in cool years (Shine et al., 2017; Olsson et al., 2018). Furthermore, studies on other oviparous lacertids show that the average time the embryo spends in utero has been estimated to be about half of total time of embryogenesis (Shine, 1983). Such retention would result in the embryo experiencing longer periods of high maternal temperatures associated with behavioural thermoregulation that easily can exceed the incubation temperatures used in this study (Bauwens et al., 1995). Incubating eggs were rotated among three shelves in each incubator weekly to minimise the effect of thermal gradients inside the incubator and were monitored daily for pipping (first sign of eggshell rupture). Hatchling morphological measurements A total of 561 eggs hatched (n(2018)=251, n(2019)=214, n(2020)=96), with a hatching success between 97–98% within and among years. On the day of hatching, hatchlings were blotted dry, brushed clean of vermiculite and weighed (±0.01 g). SVL was measured using a ruler (±1 mm). Hatchling sex was determined within hours of hatching by gently pressing on both sides of the tail base using modified forceps with tips bend to a V-shape to observe the presence or absence of hemipenes (Harlow, 1996; Olsson and Shine, 2003). This sexing method has 100% repeatability in this species (Olsson et al., 2004; 2005a, 2011b). The length of incubation was defined as the time between oviposition and hatching. MATERIALS AND METHODS Study species, collection, and husbandry The sand lizard (L. agilis) is a small (max 20 g) ground-dwelling oviparous lizard with one of the largest distributions of any reptile (Bischoff, 1984), reaching from Sweden in the north to Turkey in the south (Bülbül et al., 2019), and stretching from western France to Mongolia. This makes L. agilis Post-hatching husbandry and growth In the last 2 years of the study, we examined the effects of embryonic development and environment on subsequent juvenile growth patterns. Hatchlings were placed in cages of the same size and design as the adults, but with more rocks to allow space for thermoregulation. Initially, a maximum of ten juveniles were housed per cage, subsequently lowered to five after a growth period of about 2 weeks. Juveniles were fed daily with 5 n e p O y g o o B l i RESEARCH ARTICLE Biology Open (2023) 12, bio059967. doi:10.1242/bio.059967 small crickets (2–6 mm) and mealworms, all dusted with vitamins and minerals as for adults. The amount of food was always in excess, allowing juveniles to feed ad libitum. Water was also provided ad libitum, with cages additionally misted with water daily. We remeasured juveniles for SVL that had grown for a minimum of 20 days, which excluded seven individuals that had not reached this threshold at the end of the experiment. This lower threshold was set to minimise the influence of the differential growth and feeding ecology during the first days after hatching when residual yolk is metabolised as a source of energy (Troyer, 1987). Juveniles included in analyses were remeasured after between 20–63 days, with a total of 230 juveniles measured (n(2019)=165, n(2020)=65). The upper limit of growth duration was the logical end of the experiment when juveniles were returned to the wild given ample time to prepare for hibernation. Growth during this life stage is generally linear in reptiles (Andrews, 1982; Li et al., 2013), and thus using growth rate (see below) as a measure of growth is used to account for the variation in growth duration. Data analysis All statistical models were fitted in R v4.1.2 (R Core Team, 2021). Generalised linear mixed models (GLMM) were fitted using the glmer function from the lme4 package (Bates et al., 2014), where test statistics were estimated for significance at alfa=0.05 by likelihood ratio tests. Linear mixed models (LMM) were fitted using the lmer function from the lme4 package with denominator degrees of freedom, t-statistics and P-values for the LMMs derived from Kenward-Roger approximations using the lmerTest package (Kuznetsova et al., 2017). Maternal identity was included as a random effect in all mixed models to account for the pseudo-replication of siblings. To test the influence of incubation temperature on hatchling sex, we fitted a GLMM with binomial error distributions with sex as response and incubation temperature as fixed factor, with maternal identity as random effect. Pooling all years and examining the effects of year and year-by- treatment did not reveal any significant influence on sex. One of the years of the study (2018) was an extreme thermal outlier, and we therefore decided to analyse years separately to avoid spurious interactions with other predictors, including incubation temperature. To test whether thermal effects on hatchling sex might be sex-specific and potentially an adaptive trait, we examined sex-effects on yolk-to-tissue conversion from eggs to hatchlings. Because sand lizard eggs incubated at the same temperature can differ in incubation duration (Olsson et al., 2018), and lizards in general, including most lacertids, commonly show nonlinear developmental responses to constant incubation temperatures (Li et al., 2013), we analysed each treatment separately. We did this by fitting LMMs with hatchling mass as response and egg mass and sex, and their interaction, as fixed factors with maternal identity and year as random factors. Such sex- specific conversion efficiency is likely to cause short-term effects, like earlier hatching by completing development faster, or result in a larger body size at hatching from less thermally induced energy waste. However, long-term effects of conversion efficiency would likely have greater impact on individual fitness, and we also therefore examined whether more energy efficient embryos produced faster growing juveniles, which could have major implications on life history ontogeny and long-term evolution. Growth rate was calculated as the total SVL growth (mm) divided by the number of days of growth. Conversion efficiency was estimated from residuals from a regression of hatchling mass on egg mass. An LMM with growth rate as response, conversion efficiency as fixed effect and incubation temperature as covariate was modelled, with maternal identity and year as random factors. Acknowledgements This research was conducted at the University of Gothenburg under a scientific research permit (ldnr. 002826, Dnr. 5.8.18-06493/2020) issued by the Animal Ethics Committee at the University of Gothenburg, Sweden. We thank the two anonymous reviewers who helped to improve the quality of this manuscript. Competing interests The authors declare no competing or financial interests. draft: A.H.; Writing - review & editing: A.H., E.W., G.M.W., W.R.L., M.O.; Visualization: A.H.; Supervision: E.W., G.M.W., M.O.; Project administration: A.H., M.O., W.R.L.; Funding acquisition: M.O. Funding Funding was provided by The Swedish Science Council (VR) to M.O. Open Access funding provided by University of Gothenburg: Gö teborgs Universitet. Deposited in PMC for immediate release. Data availability All data supporting this study have been made available in the Dryad Digital Repository (https://doi.org/10.5061/dryad.8sf7m0cth). References Andrews, R. M. (1982). Patterns of growth in reptiles. In Biology of the Reptilia. Physiology D (ed. C. Gans), pp. 273-330. Academic Press: New York, USA. Angilletta, M. J. (2004). Temperature, growth rate, and body size in ectotherms: fitting pieces of a life-history puzzle. Integr. Comp. Biol. 44, 498-509. doi:10.1093/ icb/44.6.498 Bachtrog, D., Mank, J. E., Peichel, C. L., Kirkpatrick, M., Otto, S. P., Ashman, T.-L., Hahn, M. W., Kitano, J., Mayrose, I., Ming, R. et al. (2014). Sex Determination: why so many ways of doing it? PLoS Biol. 12, e1001899. Bates, D., Mä chler, M., Bolker, B. and Walker, S. (2014). Fitting linear mixed- effects models using lme4. J. Stat. Softw. 67, 1-48. Bauwens, D., Garland, T., Castilla, A. M. and Damme, R. V. (1995). Evolution of sprint speed in lacertid lizards: morphological, physiological and behavioral covariation. Evolution 49, 848-863. Ben-Ezra, N. and Burness, G. (2017). Constant and cycling incubation temperatures have long-term effects on the morphology and metabolic rate of Japanese quail. Physiol. Biochem. Zool. 90, 96-105. doi:10.1086/688383 Bischoff, B. (1984). Lacerta agilis Linnaeus 1798 – Zauneidechse. Handbuch der reptilien und amphibien Europas (ed. W. Bö hme), Band 2/1, Echsen II, pp. 23-68. Wiesbaden: AULA-Verlag. Bowden, R. M., Ewert, M. A. and Nelson, C. E. (2000). Environmental sex determination in a reptile varies seasonally and with yolk hormones. Proc. R. Soc. Lond. B 267, 1745-1749. doi:10.1098/rspb.2000.1205 Bü lbü l, U., Koç, H., Ö zkan, H., Ö ztü rk, İ . and Kutrup, B. (2019). New locality record of Lacerta agilis (Squamata: Lacertidae) in Turkey. Turk. J. Biod. 2, 52-56. doi:10.38059/biodiversity.596298 Bull, J. J. (1981). Evolution of environmental sex determination from genotypic sex determination. Heredity 47, 173-184. doi:10.1038/hdy.1981.74 Burton, T., Killen, S. S., Armstrong, J. D. and Metcalfe, N. B. (2011). What causes intraspecific variation in resting metabolic rate and what are its ecological consequences? Proc. R. Soc. B. 278, 3465-3473. doi:10.1098/rspb.2011.1778 Capel, B. (2017). Vertebrate sex determination: evolutionary plasticity of a fundamental switch. Nat. Rev. Genet. 18, 675-689. doi:10.1038/nrg.2017.60 Castelli, M. A., Georges, A., Cherryh, C., Rosauer, D. F., Sarre, S. D., Contador– Kelsall, I. and Holleley, C. E. (2020). Evolving thermal thresholds explain the distribution of temperature sex reversal in an Australian dragon lizard. Divers. Distrib. 27, 427-438. doi:10.1111/ddi.13203 Charnov, E. L. and Bull, J. (1977). When is sex environmentally determined? Nature 266, 828-830. doi:10.1038/266828a0 Conover, D. O., Van Voorhees, D. A. and Ehtisham, A. (1992). Sex ratio selection and the evolution of environmental sex determination in laboratory populations of Menidia menidia. Evolution 46, 1722-1730. doi:10.2307/2410026 Cornejo-Páramo, P., Dissanayake, D. S. B., Lira-Noriega, A., Martı́nez- Pacheco, M. L., Acosta, A., Ramı́rez-Suástegui, C., Méndez-De-La- Cruz, F. R., Székely, T., Urrutia, A. O., Georges, A. et al. (2020). Viviparous reptile regarded to have temperature-dependent sex determination has old XY chromosomes. Genome Biol. Evol. 12, 924-930. doi:10.1093/gbe/evaa104 Cunningham, G. D., While, G. M. and Wapstra, E. (2017). Climate and sex ratio variation in a viviparous lizard. Biol. Lett. 13, 20170218. doi:10.1098/rsbl.2017. 0218 Demarco, V. (1993). estimating egg retention times in sceloporine lizards. J. Herpetol. 27, 453-458. doi:10.2307/1564836 Doody, J. S., Guarino, E., Georges, A., Corey, B., Murray, G. and Ewert, M. (2006). Nest site choice compensates for climate effects on sex ratios in a lizard with environmental sex determination. Evol. Ecol. 20, 307-330. doi:10.1007/ s10682-006-0003-2 Ezaz, T., Sarre, S. D., O’meally, D., Marshall Graves, J. A. and Georges, A. (2009). Sex chromosome evolution in lizards: independent origins and rapid transitions. Cytogenet Genome Res. 127, 249-260. doi:10.1159/000300507 Frank, S. A. (1990). Sex allocation theory for birds and mammals. Annu. Rev. Ecol. Syst. 21, 13-55. doi:10.1146/annurev.es.21.110190.000305 Author contributions Conceptualization: A.H., M.O.; Methodology: A.H., M.O.; Formal analysis: A.H.; Investigation: A.H., W.R.L.; Resources: M.O.; Data curation: A.H.; Writing - original Gurley, B., Finger, J. W. and Wada, H. (2018). Sex-specific effects of incubation temperature on embryonic development of zebra finch (Taeniopygia guttata) embryos. Physiol. Biochem. Zool. 91, 1036-1045. doi:10.1086/699741 6 n e p O y g o o B l i RESEARCH ARTICLE Biology Open (2023) 12, bio059967. doi:10.1242/bio.059967 Harlow, P. S. (1996). A harmless technique for sexing hatchling lizards. Herpetol. Rev. 27, 71-72. Henderson, L. J., Evans, N. P., Heidinger, B. J., Adams, A. and Arnold, K. E. (2014). Maternal condition but not corticosterone is linked to offspring sex ratio in a passerine bird. PLoS ONE 9, e110858. doi:10.1371/ journal.pone.0110858 Hill, P. L., Burridge, C. P., Ezaz, T. and Wapstra, E. (2018). Conservation of sex- linked markers among conspecific populations of a viviparous skink, Niveoscincus ocellatus, exhibiting genetic and temperature-dependent sex determination. Genome Biol. Evol. 10, 1079-1087. doi:10.1093/gbe/evy042 Hill, P., While, G. M., Burridge, C. P., Ezaz, T., Munch, K. L., Mcvarish, M. and Wapstra, E. (2022). Sex reversal explains some, but not all, climate-mediated sex ratio variation within a viviparous reptile. Proc. R. Soc. B. 289, 20220689. doi:10. 1098/rspb.2022.0689 Holleley, C. E., O’meally, D., Sarre, S. D., Marshall Graves, J. A., Ezaz, T., Matsubara, K., Azad, B., Zhang, X. and Georges, A. (2015). Sex reversal triggers the rapid transition from genetic to temperature-dependent sex. Nature 523, 79-82. doi:10.1038/nature14574 Holleley, C. E., Sarre, S. D., O’meally, D. and Georges, A. (2016). Sex reversal in reptiles: reproductive oddity or powerful driver of evolutionary change? Sex Dev 10, 279-287. doi:10.1159/000450972 Janzen, F. J. and Phillips, P. C. (2006). Exploring the evolution of environmental sex determination, especially in reptiles. J. Evol. Biol. 19, 1775-1784. doi:10.1111/ j.1420-9101.2006.01138.x Korsten, P., Lessells, C. K. M., Mateman, A. C., Van Der Velde, M. and Komdeur, J. (2006). Primary sex ratio adjustment to experimentally reduced male UV attractiveness in blue tits. Behav. Ecol. 17, 539-546. doi:10.1093/beheco/ arj061 Krackow, S. (1995). Potential mechanisms for sex ratio adjustment in mammals and birds. Biol. Rev. 70, 225-241. doi:10.1111/j.1469-185X.1995.tb01066.x Kuznetsova, A., Brockhoff, P. B. and Christensen, R. H. B. (2017). lmerTest Package: Tests in Linear Mixed Effects Models. J. Stat. Soft. 82, 1-26. doi:10. 18637/jss.v082.i13 Li, H., Zhou, Z.-S., Ding, G.-H. and Ji, X. (2013). Fluctuations in incubation temperature affect incubation duration but not morphology, locomotion and growth of hatchlings in the sand lizard Lacerta agilis (Lacertidae): Phenotypic variation in lizards. Acta Zool. 94, 11-18. doi:10.1111/j.1463-6395.2011.00526.x Li, S., Xu, Z., Luo, L., Ping, J., Zhou, H., Xie, L. and Zhang, Y. (2022). Latitudinal temperature-dependent sex determination in the variation in the pattern of Japanese gecko, Gekko japonicus. Animals 12, 942. doi:10.3390/ani12080942 Ligon, D. B., Bidwell, J. R. and Lovern, M. B. (2009). Incubation temperature effects on hatchling growth and metabolic rate in the African spurred tortoise, Geochelone sulcata. Can. J. Zool. 87, 64-72. doi:10.1139/Z08-138 Lindsay, W. R., Madsen, T., Wapstra, E., Lillie, M., Loeb, L., Ujvari, B. (2020). Long term effects of outbreeding: experimental and Olsson, M. founding of island population eliminates malformations and improves hatching success in sand lizards. Biol. Conserv. 249, 108710. doi:10.1016/j.biocon.2020. 108710 Lisachov, A. P., Giovannotti, M., Pereira, J. C., Andreyushkova, D. A., Romanenko, S. A., Ferguson-Smith, M. A., Borodin, P. M. and (2020). Chromosome painting does not support a sex Trifonov, V. A. chromosome turnover in Lacerta agilis Linnaeus, 1758. Cytogenet Genome Res. 160, 134-140. doi:10.1159/000506321 Nord, A. and Nilsson, J.-Å. (2011). Incubation temperature affects growth and energy metabolism in blue tit nestlings. Am. Nat. 178, 639-651. doi:10.1086/ 662172 Olsson, M. and Madsen, T. (1996). Costs of mating with infertile males selects for late emergence in female sand lizards (Lacerta agilis L.). Copeia 1996, 462-464. doi:10.2307/1446866 Olsson, M. and Shine, R. (1997a). Advantages of multiple matings to females: a test of the infertility hypothesis using lizards. Evolution 51, 1684-1688. doi:10. 2307/2411220 Olsson, M. and Shine, R. (1997b). The seasonal timing of oviposition in sand lizards (Lacerta agilis): why early clutches are better. J. Evol. Biol. 10, 369-381. doi:10.1007/s000360050030 Olsson, M. and Shine, R. (2003). Female-biased natal and breeding dispersal in an alpine lizard, Niveoscincus microlepidotus: natal dispersal in a lizard. Biol. J. Linn. Soc. 79, 277-283. doi:10.1046/j.1095-8312.2003.00116.x Olsson, M., Ujvari, B., Madsen, T., Uller, T. and Wapstra, E. (2004). Haldane rules: costs of outbreeding at production of daughters in sand lizards. Ecol. Lett. 7, 924-928. doi:10.1111/j.1461-0248.2004.00652.x Olsson, M., Madsen, T., Uller, T., Wapstra, E. and Ujvari, B. (2005a). The role of Haldane’s rule in sex allocation. Evolution 59, 221-225. Olsson, M., Wapstra, E. and Uller, T. (2005b). Differential sex allocation in sand lizards: bright males induce daughter production in a species with heteromorphic sex chromosomes. Biol. Lett. 1, 378-380. doi:10.1098/rsbl.2005.0327 Olsson, M., Wapstra, E., Schwartz, T., Madsen, T., Ujvari, B. and Uller, T. (2011a). In hot pursuit: fluctuating mating system and sexual selection in sand lizards: fluctuating mating system and sexual selection. Evolution 65, 574-583. doi:10.1111/j.1558-5646.2010.01152.x Olsson, M., Pauliny, A., Wapstra, E., Uller, T., Schwartz, T. and Blomqvist, D. (2011b). Sex differences in sand lizard telomere inheritance: paternal epigenetic effects increases telomere heritability and offspring survival. PLoS ONE 6, e17473. doi:10.1371/journal.pone.0017473 Olsson, M., Loeb, L., Lindsay, W., Wapstra, E., Fitzpatrick, L. and Shine, R. (2018). Extreme plasticity in reproductive biology of an oviparous lizard. Ecol. Evol. 8, 6384-6389. doi:10.1002/ece3.4247 Pen, I., Uller, T., Feldmeyer, B., Harts, A., While, G. M. and Wapstra, E. (2010). Climate-driven population divergence in sex-determining systems. Nature 468, 436-438. doi:10.1038/nature09512 Pennell, M. W., Mank, J. E. and Peichel, C. L. (2018). Transitions in sex determination and sex chromosomes across vertebrate species. Mol. Ecol. 27, 3950-3963. doi:10.1111/mec.14540 Quinn, A. E., Georges, A., Sarre, S. D., Guarino, F., Ezaz, T. and Graves, J. A. M. (2007). Temperature sex reversal implies sex gene dosage in a reptile. Science 316, 411-411. doi:10.1126/science.1135925 Quinn, A. E., Radder, R. S., Sarre, S. D., Georges, A., Ezaz, T. and Shine, R. (2009). Isolation and development of a molecular sex marker for Bassiana duperreyi, a lizard with XX/XY sex chromosomes and temperature-induced sex reversal. Mol. Genet. Genomics 281, 665-672. doi:10.1007/s00438-009- 0437-7 Quinn, A. E., Sarre, S. D., Ezaz, T., Marshall Graves, J. A. and Georges, A. (2011). Evolutionary transitions between mechanisms of sex determination in vertebrates. Biol. Lett. 7, 443-448. doi:10.1098/rsbl.2010.1126 R Core Team (2021). R: a Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. Radder, R. S., Quinn, A. E., Georges, A., Sarre, S. D. and Shine, R. (2008). Genetic evidence for co-occurrence of chromosomal and thermal sex-determining systems in a lizard. Biol. Lett. 4, 176-178. doi:10.1098/rsbl.2007.0583 Sarre, S. D., Georges, A. and Quinn, A. (2004). The ends of a continuum: genetic and temperature-dependent sex determination in reptiles. BioEssays 26, 639-645. doi:10.1002/bies.20050 Schwanz, L. E. (2016). Parental thermal environment alters offspring sex ratio and fitness in an oviparous lizard. J. Exp. Biol. 219, 2349-2357. Schwanz, L. E. and Georges, A. (2021). Sexual development and the environment: conclusions from 40 years of theory. Sex Dev 15, 7-22. doi:10. 1159/000515221 Schwanz, L. E., Ezaz, T., Gruber, B. and Georges, A. (2013). Novel evolutionary pathways of sex-determining mechanisms. J. Evol. Biol. 26, 2544-2557. doi:10. 1111/jeb.12258 Schwanz, L. E., Georges, A., Holleley, C. E. and Sarre, S. D. (2020). Climate change, sex reversal and lability of sex–determining systems. J. Evol. Biol. 33, 270-281. doi:10.1111/jeb.13587 Shine, R. (1983). Reptilian reproductive modes: the oviparity-viviparity continuum. Herpetologica 39, 1-8. Shine, R. (1999). Why is sex determined by nest temperature in many reptiles? Trends Ecol. Evol. 14, 186-189. doi:10.1016/S0169-5347(98)01575-4 Shine, R., Elphick, M. J. and Donnellan, S. (2002). Co-occurrence of multiple, supposedly incompatible modes of sex determination in a lizard population. Ecol Letters 5, 486-489. doi:10.1046/j.1461-0248.2002.00351.x Shine, R., Elphick, M. J. and Harlow, P. S. (1995). Sisters like it hot. Nature 378, 451-452. doi:10.1038/378451a0 Shine, R., Wapstra, E. and Olsson, M. (2017). Seasonal shifts along the oviparity- viviparity continuum in a cold-climate lizard population. J. Evol. Biol. 31, 4-13. doi:10.1111/jeb.13202 Shine, R., Warner, D. A. and Radder, R. (2007). Windows of embryonic sexual lability in two lizard species with environmental sex determination. Ecology 88, 1781-1788. doi:10.1890/06-2024.1 Spencer, R.-J. and Janzen, F. J. (2014). A novel hypothesis for the adaptive maintenance of environmental sex determination in a turtle. Proc. R. Soc. B. 281, 20140831. doi:10.1098/rspb.2014.0831 Srikulnath, K., Matsubara, K., Uno, Y., Nishida, C., Olsson, M. and Matsuda, Y. (2014). Identification of the linkage group of the Z sex chromosomes of the sand lizard (Lacerta agilis, Lacertidae) and elucidation of karyotype evolution in lacertid lizards. Chromosoma 123, 563-575. doi:10.1007/s00412-014-0467-8 Steyermark, A. C. (2002). A high standard metabolic rate constrains juvenile growth. Zoology 105, 147-151. doi:10.1078/0944-2006-00055 Trivers, R. L. and Willard, D. E. (1973). Natural selection of parental ability to vary the sex ratio of offspring. Science 179, 90-92. doi:10.1126/science.179.4068.90 Troyer, K. (1987). Posthatching yolk in a lizard: internalization and contribution to growth. J. Herpetol. 21, 102-106. doi:10.2307/1564470 Valenzuela, N. (2007). Relic thermosensitive gene expression in a turtle with genotypic sex determination: relic thermal gene expression in GSD turtles. Evolution 62, 234-240. doi:10.1111/j.1558-5646.2007.00279.x Valenzuela, N., Adams, D. C. and Janzen, F. J. (2003). Pattern does not equal process: exactly when is sex environmentally determined? Am. Nat. 161, 676-683. doi:10.1086/368292 Via, S. (1993). Adaptive phenotypic plasticity: target or by-product of selection in a variable environment? Am. Nat. 142, 352-365. doi:10.1086/285542 7 n e p O y g o o B i l RESEARCH ARTICLE Biology Open (2023) 12, bio059967. doi:10.1242/bio.059967 Wada, H., Kriengwatana, B., Allen, N., Schmidt, K. L., Soma, K. K. and Macdougall-Shackleton, S. A. (2015). Transient and permanent effects of suboptimal incubation temperatures on growth, metabolic rate, immune function, and adrenocortical responses in zebra finches. J. Exp. Biol. 218, 2847-2855. Wang, C., Tang, X., Xin, Y., Yue, F., Yan, X., Liu, B., An, B., Wang, X. and Chen, Q. (2015). Identification of sex chromosomes by means of comparative genomic hybridization in a lizard, Eremias multiocellata. Zoolog. Sci. 32, 151-156. doi:10.2108/zs130246 Wapstra, E., Olsson, M., Shine, R., Edwards, A., Swain, R. and Joss, J. M. P. (2004). Maternal basking behaviour determines offspring sex in a viviparous reptile. Proc. R. Soc. Lond. B 271, S230-S232. doi:10.1098/rsbl.2003. 0152 Warner, D. A. and Shine, R. (2008a). Maternal nest-site choice in a lizard with temperature-dependent sex determination. Anim. Behav. 75, 861-870. doi:10. 1016/j.anbehav.2007.07.007 Warner, D. A. and Shine, R. (2008b). The adaptive significance of temperature- dependent sex determination in a reptile. Nature 451, 566-568. doi:10.1038/ nature06519 West, S. A., Reece, S. E. and Sheldon, B. C. (2002). Sex ratios. Heredity 88, 117-124. doi:10.1038/sj.hdy.6800018 Whiteley, S. L., Georges, A., Weisbecker, V., Schwanz, L. E. and Holleley, C. E. (2021). Ovotestes suggest cryptic genetic influence in a reptile model for temperature-dependent sex determination. Proc. R. Soc. B. 288, 20202819. doi:10.1098/rspb.2020.2819 Wiggins, J. M., Santoyo-Brito, E., Scales, J. B. and Fox, S. F. (2020). gene dose indicates presence of sex chromosomes in collared lizards (Crotaphytus collaris), a species with temperature-influenced sex determination. Herpetologica 76, 27-30. doi:10.1655/Herpetologica-D-19-00036 Zakharov, V. (1989). Future prospects for population phenogenetics. Sov. Sci. Rev. F Physiol. Genet. Biol. 4, 1-79. 8 n e p O y g o o B i l
10.1186_s40360-020-00405-6	Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 https://doi.org/10.1186/s40360-020-00405-6 R E S E A R C H A R T I C L E Open Access In vitro and in vivo effects of flubendiamide and copper on cyto- genotoxicity, oxidative stress and spleen histology of rats and its modulation by resveratrol, catechin, curcumin and α- tocopherol Rajesh Mandil1, Atul Prakash2, Anu Rahal3, S. P. Singh4, Deepak Sharma4, Rahul Kumar5 and Satish Kumar Garg2 Abstract Background: Living organisms are frequently exposed to more than one xenobiotic at a time either by ingestion of contaminated food/fodder or due to house-hold practices, occupational hazards or through environment. These xenobiotics interact individually or in combination with biological systems and act as carcinogen or produce other toxic effects including reproductive and degenerative diseases. Present study was aimed to investigate the cyto-genotoxic effects of flubendiamide and copper and ameliorative potential of certain natural phyotconstituent antioxidants. Method: In vitro cytogenotoxic effects were evaluated by employing battery of assays including Propidium iodide staining, Tunel assay, Micronuclei, DNA fragmentation and Comet assay on isolated splenocytes and their prevention by resveratrol (5 and 10 μM), catechin (10 and 20 μM), curcumin (5 and 10 μM) and α-tocopherol (5, 10 and 20 μM). In vivo study was also undertaken daily oral administration of flubendiamide (200 mg/kg) or copper (33 mg/kg) and both these in combination, and also all these concurrently with of α-tocopherol to Wistar rats for 90 days. Results: Flubendiamide and copper produced concentration-dependent cytotoxic effects on splenocytes and at median lethal concentrations, flubendiamide (40 μM) and copper (40 μM) respectively produced 71 and 81% nonviable cells, higher number of Tunel+ve apoptotic cells, 7.86 and 9.16% micronucleus and 22.90 and 29.59 comets/100 cells and DNA fragmentation. In vivo study revealed significant (P < 0.05) increase in level of lipid peroxidation (LPO) and decrease in glutathione peroxidase (GPx), glutathione-S-transferase (GST) and superoxide dismutase (SOD) activities in groups exposed to flubendiamide or copper alone or both these in combination. Histopathological examination of rat spleens revealed depletion of lymphoid tissue, separation of splenocytes and rarification in splenic parenchyma of xenobiotic(s) treated groups. (Continued on next page) Correspondence: rajesh_mandil@rediffmail.com 1Department of Veterinary Pharmacology and Toxicology, College of Veterinary and Animal Sciences, Sardar Vallabhbhai Patel University of Agriculture and Tecahnology, 250110, Meerut, India Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 2 of 17 (Continued from previous page) Conclusion: Flubendiamide and copper induce oxidative stress and produce cytogenotoxic effects along with histoarchitectural changes in spleen. All four tested natural antioxidants (resveratrol, catechin, curcumin and α-tocopherol) reduced flubendiamide and copper-induced cytotoxic effects in rat splenocytes. Rat splenocytes are very sensitive to flubendiamide and copper-induced cytogenotoxicity, therefore, these can be effectively employed for screening of compounds for their cytogenotoxic potential. α-tocopherol was effective in restoring alterations in oxidative stress biomarkers and preventing histoarchitectural lesions in spleen. Keywords: Flubendiamide, Copper, Splenocytes, Cyto-genotoxicity, Oxidative stress Background Last few decades toxicological research has revealed that immune system is the potential target for xenobiotics- induced adverse effects due to exposure to environmental pollutants, indiscriminate use of agrochemicals, metals, drugs, other chemicals and their metabolites. Therefore, the present study was undertaken to investigate the cyto- genotoxic potential of flubendiamide and copper in rat splenocytes primary cell culture following in vitro expos- ure. In vivo effect of these xenobiotics on oxidative stress biomarkers and histopathological changes in rat spleen were also studied. Ameliorative potential of α-tocopherol and other plants-based antioxidants against the adverse ef- fects of these xenobiotics was also evaluated. For in vivo study, Wistar rats were orally exposed to flubendiamide or copper alone, both these in combination, and also along with α-tocopherol for 90 days. Flubendiamide is a comparatively new insecticide and se- lectively acts on insects ryanodine receptors (RyR). It pos- sesses favourable toxicological profile due to its higher (> 2000 mg/kg) oral and dermal LD50 values in rats. Being com- paratively safe, it is being widely used on large number of crops which include fruits, vegetable crops and nuts to control insects. Therefore, human beings and ani- mals are also being indiscriminately exposed to flu- routes. direct bendiamide Genotoxicity is the primary risk factor associated with long-term exposure to environmental pollutants in- cluding insecticides and metals. Flubendiamide does not have genotoxic effects on bone marrow cells [1– 6]. But there are reports that exposure to certain xe- nobiotics, either individually or in combination, may result in gene mutation, chromosomal aberrations and DNA damage [7–9]. through indirect and Copper, being a micronutrient, is essential for life of humans and animals and is required in minute concen- trations for functioning of several metalloenzymes [10– 12]. It also possesses fungicidal, molluscicidal and weedi- cidal activities and is employed for control of bacterial and fungal diseases of fruits, vegetables, nuts and field crops, algae in domestic lakes and ponds and in garden- ing as powder and spray [13, 14]. In India, copper also enters in human body through drinking water, and inhalation of copper dust and fumes [15]. But it is toxic when present in the body in excess [10]. Environmental pollutants increase oxidative stress [16] and dietary antioxidants prevent free radicals induced tis- sue damage by preventing formation of radicals, scaven- ging them, or by promoting their decomposition [17–19]. Several natural food-derived components have received great attention in recent years as nutraceuticals due to their promising biological activities. α-tocopherol (α- TOH) is the major lipid soluble natural form of vitamin E and possesses antioxidant property. It protects cellular membrane and lipoproteins from peroxidation by reacting with lipid radicals produced in lipid peroxidation chain re- action [20–22]. Green tea is very rich in phenolic com- pounds including catechin and epigallocatechin gallate (EGCG) inhibit [23]. These are powerful antioxidant, apoptosis by inhibiting caspase 3 activity thereby prevent- ing expression of proapoptotic (Bax, Bad and Mdm2) and antiapoptotic genes (Bcl-2, Bcl-w and Bcl-xL) to protect SH-SY5Y cells from 6-OHDA-induced apoptosis [24–26] and EGCG is cancer chemopreventive also [27]. Curcumin is the main coloring agent of turmeric, used as a spice in India, and possesses number of promising pharmaco- logical activities including antioxidant [28–31] and DNA protective effect against arsenic, fluoride and chlorpyri- phos [32–34]. Phytoallexin resveratrol, found in the skin of grapes, possesses the potential to inhibit cancer initi- ation, promotion and progression, and inhibits TNFα- induced reactive oxygen intermediate generation [35–37]. In view of the sparse information on in vitro cyto- genotoxicity potential and in vivo adverse effects of fluben- diamide in mammals, and conflicting reports on genotoxic effects of copper, the present study was undertaken. We also evaluated the ameliorative potential of certain natural phyotconstituent antioxidants against these xenobiotics to explore their therapeutic and prophylactic use. Methods Experimental animals and chemicals Present study was undertaken on Wistar rats, which were procured from Laboratory Animal Resource Sec- tion, Izatnagar, India and maintained under standard managemental Indian Veterinary Research Institute, Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 3 of 17 conditions in the Departmental Experimental Animal House. Animals had free access to pelleted feed (Ashir- wad Industries, Chandigarh) and clean and deionized drinking water. Daily light and dark cycle of 12 h was en- sured. Before start of the experiment, an acclimatization period of 15 days was allowed. Whole study was carried out in two phases: Phase I - in vitro apoptosis studies while Phase II included only in vivo studies. The study was approved by the Institutional Animal Ethics Committee (IAEC; 79 IAEC/13). Flubendiamide, dexamethasone, resveratrol, catechin, curcumin, and α- tocopherol were procured from Sigma-Aldrich (USA) while copper sulphate from Sd Fine Chemical Ltd. Phase I- in vitro study Twenty adult male Wistar rats weighing 80–100 g were used for in vitro cyto-genotoxicity study on primary cell culture of isolated rat splenocytes. Isolation of splenocytes Rats were sacrificed by cervical dislocation and spleen was aseptically removed and quickly disintegrated into many pieces. Vigorous pipetting of meshed tissue was done with the help of 10 ml glass pipette to break the minced tissue and these cells were transferred to 15 ml test tubes containing chilled PBS and allowed to stand on ice for 15 min. Top 12 ml of suspension was collected into another centrifuge tube and cells were pelleted by centrifugation at 1500 rpm for 10 min. Cells pellet was re-suspended in PBS and centrifuged again at 1500 rpm for 10 min. The supernatant was discarded and pellet was treated with 5 ml of RBC lysis buffer (4.15 g NH4Cl; 0.5 g NaHCO3; 0.0186 g Na2-EDTA; 200 ml DW) and kept for 10 min in ice and centrifuged at 1500 rpm for 10 min. Then the pel- let was given two washings with PBS at 1500 rpm for 10 min. The pellet was re-suspended in 1 ml of Ros- well Park Memorial Institute (RPMI-1640; Sigma- Aldrich) medium with 10% foetal calf serum (Sigma- Aldrich). Viability count was done using 0.1% trypan blue exclusion test and the cells density was adjusted to obtain 106 cells/ml [38]. Median lethal concentrations Isolated splenocytes were seeded in 24 well culture plates containing 106 cells/ml in 10% RPMI with foetal calf serum. Different concentrations of fluben- diamide and copper i.e. 1.0, 2.5, 5, 7.5, 10, 15, 20, 40, 60, 80 and 100 μM were used. Culture plates were in- cubated for 12 h in CO2 incubator (New Brunswick Scientific, USA) at 37 °C with 5% CO2. After incuba- tion, samples were collected in 1.5 ml eppendorf tubes and centrifuged at 3200 rpm for 10 min. Supernatant was discarded and the pellet was dissolved in 0.5 ml PBS. Propidium iodide (Sigma) was added at 1 μg/ml concentration to cells and incubated for another 15 min in dark at room temperature. Cells were ob- served under fluorescent microscope (Microscan 20 PFM, Nitco) under green filter to determine the ap- proximate concentrations of test xenobiotics at which almost 50% dead splenocytes were observed. Calcula- tion of the LC50 value of flubendiamide and copper was done by subjecting the data (concentrations used versus % cell dead) of Table 1 to “Probit Analysis method” using “Graph Pad Prism software” and by plotting the log values of the concentrations of xeno- biotics used against log values of the per cent cells dead. Further we respective log values of the xenobiotics (copper and flubendiamide) at which 50% of the cells are expected to be dead, and then the antilog values of these log values were calculated. It is apparent that the interpolated LC50 value for copper was 38.90 μM and for flubendiamide, it was 37.23 μM. Both these values are very close to 40 μM and considered as median lethal concentration of flubendiamide and copper and used for further studies. interpolated the Viability of splenocytes Freshly collected splenocytes (106 cells/ml) were exposed to median lethal concentrations of flubendiamide and copper alone, and also along with the antioxidants- resver- atrol (5 and 10 μM), curcumin (5 and 10 μM), catechin (10 and 20 μM) and α-tocopherol (5, 10 and 20 μM). Solu- tions of resveratrol, catechin, curcumin, α-tocopherol, flu- bendiamide, copper sulphate and dexamethasone were prepared in dimethyl sulphoxide (DMSO). Culture plates were incubated for 12 h in CO2 incubator at 37 °C with 5% CO2 and further processed as described above to de- termine the number of nonviable cells. TUNEL assay After exposure of splenocytes to median lethal concen- trations of flubendiamide (40 μM) and copper (40 μM) for 12 h, these samples were further processed for deter- mination of apoptosis as per the protocol described in TUNEL Assay Kit (Invitrogen, USA; Ref. No. A35126). Apoptotic cells, which underwent extensive DNA deg- radation during late stages of apoptosis, were examined under blue filter of fluorescent microscope. Cells which fluoresced brightly were considered as apoptotic. Genotoxicity assays (micronucleus, DNA fragmentation and comet) Micronucleus assay Flubendiamide and copper genotoxicity potential was assessed by micronuclei assay by using the isolated splenocytes [39]. 106 cells/ml were incubated with Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 4 of 17 Table 1 Effect of different concentrations of flubendiamide and copper on per cent viability of rat splenocytes following their in vitro exposure to these xenobiotics Treatment Control Vehicle control (DMSO) 50 μl 1.0 μM Flubendiamide 2.5 μM Flubendiamide 5.0 μM Flubendiamide 7.5 μM Flubendiamide 10 μM Flubendiamide 15 μM Flubendiamide 20 μM Flubendiamide 40 μM Flubendiamide 60 μM Flubendiamide 80 μM Flubendiamide 1.0 μM Copper 2.5 μM Copper 5.0 μM Copper 7.5 μM Copper 10 μM Copper 15 μM Copper 20 μM Copper 40 μM Copper 60 μM Copper 80 μM Copper % Dead Splenocytes 4.93 ± 0.67 7.46 ± 0.83 12.98 ± 1.92 15.13 ± 1.87 22.18 ± 1.67 24.28 ± 2.24 28.09 ± 1.33 29.76 ± 1.55 32.32 ± 1.32 45.42 ± 2.50 67.89 ± 3.14 88.81 ± 5.62 6.45 ± 3.04 10.34 ± 1.63 13.88 ± 1.39 16.66 ± 1.92 26.08 ± 4.73 28.39 ± 1.74 34.95 ± 5.87 51.09 ± 2.01 61.11 ± 2.03 76.68 ± 1.71 Data presented are Mean ± SEM of three observations flubendiamide (40 μM) and copper (40 μM) alone and with different μM concentrations of resveratrol, cat- echin, curcumin and α-tocopherol and incubated for 12 h in CO2 incubator. After incubation, samples were collected in 1.5 ml eppendorf tubes and centrifuged at 3200 rpm for 10 min. Supernatants were discarded and the pellets were dissolved in 1.0 ml of Hank’s bal- anced salt solution (HBSS) having pH 7.2 and centri- fuged again for 10 min at 3200 rpm. Supernatant was removed and cells in suspension were mixed carefully in 100 μl of HBSS. A drop of cell suspension was taken on grease-free clean glass slide and smeared. The smear was air-dried and fixed with absolute methanol (100%) for 5 min and stained with acridine orange for 1 min at room temperature. The slide was rinsed in Sorensen’s buffer (pH 6.8) and kept for at least 3 min and this step was repeated three times. Slides were examined on the same day and 1000 cells (both mono- nuclear and binucleated) per slide were scored under green filter of the fluorescent microscope to determine the frequency of micronuclei formation. DNA fragmentation assay DNA ladder assay was performed according to phenol- chloroform-DNA isolation protocol [40]. After incubation of 5 X106 cells each with flubendiamide or copper alone and with antioxidants, as mentioned in micronuclei assay method, the cells were collected in 1.5 ml of eppendorf tubes and centrifuged at 3200 rpm for 10 min at 4 °C. The cells pellet was washed with PBS having pH 7.2, mixed with DNA extraction buffer (500 μl/tube) and kept in water bath for 1.0 h at 37 °C. 10% SDS was added (20 μl/ml) to the cell suspen- sion and tubes were gently mixed by inverting the tubes. Contents of the tubes appearing viscous indicated lysis of splenocytes. Proteinase K (15 μl of 20 mg Proteinase K/ml of buffer) was added to each tube in two pulses i.e. half the re- quirement was added to tube in the 1st pulse and mixed gently and kept in water bath at 50 °C. After 3–4 h, a second pulse of the remaining amount of proteinase K was added. Tubes were incubated at 50 °C overnight. Next day morning, equal amount of equilibrated phenol (Tris saturated phenol pH > 7.8) was added to each tube and mixed by gently inverting the tubes for 15 min till light coffee coloured uni- form solution was formed and centrifuged at 3400 rpm for 15 min. The upper aqueous phase containing DNA was transferred into fresh 1.5 ml clean eppendorf tube. Similar extraction was done (as in the above step) once with equal volume of phenol: chloroform: isoamyl alcohol (25:24:1) and with chloroform: isoamyl alcohol (24:1). To obtain the final aqueous phase, double the volume of chilled (− 20 °C) etha- nol was added. Tubes were mixed gently by inversion and kept at room temperature to allow precipitation of DNA. DNA pellet was washed twice with 500 μl of 70% ethanol and eppendorf tube was centrifuged at 10000 rpm for 10 min at room temperature. Finally 70% ethanol was discarded and DNA pellet was air dried by inverting tube on blotting paper so that last traces of ethanol were removed. However, it was ensured that pellet did not over-dry so to enable an easy dissolution in the following step. Approximately 50 μl of tris-EDTA buffer (TE) was added and kept in water bath at 60 °C for 2 h to inactivate DNAse and other enzymes. Eppendorf was stored at 4 °C for a week so that DNA was dissolved. DNA concentration and its purity was determined spectrophotmetrically by Biophotometer plus (Eppendorf) at 260 and 280 OD. Integrity of the DNA was examined in agarose gel (1.0%) electrophoresis and visualized under UV light in gel documentation system after staining with eth- idium bromide. Comet assay Single splenocyte cells were isolated from spleen after cervical dislocation and viability checked by Trypan blue exclusion test. 5X106 cells/well were kept for culturing and treated with flubendiamide and copper (40 μM/well) alone and with different micromolar concentrations of resveratrol, catechin, curcumin and α-tocopherol. After Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 5 of 17 incubation of 12 h in CO2 incubator, cells were collected in 1.5 ml eppendorf tubes and centrifuged at 3200 rpm for 10 min at 4 °C. Supernatant was discarded and the pellet was washed with PBS (pH 7.2). Comet assay was performed using the standard method with normal (NMA) and low melting agarsoe (LMPA) [41]. Briefly, slides were dipped in methanol and heated over blue flame to remove the grease, dust and oil. 1.5% NMA (Sigma-Aldrich) and 0.5% LMPA (Sigma–Aldrich) were prepared in PBS. LMP agarose was kept in water bath at 40 °C to cool and stabilize while NMA agarose was kept at 100 °C. First layer of agarose on the slides was prepared by dipping conventional pre-cleaned slide for few seconds in 100 ml wide mouth beaker containing 1.5% NMA up to one-third area and gently removed. Underside of the slide was wiped to remove excess agar- ose and allowed to dry in a tray. Slides were generally prepared a day earlier. Splenocyte cell pellets were uni- formly mixed with 100 μl of 0.5% LMPA and poured carefully on the first agarose layer and immediately cov- ered with a full length cover slip. Slides were kept on ice-pack for 15–20 min to allow for the 2nd agarose layer to solidify. After solidification, the cover slip was removed and the slide was kept in a coupling jar con- taining freshly prepared lysis solution (1 ml-Triton X- 100 and 10 ml DMSO was added to 89 ml stock lysing solution containing NaCl-36.52 g; EDTA disodium salt- 9.3 g; Trizma-0.3 g; NaOH-2 g- For 250 ml) at 4 °C over- night. Next day, the slide was removed from lysis solu- tion and prepared electrophoretic buffer so as to cause unwinding of DNA and expression of alkali-labile sites. Slide was run in horizontal electrophoresis (Bio Rad) chamber with the same electrophoresis buffer (pH > 13) at 25 V and 300 mA for 1 h. After running in electrophoresis chamber, the slide was gently removed and placed horizontally in a tray and covered with neutralizing buffer for 5 min and then decanted it; the same step was repeated three times to remove alkali and detergent. This step was critical to bring down the pH from 13 to 7.5. After neutralization, slides were stained by placing 3–4 drops of 100 μl work- ing ethidium bromide solution at equal distance and im- mediately covered with cover slip. Slides were examined under fluorescent microscope, individual cell/comets were observed and images were captured at 40X magni- fication using green filter and duplicate slides per treat- ment were observed. At least 50 cells from each slide were scored and a total of 100 cells/treatment was scored to get the reproducible data. 30 min in freshly kept for Phase II-in vivo chronic toxicity study Fifty four adult male Wistar rats weighing between 130 and 150 g were divided in nine groups of six animals each. Animals of six groups (IV to IX) were orally treated on daily basis with copper (33 mg/kg; group IV), flubendiamide (200 mg/kg; group V) or combination of both these (group VI), and α-tocopherol (100 mg/kg) along with these xenobiotics singly (group VII and VIII) or both these in combination (group IX) for 90 days. Groups I and II served as negative and vehicle controls (corn oil), respectively while rats of group III were ad- ministered only α-tocopherol (100 mg/kg). Solutions of copper sulphate and flubendiamide (FAME®, Bayer) were prepared in deionized water while α-tocopherol was dis- solved in corn oil. Doses of flubendiamide and copper were 1/10th of the LD50. At the end of exposure period, rats were humanely sacrificed by cervical dislocation and their spleen was collected and blotted with tissue paper. It was then used to determine its levels of different oxi- dative stress related parameters such as lipid peroxida- tion (LPO), reduced glutathione (GSH), catalase (CAT), superoxide dismutase (SOD), glutathione-S-transferase (GST) and glutathione peroxidase (GPx), along with total protein content in splenic tissue using UV- VIS spectrophotometeric methods [42–48]. 200 mg of the spleen sample was weighed and transferred in 2 ml of chilled saline. The same weight of the spleen sample was separately taken in 2 ml of 0.02 M EDTA for GSH esti- mation. Tissue homogenates were prepared by using tis- sue homogenizer (Heidolph) under cold conditions and centrifuged for 10 min at 3000 rpm. The supernatant was used for estimation of different oxidative stress bio- markers. Lipid peroxidation (LPO) and reduced glutathi- one (GSH) were assayed immediately after tissue collection. A small piece of the spleen tissue was collected in 10% formaldehyde saline solution and processed for prepar- ation of paraffin blocks as per the method described by [49]. Tissue sections of 5–6 μm thickness were cut using a microtome (Leica, Germany) and stained with haema- toxylin and eosin. Microscopic slides were examined under light microscope to observe the histoarchitecture changes in spleen. Statistical analysis of data the in vitro study has been presented as Data of Mean ± SEM of the three observations in each treat- ment group in Tables 1 and 2. Table 3 presents the Mean ± SEM data of in vivo study. Effects of different in vitro treatments were compared between the con- trol and xenobiotics alone-treated groups, and also between the xenobiotics alone and those treated con- currently with antioxidants. Statistically significant dif- groups ferences between the different observed in in vivo study were determined using one- way ANOVA followed by Tukey’s multiple post-hoc test with the help of SPSS® 16 software. Significant difference was considered at P < 0.05. treatment Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 6 of 17 Table 2 Effect of median lethal concentrations of flubendiamide and copper alone and in the presence of different concentrations of resveratrol, catechin, curcumin and α-tocopherol on viability, micronuclei and comet formation in rat splenocytes following their in vitro exposure Treatments aNonviable cells (%) aMicronuclei (%) No. of Comet/ 100 cells (%) Control DMSO (50 μl) Dexamethasone (20 μM) Flubendiamide (40 μM) Resveratrol (5 μM) + Flubendiamide (40 μM) Resveratrol (10 μM) + Flubendiamide (40 μM) Catechin (10 μM) + Flubendiamide (40 μM) Catechin (20 μM) + Flubendiamide (40 μM) Curcumin (5 μM) + Flubendiamide (40 μM) Curcumin (10 μM) + Flubendiamide (40 μM) α-tocopherol (5 μM) + Flubendiamide (40 μM) α-tocopherol (10 μM) + Flubendiamide (40 μM) α-tocopherol (20 μM) + Flubendiamide (40 μM) Copper (40 μM) Resveratrol (5 μM) + Copper (40 μM) Resveratrol (10 μM) + Copper (40 μM) Catechin (10 μM) + Copper (40 μM) Catechin (20 μM) + Copper (40 μM) Curcumin (5 μM) + Copper (40 μM) Curcumin (10 μM) + Copper (40 μM) α-tocopherol (5 μM) + Copper (40 μM) α-tocopherol(10 μM) + Copper (40 μM) α- tocopherol (20 μM) + Copper (40 μM) aData presented are Mean + SEM of three observations 5.41 ± 0.33 8.59 ± 0.88 – 71.88 ± 2.90 50.00 ± 1.85 24.36 ± 0.88 53.66 ± 1.76 52.46 ± 2.33 56.25 ± 3.05 38.24 ± 3.18 55.00 ± 0.33 40.26 ± 2.02 17.65 ± 0.57 81.11 ± 6.06 76.54 ± 4.84 30.43 ± 4.04 72.13 ± 3.71 65.82 ± 1.41 64.18 ± 3.84 59.68 ± 4.33 59.72 ± 5.0 54.73 ± 4.91 51.06 ± 4.18 0.96 ± 0.08 1.36 ± 0.08 7.60 ± 0.20 7.86 ± 0.17 1.20 ± 0.17 1.10 ± 0.05 1.43 ± 0.24 1.33 ± 0.08 3.13 ± 0.12 1.40 ± 0.15 2.10 ± 0.40 1.93 ± 0.29 3.30 ± 0.26 9.16 ± 0.21 4.80 ± 0.20 1.90 ± 0.32 5.20 ± 0.20 2.10 ± 0.36 3.47 ± 0.14 2.63 ± 0.21 3.33 ± 0.24 1.93 ± 0.18 3.16 ± 0.26 3.09 ± 0.31 4.58 ± 0.28 27.69 ± 0.87 22.90 ± 0.90 20.15 ± 1.91 15.44 ± 1.47 14.80 ± 1.25 12.64 ± 0.57 7.58 ± 0.89 7.20 ± 0.32 11.56 ± 0.33 6.96 ± 0.30 4.89 ± 0.33 29.59 ± 1.76 25.33 ± 0.47 9.69 ± 0.66 15.12 ± 0.32 12.41 ± 1.20 16.80 ± 0.87 12.71 ± 1.32 10.33 ± 1.20 15.20 ± 1.45 10.61 ± 0.66 Results Phase I- in vitro study Median lethal concentrations Data on in vitro effect of different concentrations of flu- bendiamide (1.0–80 μM) and copper (1.0–80 μM) on rats splenocytes revealed concentration-dependent lethal effect of these xenobiotics. There was dose-dependent increase in percentage of the nonviable splenocytes and nearly 50 % nonviable splenocytes were observed be- tween 40 μM and 60 μM concentrations of these xenobi- otics (Table 1). Therefore, 40 μM was considered the approximate median lethal concentration both for flu- bendiamide and copper. Viability of splenocytes Fluorescent microscopic examination of flubendiamide (40 μM) and copper (40 μM) alone-treated splenocytes re- spectively showed 71.88 and 81.11% nonviable cells com- pared to 5.41% in control and 8.59% in DMSO-treated cells (Table 2). Following concomitant in vitro treatment and antioxidants- of splenocytes with xenobiotics resveratrol, catechin, curcumin and α-tocopherol, the per- centage of the nonviable splenocytes was found to de- crease and effect of all these four antioxidants was concentration-dependent (Table 2). Out of these tested antioxidants, based on their comparative efficacy on equi- molar concentration basis (10 μM), resveratrol was found to be the most effective against flubendiamide in reducing the percentage of nonviable splenocytes, and the order of ameliorative potential of these antioxidants was: resvera- trol > curcumin ≈ α-tocopherol > catechin (Table 2). Similarly, resveratrol was also found to be the most effect- ive against copper-induced viability losses in splenocytes; and the order of ameliorative potential against copper was: resveratrol > α-tocopherol > curcumin > catechin. Tunel assay Splenocytes exposed to 40 μM flubendiamide or copper showed higher number of Tunel-positive (Tunel+ve) cells compared to those in negative or vehicle control (DMSO) groups as shown in Figs. 1 and 2, respectively. Compared to flubendiamide, copper was more potent in Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 7 of 17 ) i n e t o r p i g m / n m / d e z i l i t u H P D A N M n ( x P G a 5 3 0 . ± 3 9 1 . b a 7 3 0 . ± 6 9 0 . b a 6 0 0 . ± 9 5 0 . b 4 0 0 . ± 9 4 0 . b 6 0 0 . ± 7 3 0 . b 1 1 0 . ± 3 4 0 . d n a ) g k / g m 3 3 ( r e p p o c o t s t a r f o e r u s o p x e l a r o s y a d 0 9 i g n w o l l o f s r e k r a m o b i s s e r t s e v i t a d x o i i c n e p s l i n a t r e c n o ) g k / g m 0 0 1 ( l o r e h p o c o t - α f o n o i t a r t s i n m d a i l a r o f o t c e f f E 3 e l b a T ) g k / g m 0 0 2 i i e d m a d n e b u l f + g k / g m 3 3 ( n o i t a n b m o c i n i h t o b d n a e n o a l ) g k / g m 0 0 2 ( i i e d m a d n e b u l f H S G - B N D C f o M μ ( a T S G 1 - g m 1 i - n m e t a g u n o c j ) i n e t o r p g / H S G M m ( ) e u s s i t H S G ) i n e t o r p f o g m i n m / d e z i l i t u 2 O 2 H M m ( l e s a a t a C f o g m / U ( ) i n e t o r p a D O S g / A D M M n ( O P L ) e u s s i t i c n e p s l n i i n e t o r P ) l m / g m ( e u s s i t t n e m t a e r T s p u o r G b a 5 4 0 . ± 6 1 1 . a 9 0 0 . ± 0 5 1 . c b 9 0 0 . ± 7 8 1 . . a 9 7 8 ± 8 4 3 8 . b a 5 2 0 . ± 2 0 3 . b 4 7 0 1 . ± 3 1 7 7 . b a 6 3 0 . ± 8 6 0 . a 5 1 0 . ± 3 4 1 . c b a 9 1 0 . ± 9 1 2 . . a 5 5 0 1 ± 4 7 1 6 . a 0 4 0 . ± 2 4 4 . b a 0 5 6 . ± 4 1 1 6 . b a 8 4 0 . ± 9 0 1 . a 0 3 0 . ± 5 9 . 1 c b a 0 1 0 . ± 4 2 2 . . a 9 1 7 ± 0 4 1 8 . b a 7 3 0 . ± 5 0 3 . b a 4 8 4 . ± 6 9 8 6 . a 2 1 0 . ± 2 6 1 . c b a 1 2 0 . ± 8 3 2 . . a 5 3 7 ± 2 3 5 8 . a 5 3 0 . ± 3 9 3 . a 0 6 1 . ± 6 5 0 5 . a 4 1 0 . ± 4 7 1 . b a 7 1 0 . ± 8 4 2 . . a 1 6 0 1 ± 7 0 9 7 . a 6 3 . . 0 ± 6 1 4 . b a 2 1 2 . ± 8 7 1 6 . a 5 1 0 . ± 6 6 1 . a 8 0 0 . ± 9 4 2 . . a 5 1 9 ± 4 9 7 8 . b a 0 4 0 . ± 3 3 3 . b a 8 0 8 . ± 6 5 8 6 . a 0 2 0 . ± 5 6 1 . c b a 4 0 0 . ± 1 9 1 . . a 9 2 7 ± 4 5 0 8 . b 5 2 0 . ± 2 8 1 . b 2 1 0 . ± 1 6 0 . c 0 0 0 . ± 4 8 1 . . a 9 9 7 ± 7 3 0 6 . a 8 3 0 . ± 0 4 3 . a 5 2 0 . ± 4 3 1 . c b a 2 1 0 . ± 6 9 1 . . a 0 8 0 1 ± 6 7 8 9 . b a 4 3 0 . ± 7 2 3 . b 8 0 4 . ± 5 6 1 8 . b 7 5 5 . ± 0 7 1 8 . b 7 6 7 . ± 6 0 5 7 . a a a a a a a a a . 6 5 0 ± 2 4 3 . . 2 0 1 ± 5 0 4 . . 1 4 0 ± 1 2 3 . . 2 3 0 ± 9 5 2 . . 1 6 0 ± 9 6 3 . . 4 3 0 ± 0 3 2 . i i e d m a d n e b u F + l ) g k / g m 3 3 ( e t a h p u s l r e p p o C ) g k / g m 0 0 2 ( ) g k / g m 3 3 ( e t a h p u s l r e p p o C ) g k / g m 0 0 1 ( l o r e h p o c o t - α ) l i o n r o C ( l o r t n o c i l e c h e V l o r t n o C ) g k / g m 0 0 2 ( i i e d m a d n e b u F l I I I I I I V I V I V . 7 3 0 ± 1 8 2 . l o r e h p o c o t - α + ) g k / g m 3 3 ( e t a h p u s l r e p p o C I I V ) g k / g m 0 0 1 ( . 2 5 0 ± 6 1 4 . l o r e h p o c o t - α + ) g k / g m 0 0 2 ( i i e d m a d n e b u F l I I I V ) g k / g m 0 0 1 ( . 7 2 0 ± 5 8 2 . i i e d m a d n e b u F + l ) g k / g m 3 3 ( e t a h p u s l r e p p o C X I ) g k / g m 0 0 1 ( l o r e h p o c o t - α + ) ) g k / g m 0 0 2 ( . ) 5 0 0 < P ( y l t n a c i f i n g i s d e r e f f i d n m u o c l e m a s e h t n i s t p i r c s r e p u s t n e r e f f i d g n i r a e b ) 6 = n ; M E S ± n a e M ( s e u a V l l y n o s l a m n a i e v i f f o M E S ± n a e M e r a s e u a V a l (cid:129) (cid:129) Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 8 of 17 Fig. 1 Representative photographs of rat splenocytes showing TUNEL + ve cells (40 X) following in vitro exposure to median lethal concentration of flubendiamide alone (40 μM) and in the presence of different concentrations of natural antioxidants-resveratrol, catechin, curcumin and α-tocopherol producing Tunel+ve splenocytes, and compared to the flu- bendiamide or copper-alone treated splenocytes, marked reduction in Tunel+ve cells was observed in the spleno- cytes treated concurrently with either of the xenobiotic (flubendiamide or copper) and different antioxidants (res- veratrol 5 and 10 μM, catechin 10 and 20 μM, curcumin 5 and 10 μM or α-tocopherol 5, 10 and 20 μM) as shown in Figs. 1 and 2. However, based on the efficacy of different antioxidants at equimolar concentration basis i.e. 10 μM, resveratrol was most effective in reducing the number of Tunel+ve cells induced by flubendiamide (Fig. 1) and the overall order of efficacy of different antioxidants was res- veratrol > curcumin >α-tocopherol > catechin. Just like their efficacy against flubendiamide, all these were effect- ive in reducing copper-induced increase in number of Tunel+ve cells and the overall order of efficacy of different antioxidants was curcumin > catechin ≥ α-tocopherol ≥ resveratrol (Fig. 2). However, contrary to resveratrol, cur- cumin was most effective against copper. Micronuclei formation Flubendiamide and copper alone treated splenocytes showed micronuclei formation in 7.86 and 9.16% cells re- spectively compared to 0.96% in negative control and 1.36% in DMSO-treated splenocytes (Table 2; Fig. 3). Dexamethasone-induced micronuclei formation (7.6%) was much higher compared to that in negative control and DMSO-treated splenocytes. Almost a similar percentage of micronuclei were observed in splenocytes treated with flu- bendiamide (7.86%) or copper (9.16%) as summarized in Table 2. Ameliorative efficacy studies with resveratrol, cat- echin, curcumin and α-tocopherol against flubendiamide Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 9 of 17 Fig. 2 Representative photographs of rat splenocytes showing TUNEL + ve cells (40 X) following in vitro exposure to median lethal concentration of copper alone (40 μM) and in the presence of different concentrations of natural antioxidants-resveratrol, catechin, curcumin and α-tocopherol or copper-induced micronuclei formation revealed marked reduction in micronuclei formation by all four test antioxi- dants. The order of ameliorative efficacy of these antioxi- dants on equimolar basis (10 μM) against flubendiamide was resveratrol > curcumin ≈ catechin > α-tocopherol while resveratrol ≈ α-tocopherol > curcumin > catechin against copper-induced micronuclei formation (Table 2). DNA fragmentation DNA of the flubendiamide, copper and dexamethasone treated splenocytes showed more shearing compared to the DNA of untreated splenocytes. DNA of the spleno- cytes treated concurrently with flubendiamide and equi- molar concentration (10 μM) of resveratrol, catechin or α-tocopherol also showed almost similar pattern of DNA shearing as observed in the DNA of flubendiamide alone treated splenocytes (Fig. 4). But DNA samples from curcumin (10 μM) + flubendiamide treated spleno- cytes showed less shearing compared to those treated with resveratrol + flubendiamide, catechin + flubendiamide or α-tocopherol + flubendiamide. Just like flubendiamide and curcumin treated splenocytes, DNA samples from copper + curcumin treated splenocytes also showed compara- tively less shearing than in the DNA from splenocytes treated with copper and other antioxidants (resveratrol, catechin, α-tocopherol) as shown in Fig. 5. Comet formation Comet formation data in splenocytes following their exposure to flubendiamide (40 μM), copper (40 μM) and dexamethasone (20 μM) alone revealed 22.90, formation compared to 29.59 and 27.69% comets Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 10 of 17 Fig. 3 Representative photographs of rat splenocytes showing micronuclei formation (100 X) following in vitro exposure to median lethal concentrations of flubendiamide and copper alone (40 μM) and in the presence of dimethyl sulphoxide (DMSO) and dexamethasone 3.09% in negative control and 4.58% in DMSO-treated splenocytes (Table 2; Fig. 6). Resveratrol, catechin, curcumin and α-tocopherol (10 μM each) were found formed in to reduce the percentage of comets flubendiamide and copper-treated splenocytes and the effect of all these agents was concentration-dependent (Table 2). Further, the ameliorative efficacy potential these antioxidants on equimolar basis against of Fig. 4 In vitro effect of median lethal concentration of flubendiamide and natural antioxidants at different concentrations on DNA fragmentation pattern in rat splenocytes. RV: Resveratrol (5 and 10 μM), Cath: Catechin (10 and 20 μM), A-T: α-tocopherol (5, 10 and 20 μM), Cur: Curcumin (5 and 10 μM), Flb: Flubendiamide, Dexa: Dexamethasone,Cont: Control Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 11 of 17 Fig. 5 In vitro effect of median lethal concentration of copper and natural antioxidants at different concentrations on DNA fragmentation pattern in rat splenocytes. RV: Resveratrol (5 and 10 μM), Cath: Catechin (10 and 20 μM), A-T: α-tocopherol (5, 10 and 20 μM). Cur: Curcumin (5 and 10 μM), Cu: Copper, Dexa: Dexamethasone, Cont: Control Fig. 6 Representative photographs of rat splenocytes showing comet formation following their in vitro exposure to median lethal concentrations of flubendiamide and copper Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 12 of 17 flubendiamide was α-tocopherol ≈ curcumin > cat- echin > resveratrol while resveratrol > curcumin > catechin > α-tocopherol against copper (Table 2). Phase II-in vivo chronic toxicity study Oxidative stress biomarkers data of rat spleens after 90 days of daily oral exposure to copper and flubendiamide alone and both these in combination (copper + fluben- diamide) and those treated simultaneously with α- tocopherol and test xenobiotics are presented in Table 3. Lipid peroxidation levels in rats of the groups exposed to flubendiamide or copper alone and copper + fluben- diamide were significantly (P < 0.05) higher, while re- duced glutathione (GSH) levels in flubendiamide and copper alone and copper + flubendiamide treated groups were moderately decreased when compared with the control group. Similarly, glutathione peroxidase (GPx) activity was found to be significantly (P < 0.05) decreased in rat groups exposed to copper, flubendiamide and cop- rats. per + flubendiamide Glutathione-S-transferase (GST) activity in flubendia- mide alone group was significantly (P < 0.05) lower (0.61 ± 0.12 μM of CDNB-GSH conjugate min-1 mg-1 protein) than in rest of the groups (Table 3). Further sig- nificant (P < 0.05) decrease in SOD activity was also ob- served in copper alone exposed group. Total protein content and catalase activity did not differ significantly between the control and any of the xenobiotics-treated groups. On simultaneous exposure of rats to xenobiotics and α-tocopherol, decrease in lipid peroxidation level and improvement in antioxidants (SOD, GST, GPx, cata- lase and GSH) cellular defense in splenic tissue of rats were observed compared to the rats exposed to xenobi- otics alone. compared to group I Fig. 8 Spleen section of copper sulphate (33 mg/kg) exposed group (IV) showing mild depletion of lymphoid tissue from white pulp (arrow) (10 X H&E stain) Rat spleens from the control groups (I, II and III) ex- hibited normal histoarchitecture characterized by nor- mal red and white pulps (Fig. 7). Spleen sections of copper sulphate group (IV) rats showed mild depletion of the lymphoid tissue from the white pulp (Fig. 8). Flu- bendiamide alone (V) group spleen showed separation of splenocytes and rearification in splenic parenchyma (Fig. 9). But spleen sections of copper + flubendiamide treatment group (VI) exhibited separation of splenocytes and rearification in splenic parenchyma (Fig. 10). Con- current treatment of the rats of groups VII, VIII and IX with α-tocopherol and copper sulphate, α-tocopherol and flubendiamide, and α-tocopherol and combination of flubendiamide and copper, respectively, showed al- most normal histoarchitecture of spleens as shown in Figs. 11, 12 and 13, respectively. Fig. 7 Section of spleen of rat from control group showing healthy histoarchitecture with red (arrow head) and white pulp (arrow) in splenic parenchyma (10X H&E stain) Fig. 9 Spleen section of flubendiamide (200 mg/kg) exposed group (V) showing separation of splenocytes and rarefication (arrow) in splenic parenchyma (10 X H&E stain) Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 13 of 17 Fig. 10 Spleen section of flubendiamide (200 mg/kg) + copper sulphate (33 mg/kg)-exposed group (VI) showing separation of splenocytes and rarefication (arrow) in splenic parenchyma (40 X H&E stain) Fig. 12 Spleen section of rats treated with α-tocopherol (100 mg/ kg) + flubendiamide (200 mg/kg) of group (VIII) showing normal histoarchitecture, abundant lymphoid tissue in the white pulp (arrow) suggestive of amelioration (40 X H&E stain) Discussion Humans and animals are being continuously exposed to mixture of agrochemicals and metals due to agricultural practices, working in heavy metals infested environment and use of ectoparasiticides and other chemicals in household practices [50, 51]. Therefore, an interaction study between different xenobiotics in human and ani- mal systems due to concurrent exposure to these chem- ical moieties along with their remedial measures seems very important. No information is available on the cyto-genotoxic ef- fects of flubendiamide and its possible mechanism. Markedly higher percentage of nonviable splenocytes in flubendiamide and copper treated groups evidently sug- gests cytotoxic effects of flubendiamide and copper in rat splenocytes similar to those reported with certain neonicotinoid insecticides in human peripheral blood lymphocytes [52]. Increase in the number of Tunel+ve cells in the present study are in agreement with increase in number of Tunel+ve germ cells in seminiferous tu- imidacloprid-treated rats [53] and Tunel+ve bules of fragmented DNA in brain and hippocampus of copper- treated mice and rats [54, 55]. Our findings suggest the ability of flubendiamide and copper to interact with double-stranded DNA (dsDNA) and induce cellular damage which enables TdT to bind with 3’OH label blunt ends of dsDNA and serve as a marker of apoptosis. Micronuclei assay is one of the most sensitive DNA damage indicator tests and is widely used for evaluation Fig. 11 Spleen section of rats treated with α-tocopherol (100 mg/ kg) + copper sulphate(33 mg/kg) of group (VII) showing normal histo-architecture with abundant lymphoid tissue in the white pulp (arrow) suggestive of amelioration (40 X H&E stain) Fig. 13 Spleen of rats treated with α-tocopherol (100 mg/kg) + flubendiamide (200 mg/kg) + copper sulphate (33 mg/kg) of group (IX) showing apparently healthy histoarchitecture with ample red (arrow head) and white pulp (arrow) in splenic parenchyma suggestive of amelioration (10 X H&E stain) Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 14 of 17 of genotoxic potential of environmental contaminants [56]. Micronuclei assay is employed to detect clastogen and aneugen properties of xenobiotics and determine mitotic delay, apoptosis, chromosome breakage, chromo- some loss and non-disjunction potential of xenobiotics [57, 58]. Increase in frequency of micronuclei formation in flubendiamide-treated splenocytes was almost com- induced by dexamethasone. Similar parable to that micronuclei forming effect of chlorpyrifos in fish eryth- rocytes [59] and imidacloprid in human peripheral blood lymphocytes has been reported [60]. Copper-induced in- crease in frequency of micronuclei formation in rat sple- nocytes is also in agreement with the observations in bone marrow cells of mice following exposure to copper, and erythrocytes [61, 62], gill and liver cells of fish fol- lowing exposure to cadmium and copper [63]. Other genotoxicity studies have also suggested that copper is a clastogenic agent [62, 64]. Apart from increase in number of Tunel+ cells and micronuclei formation, DNA fragmentation and comet assay studies also revealed that interaction of flubendia- mide or copper with splenocyte cells resulted in DNA damage which is manifested in the form of DNA strand breaks, laddering appearance of DNA in electrophoretic field and comet formation in alkaline conditions. Comet may be formed due to DNA single strand breaks, DNA double strand breaks, DNA adduct formations, DNA- DNA and DNA-proteincross-links or due to interaction of these xenobiotics with DNA [65, 66]. Similar DNA fragmentation in human peripheral blood mononuclear cells [67] and DNA fragmentation along with decrease in cell viability in HepG2 cells following exposure to copper has also been documented [68]. the against against adverse catalase effects of ROS the main defense Superoxide dismutase, are and glutathione free peroxidase radicals-induced oxidative stress and these act in concert with reduced glutathione and other antioxi- dants such as α-tocopherol and selenium that pro- tect [69]. Increased lipid peroxidation, a decrease in activities of antioxidant enzymes (SOD, GST, GPx) and GSH, separation of splenocytes, and rearification of splenic parenchyma revealed cellular damaging effects of flu- bendiamide and copper due to generation of oxida- tive stress. Depletion of GSH occurs as a result of excessive GSH consumption during oxidative stress [70, 71]. Further, GSH is not only a substrate for GPx, but is also involved in electrophile detoxifica- scavenging, α-tocopherol gener- tion, ation, phase II conjugation and other reactions [72, 73]. Glutathione-S-transferase catalyzes conjugation of glutathione with a number of electrophilic xenobi- otics and prevents their interaction with cellular pro- teins and nucleic acids, and plays an important role free radical in cellular defense against these xenobiotics [74, 75]. Therefore, inadequate detoxification of flubendiamide or copper or both these in combination, which amp- lified ROS generation and resulted in oxidative dam- age, may be responsible for these test compounds- induced decrease in membrane potential and in- crease in permeability to H+ and other ions, and eventually the cell contents release [76]. Copper-induced cyto-genotoxicity in the present study seems to be due to propensity of free Cu ions to partici- pate in formation of ROS by redox cycling and copper- induced formation of hydroxyl radicals from hydrogen peroxide via Haber-Weiss reaction [77–79]. Lipid peroxy radicals damage cells by changing the fluidity and per- meability of cell membrane or by attacking the cellular DNA molecule, leading to DNA strand brakes, oxidation of its bases and other intracellular molecules such as proteins [80, 81]. Copper-induced oxidative stress and apoptosis in kidney via intrinsic and extrinsic apoptotic pathways is also well documented [82]. in percentage of non-viable Simultaneous treatment of splenocytes with flubendia- mide or copper and either of these along with resveratrol, curcumin, catechin or α-tocopherol resulted in marked decrease splenocytes, Tunel+ve cells, and micronuclei and comet formation in splenocytes. Thus evidently suggests the ameliorative po- tential of these natural antioxidants against flubendiamide and copper-induced cytogenotoxic effects. The protective effect of resveratrol against xenobiotics is linked to de- crease in intracellular ROS accumulation, reactive oxygen intermediate (ROI) generation and lipid peroxidation [83, 84]. Attenuation of pyrogallol-induced hepatic toxicity and oxidative stress changes in hepatic damage and alter- ations in xenobiotic metabolizing enzymes by resveratrol has also been reported in Swiss mice [85]. Antiapoptotic property of catechin against copper is linked to chelation of Cu2+ and formation of an inactive complex with this metal, and thus prevention of gener- ation of potentially damaging free radicals [86, 87]. Simi- lar antiapoptotic, antioxidant and neuroprotective action of green tea extract, rich in various polyphenols, includ- ing catechin, against deltamethrin-induced neurotoxicity by improving oxidative status and DNA fragmentation, and suppressing the expression of apoptotic TP53 and COX2 genes has been reported in male rats [88]. Possi- bility of involvement of similar protective mechanisms of action of the test antioxidants against copper and flubendiamide-induced cytotoxic effects cannot be ruled out. Curcumin has been reported to ameliorate the arsenic and fluoride-induced genotoxicity in human peripheral blood lymphocytes [33]. Even curcumin has been dem- onstrated to be effective against radiations-induced haz- ards [89]. Protective effect of curcumin against different Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 15 of 17 xenobiotics has been attributed to its ability to decrease ROS generation, apoptosis, DNA fragmentation, and cell cycle arrest [33]. Anti-cytogenotoxic effect of curcumin in the present study against flubendiamide and copper could be attributed to its unique conjugated structure which facilitates the coupling reaction at 3′ position of the curcumin with lipids or due to its typical radical trapping ability as a chain-breaking antioxidant that in- hibits lipid peroxidation and reduces oxidative stress [90–93]. α-tocopherol is lipophilic in nature which facilitates its entry through cell membrane and thereby quenches free radical species, terminates lipid peroxidation chain reac- tion and thus interferes with initiation and progression of xenobiotics-induced oxidative damage [94, 95]. α- tocopherol produced protective effect against flubendia- mide and copper-induced oxidative stress in splenic tis- sues by modulating the oxidant-antioxidant mechanisms as substantiated by the altered values of different oxida- tive stress biomarkers. It also normalized the spleen his- toarchitecture towards almost normal as observed in rats of control groups. This evident ameliorative poten- tial of α-tocopherol is in agreement with our previous findings that studied flubendiamide and copper induced testicular injury [96]. Similar preventive effects of α- tocopherol against copper and cadmium-induced cyto- toxicity in COS-7 cells [81] and carbofuran-induced gen- otoxicity in human lymphocytes has also been reported [97]. Conclusions Summing up the findings of our in vitro and in vivo studies, it is apparent that flubendiamide and copper- induced alterations in oxidative stress biomarkers inter- act with cellular subcomponents, especially DNA and re- to splenocytes and spleen- in cytotoxic-insult sult histoarchitecture. Resveratrol is most effective against flubendiamide and curcumin against copper-induced cytotoxic effects, therefore, both these natural phyotcon- stituent antioxidants hold promising potential for their use in fortifying the conventional food-ingredients to prevent the adverse effects of xenobiotics on human and animals health. However, further studies on signaling intermediary steps and alterations in gene-expression are also warranted. Acknowledgements The authors are thankful to the Dean, College of Veterinary Science and Animal Husbandry, and Head of the Pharmacology Department for providing the necessary facilities in the laboratories established under Niche Area of Excellence Programme of ICAR and Rashtriya Krishi Vikas Yojana of Govt. of India. Routine financial assistance provided by the University for Doctoral Research to the first author is also duly acknowledged. Authors’ contributions RM searching of literature, conceiving of research plan and execution of research work. AP planning of the experimental design for in vitro study. AR supervision of the oxidative stress biomarkers study and assistance in laboratory work. SPS DNA isolation from splenocytes and gel-electrophoresis; and statistical analysis of data. DS DNA fragmentation assay and interpret- ation of results of DNA fragmentation assay. RK assistance and guidance in histopathological studies including interpretation. SKG planning and supervi- sion of research as Guide, interpretation of data and manuscript preparation. All the authors have read and approved the final manuscript. Funding Part of the PhD thesis of the first author. No additional funding was obtained for this study. Availability of data and materials Data-sets generated and/or analyzed during the current study are available in the thesis submitted by the first author in the University library and also available with the corresponding author on reasonable request. Ethics approval and consent to participate Present study was conducted on healthy male Wistar rats and the experimental protocol was dully approved by Institutional Animal Ethics Committee (IAEC) vide communication No. 79 IAEC/13 dated 16.07.13 (U.P. Pandit Deen Dayal Upadhyaya Pashu Chikitsa Vigyan Vishwavidyalaya Evam Go-Anusandhan Sansthan (DUVASU), Mathura-281001, India. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1Department of Veterinary Pharmacology and Toxicology, College of Veterinary and Animal Sciences, Sardar Vallabhbhai Patel University of Agriculture and Tecahnology, 250110, Meerut, India. 2Department of Veterinary Pharmacology and Toxicology, College of Veterinary Science and Animal Husbandry, U.P. Pt. Deen Dayal Upadhyay Pashu Chikitsa Vigyan Vishwavidyalaya Evam Go- Anusandhan Sansthan (DUVASU), -281001, Mathura, India. 3Division of Goat Health, Central Institute for Research on Goat (CIRG), Makhdoom, Farah, Mathura, Uttar Pradesh 281122, India. 4Department of Animal Genetics & Breeding, College of Veterinary Science and Animal Husbandry, U.P. Pt. Deen Dayal Upadhyay Pashu Chikitsa Vigyan Vishwavidyalaya Evam Go-Anusandhan Sansthan (DUVASU), 281001, Mathura, India. 5Department of Veterinary Pathology, College of Veterinary Science and Animal Husbandry, U.P. Pt. Deen Dayal Upadhyay Pashu Chikitsa Vigyan Vishwavidyalaya Evam Go-Anusandhan Sansthan (DUVASU), 281001, Mathura, India. Received: 29 September 2018 Accepted: 20 March 2020 Abbreviations LPO: Lipid peroxidation; GSH: Reduced glutathione; CAT: Catalase; GPx: Glutathione peroxidase; GST: Glutathione-S-transferase; SOD: Superoxide dismutase; ROS: Reactive oxygen species; IAEC: Institutional Animals Ethics Committee; PBS: Phosphate buffer saline; RPMI: Roswell Park Memorial Institute; HBSS: Hank’s balanced salt solution; SDS: Sodium dodecyl sulfate; NMA: Normal melting agarose; LMPA: Low melting point agarose; DMSO: Dimethyl sulfoxide References 1. Ebbinghaus-Kintscher U, Luemmen P, Lobitz N, et al. Phthalic acid diamides activate ryanodine sensitive Ca2 C release channels in insect. Cell Calcium. 2006;39:21–33. 2. Masaki T, Yasokawa N, Tohnishi M. Phthalic acid diamides activate ryanodine sensitive Ca2_ release channels in insects. Cell Calcium. 2006;39: 21–33. JMPR (Joint Meeting on Pesticide Residues). Report of the joint meeting of the FAO panel of experts on pesticide residues in food and the 3. Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 16 of 17 environment and the WHO core assessment group on pesticide residues at the request of codex committee on pesticide residues Italy; 2010. p. 345–82. 4. Miyahana K. Micronucleus test of NNI-0001 in mice. Unpublished report No. 5. 6. 7. 8. 9. LSRC-T02-089A from Nihon Nohyaku Co., Ltd, Osaka, Japan. Submitted to WHO by Bayer CropScience AG, Monheim, Germany. (2003). Herbold B. NNI-0001, micronucleus-test on the male mouse. Unpublished report No. AT01775 from Bayer Health Care, Wuppertal, Germany. Submitted to WHO by Bayer CropScience AG, Monheim, Germany. (2005). Tohnishi M, Nakao H, Furuya T, et al. Flubendiamide, a novel insecticide highly active against lepidopterous insect pests. J Pestic Sci. 2005;30:354–60. Kaur K, Kaur R. Occupational Pesticide Exposure, Impaired DNA Repair, and Diseases. Ind J Occup Environ Med. 2018;22(2):74–81. EC. Opinion of the Scientific Committee on Food (SCF) on the upper intake level of copper (27 March 2003). (2003). http://ec.europa.eu/food/fs/sc/scf/ out176_en.pdf. Shen HM, Lin Y, Choksi S, et al. Essential roles of receptor-interacting protein and TRAF2 in oxidative stress-induced cell death. Mol Cell Biol. 2004;24: 5914–22. 10. Gaetke LM, Chow CK. Copper toxicity, oxidative stress, and antioxidant 11. nutrients. Toxicology. 2003;189:147–63. Toplan S, Dariyerli N, Ozcelik D, Akyolcu MC. The effects of copper application on oxidative and antioxidant systems in rats. Trace Elem Elect. 2005;22:178–81. 12. Zhang SS, Noordin MM, Rahman SO, Haron J. Effects of copper overload on hepatic lipid peroxidation and antioxidant defense in rats. Vet Hum Toxicol. 2000;42:261–4. 13. Boone C, Gervais JA, Luukinen B, Buhl K, Stone D. Copper Sulfate Technical Fact Sheet; National Pesticide Information Center, Oregon State University Extension Services. 2012. http://npic.orst.edu/factsheets/cuso4tech.pdf. Oregon State University and the U.S. Environmental Protection Agency (cooperative agreement #X8-83947901). 14. Cano AP, Gillado AV, Montecillo AD, Herrera MU. Copper sulfate-embedded and copper oxide-embedded filter paper and their antimicrobial properties. Mater Chem Phys. 2018;207:147–53. 16. 15. Central Water Commission, Ministry of Water Resources, River Development and Ganga Rejuvnation, Government of India (CWC, GOI). Status of trace and toxic metals in Indian rivers. River Data Compilation −2 Directorate Planning and Development Organisation New Delhi 110066. 2018. Filho DW, vila Júnior SÁ, Possamai FP, Possamai FP, Parisotto EB, Moratelli AM, Garlet TR, Ina’cio DB, Torres MA, Colepicolo P, Dal-Pizzol F. Antioxidant therapy attenuates oxidative stress in the blood of subjects exposed to occupational airborne contamination from coal mining extraction and incineration of hospital residues. Ecotoxicology. 2010;19:1193–200. Lobo V, Patil A, Pathak A, Chandra N. Free radicals, antioxidants and functional foods: impact on human health. Pharmacogn Rev. 2010;8:118–26. 17. 18. Halliwell B. How to characterize an antioxidant- an update. Biochem Soc Symp. 1995;61:73–101. 19. Wolf G. The discovery of the antioxidant function of vitamin E: the contribution of Henry a. Mattill J Nutr. 2005;135:363–6. 20. Yavuz T, Delibas N, Yildirim B, et al. Vascular wall damage in rats induced by methidathion and ameliorating effect of vitamins E and C. Arch Toxicol. 2004;78:655–9. 21. Herrera E, Barbas C. Vitamin E: action, metabolism and perspectives. J 24. 22. 23. Physiol Biochem. 2001;57(2):43–56. Traber MG, Atkinson J. Vitamin E, antioxidant and nothing more. Free Radic Biol Med. 2007;43(1):4–15. Lorenzo JM, Munekata PES. Phenolic compounds of green tea: health benefits and technological application in food. Asian Pac J Trop Biomed. 2016;6(8):709–19. Kagaya N, Tagawa Y, Nagashima H, et al. Suppression of cytotoxin-induced cell death in isolated hepatocytes by tea catechins. Eur J Pharmacol. 2002; 450:231–6. Jeong JH, Kim HJ, Lee TJ, et al. Epigallocatechin 3-gallate attenuates neuronal damage induced by 3-hydroxykynurenine. Toxicology. 2004;195:53–60. Levites Y, Amit T, Youdim MB, Mandel S. Involvement of protein kinase C activation and cell survival/cell cycle genes in green tea polyphenol (_)-epigallocatechin 3- gallate neuroprotective action. J Biol Chem. 2002;277:30574–80. 27. Du GJ, Zhang Z, Wen XD, Yu C, Calway T, Yuan CS, Wang CZ. 25. 26. Epigallocatechin Gallate (EGCG) is the most effective cancer chemopreventive polyphenol in green tea. Nutrients. 2012;4(11):1679–91. https://doi.org/10.3390/nu4111679. 28. Luthra PM, Singh R, Chandra R. Therapeutic use of Curcumin longa. Ind J Clin Biochem. 2001;16:153–60. 29. Ak T, Gülçin I. Antioxidant and radical scavenging properties of curcumin. 30. 31. Chem Biol Interact. 2008;174:27–37. Shukla PK, Khanna VK, Khan MY. Protective effect of curcumin against lead neurotoxicity in rats. Hum Exp Toxicol. 2003;22:653–8. Eybl V, Kotyzova D, Leseticky L, et al. The influence of curcumin and manganese complex of curcumin on cadmium-induced oxidative damage and trace element status in tissues of mice. J Appl Toxicol. 2006;26:207–21230. 32. Oda Y. Inhibitory effect of curcumin on SOS functions induced by UV 33. irradiation. Mutat Res. 1995;348:67–73. Tiwari T, Rao MV. Curcumin supplementation protects from genotoxic effects of arsenic and fluoride. Food Chem Toxicol. 2010;48:1234–8. 34. Prakash A, Khan S, Kumar D, et al. Concurrent administration of curcumin mitigates arsenic – and chlorpyrifos – induced apoptosis in rat thymocytes. Adv Anim Vet Sci. 2014;7:407–13. 35. Gatz SA, Wiesmuller L. Take a break-resveratrol in action on DNA. Carcinogenesis. 2008;29:321–32. 36. Mehendale SR, Wang CZ, Shao ZH, et al. Chronic pretreatment with American ginseng berry and its polyphenolic constituents attenuate oxidant stress in cardiomyocytes. Eur J Pharmacol. 2006;553:209–14. 37. Du Y, Guo H, Lou H. Grape seed polyphenols protect cardiac cells from apoptosis via induction of endogenous antioxidant enzymes. J Agric Food Chem. 2007;55:1695–701. 38. Boyse EA, Miyazawa M, Aoki T, et al. Ly-A and Ly-B: two systems of lymphocyte antigens in the mouse. Proc roy Soc B. 1968;170:175. 39. Hayashi M, Sofuni T, Ishidata MJ. An application of acridine orange 40. fluorescent staining to the micronucleus test. Mutat Res. 1983;120:241–7. Sambrook J, Russel DW. Molecular cloning: a laboratory manual. New York: Cold/Spring Harbor Laboratory Press; 2001. p. A1–A12. 42. 41. Dhawan A, Bajpayee M, Pandey AK, et al. Protocol for the single cell gel electrophoresis/comet assay for rapid genotoxicity assessment. Lucknow: ITRC; 2009. p. 1–10. Shafiq-U-Rehman. Lead-induced regional lipid peroxidation in brain. Toxicol Lett. 1984;21:333–7. Sedlak J, Lindsay RH. Estimation of total protein bound NPSH groups in tissues with Ellaman’s reagent. Anal Biochem. 1968;25:192–205. 44. Bergmeyer HU. U.V. method of catalase assay. In: Methods of enzymatic 43. analysis, vol. III. 3rd ed. Weinheim, Dacrfield Beach, Florida, Basal; 1983. p. 273. 45. Madesh M, Balasubramanian KA. Microtiter plate assay for superoxide dismutase using MTT reduction by superoxide. Indian J Biochem Biophys. 1998;35:184–8. 46. Habig WH, Pabst MJ, Jacokoby WB. Gluuathione-S-transferase: the first enzymatic step in mercapturic acid formation. J Biol Chem. 1974;249:7130–9. 47. Paglia DE, Valantine WN. Studies on the quantitative and qualitative characterization of erythrocyte glutathione peroxidase. J Lab Clin Med. 1967; 70:158–69. Lowery OH, Rosenbrough NJ, Farr AL, Randall RJ. Protein measurement with the Folin phenol reagent. J Biol Chem. 1951;193:265–75. Luna LG. Manual of histology staining methods of the armed forces Institute of Pathology. 3rd ed. New York: Mcgraw-Hill Book Co; 1968. International Agency for Research on Cancer. Monographs-cadmium. Lyon: IARC; 2003. 48. 49. 50. 51. Agency for Toxic substance and Disease Registry. Draft Toxicological profile 52. for cadmium. Atlanta: ATSDR; 2007. Segura MEC, Gomez-Arroyo S, Villalobos-Pietrini R, et al. Evaluation of genotoxic and cytotoxic efffects in human peripheral blood lymphocytes exposedi n vitro to neonicotinoid insecticides news. J Toxicology. 2012; 2012:1–11. https://doi.org/10.1155/2012/612647. 54. 53. Bal R, Türk G, Tuzcu M, Yilmaz O, et al. Assessment of imidacloprid toxicity on reproductive organ system of adult male rats. J Envi Sci Health, Part B. 2012;47:434–44. Lu J, Yaun-lin Z, Dong-Mei W, et al. Trace amounts of copper induce neurotoxicity in the cholesterol-fed mice through apoptosis. FEBS Lett. 2006;580:6730–40. Tufan AC, Akdogan I, Turgut G, et al. Increased Tunel positive cells in CA1, CA2, and CA3 subfields of rat hippocampus due to copper and ethanol co- exposure. International J Neurosci. 2008;118:647–56. Lemos AO, Oliviera NCD, Lemos CT. In vitro micronuclei test to evaluate the genotoxicity of surface water under the influence of tanneries. Toxicol in Vitro. 2011;25:761–6. 56. 55. Mandil et al. BMC Pharmacology and Toxicology (2020) 21:29 Page 17 of 17 82. Sarkar A, Das J, Manna P, Sil PC. Nano copper induces oxidative stress and apoptosis in kidney via both extrinsic and intrinsic pathways. Toxicol. 2011; 290:208–17. 83. Cao Z, Li Y. Potent induction of cellular antioxidants and phase 2 enzymes by resveratrol in cardiomyocytes: protection against oxidative and electrophilic injury. Eur J Pharmacol. 2004;489:39–48. 84. Wu MX, Ao Z, Prasad KV, Wu R, Schlossman SF. Science. 1998;281:998–1001. 85. Upadhyay G, Singh AK, Kumar A, Prakash O, Singh MP. Resveratrol modulates pyrogallol-induced changes in hepatic toxicity markers, xenobiotic metabolizing enzymes and oxidative stress. Eur J Pharmacol. 2008;596(1–3):146–52. Kashima M. Effects of catechins on superoxide and hydroxyl radical. Chem Pharm Bull. 1999;47:279–83. Seeram NP, Nair MG. Inhibition of lipid peroxidation and structure–activity- related studies of the dietary constituents anthocyanins, anthocyanidins, and catechins. J Agric Food Chem. 2002;50:5308–12. 87. 86. 88. Ogaly HA, Khalaf AA, Ibrahim MA, Galal MK, Abd-Elsalam RM. Influence of green tea extract on oxidative damage and apoptosis induced by deltamethrin in rat brain. Neurotoxicol Teratol. 2015;50:23–31. Jagetia GC. Radioprotection and radiosensitization by curcumin. Adv Exp Med Biol. 2007;595:301–20. 89. 90. Masuda T, Maekawa T, Hidaka K, Bando H, Takeda Y, Yamaguchi H. Chemical studies on antioxidant mechanisms of curcumin: analysis of oxidative coupling products from curcumin and linoleate. J Agric Food Chem. 2001;49:2539–47. Sreejayan N, Rao MN. Curcuminoids as potent inhibitors of lipid peroxidation. J Pharm Pharmacol. 1994;46:1013–6. 91. 92. Gutteridge JM, Halliwell B. Free radicals and antioxidants in the year 2000. A historical look to the future. Ann NY Acad Sci. 2000;899:136–47. 93. Chan MM, Huang HI, Fenton MR, Fong D. In vivo inhibition of nitric oxide synthase gene expression by curcumin, a cancer preventive natural product with anti-inflammatory properties. Biochem Pharmacol. 1998;55:1955–62. 94. Bisby RH, Johnson SA, Parker AW. Quenching of reactive oxidative species by probucol and comparison with other antioxidants. Free Radic Biol Med. 1996;20:411–20. Kalender S, Kalender Y, Ogutcu A, Uzunhisarcikli M, Durak D, Acikgoz F. Endosulfan-induced cardiotoxicity and free radical metabolism in rats: the protective effect of vitamin E. Toxicology. 2004;202:227–35. 95. 96. Mandil R, Rahal A, Prakash A, Garg SK, Gangwar NK, Swain DK. Ameliorative potential of α-tocopherol against flubendiamide and copper-induced testicular-insult in Wistar rats. Chem Biol Interact. 2016;260:91–101. Sharma RK, Sharma B. In-vitro carbofuran induced genotoxicity in human lymphocytes and its mitigation by vitamins C and E. Dis Markers. 2012;32: 153–63. 97. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 57. Garriott ML, Phelps JB, Hoffman WP. A protocol for the in-vitro micronucleus test I. contribution to the development of a protocol suitable for regulatory submissions from an examination of 16 chemicals with different mechanism of action and level of activity. Mutation Res. 2002;517: 123–34. 58. Corvi R, Albertini S, Hartung T, et al. ECVAM retrospective validation of in vitro micronucleus test (MNT). Mutagenesis. 2008;7:1–13. 59. Ali D, Nagpure NS, Kumar S, et al. Assessment of genotoxic and mutagenic effects of chlorpyrifos in freshwater fish Channa punctatus (Bloch) using micronucleus assay and alkaline single-cell gel electrophoresis. Food Chem Toxicol. 2009;47:650–6. 60. Costa C, Silvari V, Melchini A, et al. Genotoxicity of imidacloprid in relation to metabolic activation and composition of the commercial product. Mutat Res-Genet Toxicol Environ Mutagenesis. 2009;672:40–4. 61. Bhunya SP, Pati PC. Genotoxicity of an inorganic pesticide, copper sulphate in mouse in vivo test system. Cytologia. 1987;52:801–8. 62. Agarwal K, Sharma A, Talukder G. Clastogenic effects of copper sulfate on the bone marrow chromosomes of mice in vivo. Mutat Res. 1990;243:1–6. 63. Cavas T, Garanko NN, Arkhipchuk VV. Induction of micronuclei and binuclei in blood, gill and liver cells of fishes subchronically exposed to cadmium chloride and copper sulphate. Food Chem Toxicol. 2005;43:569–74. 64. Bhunya SP, Jena GB. Clastogenic effect of copper sulphate in chick in vivo 65. test system. Mutat Res. 1996;367:57–63. Fairbairn DW, Olive PL, O’Neil KL. The comet assay comprehensive review. Mutat Res. 1995;399:32–59. 66. Mitchelmore CL, Chipman JK. DNA strand breakage in aquatic organisms and the potential value of the comet assay in environmental monitoring. Mutat Res. 1998;399:135–47. Singh RP, Kumar S, Nada R, et al. Evaluation of Cu toxicity in isolated human peripheral blood mononuclear cells and it's attenuation by zinc: Ex Vivo. Mol Cell Biochem. 2006;282:13–21. 67. 68. Vidyashankar S, Patki PS. Liv.52 attenuate copper induced toxicity by inhibiting glutathione depletion and increased antioxidant enzyme activity in HepG2 cells. Food Chem Toxicol. 2010;48:1863–8. 69. Halliwell B, Gutteridge JMC. Free radicals, other reactive species in: free 70. radicals in biology and Madicine. Oxford: Oxford University Press; 1999. p. 617–783. Lieber CS. S-Adenosyl-L-methionine and alcoholic liver disease in animal models: implications for early intervention in human beings. Alcohol. 2002; 27:173–7. 71. Hung YC, Sava VM, Blagodarsky VA, et al. Protection of tea melanin on hydrazineinduced liver injury. Life Sci. 2003;72:1061–71. 72. Ookhtens M, Kaplowitz N. Role of the liver in interorgan homeostasis of glutathione and cyst(e)ine. Semin Liver Dis. 1998;18:313–29. 73. Mahboob M, Siddiqui MK. Alteration in hepatic detoxifying enzymes by new organophosphorous insecticides following subchronic exposure in rats. JAppl Toxicol. 2001;21:5015–505. 74. Hayes JD, Pulford D. The glutathione- S-transferase supergene family: regulation of GST and the contribution of the isoenzymes to cancer chemoprotection and drug resistance. Crit Rev Biochem Mol Biol. 1995;30: 445–600. Josephy PD. Genetic variations in human glutathione transferase enzymes: significance for pharmacology and toxicology. Hum. Genomics Proteomics. 2010;2010:876940–876953.1–14. https://doi.org/10.4061/2010/876940. 76. Halliwell B, Gutteridge JMC. Free radicals in biology and medicine. 4th ed. 75. 77. Oxford: Clarendon Press; 2001. Kadiiska MB, Hanna PM, Jordan SJ, Mason RP. Electron spin resonance for free radical generation in copper-treated vitamin-E and selenium-deficient rats: in vivo spin-trapping investigation. Mol Pharmacol. 1993;44:222–7. 78. Bremner I. Manifestations of copper excess. Am J Clin Nutr. 1998;67:1069S– 73S. 79. Pourahmad J, O’Brien PJ. A comparison of hepatocytes cytotoxic mechanisms for Cu_2 and Cd_2. Toxicology. 2000;143:263–73. Kawanishi S, Inoe S, Yamamoto K. Hydroxyl radical and singlet oxygen production and DNA damage induced by carcinogenic metal compounds and hydrogen peroxide. Biol Trace Elem Res. 1989;21:367–72. 80. 81. Mattie MD, Freedman JH. Protective effects of aspirin and vitamin E (alpha- tocopherol) against copper and cadmium-induced toxicity. Biochem Biophys Res Commun. 2001;285:921–5.
10.1371_journal.pcbi.1009407	RESEARCH ARTICLE Thalamic bursts modulate cortical synchrony locally to switch between states of global functional connectivity in a cognitive task Oscar PortolesID P. BorstID 1* 1,2, Manuel Blesa3, Marieke van VugtID 1, Ming CaoID 2, Jelmer a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Portoles O, Blesa M, van Vugt M, Cao M, Borst JP (2022) Thalamic bursts modulate cortical synchrony locally to switch between states of global functional connectivity in a cognitive task. PLoS Comput Biol 18(3): e1009407. https://doi. org/10.1371/journal.pcbi.1009407 Editor: Daniel Bush, University College London, UNITED KINGDOM Received: August 23, 2021 Accepted: February 16, 2022 Published: March 9, 2022 Peer Review History: PLOS recognizes the benefits of transparency in the peer review process; therefore, we enable the publication of all of the content of peer review and author responses alongside final, published articles. The editorial history of this article is available here: https://doi.org/10.1371/journal.pcbi.1009407 Copyright: © 2022 Portoles et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All code written in support of this publication, processed data, simulation input files and simulation out put files are publicly available at https://gin.g-node.org/ 1 Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence, Faculty of Science and Engineering, University of Groningen, Groningen, The Netherlands, 2 Engineering and Technology Institute Groningen, Faculty of Science and Engineering, University of Groningen, Groningen, The Netherlands, 3 MRC Centre for Reproductive Health, University of Edinburgh, Edinburgh, United Kingdom o.portoles.marin@rug.nl (OP); j.p.borst@rug.nl (JP) Abstract Performing a cognitive task requires going through a sequence of functionally diverse stages. Although it is typically assumed that these stages are characterized by distinct states of cortical synchrony that are triggered by sub-cortical events, little reported evidence supports this hypothesis. To test this hypothesis, we first identified cognitive stages in sin- gle-trial MEG data of an associative recognition task, showing with a novel method that each stage begins with local modulations of synchrony followed by a state of directed func- tional connectivity. Second, we developed the first whole-brain model that can simulate cor- tical synchrony throughout a task. The model suggests that the observed synchrony is caused by thalamocortical bursts at the onset of each stage, targeted at cortical synapses and interacting with the structural anatomical connectivity. These findings confirm that cog- nitive stages are defined by distinct states of cortical synchrony and explains the network- level mechanisms necessary for reaching stage-dependent synchrony states. Author summary A novel machine-learning method was applied to unveil the dynamics of local and cortex- wide neural coordination underlying the fundamental cognitive processes involved in a memory task. To explain how neural activity–and ultimately behavior–was coordinated throughout the task, we developed a whole-brain model that incorporates cognitive mech- anisms, anatomy, and neural biophysics. Similar models are regularly used with resting state data, but simulating a cognitive task remained elusive. By using hidden semi-Markov models to divide the task into stages with separate connectivity patterns, we were able to generalize the whole brain model from resting state to cognitive task data. The model showed that sub-cortical pulses at the onset of cognitive processes–as hypothesized by cognitive and neurophysiological theories–were sufficient to switch between the states of neural coordination observed. These findings have implications for understanding goal- PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 1 / 20 PLOS COMPUTATIONAL BIOLOGYoportoles/SwitchingFCalongAtask. Raw data is available at https://www.jelmerborst.nl/models/. Funding: MC was supported in part by the European Research Council (ERC-CoG-771687); https://erc.europa.eu/. The PhD of OP was largely funded by the Data and Systems Complexity Centre of the University of Groningen. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. Switching between functional connectivity states along a cognitive task directed cognitive processing and the mechanisms needed to reach states of neural coordination. Introduction Already in the 19th century, Donders hypothesized that information processing in the brain proceeds through a sequence of fundamental cognitive stages with different functions such as visual encoding, memory retrieval, and decision making [1]. Initially, cognitive stages were investigated with behavioral metrics like reaction time (e.g., [2]). Over the past decade, neuro- imaging analyses have begun to uncover the neural correlates of these cognitive stages (e.g., [3]). The dominant view is that cognitive stages require specific patterns of neural coordination across the cortex [3–5]. The transition from one cognitive stage to the next is thought to be driven by the basal-ganglia-thalamus (BGT) system which sets new states of cortical coordina- tion [6–8]. The striatum monitors the current state of the cortex, and based on a comparison to predefined states, selects and triggers the next cognitive stage. The role of the BGT system modulating cortical coordination is supported by animal studies, intracranial recordings, and neural models [9–14]. However, the network-level mechanisms required to reach a new state of cortical coordination from subcortical inputs are poorly understood. To give a detailed account of these mechanisms, one first needs to characterize the different states of neural coordination within the sequence of cognitive stages. We measured neural activity with cortically-projected magnetoencephalographic (MEG) recordings as these have a sufficiently fine temporal resolution to measure cognitive stages, as well as adequate spatial res- olution [5]. However, cognitive stages have high temporal variability–that is, stages typically have a different duration on each trial of an experimental task–which makes it difficult to mea- sure neural coordination. To overcome this problem, we used a machine learning method that identifies the onsets of cognitive stages on a trial-by-trial basis [15]. Afterwards, the identified stage onsets were used to time-lock the measures of neural coordination within regions (local synchrony) and between regions (functional connectivity, FC), as there are concurrent changes at both spatial scales [16,17]. Specifically, we focused on coordination of theta band oscillations as the thalamus is thought to modulate local theta activity [18] which may change theta band FC [19], and which in turn may modulate the activity in higher frequency bands [20]. The machine learning method that we used to identify cognitive stages combines multivari- ate pattern analysis with a Hidden semi-Markov Model (HSMM-MVPA). The HSMM-MVPA method searches in each trial for a sequence of short-lived modulations of MEG amplitude (hereafter called bumps, following the original paper [15]) that have a consistent topology across trials. These bumps signify the onset of cognitive stages, and are thought to be triggered by the BGT system. Previously this method has been used successfully to, for example, identify the cognitive stages that are affected by task manipulations such as difficulty, stage insertion, and evidence accumulation for decisions [4,5,15,21]. To understand how events from the BGT system can cause switches between states of neu- ral coordination–and thus between cognitive stages–we build upon generative whole-brain biophysical models of large-scale activity (GWBM) that have been used to explain the dynam- ics of neural coordination at rest [22]. GWBMs reduce the whole-brain network of neurons and synapses to a smaller network that still incorporates the most relevant principles of neural PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 2 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task dynamics. The nodes of such a network describe the macroscopic activity within a region, while the links reflect the neural fibers that connect these regions (i.e. structural connectivity). GWBMs of resting state indicate that time-resolved patterns of neural coordination depend on the anatomical structure of the brain and that these patterns evolve without requiring any input (a phenomenon referred to as metastable coordination; [17,22]. Such coordination dynamics are thought to provide an optimal mechanism for simultaneously integrating and segregating information that allows the system to adapt quickly or alternatively, to persist in a given state [23]. While this is sufficient to explain resting-state data, cognitive tasks require specific, controlled sequences of coordination states. Here, we explored a GWBM in which inputs from the BGT system modulated local connec- tivity strength briefly at the onset of cognitive stages, as suggested by cognitive theories and electrophysiology measurements [6–14]. In other complex networks with similar dynamics as the brain, such local perturbations can, in turn, produce controlled switches between global states [24]. Similarly, even though the inputs from the BGT system only triggered direct changes in local connectivity strength, we observed transient modulations of local synchrony and switches to the targeted states of directed functional connectivity that lasted until the next input. When there were no further inputs from the BGT system, neural coordination returned to resting-state patterns after tens of seconds. These results matched the observed neural coor- dination throughout the cognitive stages in the empirical data. Finally, we used the GWBM to determine the importance of each brain region in facilitating the switches between states of coordination. Results Five cognitive stages in an associative memory task We re-analyzed MEG data from an associative memory recognition task with 18 participants [3]. We chose this task because associative recognition memory involves a rich variety in cog- nitive stages that have also been widely studied [3,5,15,25,26]. The task consisted of a self- directed learning phase during which participants memorized 32 word pairs and a test phase. In the test phase–which we analyzed here–participants were again presented with word pairs. These could be target pairs from the learning phase or re-paired foil pairs, which consisted of the same words paired differently (e.g., if the participants learned apple-tree and month- house, a foil pair could be apple-house). Participants were asked to indicate as quickly and accurately as possible with a key press if it was a learned pair or a re-paired foil. Only correct responses were included in our analysis. We were interested in the evolution of neural coordi- nation along with the cognitive stages involved in performing the task, and in particular in how the brain switches between these consecutive states of functional neural coordination. As the goal is to develop a cortical model, the MEG signals were projected onto 5,124 corti- cal sources using the structural MRI of each participant with minimum-norm estimation [3]. The resulting cortical activity was aggregated into 68 cortical regions following the Desikan- Killiany atlas [27]. Each cortical region in the atlas contained the average activity of the cortical sources within that region. Next, HSMM-MVPA was used to estimate the timing of bumps that indicate the onset of cognitive stages in each trial. All trials were assumed to go through the same sequence of stages as in previous studies [3,4,15]. Thus, bumps were assumed to have the same spatial topology across trials, but trial-to-trial variable temporal location. Neverthe- less, the HSMM-MVPA can cope relatively well with extra bumps in some trials [15]. The intervals between stimulus-onset-to-bump, bump-to-bump, and bump-to-response constitute the cognitive stages. A leave-one-subject-out cross validation method showed that the MEG PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 3 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task Fig 1. Theta-band MEG local synchrony and directed functional connectivity by cognitive stages. (A) Cognitive stages derived with the HSMM-MVPA along with their median durations. (B) Significant directed functional connectivity throughout the stage (within-stage dpFC). Links go from phase-ahead to phase-behind regions. The nodes represent the nodal degree (size) and the difference between phase-ahead and phase-behind links (color). (C) Directed functional connectivity at every sample time-locked at the onset of the stages (across-trials dpFC). Colored (dark gray) line: average across links with (without) significant across-trial dpFC at the current stage; Shading: standard error of the mean across subjects. Black vertical lines indicate the onsets of the stages. The white background spans the median stage duration. Retrieval and response insets: Directed functional connectivity time-locked to the onsets of the decision stage and to the end-of-trial response, respectively (D) Across-trials averaged local synchrony (z-scored envelope of amplitudes) time-locked at the onset of the stages. Y-axis represents cortical regions – blue: temporal, orange: occipital, red: parietal, and green: frontal. Magenta lines define the time windows used to measure the relative change in local coordination at the onset of the stages. First and second windows span -60 to -10 ms, and 0 to 50 ms with respect to stage onset. (E) Histogram of stage durations derived with the HSMM-MVPA. https://doi.org/10.1371/journal.pcbi.1009407.g001 data were best explained by a HSMM-MVPA model with four bumps, which corresponds to five cognitive stages (Fig 1A). Following previous work on associative recognition [3,5,15], we interpreted the five cogni- tive stages as follows: pre-encoding, encoding of visual information, memory retrieval, deci- sion making, and motor response. We did not analyze the pre-encoding stage as it is mostly driven by the task stimulus and not by events from the BGT system that produce the transi- tions between cognitive stages. The retrieval and motor stages were longer and had larger across-trial variability than the encoding and decision stages (Fig 1A and 1E). Different local synchrony and directed functional connectivity states between stages Next, we measured neural coordination in the discovered stages. We focused on coordination of theta band oscillations (4–8 Hz), for several reasons: we previously found synchrony pat- terns in this frequency band to vary across task stages [4]; theta oscillations have been related PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 4 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task to cognitive processes such as attention, memory, control, and decision making [28–31]; the phase of theta oscillations is known to modulate the activity in higher frequency bands [20,29]; local modulations of theta-band activity are hypothesized to mediate changes in long-range functional connectivity [19]; and thalamic activity modulates cortical theta band activity [18]. Directed FC in the theta band was operationalized by means of the directed phase-lag index [32] (dpFC). The directed phase-lag index measures the consistency of the sign of the differ- ence between the phases of two signals. Such consistency can exist either over a period of time or across trials at a given time point. We measured first within-stage dpFC to capture directed FC states that are constant from the start to the end of a cognitive stage. Fig 1B shows the links with significant within-stage dpFC, as well as the local difference between phase-ahead and phase-behind links (node color) and the total number of links regardless of their direction (node size). The significance of dpFC was obtained using a permutation in which we created 200 surrogate data sets with random circular shifts of the original phases. Such circular shifts keep the structure of the phases, while they destroy the temporal relationship between a pairs of phase signals. We then examined whether the empirically observed phase lags were more consistent than this population of dpFCs from randomly shifted signals. A significant dpFC indicates that the phase in one region is consistently ahead or behind another region. Next, we measured across-trial dpFC to reveal, sample-by-sample, the temporal evolution of sign-con- sistent phase differences across trials during each stage (Fig 1C). Across-trial dpFC was calcu- lated at every sample with the trials time-locked to the onset of each of the stages. Time- locking to the onset of stages rather than an absolute time relative to stimulus onset allows for aligning cognitive processes that start at different times on each trial. Across-trial dpFC revealed that functional states of directed FC switch at the transition between cognitive stages. These switches are visible because across-trial dpFC takes into account the trial-to-trial temporal variability of the cognitive stages as revealed by the HSMM-MVPA analysis. For example, for the memory retrieval stage, across-trial dpFC seems to fade halfway through. However, when across-trial dpFC is time-locked to the onset of the next stage–the decision stage–dpFC for memory retrieval materializes until shortly before the decision stage (see the insets in Fig 1). This illustrates why the HSMM-MVPA analysis is cru- cial: otherwise dpFC would appear to fade quickly after stimulus onset, while that is not the case when first isolating cognitive stages. Local synchrony was operationalized as the envelope of the theta band analytic signals in each region, which indicates the degree of synchronous neural activity within a region. The envelopes were z-scored over time and then averaged across trials and participants. Across- trial averages were time-locked to the onset of cognitive stages which gave a time course of local synchrony for each stage (Fig 1D). This showed that the local modulations of synchrony occurred only briefly at the start of each stage, and involved different regions depending on the cognitive operations involved in that stage. As expected, each stage had a different neural coordination pattern. In the visual encoding stage, occipital and left-temporal regions showed local synchrony and dpFC which might facil- itate the transfer of visual information to the medial temporal lobe and the hippocampus to start a retrieval process [26]. The encoding of information is controlled by a large fronto-poste- rior, fronto-lateral network [28,29]. During the memory retrieval stage, local synchrony at occipital and temporal regions is reduced. dpFC now happens mostly between left-medial- temporal and frontal regions, whose coordination is required for memory tasks [26,33]. At the onset of decision making the frontal regions begin to synchronize locally. The decision process is mediated by fronto-parietal dpFC [30], and dpFC between temporal and parietal regions dpFC to reinsert the memory retrieved into the left-parietal cortex [26]. Finally, at the motor stage a large dpFC network appears between motor, temporal, left-parietal, and pre-frontal PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 5 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task regions. This complex network has previously been associated with motor preparation, action reevaluation, decision, and cognitive control [30,31,34], in line with the idea that the action is reevaluated during the motor response [35]. Together, these analyses unveiled that right at the onset of a cognitive stage there is a reor- ganization of neural coordination in the cortex. Whereas the change in local synchrony was only brief, dpFC lasted throughout the cognitive stage, indicating that short modulations of local synchrony can have persistent global effects. Next, we used a GWBM to investigate the mechanism underlying this. Generative large-scale whole-brain model (GWBM) In order to integrate cognitive stages and neural coordination into one framework along with neural anatomy and neural dynamics, we used a parsimonious GWBM that describes within- and between-region modulations of synchrony. Previously, we have used this model to dem- onstrate that modulations of local synchrony are related to time-resolved FC during resting state [17]. This GWBM is a low-dimensional reduction of a network-of-networks of Kuramoto oscil- lators [36]. Kuramoto oscillators describe the dynamics of synchrony in biological systems including neural networks [22,37,38]. Each sub-network in this study represents a cortical region from the Desikan-Killiany atlas. All units in a region are assumed to be fully and instantly connected, while connections between regions are weighted and delayed by the den- sity and length of the neural fibers in MRI-derived structural connectivity networks. The regions in the GWBM were defined with the same parcellation atlas as the MEG data that we sought to model. First, we set default values for local connectivity strength (L in Eq 3, identical for all regions), and global scaling (G in Eq 3 and Eq 4) of the structural connectivity such that the model simulated resting-state coordination dynamics in the theta-band [22,39]. Resting-state dynamics are characterized by fluctuations over time of the local and global synchrony as well as time-resolved FC patterns (i.e. local and global metastability) [17,22,39]. These dynamical properties of resting state neural coordination were identified with GWBMs simulated over a grid of L and G values. The identified L and G values displayed the most similar dynamics to local and global metastability in the grid search. (see S1A Fig). Next, we simulated the switching between cognitive states by adding short inputs (30 milliseconds) from the BGT system–specifically from the thalamus to the cortex–at the onset of cognitive stages. The rationale for using this mechanism derives from theories of cognition and data derived from electrophysiology. Specifically, cognitive theories state that the BGT system modulates cortical synchrony at the onset of cognitive stages via tha- lamocortical signals [6,25]. Electrophysiology has shown that thalamocortical neurons can indeed drive cortical activity [10,13] and establish FC [40,41]. These thalamocortical neurons tend to produce short burst of activity [42], which target pools of either excitatory or inhibitory cortical neurons specifically [12,43]. Therefore, our model simulated thala- mocortical inputs as short pulses of increased or decreased local connectivity strength (L in Eq 3) that represent transient modulations of excitatory or inhibitory synaptic activity [38]. To simulate the sequence of neural coordination states found in the MEG data, we esti- mated the magnitude of the required activity pulses simultaneously in all regions, stage-by- stage. Each cortical region received one pulse at the onset of each processing stage. The optimi- zation scheme to obtain the magnitude of the pulses maximized concurrently the fitness of local synchrony and within-stage dpFC, while minimizing the total magnitude of the pulses. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 6 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task The optimization was accomplished with the generalized island model for distributed evolu- tionary optimization which in relatively short time explores and exploits different areas of the parameter space simultaneously [44]. Changes in local connectivity cause switches between global states of cognitive coordination To assess how well the model simulated local synchrony, we measured the relative change in theta envelope before and after stage onset (magenta lines in Fig 1D). All model results were computed from 1,000 models randomly selected from the top one percentile of models after the optimization. We used 1,000 GWBMs to derive our conclusions in order to obtain the gen- eral behavior of the GWBM and not the behavior of a single parametrization of the model that could reflect local optima. Relative changes in simulated and MEG envelopes were correlated significantly across different cortical regions (Spearman’s ρ–encoding: 0.552 ± 0.00158 SEM; retrieval: 0.702 ± 0.000434 SEM; decision: 0.743 ± 0.000683 SEM; motor 0.477 ± 0.00151 SEM; all p-values < 0.05). Fig 2A compares the within-stage dpFC fitness of the worst model in the top one percentile to a distribution of the same fitness metric obtained with 20,000 random within-stage dpFCs, and shows that the model performs much better than chance. Fig 2B shows the fitness of within-stage dpFC at individual links. The fitness was quantified as the proportion of links with the same phase-lag direction as in the MEG within-stage dpFC (encoding: 0.697 ± 0.00014 SEM; retrieval: 0.837 ± 0.0016 SEM; decision: 0.749 ± 0.00092 SEM; motor: 0.758 ± 0.001 SEM). Fig 2C compares the across-trial dpFC of the model to the MEG data over time. Each state of dpFC begins after the pulse that modulates local connectivity strength at the onset of the stage, and vanishes with the next onset (Fig 2C). The last state of dpFC–the Fig 2. Simulated directed functional connectivity. (A) Blue histograms show the fitness between 20,000 randomly generated within-stage dpFCs and the MEG within-stage dpFC. The red line indicates the within-stage dpFCs fitness of the model with the lowest fitness index within the top 1 percentile of the optimized models. (B) Fitness of simulated-to-MEG within-stage dpFC is shown in cyan-purple grading over MEG links with significant within-stage dpFC (same as Fig 1B). The nodes indicate the relevance of a region for reaching a state of within-stage dpFC (size), and the pulse of local connectivity strength at the onset of the stage due to sub-cortical inputs (colors). See S4 Fig for the standard error of the means. These results show the averages of 1000 random picks from top ~1% of the optimizations with the best fitness index. (C) Temporal evolution of simulated-to-MEG fitness of within-stage dpFC for the current stages (solid lines) compared to other stages (dashed). The white background spans the median stage duration. The colors of the lines represent the different stages and follows Fig 1C. https://doi.org/10.1371/journal.pcbi.1009407.g002 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 7 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task motor response–vanishes slowly (in ~10 seconds), and the GWBM returns to resting-state coordination dynamics (S2 Fig). Taken together, the GWBM showed that a short pulse of local connectivity strength at the onset of a cognitive stage can first cause a modulation of local synchrony and then initiate a new state of dpFC that lasts until the onset of the next stage (Fig 2C). If there is no subsequent cognitive stage, the GWBM returns to the coordination dynamics that are characteristic of the resting state. Relevance of regions to switch between functional states of coordination Not all regions in the GWBM are equally important for switching between states of dpFC. The relevance of a region increases with the size of the pulses and the strength of structural connec- tivity with other regions. The size of the nodes in Fig 2B indicates the relevance of a region for switching between states of dpFC. The relevance of each node is the average of the 1,000 GWBMs randomly picked from the best-fitting ~1% GWBMs in the optimization process. Although each of the 1,000 GWBM had slightly different parameters (S3 Fig), the GWBMs had similar dynamics and gave consistent results as the small SEM show here and in S4 Fig. The absolute size of the pulses from the BGT predicts 22.52% (± 0.096 SEM) of the variance in the relevance of the nodes, while the interaction between the absolute size of these BGT pulses and the log-scaled strength of structural connectivity predicts 25.98% (± 0.12 SEM) of the same variance. This analysis shows that there are regions such as the left superior frontal region in the last stage that do not show dpFC, but that are still highly relevant for entering a state of high dpFC between other regions. This supports the mechanistic role of the superior frontal regions in exerting cognitive control [28,30] and highlights the complexity of interactions required for implementing changes in FC patterns. Discussion In this paper, we first analyzed the evolution of macroscopic neural coordination states across the cortex during an associative recognition memory task. Our analysis of MEG data showed that at the onset of fundamental cognitive stages there are transient modulations of local syn- chrony, which are directly followed by a new state of dpFC that persists until the next cognitive stage. Next, we used a generative model of whole brain activity (GWBM) with inputs from the basal-ganglia-thalamus system to explain these findings. The GWBM showed that short pulses that strengthen or weaken local connectivity strength at the onset of cognitive stages were suf- ficient to cause the switch between states of neural coordination consistent with empirical data. In addition, the GWBM indicated which individual regions were most relevant for caus- ing these switches. GWBM and the Basal Ganglia-Thalamus Circuit The GWBM that we developed in this paper has shown that inputs from the basal-ganglia-thal- amus system at the onset of cognitive stages are sufficient to cause switches between cognitive stages. This role of the BGT system had been hypothesized by cognitive theories [6,25]. Direct evidence for thalamic modulations of cortical activity is limited to some cortical regions due to methodological constraints, such as the fact that it is challenging to record simultaneously from many sub-cortical and cortical areas with high temporal resolution [10,12,13,40]. Never- theless, a recent meta-analysis has shown that the thalamus plays a critical role as a central hub that connects neighboring and distant regions to allow for cognitive functions [9], which is in line with the hypothesis that local thalamocortical inputs can mediate FC [14]. In addition, PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 8 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task there is evidence for neural fibers connecting the thalamus with most cortical regions [45]. Our model provides additional support for both hypotheses: the BGT systems can trigger a switch between fundamental stages of cognition [6,25] and thalamic input modulates coordi- nation of cortical activity according to cognitive demands [14]. Given our limited understanding of how the thalamus modulates cortical activity, we opted for a very simple representation of thalamocortical input. These inputs were short [40–42], tar- geted excitatory or inhibitory local connections [12,43], and came at the onset of cognitive stages [6,25]. Such inputs drove the GWBM throughout the sequence of empirical local syn- chrony and dpFC states. Afterwards, the GWBM returned to resting state dynamics. In other words, a short modulation of local excitation/inhibition modified the local synchrony and the phase of the local mean-field oscillation. This change in local mean-field phase set a new phase-lag relationship with other regions that vanished over time due to structural cortical interactions. This response to local perturbations suggests that cortical dynamics are metasta- ble as many states of coordination can be reached, and the brain does not remain into a partic- ular state in the absence of perturbations. Metastable dynamics are thought to allow for integrating and segregating information simultaneously, as well as for the flexibility of cogni- tive functions and behaviors [23]. Importantly, dpFC was not driven by thalamocortical inputs exclusively. Instead, the mac- roscopic connectivity structure of the brain also played an important role. The importance of the structural connectivity was highlighted by the presence of regions with very low dpFC that turn out to be very important for coordinating other pairs of regions as the analysis of the rele- vance of single regions for switching between states of dpFC shows. One example of such regions is the left superior frontal region during the motor response stage, a region that has been related to cognitive control, attention, and decision making [28,30,31]. The role of struc- tural connectivity in generating specific coordination patterns was first brought to light by GWBMs of resting-state dynamics [22]. In a previous study we have shown analytically that the strength of structural connectivity plays an important role in selectively coordinating regions by means of modulations of local connectivity strength [46]. Additionally, structural symmetries and time delays might have influenced dpFC in our simulations [22,47,48]. There are other biological aspects that might be relevant for coordination of cortical activity that were not included here, including the delay over thalamocortical neurons [49], the dynamics of the synapses targeted by thalamocortical inputs [12], tonic activity in the thalamus [42], noise, or the state of cortical oscillations at the time of a thalamic input. Moreover, our measurement of directed FC has neglected zero-phase-lag coordination which can emerge from thalamocortical and cortico-cortical loops [49]. However, while including these addi- tional aspects might improve the fit of the model, the current model could already account for the data surprisingly well. Additionally, we have assumed that perturbations of cortical dynam- ics at the onset of cognitive stages come exclusively from the thalamus. However, there might be other regions such as the hypothalamus that modulate cortical activity in the same or another way that is not included in our model. This is the first GWBM that can simulate local and cortex-wide neural coordination throughout a cognitive task. In addition to modeling the transition between cognitive stages, this study overcame several difficulties to make it feasible to simulate different neural coordi- nation states occurring over the course of a cognitive task, some of which may be applied to simulating resting state as well. Mainly, we developed a method to derived the initial condi- tions of the GWBM (phases and amplitudes) at the beginning of a task based on the state of neural coordination at that time. Second, we defined a fitness function that incorporates local and cortex-wide synchrony and focuses on relevant properties of the data (i.e., significant dpFC and modulations of local synchrony). Third, the optimization algorithm in combination PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 9 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task with the fitness function made computationally feasible to reliably estimate 68�4+1 parameters. Neural coordination across the cortex along a sequence of cognitive stages Our novel approach to measuring local synchrony and dpFC time-locked to the onset of cog- nitive stages revealed with high temporal resolution that each cognitive stage has a particular pattern of dpFC. This stage-dependent pattern of dpFC starts at the onset of the cognitive stage and vanishes at the end of the same cognitive stage. Furthermore, a switch between con- secutive states of dpFC has modulations of local synchrony in-between both states of dpFC– the onset of the cognitive stage. Our stage-by-stage analyses of neural coordination are consis- tent with the hypothesis that the local modulations of phase synchrony in the theta-band mark a change in long-range functional connectivity and enable a new cognitive function [19]. Moreover, our results support the hypothesis that a new state of neural coordination is estab- lished at the onset of cognitive stages [6,25]. Our previous research has shown that alpha band FC also varies across cognitive stages [4], but cortical alpha has been found to lead thalamic activity rather than being caused by it [50], as is the case with theta [13]. To uncover neural coordination stage-by-stage it was crucial to account for the temporal variability of cognitive stages across trials using the HsMM-MVPA analysis. Only after correct- ing for this variability, our analyses showed that dpFC lasts throughout a cognitive stage and differs across stages. The corresponding states of dpFC had different length, strength, and topology. This diversity of properties might have biased some traditional metrics of neural coordination. For example, if one were interested in the FC at the interval between 250 and 600 milliseconds after stimulus onset–roughly the period of memory retrieval, this interval would contain elements of the encoding or decision stages. The first reason for this is the trial- by-trial variability in stage durations: in one trial encoding might last till 400 ms, while in another trial memory retrieval might already have finished by 400 ms. Secondly, the retrieval stage has fewer and weaker connections than the encoding and decision stages in our study, which mean that these connections might have been missed altogether. These effects are worse the further one moves away from fixed time points (trial onset/response), which is one of the reasons that M/EEG studies have had severely limited trial lengths traditionally. Furthermore, our stage-by-stage analysis might contribute to disentangling competing the- ories. For example, our results suggest that the decision is made and evaluated in the last two stages. We interpreted the penultimate stage as a decision process in which memories are transferred to parietal areas by coordinating left-temporal regions with parietal regions, medi- ated by local frontal and fronto-parietal coordination [15,26,28,30]. The last stage has been tra- ditionally related with a pure motor response. However, our results indicate that the motor stage has elements associated with motor preparation, action reevaluation, decision, and cog- nitive control [29,31,34]. This functional network in the last stage suggests that during the motor stage the decision is reevaluated, and it supports the line of thought in which respond- ing is a process that is not independent from decision making (e.g., [31,35]). Conclusion To the best of our knowledge we have developed the first generative large-scale brain model that simulates the dynamics of the states of neural coordination along the fundamental cogni- tive stages in a task. In this model we have integrated structural connectivity, macroscopic neu- ral dynamics, sub-cortical inputs, and the cognitive theories of associative recognition memory. The model has multiple simplifying assumptions which made it feasible to simulate and optimize the model while taking into account the macroscopic properties of neural PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 10 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task anatomy and dynamics. This work opens up the way for considering other tasks in similarly integrated and multidimensional manners to better understand how the brain implements cognition through cortical coordination. Methods Experimental paradigm We re-analyzed MEG data from an associative memory task [3]. We combined the trials with correct responses from all experimental conditions, as we were interested in the transition between fundamental cognitive stages and not in the differences between conditions (which did all proceed through the same stages; [15]). All 18 participants were right-handed (6 males and 12 females with a mean age of 23.6 years). First, participants studied 32 pairs of words until they knew them well [3]. This was fol- lowed by a test session in which MEG was recorded. In the test session participants were pre- sented with pairs of words which were either the same as in the study season (targets) or paired differently (re-paired foils). The pairs of words remained on the screen until the participant responded, and were followed by 1-sec feedback and a brief inter-trial interval. A full descrip- tion of the task and the recording procedure can be found in [3]. MEG data preprocessing MEG data was preprocessed and source-reconstructed following the analysis pipeline of the original manuscript [3]. After artifact rejection there were 6,708 trials left. The MEG data of each participant was combined with their own structural MRI to obtain the cortical sources of MEG data. MEG sources consisted of 5,124 dipoles estimated with cortically constrained mini- mum norm estimates [3,51]. Source estimates were then morphed onto the standard MNI brain and parcellated into 68 cortical regions with the Desikan-Killiany atlas [27,52]. Each par- cel contained the average activity of all dipoles within the region with a 100 Hz sampling rate. Identification of cognitive stages To find the onset of cognitive stages the data were bandpass filtered (1–30 Hz, which are default values in Field Trip [53]) and epoched from trial onset to response. Single trials were baseline corrected (-0.4 to 0 seconds), and transformed to one covariance matrix per subject. The average covariance matrix across subjects was used to reduce the dimensionality of the data to 33 principal components (which together accounted for 90% of variance). These prin- cipal components were z-scored and fed into the HSMM-MVPA. The HSMM-MVPA first applies a half-sine window function to increase the signal-to-noise ratio of the bumps, the cor- tical response to sub-cortical input. The bumps are assumed to be 50-millisecond modulations of amplitude at the onset of cognitive stages with the same topology across trials. The signals from the end of a bump to the next bump are assumed to have zero-mean amplitude, a flat. The duration of a given stage (bump + flat) is assumed to come from a gamma distribution with shape parameter equal to two, which turns out to be the most suitable shapes for model- ing the durations of cognitive processes [15] Consequently, a stage is modeled as a bump of a certain amplitude followed by a zero-mean amplitude flat and a duration given by a gamma-2 distribution. There is one exception and this is the first stage (pre-visual encoding here) which does not start with a bump. With this stage model and a predefined number of stages, the Baum-Welch algorithm for HSMMs searches the amplitude and location of bumps that explain the z-scored principal components best [54]. The bump amplitudes (for the 33 PCA components) are the same for all trials and vary across stages. The temporal location of the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 11 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task bumps also varies across trials, but the resulting stage durations are constrained to gamma-2 distributions with one scale parameter per stage. We explored models with 3 to 7 cognitive stages as previous studies have shown that this memory task consists of 5 to 6 stages [4,5,15]. For a model with N stages we ran the HSMM-MVPA 200 times with random initial parameters to avoid converging in local max- ima. To select the most representative number of stages for all subjects we used a leave-one- subject-out cross-validation to obtain the likelihood of fitting the MEG data of a subject not used to train the HSMM-MVPA. Next we used a sign-test to assess whether a HSMM-MVPA with N+1 stages could explain the MEG data of more subjects significantly better than a HSMM-MVPA with N stages [15]. The final model was the simplest one that generalized across subjects–a five-stage model. Then, we allowed one stage to have different gamma-scale parameters across experimental conditions, and we used leave-one-subject-out cross-valida- tion to decide on the best model. As in previous studies [4,15], a model with different gamma distributions in the retrieval stage explained the MEG data best. Measurements of neural coordination To measure neural coordination–local synchrony and directed functional connectivity–we used the analytic signal of theta band oscillations. The parcellated MEG data were band pass filtered (cut-off frequencies: 3.8, and 8.5 Hz; forward-backward IIR Butterworth filter of order 4) and epoched from -0.4 seconds before stimulus onset to 0.4 seconds after the response. Epochs were Hilbert transformed to the analytic signal using a symmetric padding of 0.4 sec- onds to avoid edge artifacts. The analytic signal was transformed into phase and envelope val- ues to compute dpFC and local synchrony, respectively. Directed functional connectivity between regions i and j was measured with the directed phase-lag index (dpFC) [32] as follows: within (cid:0) stagesdpFCij ¼ PN n¼1 1 N 1 Ts n PTs (cid:0) t¼bsþ1sgn ImðSij t Þ n across (cid:0) trialss tdpFCij ¼ 1 N XN n¼1 (cid:0) sgn ImðSij ntÞ � � ð1Þ ð2Þ In Eqs 1 and 2 Im(Sij) is the imaginary part of the cross-spectral density between regions i and j, sgn is the sign function, N is the number of trials, and n is each individual trial. Then to compute within-stage dpFC at stage s (Eq 1), bs +1 is the first sample after the bump at the onset of the stage s, and Ts fore, Sij trials dpFC (Eq 2) was computed at every individual sample t of stage s time-locked to the onset of the very same stage s. Therefore Sij nt is the cross-spectral density at sample t of stage s over the N trials of a subject, and nt is the trial index. Both within-stage and across-trial dpFC were later averaged across subjects. t is the cross-spectral density over the time points t of the flat in stage s of trial n. Across- n is the number of samples of the flat of the stage s in trial n. There- Generative whole-brain model The generative whole-brain model (GWBM) was derived with the Ott-Antonsen ansatz [55] from a network-of-networks of Kuramoto oscillators [36]. See [17] for a step-by-step deriva- tion. The dynamics of synchrony in a region are given by the Kuramoto order parameter (KOP) which describes the dynamics of synchrony in in biological systems as well as a pool of neurons [56]. The KOP is a complex number (KOP = reiψ) with the modulus bound by zero PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 12 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task (asynchrony) and one (full synchrony). Here, the KOP simulated the analytic signal of the MEG sources. MEG sensors record the aggregated neural activity generated by millions of neurons in a cortical region. Each active neuron in this region generates an electric field. The measurable electric potential depends on the alignment of active neurons and the temporal synchrony of the dipole moments generated by the electric fields. The neurons that contribute to the MEG signal are parallel to each other. Therefore, the strength of the measurable electric potential in the region is proportional to the synchrony of dipole moments [57,58]. In our model, the KOP represents the synchrony of dipole moments that generate the MEG signals. Beforehand we set the natural frequencies of the oscillators to a Lorentzian distribution centered in the theta band (center, O: 6 Hz, spread, Δ: 1), and the spike-propagation velocity along the struc- tural fibers to 5 m/s. The equations of the KOP in on region, i, of the GWBM are as follows: � � � � _ri ¼ (cid:0) Diri þ 1 (cid:0) r2 i ri þ 1 (cid:0) r2 i j¼1;j6¼iAijrj t (cid:0) tij cos cjðt (cid:0) tijÞ (cid:0) ci ð3Þ �PR � (cid:0) Li 2 (cid:0) G 2R _ci ¼ Oi þ � G 2R ri þ � 1 ri PR j¼1;j6¼iAijrj � � � � t (cid:0) tij sin cjðt (cid:0) tijÞ (cid:0) ci ð4Þ The time dependency has been removed in variables without time delays; τ are the time delays between regions (fiber length x spike-propagation velocity); A is the coupling strength between regions (density of structural fibers); and R is the number of regions. To simulate rest- ing state dynamics we explored parameters G (global scaling of structural connectivity) and L (local connectivity strength, same in all regions) with 25 randomly initialized models. The results of this exploration are shown in S1 Fig. With G and L set to correctly reproducing the resting state, the thalamocortical inputs were simulated as 0.03 second increases/decreases of L at each region and stage onset independently. Simulated dpFC was measured with Eq 1 (within-stage dpFC), but here N represented 25 models with different initial conditions and Tn was the median duration of the MEG stages. The initial conditions for the first stage were the MEG phases and amplitudes at the pre-encoding stage plus random noise. More details of the simulations are reported in the SI. Generative whole-brain model: Resting-state To identify a GWBM that simulated resting-state dynamics we performed a grid-search over the global and local coupling parameter space. The local couplings were assumed to be identi- cal for all regions. Resting-state dynamics are characterized by temporal fluctuations of global and local synchrony, and time-resolved patterns of functional connectivity (i.e. metastable dynamics). Metastability was measured as the standard deviation of the modulus of the KOP over time at local and global levels [59,60]. At the local level, the metastabilities were averaged across regions. To obtain the global KOP over time we averaged the phases of the local KOPs across regions (ψ in Eq 3 of the main text). To assess the temporal structure of the global meta- stability we computed the mean of the absolute values of its autocorrelation function. To avoid the influence of the initial conditions on the simulations we ran twenty-five GWBMs with ran- dom initial conditions for each combination of parameters. The simulations were run for 1000 seconds, but the initial 200 seconds were removed to discard initial transients. All simulations were performed with a time-delayed first-order Euler method and an integration step of 1 mil- lisecond. We ended up with a global coupling of 0.15 and a local coupling of 0.7, which had the best trade-off between high metastability and low autocorrelation of global KOP, and therefore were chosen as the default values for the following GWBMs. S1 Fig shows the grid search results of resting-state dynamics. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 13 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task Generative whole-brain model: Cognitive task To simulate the sequence of cognitive stages and their associated neural coordination patterns, we initialized 25 models informed by the theta-band phases and envelope amplitudes observed at the pre-encoding stage. The MEG envelopes were measured 0.1 and 0.05 seconds before the onset of the encoding stage. Then, the initial history of the KOP modulus was a straight line that joined the mean of these amplitudes across trials plus Gaussian noise (σ = 0.01). To choose the initial history of phases we measured inter-trial phase consistency, and within-stage dpFC at the pre-encoding stage. There were 10 regions (mostly occipital and parietal) that showed significant inter-trial phase consistency. The initial history of phases at these regions were set to the average MEG phases across trials at 0.05 seconds before the onset of the first stage plus Gaussian noise (σ = 0.01). The phases of these regions were used as a referent point for the remaining regions. The initial phases of the remaining regions were set by an optimization algorithm (CMAES [61]) which tried to establish a phase-lag relationship between regions as in the empirical within-stage dpFC. The dpFC of the initial history of phases had an average similarity to empirical within-stage dpFC of 78%. The 25 GWBMs of the later stages were ini- tialized with the last simulated samples of the previous stage in the best individual of the opti- mization process (see section Optimization of thalamocortical inputs). The model with the best fitting sequence of parameters was left to run 400 seconds after the last stage. S2 Fig shows that the model neither remained trapped into the functional connectiv- ity state of the last stage, nor did it return to any of the previous states (S2 Fig, bottom). Instead, the model returned to resting state patterns of global and local synchrony for which the functional connectivity fluctuated over time (i.e. metastable dynamics; S2 Fig, top & middle). Optimization of thalamocortical inputs To find the optimal thalamocortical inputs for reproducing the observed connectivity patterns, we used the generalized island model for evolutionary optimization [44]–algorithm DE1220 as implemented in the pagmo toolbox [62]. The generalized island model optimized in parallel ten islands connected in a ring. Each island consisted of 50 individuals and had a particular parametrization of a differential evolution algorithm (see S1 Table). The islands occasionally exchanged their best-fitted individuals. This configuration allowed for simultaneously explor- ing and exploiting multiple areas of the parameter space. Their fitness function had three objectives that were combined into one index of fitness. The dominant objective was to maxi- mize the similarity of simulated and empirical within-stage dpFC, f1: f 1 ¼ PE i¼1xi � yið PE i¼1jxijÞ(cid:0) 1 ð5Þ The links, E, in the empirical dpFC, x, were either 0 (not significant), 1 (lag-ahead) or -1 (lag-behind). Simulated dpFC links, y, were either -1 or 1. The objective f1 gave discrete values which interval was used by the other two objectives. The second objective, f2, maximized the topological similarity of the relative change in envelope amplitude at the onset of each stage. This similarity was measured with the Spearman rank-correlation between MEG and simu- lated relative amplitudes. The third objective, f3, minimized the absolute size of the thalamic pulses as f 3 ¼ 1 Lmax PR PE j¼1jLjjð i¼1jxijÞ(cid:0) 1 ð6Þ where Li are the local connectivities, and Lmax is the largest absolute pulse allowed to the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 14 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task optimizers. The combined fitness index was f = f1 –(1-f2)f3. The best individual had the mini- mal (1-f2)f3 among the 5000 individuals with the highest f in order to avoid a GWBM with low f2 and f3. The last simulated samples of this individual were used to initialize the simulations of the next stage (see section Generative whole-brain model: Cognitive task). S3 Fig shows the parameters of the individuals and their fitness along the evolution in one island as example. This figure shows how the cost function could simultaneously maximize the fitness of within-stage dpFC and relative local synchrony at the onset (Spearman correlation), while the change in local coupling was minimized. The optimization of the four stages took approximately 4 days using 10 CPUs, one for each island. Relevance of individual regions for switches To assess the relevance of a region for switching between states of dpFC, a GWBM was lesioned by setting the thalamocortical pulse in this region to zero while the remaining regions were left untouched. Then, the fitness of the lesioned GWBM (Eq 5) was compared to the fit- ness achieved by the original GWBM. The relevance of a region was measured as the number of within-stage dpFC links in the lesioned model that were not matching MEG data relative to the number of links matching MEG data in the full model. This process for measuring rele- vance was repeated for the 68 regions in the GWBM and the four transitions between stages. To obtain a measure of relevance that was not dependent on a single GWBM, relevance was evaluated in 1,000 GWBMs randomly picked from among the models in the top one percentile after optimization. Next, we used linear regression models with one independent variable to explain the relevance of regions. Each linear model included as dependent variable the rele- vance of the 68 regions and four stages in a lesioned GWBM. A linear model was fitted for each of the 1,000 lesioned models independently. Structural connectivity, MRI acquisition and processing The density and the length of the neural fibers that anatomically connect cortical region was obtained from 45 subjects in the test-retest dataset of the Human Connectome Project (HCP) 3T. This data set consisted of T1-weighted and multi-shell diffusion MRI data. T1-weighted data were acquired with 0.7 mm isotropic voxel size, TE = 2.14 ms, and TR = 2400 ms. Diffu- sion MRI data were acquired with a 1.25-mm isotropic voxel size, TE = 89.5 ms, and TR 5520 ms, with three shells with b = 1000, 2000, and 3000 s/mm2, each shell with 90 diffusion weighted volumes and 6 non-weighted images [63]. The diffusion MRI data was already pre- processed as described in [64]. In short, diffusion MRI data were corrected for head motion and geometrical distortions arising from eddy currents and susceptibility artifacts [65]. Finally, the diffusion MRI images were aligned to the structural T1 image. The T1w image was parcel- lated using the Desikan–Killany parcellation [27], resulting in 68 cortical ROIs. Using the T1w image, the probability maps of the different tissues were obtained to create the five tissue-type files [66,67]. Tractography was carried out with constrained spherical deconvolution [68,69]. A multi- tissue response function was calculated [70] and the average response functions were calcu- lated. The multi-tissue fiber orientation distribution was calculated [71] with the average response function (Lmax = 8). The fiber orientation distribution images had a joint bias field correction and a multi-tissue informed log-domain intensity normalization [72]. Then, tracto- graphy was performed with the iFOD2 algorithm [73] using anatomically constrained tracto- graphy [74]; generating 10 million streamlines (cutoff at 0.05, default); and using backtracking [74] and a dynamic seeding [75]. The length of the fibers was set to a minimum of 20 mm and PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 15 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task a maximum of 250 mm [74]. To be able to quantitatively assess the connectivity, SIFT2 was applied to the resulting tractograms [75]. The connectivity matrix was built with a robust approach. In particular a 2-mm radial search at the end of the streamline was performed to allow the tracts to reach the gray matter parcellation [76]. Each connectivity matrix was multiplied by its μ coefficient obtained from the SIFT2 process, as the sum of the streamline weights needs to be proportional to the units of fiber density for each subject [77]. Connectivity matrices were averaged across subjects, and the 10% of links with the highest coefficient of variation across subjects were set to zero [78]. Finally, the averaged and thresholded structural connectivity matrix was normalized to have an average value of one. Supporting information S1 Table. Parameters of DE1220 algorithm on each island. (PDF) S1 Fig. Resting state neural coordination dynamics. The green dot indicates the parametriza- tion of the model. The location of the green dot was based on the idea that resting state dynam- ics should have simultaneously the lightest color in the three panels and the weakest coupling parameters. (PDF) S2 Fig. Return to resting-state after cognitive stages. (top) Modulus of the global KOP. (mid- dle) Modulus of the local Kuramoto order parameter (KOP) for the cortical 68 ROIs. (bottom) Temporal evolution of simulated-to-MEG fitness of within-stage dpFC for the four cognitive stages. This is similar to Fig 2B but for a much longer period of time. The MEG within-stage dpFC of each stage (Fig 1B) were compared (Eq 5) with the simulated dpFC sample-by-sample (Eq 2). (PDF) S3 Fig. Individuals and their fitness along the optimization in one island. (A) fitness index f. (B) Spearman correlation, objective f1. (C) Sum of the absolute change in local coupling at the onset of the stage. Blue dots are the A, B and C values in the order that they were evaluated along the optimization process. Orange dots are the same values but sorted by the Fit Index (A). (D) Change in local coupling (thalamic input) at the onset that produces the blue dots in A, B, C. (E) Same as (D) but sorted by their Fit Index. (PDF) S4 Fig. Mean and standard error of the mean of the relevance of each region for switching between two states of within-stage dpFC. Average values are obtained for the 1,000 GWBM randomly picked from the 10,000 GWBMs with the best fitness index. (PDF) Acknowledgments HCP datasets were provided by the Human Connectome Project, WU-Minn Consortium (Principal Investigators: David Van Essen and Kamil Ugurbil; 1U54MH091657) funded by the 16 NIH Institutes and Centers that support the NIH Blueprint for Neuroscience Research; and by the McDonnell Center for Systems Neuroscience at Washington University. We thank Lio- nel Newman from the University of Groningen for proofreading the manuscript. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 16 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task Author Contributions Conceptualization: Oscar Portoles. Data curation: Oscar Portoles, Manuel Blesa, Jelmer P. Borst. Formal analysis: Oscar Portoles. Methodology: Oscar Portoles. Software: Oscar Portoles. Supervision: Jelmer P. Borst. Visualization: Oscar Portoles. Writing – original draft: Oscar Portoles, Jelmer P. Borst. Writing – review & editing: Manuel Blesa, Marieke van Vugt, Ming Cao, Jelmer P. Borst. References 1. Donders FC. On the speed of mental processes. Acta Psychol (Ams). 1969; 30: 412–431. https://doi. org/10.1016/0001-6918(69)90065-1 PMID: 5811531 2. Sternberg S. The discovery of processing stages: Extensions of Donders’ method. Acta Psychologica. 1969; 30: 276–315. https://doi.org/10.1016/0001-6918(69)90055-9 3. Borst JP, Ghuman AS, Anderson JR. Tracking cognitive processing stages with MEG: A spatio-tempo- ral model of associative recognition in the brain. NeuroImage. 2016; 141: 416–430. https://doi.org/10. 1016/j.neuroimage.2016.08.002 PMID: 27498135 4. Portoles O, Borst JP, van Vugt MK. Characterizing synchrony patterns across cognitive task stages of associative recognition memory. Eur J Neurosci. 2018; 48: 2759–2769. https://doi.org/10.1111/ejn. 13817 PMID: 29283467 5. Anderson JR, Borst JP, Fincham JM, Ghuman AS, Tenison C, Zhang Q. The Common Time Course of Memory Processes Revealed. Psychol Sci. 2018; 29: 1463–1474. https://doi.org/10.1177/ 0956797618774526 PMID: 29991326 6. Stocco A, Lebiere C, Anderson JR. Conditional routing of information to the cortex: A model of the basal ganglia’s role in cognitive coordination. Psychol Rev. 2010; 117: 541–574. https://doi.org/10.1037/ a0019077 PMID: 20438237 7. O’Reilly RC, Frank MJ. Making working memory work: a computational model of learning in the prefron- tal cortex and basal ganglia. Neural Comput. 2006; 18: 283–328. https://doi.org/10.1162/ 089976606775093909 PMID: 16378516 8. Redgrave P, Prescott TJ, Gurney K. The basal ganglia: a vertebrate solution to the selection problem? Neuroscience. 1999; 89: 1009–1023. https://doi.org/10.1016/s0306-4522(98)00319-4 PMID: 10362291 9. Hwang K, Bertolero MA, Liu WB, D’Esposito M. The Human Thalamus Is an Integrative Hub for Func- tional Brain Networks. J Neurosci. 2017; 37: 5594–5607. https://doi.org/10.1523/JNEUROSCI.0067- 17.2017 PMID: 28450543 10. Bruno RM, Sakmann B. Cortex Is Driven by Weak but Synchronously Active Thalamocortical Synap- ses. Science. 2006; 312: 1622–1627. https://doi.org/10.1126/science.1124593 PMID: 16778049 11. Rektor I, Baresˇ M, Bra´zdil M, Kaňovsky´ P, Rektorova´ I, Sochı¯rkova´ D, et al. Cognitive- and movement- related potentials recorded in the human basal ganglia. Mov Disord. 2005; 20: 562–568. https://doi.org/ 10.1002/mds.20368 PMID: 15666424 12. Cruikshank SJ, Lewis TJ, Connors BW. Synaptic basis for intense thalamocortical activation of feedfor- ward inhibitory cells in neocortex. Nat Neurosci. 2007; 10: 462–468. https://doi.org/10.1038/nn1861 PMID: 17334362 13. Malekmohammadi M, Elias WJ, Pouratian N. Human thalamus regulates cortical activity via spatially specific and structurally constrained phase-amplitude coupling. Cereb Cortex. 2015; 25: 1618–1628. https://doi.org/10.1093/cercor/bht358 PMID: 24408958 14. Saalmann YB. Intralaminar and medial thalamic influence on cortical synchrony, information transmis- sion and cognition. Front Syst Neurosci. 2014; 8. https://doi.org/10.3389/fnsys.2014.00083 PMID: 24847225 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 17 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task 15. Anderson JR, Zhang Q, Borst JP, Walsh MM. The discovery of processing stages: Extension of Stern- berg’s method. Psychol Rev. 2016; 123: 481–509. https://doi.org/10.1037/rev0000030 PMID: 27135600 16. Tewarie P, Hunt BAE, O’Neill GC, Byrne A, Aquino K, Bauer M, et al. Relationships Between Neuronal Oscillatory Amplitude and Dynamic Functional Connectivity. Cerebral Cortex. 2019; 29: 2668–2681. https://doi.org/10.1093/cercor/bhy136 PMID: 29897408 17. Portoles O, Qin Y, Hadida J, Woolrich M, Cao M, Vugt M van. Modulations of local synchrony over time lead to resting-state functional connectivity in a parsimonious large-scale brain model. bioRxiv. 2021; 2021.01.20.427443. https://doi.org/10.1101/2021.01.20.427443 18. Ketz NA, Jensen O, O’Reilly RC. Thalamic pathways underlying prefrontal cortex-medial temporal lobe oscillatory interactions. Trends Neurosci. 2015; 38: 3–12. https://doi.org/10.1016/j.tins.2014.09.007 PMID: 25455705 19. Voloh B, Womelsdorf T. A Role of Phase-Resetting in Coordinating Large Scale Neural Networks Dur- ing Attention and Goal-Directed Behavior. Front Syst Neurosci. 2016; 10. https://doi.org/10.3389/fnsys. 2016.00018 PMID: 27013986 20. Canolty RT, Edwards E, Dalal SS, Soltani M, Nagarajan SS, Kirsch HE, et al. High Gamma Power Is Phase-Locked to Theta Oscillations in Human Neocortex. Science. 2006; 313: 1626–1628. https://doi. org/10.1126/science.1128115 PMID: 16973878 21. Zhang Q, Walsh MM, Anderson JR. The Impact of Inserting an Additional Mental Process. Comput Brain Behav. 2018; 1: 22–35. https://doi.org/10.1007/s42113-018-0002-8 22. Cabral J, Kringelbach ML, Deco G. Functional connectivity dynamically evolves on multiple time-scales over a static structural connectome: Models and mechanisms. NeuroImage. 2017; 160: 84–96. https:// doi.org/10.1016/j.neuroimage.2017.03.045 PMID: 28343985 23. Tognoli E, Kelso JAS. The Metastable Brain. Neuron. 2014; 81: 35–48. https://doi.org/10.1016/j. neuron.2013.12.022 PMID: 24411730 24. Wang L-Z, Su R-Q, Huang Z-G, Wang X, Wang W-X, Grebogi C, et al. A geometrical approach to con- trol and controllability of nonlinear dynamical networks. Nat Commun. 2016; 7: 11323. https://doi.org/ 10.1038/ncomms11323 PMID: 27076273 25. Anderson JR. How can the human mind occur in the physical universe? Oxford; New York: Oxford Uni- versity Press; 2007. 26. Staresina BP, Wimber M. A Neural Chronometry of Memory Recall. Trends Cogn Sci. 2019; 23: 1071– 1085. https://doi.org/10.1016/j.tics.2019.09.011 PMID: 31672429 27. Desikan RS, Se´ gonne F, Fischl B, Quinn BT, Dickerson BC, Blacker D, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage. 2006; 31: 968–980. https://doi.org/10.1016/j.neuroimage.2006.01.021 PMID: 16530430 28. Cavanagh JF, Frank MJ. Frontal theta as a mechanism for cognitive control. Trends Cogn Sci. 2014; 18: 414–421. https://doi.org/10.1016/j.tics.2014.04.012 PMID: 24835663 29. Sauseng P, Griesmayr B, Freunberger R, Klimesch W. Control mechanisms in working memory: a pos- sible function of EEG theta oscillations. Neurosci Biobehav Rev. 2010; 34: 1015–1022. https://doi.org/ 10.1016/j.neubiorev.2009.12.006 PMID: 20006645 30. Rajan A, Siegel SN, Liu Y, Bengson J, Mangun GR, Ding M. Theta Oscillations Index Frontal Decision- Making and Mediate Reciprocal Frontal–Parietal Interactions in Willed Attention. Cereb Cortex. 2019; 29: 2832. https://doi.org/10.1093/cercor/bhy149 PMID: 29931088 31. Pellegrino G, Tomasevic L, Herz DM, Larsen KM, Siebner HR. Theta Activity in the Left Dorsal Premo- tor Cortex During Action Re-Evaluation and Motor Reprogramming. Front Hum Neurosci. 2018; 12. https://doi.org/10.3389/fnhum.2018.00364 PMID: 30297991 32. Stam CJ, van Straaten ECW. Go with the flow: Use of a directed phase lag index (dPLI) to characterize patterns of phase relations in a large-scale model of brain dynamics. NeuroImage. 2012; 62: 1415– 1428. https://doi.org/10.1016/j.neuroimage.2012.05.050 PMID: 22634858 33. Reinhart RMG. Working memory revived in older adults by synchronizing rhythmic brain circuits. Nat Neurosci.: 16. https://doi.org/10.1038/s41593-019-0371-x PMID: 30962628 34. Hartwigsen G, Siebner HR. Joint Contribution of Left Dorsal Premotor Cortex and Supramarginal Gyrus to Rapid Action Reprogramming. Brain Stimul. 2015; 8: 945–952. https://doi.org/10.1016/j.brs.2015.04. 011 PMID: 26028563 35. Gallivan JP, Chapman CS, Wolpert DM, Flanagan JR. Decision-making in sensorimotor control. Nat Rev Neurosci. 2018; 19: 519–534. https://doi.org/10.1038/s41583-018-0045-9 PMID: 30089888 36. Skardal PS, Restrepo JG. Hierarchical synchrony of phase oscillators in modular networks. Phys Rev E. 2012; 85: 016208. https://doi.org/10.1103/PhysRevE.85.016208 PMID: 22400644 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 18 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task 37. Breakspear M, Heitmann S, Daffertshofer A. Generative Models of Cortical Oscillations: Neurobiologi- cal Implications of the Kuramoto Model. Front Hum Neurosci. 2010; 4. https://doi.org/10.3389/fnhum. 2010.00190 PMID: 21151358 38. Maistrenko YL, Lysyansky B, Hauptmann C, Burylko O, Tass PA. Multistability in the Kuramoto model with synaptic plasticity. Phys Rev E. 2007; 75: 066207. https://doi.org/10.1103/PhysRevE.75.066207 PMID: 17677340 39. Deco G, Cabral J, Woolrich MW, Stevner ABA, van Hartevelt TJ, Kringelbach ML. Single or multiple fre- quency generators in on-going brain activity: A mechanistic whole-brain model of empirical MEG data. NeuroImage. 2017; 152: 538–550. https://doi.org/10.1016/j.neuroimage.2017.03.023 PMID: 28315461 40. Saalmann YB, Pinsk MA, Wang L, Li X, Kastner S. The Pulvinar Regulates Information Transmission Between Cortical Areas Based on Attention Demands. Science. 2012; 337: 753–756. https://doi.org/10. 1126/science.1223082 PMID: 22879517 41. Sherman SM. The thalamus is more than just a relay. Curr Opin Neurobiol. 2007; 17: 417–422. https:// doi.org/10.1016/j.conb.2007.07.003 PMID: 17707635 42. Ramcharan EJ, Gnadt JW, Sherman SM. Higher-order thalamic relays burst more than first-order relays. Proc Natl Acad Sci USA. 2005; 102: 12236–12241. https://doi.org/10.1073/pnas.0502843102 PMID: 16099832 43. Kloc M, Maffei A. Target-Specific Properties of Thalamocortical Synapses onto Layer 4 of Mouse Pri- mary Visual Cortex. J Neurosci. 2014; 34: 15455–15465. https://doi.org/10.1523/JNEUROSCI.2595- 14.2014 PMID: 25392512 44. Izzo D, Ruciński M, Biscani F. The Generalized Island Model. Parallel Architectures and Bioinspired Algorithms. 2012; 151–169. https://doi.org/10.1007/978-3-642-28789-3_7 45. Behrens TEJ, Johansen-Berg H, Woolrich MW, Smith SM, Wheeler-Kingshott C a. M, Boulby PA, et al. Non-invasive mapping of connections between human thalamus and cortex using diffusion imaging. Nat Neurosci. 2003; 6: 750–757. https://doi.org/10.1038/nn1075 PMID: 12808459 46. Qin Y, Kawano Y, Portoles O, Cao M. Partial Phase Cohesiveness in Networks of Networks of Kura- moto Oscillators. IEEE Trans Automat Contr. 2021; 1–1. https://doi.org/10.1109/TAC.2021.3062005 47. Nicosia V, Valencia M, Chavez M, Dı´az-Guilera A, Latora V. Remote Synchronization Reveals Network Symmetries and Functional Modules. Phys Rev Lett. 2013; 110: 174102. https://doi.org/10.1103/ PhysRevLett.110.174102 PMID: 23679731 48. Petkoski S, Spiegler A, Proix T, Aram P, Temprado J-J, Jirsa VK. Heterogeneity of time delays deter- mines synchronization of coupled oscillators. Phys Rev E. 2016; 94: 012209. https://doi.org/10.1103/ PhysRevE.94.012209 PMID: 27575125 49. Vicente R, Gollo LL, Mirasso CR, Fischer I, Pipa G. Dynamical relaying can yield zero time lag neuronal synchrony despite long conduction delays. Proc Natl Acad Sci USA. 2008; 105: 17157–17162. https:// doi.org/10.1073/pnas.0809353105 PMID: 18957544 50. Halgren M, Ulbert I, Bastuji H, Fabo´ D, Erőss L, Rey M, et al. The generation and propagation of the human alpha rhythm. Proc Natl Acad Sci USA. 2019; 116: 23772–23782. https://doi.org/10.1073/pnas. 1913092116 PMID: 31685634 51. Gramfort A, Luessi M, Larson E, Engemann DA, Strohmeier D, Brodbeck C, et al. MNE software for pro- cessing MEG and EEG data. NeuroImage. 2014; 86: 446–460. https://doi.org/10.1016/j.neuroimage. 2013.10.027 PMID: 24161808 52. Fischl B. FreeSurfer. NeuroImage. 2012; 62: 774–781. https://doi.org/10.1016/j.neuroimage.2012.01. 021 PMID: 22248573 53. Oostenveld R, Fries P, Maris E, Schoffelen J-M. FieldTrip: Open Source Software for Advanced Analy- sis of MEG, EEG, and Invasive Electrophysiological Data. In: Computational Intelligence and Neurosci- ence [Internet]. Hindawi; 23 Dec 2010 [cited 10 Feb 2021] p. e156869. https://doi.org/https%3A//doi. org/10.1155/2011/156869 PMID: 21253357 54. Yu S-Z. Hidden semi-Markov models. Artificial Intelligence. 2010; 29. 55. Ott E, Antonsen TM. Low dimensional behavior of large systems of globally coupled oscillators. Chaos. 2008; 18: 037113. https://doi.org/10.1063/1.2930766 PMID: 19045487 56. Montbrio´ E, Pazo´ D, Roxin A. Macroscopic Description for Networks of Spiking Neurons. Phys Rev X. 2015; 5: 021028. https://doi.org/10.1103/PhysRevX.5.021028 57. Lopes da Silva F. EEG and MEG: Relevance to Neuroscience. Neuron. 2013; 80: 1112–1128. https:// doi.org/10.1016/j.neuron.2013.10.017 PMID: 24314724 58. Buzsa´ki G, Anastassiou CA, Koch C. The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nat Rev Neurosci. 2012; 13: 407–420. https://doi.org/10.1038/nrn3241 PMID: 22595786 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 19 / 20 PLOS COMPUTATIONAL BIOLOGYSwitching between functional connectivity states along a cognitive task 59. Wildie M, Shanahan M. Metastability and chimera states in modular delay and pulse-coupled oscillator networks. Chaos. 2012; 22: 043131. https://doi.org/10.1063/1.4766592 PMID: 23278066 60. Cabral J, Luckhoo H, Woolrich M, Joensson M, Mohseni H, Baker A, et al. Exploring mechanisms of spontaneous functional connectivity in MEG: how delayed network interactions lead to structured ampli- tude envelopes of band-pass filtered oscillations. Neuroimage. 2014; 90: 423–435. https://doi.org/10. 1016/j.neuroimage.2013.11.047 PMID: 24321555 61. Hansen N. The CMA Evolution Strategy: A Comparing Review.: 28. 62. Biscani F, Izzo D. A parallel global multiobjective framework for optimization: pagmo. Journal of Open Source Software. 2020; 5: 2338. https://doi.org/10.21105/joss.02338 63. Van Essen DC, Ugurbil K, Auerbach E, Barch D, Behrens TEJ, Bucholz R, et al. The Human Connec- tome Project: a data acquisition perspective. Neuroimage. 2012; 62: 2222–2231. https://doi.org/10. 1016/j.neuroimage.2012.02.018 PMID: 22366334 64. Glasser MF, Sotiropoulos SN, Wilson JA, Coalson TS, Fischl B, Andersson JL, et al. The minimal pre- processing pipelines for the Human Connectome Project. Neuroimage. 2013; 80: 105–124. https://doi. org/10.1016/j.neuroimage.2013.04.127 PMID: 23668970 65. Sotiropoulos SN, Jbabdi S, Xu J, Andersson JL, Moeller S, Auerbach EJ, et al. Advances in diffusion MRI acquisition and processing in the Human Connectome Project. Neuroimage. 2013; 80: 125–143. https://doi.org/10.1016/j.neuroimage.2013.05.057 PMID: 23702418 66. Zhang Y, Brady M, Smith S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Trans Med Imaging. 2001; 20: 45–57. https:// doi.org/10.1109/42.906424 PMID: 11293691 67. Patenaude B, Smith SM, Kennedy DN, Jenkinson M. A Bayesian model of shape and appearance for subcortical brain segmentation. Neuroimage. 2011; 56: 907–922. https://doi.org/10.1016/j.neuroimage. 2011.02.046 PMID: 21352927 68. 69. Tournier J-D, Calamante F, Connelly A. Robust determination of the fibre orientation distribution in diffu- sion MRI: Non-negativity constrained super-resolved spherical deconvolution. NeuroImage. 2007; 35: 1459–1472. https://doi.org/10.1016/j.neuroimage.2007.02.016 PMID: 17379540 Tournier J-D, Smith R, Raffelt D, Tabbara R, Dhollander T, Pietsch M, et al. MRtrix3: A fast, flexible and open software framework for medical image processing and visualisation. Neuroimage. 2019; 202: 116137. https://doi.org/10.1016/j.neuroimage.2019.116137 PMID: 31473352 70. Dhollander T, Raffelt D, Connelly A. Unsupervised 3-tissue response function estimation from single- shell or multi-shell diffusion MR data without a co-registered T1 image. 2016. 71. Jeurissen B, Tournier J-D, Dhollander T, Connelly A, Sijbers J. Multi-tissue constrained spherical deconvolution for improved analysis of multi-shell diffusion MRI data. Neuroimage. 2014; 103: 411– 426. https://doi.org/10.1016/j.neuroimage.2014.07.061 PMID: 25109526 72. Raffelt D, Dhollander T, Tournier J-D, Tabbara R, Smith R, Pierre E, et al. Bias Field Correction and Intensity Normalisation for Quantitative Analysis of Apparent Fibre Density. 2017. 73. Tournier J-D, Calamante F, Connelly A. Improved probabilistic streamlines tractography by 2 nd order integration over fibre orientation distributions. 2009 [cited 10 Feb 2021]. Available: /paper/Improved- probabilistic-streamlines-tractography-by-Tournier-Calamante/ b4ffcb9ec889a8a68bffc46387a96b78a50ef94a 74. Smith RE, Tournier J-D, Calamante F, Connelly A. Anatomically-constrained tractography: improved diffusion MRI streamlines tractography through effective use of anatomical information. Neuroimage. 2012; 62: 1924–1938. https://doi.org/10.1016/j.neuroimage.2012.06.005 PMID: 22705374 75. Smith RE, Tournier J-D, Calamante F, Connelly A. SIFT2: Enabling dense quantitative assessment of brain white matter connectivity using streamlines tractography. Neuroimage. 2015; 119: 338–351. https://doi.org/10.1016/j.neuroimage.2015.06.092 PMID: 26163802 76. Smith RE, Tournier J-D, Calamante F, Connelly A. The effects of SIFT on the reproducibility and biologi- cal accuracy of the structural connectome. Neuroimage. 2015; 104: 253–265. https://doi.org/10.1016/j. neuroimage.2014.10.004 PMID: 25312774 77. Smith R, Raffelt D, Tournier J-D, Connelly A. Quantitative streamlines tractography: methods and inter- subject normalisation. OSF Preprints; 2020. https://doi.org/10.31219/osf.io/c67kn 78. Roberts JA, Perry A, Roberts G, Mitchell PB, Breakspear M. Consistency-based thresholding of the human connectome. NeuroImage. 2017; 145: 118–129. https://doi.org/10.1016/j.neuroimage.2016.09. 053 PMID: 27666386 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009407 March 9, 2022 20 / 20 PLOS COMPUTATIONAL BIOLOGY
10.1186_s13756-019-0596-1	Yuan et al. Antimicrobial Resistance and Infection Control (2019) 8:140 https://doi.org/10.1186/s13756-019-0596-1 R E S E A R C H Open Access blaNDM-5 carried by a hypervirulent Klebsiella pneumoniae with sequence type 29 Yi Yuan1†, Ying Li2†, Guangxi Wang3, Chengwen Li2, Yung-Fu Chang4, Wenbi Chen3, Siji Nian3, Yingyu Mao3, Jinping Zhang3, Fangcai Zhong1 and Luhua Zhang3* Abstract Background: A carbapenem-resistant hypermucoviscous Klebsiella pneumoniae isolate was recovered from human sputum. Methods: Whole genome sequencing of this isolate was carried out to reveal its clonal background, antimicrobial resistance determinants and virulence factors. Virulence assays were performed using wax moth larvae. The transfer of blaNDM-5 between bacterial strains was tested using conjugation. 59 genome assemblies of ST29 K. pneumoniae and 230 IncX3 plasmids regardless of the carriage of resistance gene were employed for phylogenetic analysis, respectively. Results: The strain carried a virulence plasmid pVir-SCNJ1 bearing the virulence gene rmpA and exhibited a high virulence in wax moth. This hypervirulent strain belongs to sequence type 29 and carries blaNDM-5, which is located on a conjugative plasmid, designated pNDM5-SCNJ1, belonging to type IncX3. pNDM5-SCNJ1 was fully sequenced and shows high similarity with pNDM_MGR194, except some deletion inside the ISAba125 region. Phylogenetic analysis of IncX3 plasmids revealed that although blaNDM-5 can be evolved from blaNDM-1 via point mutations within some IncX3 plasmids, most of blaNDM-5-carrying IncX3 plasmids probably have acquired blaNDM-5 in multiple events. Conclusions: In this study, we characterized a blaNDM-5-positive hypervirulent K. pneumoniae of sequence type 29 in China. Our results highlight the need for active surveillance on this lineage of carbapenem-resistant K. pneumoniae. Keywords: Carbapenem resistance, blaNDM-5, Hypervirulent, IncX3 Introduction Hypervirulent Klebsiella pneumoniae (hvKP) is a world- wide concern due life- threatening, community-acquired infections in healthy individuals with high morbidity and mortality [1, 2]. hvKP strains are usually less resistant to most antimicro- bials than classic K. pneumoniae [3], but the increasing to its capacity to cause * Correspondence: zhluhua@swmu.edu.cn †Yi Yuan and Ying Li contributed equally to this work. 3Department of Pathogenic Biology, School of Basic Medical Sciences, Southwest Medical University, No.1 Section 1, Xiang Lin Road, Longmatan District, Luzhou 646000, Sichuan, China Full list of author information is available at the end of the article emergence of carbapenemase-producing hypervirulent K. pneumoniae (CP-hvKP) compromises options of anti- microbial agents for infection control and drives a global crisis [2, 4]. These CP-hvKP strains are thought to be the result of acquiring plasmid-mediated resistance and virulence markers, either by transferring of resistance plasmids into hvKP strains or virulence plasmids into carbapenem-resistant strains [5]. There have now been several reports of infections caused by carbapenemase- producing hypervirulent K. pneumoniae strains [2, 4, 6– 9]. These CP-hvKP isolates mainly produce KPC (a © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Yuan et al. Antimicrobial Resistance and Infection Control (2019) 8:140 Page 2 of 9 group of serine-lactamases), followed by IMP (a group of metallo-β-lactamases), and they belong to the widely dis- tributed sequence type (ST) 11 [2, 8] and several other STs, e.g., ST25, ST65 [6] and ST36 [4]. Here, we identified an ST29 CR-hvKP clinical strain with K54 serotype carry- ing blaNDM-5 gene and reported on its characterization. Materials and methods Bacterial identification and PCR analysis The strain SCNJ1 was recovered from the sputum of a patient with an acute bronchiolitis in a hospital of Sichuan Province in November 2018. The initial species identification was performed using the Vitek-2 compact system (bioMérieux, Marcy-l’Étoile, France). A further species confirmation was performed by PCR amplifying of the 16S rRNA gene using the primer pair 27F/1492R [10]. PCR products were purified and then sequenced by Sanger sequencing. The resulting 16S rRNA gene se- quences were compared with sequences in GenBank (NCBI) database using BLAST software. The presence of the acquired carbapenemase genes blaKPC, blaNDM, blaGES, blaIMP, blaOXA-48, and blaVIM in this isolate was screened via PCR using primers as previously described [11–14]. Antimicrobial susceptibility tests In vitro susceptibility tests of cefepime, piperacillin- tazobactam, ampicillin, ampicillin/sulbactam, cefote- tan, ceftriaxone, cefazolin, nitrofurantoin, ceftazidime, tobramycin, ciprofloxacin, trimethoprim/sulfamethoxa- zole, aztreonam, amikacin, gentamicin and levofloxa- cin were performed using Vitek-2 system. The minimum inhibitory concentrations (MICs) of imipe- nem, meropenem and colistin against the isolate were determined using the microdilution broth method fol- lowing recommendations of the Clinical Laboratory Standards Institute (CLSI) [15]. Breakpoints of colistin was defined by the European Committee on Anti- microbial Susceptibility Testing (EUCAST) (http:// www.eucast.org/), otherwise, we applied those defined by the CLSI. String test String test was performed by stretching a mucoviscous string from the colony using a standard bacteriologic loop as described previously [16]. Strains that formed strings > 5 mm in length were defined as viscous hypermucoviscous. Conjugation Conjugation experiments were carried out using broth-based methods with the azide-resistant Escheri- chia coli strain J53 as the recipient and transconju- gants were selected using 2 μg/ml meropenem plus 150 μg/ml sodium azide. The presence of blaNDM-5 in transconjugants was and sequencing. confirmed PCR by Virulence assay The virulence potential of the SCNJ1 strain was assessed using wax moth (Galleria mellonella) larvae weighing 250 to 350 mg (Tianjin Huiyude Biotech Company, Tianjin, China) with method described previously [17]. Overnight cultures of K. pneumoniae strains were ad- justed with phosphate-buffered saline (PBS) to concentra- tions of 1 × 104 CFU/ml, 1 × 105 CFU/ml, 1 × 106 CFU/ml, 1 × 107 CFU/ml after being washed with PBS. 10 μl of inoculum was injected into the hemocoel of sixteen larvae using a 25-μl Hamilton syringe via the last left proleg. The larvae were then incubated at 37 °C and the number of live larvae was counted at 12 h inter- vals for 3 days. Two blaKPC-2-carrying carbapenem- resistant K. pneumoniae clinical isolates of ST11:K47, KPNJ2 and KPLZ1050, without rmpA and rmpA2, were used as the control. Genome sequencing and analysis The strain was subjected to whole genomic sequencing using an Illumina HiSeq 2000 system with the 150-bp paired-end approach and 150 × coverage. Reads were trimmed using Trimmomatic [18]. Draft genome was then assembled using the SPAdes program [19]. Anno- tation was carried out using Prokka [20]. Sequence type and capsular type of this strain were determined using the assembled contigs to query the Multi-Locus Se- quence Typing (MLST) v 2.0 (https://cge.cbs.dtu.dk/ services/MLST/) and wzc genotyping system as previ- ously described [21], respectively. In addition, the wzi genotyping system [22] and KLeborate (https://github. com/katholt/Kleborate/) were employed to confirm the sequence type and capsular type. Clonal complexes (CCs) were determined by using eBURST v3 based on K. pneumoniae MLST data (https://eburst.mlst.net). Virulence genes were identified using the Virulence Factors Database (VFDB) available at http://www.mgc. resistance genes ac.cn/VFs/main.htm. Antimicrobial were identified using the ResFinder v3.1 software of the Center for Genomic Epidemiology (CGE, http://geno- micepidemiology.org/). Plasmids pVir-SCNJ1 and pNDM5-SCNJ1 were completely circularized with gaps between the contigs closed by PCR and respective amplicons sequenced using Sanger sequencing, respectively. Identification of types were performed on plasmid incompatibility complete sequences of plasmids via the online service PlasmidFinder v2.0 at CGE (https://cge.cbs.dtu.dk/ser- vices/PlasmidFinder/). The annotations of the plasmid sequences were conducted using the RAST tools and Yuan et al. Antimicrobial Resistance and Infection Control (2019) 8:140 Page 3 of 9 edited manually [23]. Sequence alignment of blaNDM-5- carrying plasmids was performed using BLAST and vi- sualized with Easyfig v 2.2.3 [24]. Alignments with highly homologous complete plasmid sequences of pVir-SCNJ1 available in NCBI were performed by using the BRIG tool [25]. The circular map of pNDM5- SCNJ1 was also generated using BRIG [25]. 1.4 Phylogenetic analysis All assembled K. pneumoniae genomes (n = 6823; accessed by March 1, 2019) were retrieved from Gen- Bank. MLST typing was performed using the script (https://github.com/tseemann/mlst). A total of 59 as- semblies of ST29 K. pneumoniae were included and aligned with that of strain SCNJ1 using CSI Phyl- (https://cge.cbs.dtu.dk/services/CSIPhylo- ogeny geny/) (Additional file 1 :Table S2). Gubbins (version 2.3.4) was used to remove single nucleotide polymor- phisms (SNPs) on recombination sites [26]. The fil- tered SNPs were then used as input for inferring a phylogenetic tree using RAxML with the GTRG AMMA model and 1000 bootstraps [27]. ABRicate (https://github.com/tseemann/abricate) was used to identify antimicrobial resistance genes in these ge- nomes and the capsular typing of K. pneumoniae was performed with the wzc genotyping system. The sequence of all available IncX3 plasmids regard- less of the carriage of resistance gene (n = 230; accessed by March 24, 2019) were retrieved from the GenBank (Additional file 1 :Table S4). Orthogroups were identi- fied using OrthoFinder [28] and used for multiple se- quence alignments (MSA) with MAFFT [29]. The species tree was inferred from the concatenated MSA using FastTree [30]. The STRIDE algorithm (Specie Tree Root Inference from Duplication Events) was used to root the species tree in OrthoFinder. Results and discussion Antimicrobial susceptibility test showed that the K. pneu- moniae SCNJ1 strain was resistant to imipenem (MIC, > 256 μg/ml) and meropenem (MIC, > 256 μg/ml) cefepime (MIC, 16 μg/ml), piperacillin-tazobactam (MIC, ≥ 128 μg/ ml), ampicillin (MIC, ≥ 32 μg/ml), ampicillin/sulbactam (MIC, ≥ 32 μg/ml), cefotetan (MIC, ≥ 64 μg/ml), ceftriax- one (MIC, ≥ 64 μg/ml), cefazolin (MIC, ≥ 64 μg/ml), nitro- furantoin (MIC,128 μg/ml) and ceftazidime (MIC, ≥ 64 μg/ml), but was susceptible to colistin (MIC, 2 μg/ml), tobramycin (MIC, ≤ 1 μg/ml), ciprofloxacin (MIC, ≤ (MIC, ≤ 0.25 μg/ml), trimethoprim/sulfamethoxazole 20 μg/ml), aztreonam (MIC, ≤ 1 μg/ml), amikacin (MIC, ≤2 μg/ml), gentamicin (MIC, ≤1 μg/ml) and levofloxacin (MIC, ≤ 0.25 μg/ml). Strain SCNJ1 showed hypermucov- iscosity phenotype as evidenced by forming a viscous string about 35 mm, which is beyond the > 5 mm to define hypermucoviscous. PCR and sequencing showed that blaNDM-5 was the only carbapenemase-encoding gene carried by the strain SCNJ1. NDM-5, a variant of NDM (New Delhi Metallo-β-lactamase), was first identified in an E. coli ST648 isolate (EC045) in the UK in 2011 from a pa- tient with a recent hospitalization history in India [31]. Although blaNDM-5 has been widely found in K. pneumoniae strains first discovery (Add- since its itional file 1 :Table S1), blaNDM-5-carrying hypermu- coviscous K. pneumoniae remains uncommon. We found one publication that described a blaNDM-5-posi- tive K. pneumoniae isolate (K2/ST14) in China in 2015 [32], which was speculated to be hypermucovis- cous on the basis of genome analysis. However, no experimental data was included. Draft genome sequences of SCNJ1 was assembled into 29 contigs (28 were > 1000 bp in length), which comprises 5,474,953 bp, with a 57.29% GC content. SCNJ1 was assigned to capsular type wzi115-K54 and sequence type ST29 (gapA-infB-mdh-pgi-phoE-rpoB- tonB allele number 2–3–2-3-6-4-4). K54 is a hypervir- ulent member of K. pneumoniae [16] and has been described in several previous publications as linked to ST29 [33–37]. To date, K. pneumoniae strains with the ST29 group has a worldwide distribution and has been found carrying a variety of carbapenem genes, including blaNDM-1 [38–40], blaKPC [41], blaOXA-48 [42] and blaOXA-181 [38], as well as several extended- spectrum β-lactamases (ESBLs) genes [43, 44] in vari- ous countries. However, the currently available evi- dence is insufficient to demonstrate whether ST29, a member of CC29, is an epidemic clone mediating the spread of specific and clinically relevant antibiotic re- sistance genes. analysis also showed that The gene blaNDM-5 is described for the first time in a strain of K. pneumoniae ST29 in our work, as demon- strated by the phylogenetic tree based on filtered SNPs of all available ST29 K. pneumoniae strains (Fig. 1). The phylogenetic strain SCNJ1 was clustered with four isolates recovered in China and was closest to strain SCLZ15–011 (GCA_ 001630805, carrying no carbapenemase gene, recovered in 2016 in China) with 198 SNPs difference (Fig. 1). It should be noted that sporadic cases due to ST29 K. pneumoniae were frequently detected, mainly from liver abscess patients [37, 43, 45, 46], and multidrug-resistant ST29 hvKP strains have been reported in different loca- tions of China [5, 33, 47]. This highlights the need to the ST29 clones of K. monitor the epidemiology of pneumoniae isolates in China. We found that strain SCNJ1 harbored genes encoding regulators of the mucoid phenotype (rmpA), aerobactin Yuan et al. Antimicrobial Resistance and Infection Control (2019) 8:140 Page 4 of 9 Fig. 1 Phylogenetic tree and resistance gene profile of K. pneumoniae strain SCNJ1 with other 59 ST29 K. pneumoniae genomes available Isolates from China are marked with an asterisk and resistance gene blaNDM-5 is marked with a from GenBank. SCNJ1 is indicated in red. red rectangle (iucABCD and iutA), ent siderophore (entABCDEFS, salmochelin (iroBCDEN), yersiniabactin fepABCDG), (fyuA, irp1, irp2 and ybtAEPQSTUX) and type 3 fim- briae (mrkABCDFHIJ) etc. These genes are frequently associated with hypervirulence phenotype of K. pneumo- niae [16, 48]. The rmpA, iutA, iucABCD and iroBCDN genes were carried by a 211,807-bp plasmid, designated pVir-SCNJ1. The rmpA2 (another regulator of mucoid phenotype) gene on the pVir-SCNJ1 plasmid was trun- cated, due to a frameshift mutation introducing an in- ternal stop codon. pVir-SCNJ1 was an IncHI1/IncFIB- type plasmid and was 99.71% identical to the known virulence plasmid pLVPK (219,385 bp, GenBank acces- sion no. NC_005249) at 93% coverage [49] (Fig. 2). It was notable that pVir-SCNJ1 was highly similar (99% coverage and 99.99% identity, Fig. 2) to the recently- identified plasmid pL22–1 (212, 635 bp, GenBank ac- cession no. NZ_CP031258) recovered from a Klebsiella quasipneumoniae strain L22, which suggests that the pLVPK-like plasmid has the potential to me- diate inter- and intra-species transfer of virulence genes [50]. Our virulence assays showed that survival of G. mellonella was 0% with strains SCNJ1, while that survival was 56.2 and 50.0% with the control strains KPLZ1050 and KPNJ2, at an inoculum of 1 × 105 cfu/ mL at 72 h after (Fig. 3, infection, Additional file 1 :Table S3). These findings suggest that strain SCNJ1 was hypervirulent. respectively Conjugation assays showed that strain SCNJ1 trans- ferred a plasmid carrying blaNDM-5 to E. coli J53 at a − 6 (transconjugant/recipient) by mating, frequency of 10 suggesting that blaNDM-5 was carried on a self- transmissible plasmid, which was assigned pNDM5- SCNJ1. In addition to the blaNDM-5, strain SCNJ1 had a few chromosomal resistance genes, including the ESBL gene blaSHV-187, fluoroquinolone-resistance genes oqxA and oqxB, and fosfomycin-resistance gene fosA. pNDM5-SCNJ1 was a 45,255-bp IncX3 plasmid, with an average GC content of 46.83% and had no other known antimicrobial resistance genes except blaNDM-5. pNDM5-SCNJ1 consists of a 30-kb back- bone comprising several sets of genes (pir and bis en- coding replication initiation protein, parA for plasmid partitioning, hns and topB for maintenance and a gene cluster responsible for conjugation) and a gen- etic load region with high GC content between the Yuan et al. Antimicrobial Resistance and Infection Control (2019) 8:140 Page 5 of 9 Fig. 2 Alignment of pVir-SCNJ1 with 3 hypervirulence-encoding plasmids. The alignment was performed using BRIG and pLVPK was used as a reference. Accession numbers for the plasmids are NZ_CP031258 (pL22–1), NC_006625 (pK2044), NC_005249 (pLVPK). The locations of virulence genes rmpA2, iutA, iucDCBA, rmpA and iroNDCB are indicated resolvase and the hns gene, which are typical of IncX3 plasmids (Fig.4a). BLASTn revealed that the sequence of pNDM5-SCNJ1 was highly similar (100% coverage and 99.99% identity) to the plasmid pNDM_MGR194 (GenBank accession no. KF220657) recovered from a K. pneumoniae isolate in India, as well as a number of pre- viously described IncX3 plasmids carrying blaNDM-5 in China. In the genetic load region of pNDM5-SCNJ1(Fig. 4b), the umuD gene was split into two fragments at the Fig. 3 Virulence potential of K. pneumoniae strains in a G. mellonella infection model. The effect of 1 × 105 CFU/ml of each isolate on survival of G. mellonella is shown. The results for other doses of each K pneumoniae strain are shown in Supplementary Table S3. KPLZ1050 and KPNJ2, two blaKPC − 2-carrying K pneumoniae clinical isolates of ST11 that did not harbour a virulence plasmid, were used as the control Yuan et al. Antimicrobial Resistance and Infection Control (2019) 8:140 Page 6 of 9 Fig. 4 Plasmid analysis of pNDM5-SCNJ1. (a) Genetic structure of IncX3 plasmid pNDM5-SCNJ1. This map was used to illustrate the backbone and the location of the genetic load region of pNDM5-SCNJ1. Genes are denoted by arrowheads and colored based on gene function classification. (b) Comparative analysis of the genetic load region of pNDM5-SCNJ1. Genes and insertion sequences are indicated by arrows. Light gray shades denote shared regions with a high degree of homology. The accession numbers were: pEC14_35 (JN935899), pBJ01(JX296013), pNDM_MGR194 (KF220657), pP744T (MF547511), pTK1044 (LC000627). The alignment is a pairwise BLASTn alignment performed using Easyfig [24] nucleotide position 336 bp by the blaNDM-5-containing (IS26-ΔctuA1-tat-trpF-bleMBL-blaNDM-5-ΔIS- structure Aba125-IS5-ΔISAba125-IS3000-ΔTn2), resulting in a pair of 3-bp direct repeats (TGT). In such a genetic context, an IncX3 plasmid pEC14_35 (GenBank accession no. JN935899) without any antibiotic-resistance gene, which was isolated from a patient in the USA in 1989, was likely to be the ancestral vector. It is also likely that pNDM5- SCNJ1 has diverged recently from blaNDM-1-positive JX296013) by plasmids pBJ01 (GenBank accession no. sequential mutations (Fig. 4b). In the subsequent genetic variant, the ISAba125 was truncated by the insertion of IS5 element (at 166 bp upstream blaNDM-5 start codon) and a identified. 4-bp flanking direct the genetic contexts of blaNDM-5 in Comparisons of pNDM_MGR194, pP744T, pTK1044 and pNDM5-SCNJ1 showed that the remnant of ISAba125 (73 bp of 1087 bp) upstream of blaNDM-5 was conserved, but the length of the remnant of ISAba125 between IS3000 and IS5 differed (pNDM_MGR19: 1002 bp, pP744T: 404 bp, pNDM-SCNJ1: 112 bp, pTK1044: 0 bp), suggesting that IS5 has inserted into ISAba125 at the same position in these plasmids and that gene deletions caused by homologous recombination plays a possible role in the formation of diversified ΔIS- Aba125 region. (CTAA) was repeats IncX3 plasmids are narrow-host-range vectors of the Enterobacteriaceae [51, 52]. Searches on IncX3 plasmids in NCBI showed that they were recovered from various species of Enterobacteriaceae (Additional including E. coli, K. pneumoniae, file 1 :Table S4), Citrobacter freundii, Enterobacter cloacae, Klebsiella oxytoca, Enterobacter hormaechei, Salmonella enterica, Kluyvera intermedia, Morganella morganii, Raoultella planticola and Raoultella ornithinolytica from differ- ent countries, suggesting a wide distribution of IncX3 plasmids. Three kinds of carbapenemase genes were including found to be carried by IncX3 plasmids, blaNDM, blaOXA-181 and blaKPC. The carriage rate of blaNDM was significantly higher (n = 150, 64.94%) than those of blaKPC (n = 18, 7.79%) and blaOXA-181 (n = 17, 7.35%) (Additional file 1 :Table S4). Of note, IncX3 plasmids were found to carry many different blaNDM alleles, including NDM-1, 4, 5, 6, 7, 13, 19, 20, 21, which were mainly recovered from China, confirming that IncX3 plasmids function as a common vehicle in facilitating the rapid dissemination of NDM-type MBLs among Enterobacteriaceae in China. accession pQDE2-NDM (GenBank Phylogenetic analysis based on concatenated MSA of revealed that pNDM5-SCNJ1 was IncX3 plasmids to closely related (100% coverage, 99.99% identity) plasmid no. MH917280), which also carried blaNDM-5 that was re- covered from a K. pneumoniae isolate in Shandong, in 2015 (Fig. 5). The phylogenetic tree also China, showed that most of blaNDM-5-carrying IncX3 plasmids are tightly clustered with each other and formed a rela- tively distinct clade, with only sporadic ones clustered into clades with blaNDM-1-carrying plasmids, suggesting that although blaNDM-5 likely to evolve from is blaNDM-1 via point mutations on some IncX3 plasmids, most of acquired IncX3 plasmids probably have blaNDM-5 in multiple events. Yuan et al. Antimicrobial Resistance and Infection Control (2019) 8:140 Page 7 of 9 Fig. 5 Phylogenetic analysis of IncX3 plasmids. Detailed information of the plasmids is shown in Supplementary Table 4. Those carrying blaNDM-5 are indicated in red, while pNDM5-SCNJ1 is shown in purple. The species tree was inferred from the concatenated multiple sequence alignments using FastTree Conclusion In conclusion, our work identified an ST29 CP-hvKP carry- ing the carbapenemase gene blaNDM-5 and provided add- itional evidence of the rapid dissemination of blaNDM-5 by pNDM-MGR194-like plasmid among Enterobacteriaceae in China. The association of the epidemic IncX3 plasmid carrying blaNDM-5 with a hypervirulent K. pneumoniae lineage, ST29/K54 in this case, is quite worrisome and may pose a great threat to humans. More extensive surveillance and effective action to control its further dissemination are urgently required. Additional file Additional file 1: Table S1. Background information on the blaNDM-5- positive K. pneumoniae isolates. Table S2. ST29 K. pneumoniae strains with genome sequences available in the GenBank. Table S3. Survival (number of larvae) of G. mellonella after infection by K. pneumoniae strain SCNJ1. Table S4. The names, host species, accession numbers, carbapenemase genes and locations of IncX3 plasmids. (DOCX 109 kb) Acknowledgements Not applicable. Yuan et al. Antimicrobial Resistance and Infection Control (2019) 8:140 Page 8 of 9 Authors’ contributions LZ designed the experiments. GW, CL, FZ and WC performed the experiments. SN, YM and JZ analyzed the data. YL analyzed the data and wrote the manuscript. YY wrote the manuscript. YC edited the original draft. All authors read and approved the final manuscript. Authors’ information Not applicable Funding This work was supported by Project of Education Department in Sichuan, China (18ZB0633), Natural Science Foundation of Southwest Medical University (No. 2017-ZRZD-022 and 2018-ZRZD-011), and National Under- graduate Innovation and Entrepreneurship Project (No.201816032021). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Availability of data and materials Draft whole-genome sequences of the SCNJ1 strain has been deposited into GenBank under the accession no. SPSD00000000. The complete sequences of pVir-SCNJ1 and pNDM5-SCNJ1 have been deposited into GenBank under accession no. MK715436 and MK715437, respectively. Ethics approval and consent to participate The current study was approved by the Ethics Committee of Southwest Medical University (No.201903–194) and were carried out in accordance with the approved guidelines. Written informed consent was exempted, since this retrospective study mainly focused on bacteria and patient intervention was not required. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1Department of Laboratory Medicine, The First People’s Hospital of Neijiang, Neijiang, Sichuan, China. 2Department of Immunology, School of Basic Medical Sciences, Southwest Medical University, Luzhou, Sichuan, China. 3Department of Pathogenic Biology, School of Basic Medical Sciences, Southwest Medical University, No.1 Section 1, Xiang Lin Road, Longmatan District, Luzhou 646000, Sichuan, China. 4Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, USA. Received: 8 May 2019 Accepted: 12 August 2019 References 1. Zhang Y, Zhao C, Wang Q, Wang X, Chen H, Li H, Zhang F, Li S, Wang R, Wang H. High prevalence of hypervirulent Klebsiella pneumoniae infection in China: geographic distribution, clinical characteristics, and antimicrobial resistance. Antimicrob Agents Chemother. 2016;60(10):6115–20. Gu D, Dong N, Zheng Z, Lin D, Huang M, Wang L, Chan EW-C, Shu L, Yu J, Zhang R, et al. A fatal outbreak of ST11 carbapenem-resistant hypervirulent Klebsiella pneumoniae in a Chinese hospital: a molecular epidemiological study. Lancet Infect Dis. 2018;18(1):37–46. Liu YM, Li BB, Zhang YY, Zhang W, Shen H, Li H, Cao B. Clinical and molecular characteristics of emerging hypervirulent Klebsiella pneumoniae bloodstream infections in mainland China. Antimicrob Agents Chemother. 2014;58(9):5379–85. Feng Y, Lu Y, Yao Z, Zong Z. Carbapenem-resistant hypervirulent Klebsiella pneumoniae of sequence type 36. Antimicrob Agents Chemother. 2018; 62(7):e02644–17. Chen L, Kreiswirth BN. Convergence of carbapenem-resistance and hypervirulence in Klebsiella pneumoniae. Lancet Infect Dis. 2018;18(1):2–3. Yao B, Xiao X, Wang F, Zhou L, Zhang X, Zhang J. Clinical and molecular characteristics of multi-clone carbapenem-resistant hypervirulent (hypermucoviscous) Klebsiella pneumoniae isolates in a tertiary hospital in Beijing, China. Int J Infect Dis. 2015;37:107–12. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. Zhang R, Lin D, Chan EW, Gu D, Chen GX, Chen S. Emergence of carbapenem-resistant serotype K1 hypervirulent Klebsiella pneumoniae strains in China. Antimicrob Agents Chemother. 2016;60(1):709–11. Zhan L, Wang S, Guo Y, Jin Y, Duan J, Hao Z, Lv J, Qi X, Hu L, Chen L, et al. Outbreak by Hypermucoviscous Klebsiella pneumoniae ST11 isolates with carbapenem resistance in a tertiary hospital in China. Front Cell Infect Microbiol. 2017;7:182. Zhang Y, Zeng J, Liu W, Zhao F, Hu Z, Zhao C, Wang Q, Wang X, Chen H, Li H, et al. Emergence of a hypervirulent carbapenem-resistant Klebsiella pneumoniae isolate from clinical infections in China. J Infect. 2015;71(5):553–60. Lane DJ. 16S/23S rRNA sequencing. Nucleic acid techniques in bacterial systematics. 1991:115–75. Szekely E, Damjanova I, Janvari L, Vas KE, Molnar S, Bilca DV, Lorinczi LK, Toth A. First description of bla(NDM-1), bla(OXA-48), bla(OXA-181) producing Enterobacteriaceae strains in Romania. Int J Med Microbiol. 2013;303(8):697–700. 12. Dallenne C, Da Costa A, Decre D, Favier C, Arlet G. Development of a set of multiplex PCR assays for the detection of genes encoding important beta- lactamases in Enterobacteriaceae. J Antimicrob Chemother. 2010;65(3):490–5. Ellington MJ, Kistler J, Livermore DM, Woodford N. Multiplex PCR for rapid detection of genes encoding acquired metallo-beta-lactamases. J Antimicrob Chemother. 2007;59(2):321–2. 13. 14. Poirel L, Le Thomas I, Naas T, Karim A, Nordmann P. Biochemical sequence analyses of GES-1, a novel class a extended-spectrum beta-lactamase, and the class 1 integron In52 from Klebsiella pneumoniae. Antimicrob Agents Chemother. 2000;44(3):622–32. 16. 15. Clinical and Laboratory Standards Institute. M100S. Performance standards for antimicrobial susceptibility testing. 29th edition. USA: Wayne, PA; 2019. Shon AS, Bajwa RP, Russo TA. Hypervirulent (hypermucoviscous) Klebsiella pneumoniae: a new and dangerous breed. Virulence. 2013;4(2):107–18. Lu Y, Feng Y, McNally A, Zong Z. The Occurence of colistin-resistant hypervirulent Klebsiella pneumoniae in China. Front Microbiol. 2018;9:2568. 18. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina 17. sequence data. Bioinformatics. 2014;30(15):2114–20. 19. Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, Lesin VM, Nikolenko SI, Pham S, Prjibelski AD, et al. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J Comput Biol. 2012;19(5):455–77. Seemann T. Prokka: rapid prokaryotic genome annotation. Bioinformatics. 2014;30(14):2068–9. 20. 21. Pan YJ, Lin TL, Chen YH, Hsu CR, Hsieh PF, Wu MC, Wang JT. Capsular types of Klebsiella pneumoniae revisited by wzc sequencing. PLoS One. 2013;8(12): e80670. 22. Brisse S, Passet V, Haugaard AB, Babosan A, Kassis-Chikhani N, Struve C, Decre D. wzi gene sequencing, a rapid method for determination of capsular type for Klebsiella strains. J Clin Microbiol. 2013;51(12):4073–8. 23. Aziz RK, Bartels D, Best AA, DeJongh M, Disz T, Edwards RA, Formsma K, Gerdes S, Glass EM, Kubal M, et al. The RAST server: rapid annotations using subsystems technology. BMC Genomics. 2008;9:75. Sullivan MJ, Petty NK, Beatson SA. Easyfig: a genome comparison visualizer. Bioinformatics. 2011;27(7):1009–10. 24. 25. Alikhan NF, Petty NK, Ben Zakour NL, Beatson SA. BLAST ring image generator (BRIG): simple prokaryote genome comparisons. BMC Genomics. 2011;12:402. 26. Croucher NJ, Page AJ, Connor TR, Delaney AJ, Keane JA, Bentley SD, Parkhill J, 27. 28. 29. Harris SR. Rapid phylogenetic analysis of large samples of recombinant bacterial whole genome sequences using Gubbins. Nucleic Acids Res. 2015;43(3):e15. Stamatakis A. RAxML version 8: a tool for phylogenetic analysis and post- analysis of large phylogenies. Bioinformatics. 2014;30(9):1312–3. Emms DM, Kelly S. OrthoFinder2: fast and accurate phylogenomic orthology analysis from gene sequences. bioRxiv. 2018:466201. Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013; 30(4):772–80. 30. Price MN, Dehal PS, Arkin AP. FastTree 2--approximately maximum- likelihood trees for large alignments. PLoS One. 2010;5(3):e9490. 31. Hornsey M, Phee L, Wareham DW. A novel variant, NDM-5, of the New Delhi metallo-beta-lactamase in a multidrug-resistant Escherichia coli ST648 isolate recovered from a patient in the United Kingdom. Antimicrob Agents Chemother. 2011;55(12):5952–4. Yuan et al. Antimicrobial Resistance and Infection Control (2019) 8:140 Page 9 of 9 and typing of novel plasmids in drug-resistant Enterobacteriaceae. Plasmid. 2012;68(1):43–50. 52. Wu WJ, Feng Y, Tang GM, Qiao F, McNally A, Zong ZZ. NDM metallo-β- lactamases and their bacterial producers in health care settings. Clin Microbiol Rev. 2019;32(2):e00115–8. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 32. Liu PP, Liu Y, Wang LH, Wei DD, Wan LG. Draft genome sequence of an NDM-5-Producing Klebsiella pneumoniae sequence type 14 strain of serotype K2. Genome Announc. 2016;4(2):e01610–15. 33. Chuang YC, Lee MF, Yu WL. Mycotic aneurysm caused by 34. hypermucoviscous Klebsiella pneumoniae serotype K54 with sequence type 29: an emerging threat. Infection. 2013;41(5):1041–4. Iwasaki Y, Inokuchi R, Harada S, Aoki K, Ishii Y, Shinohara K. Bacterial meningitis caused by hypervirulent Klebsiella pneumoniae capsular genotype K54 with development of granuloma-like nodal enhancement in the brain during the subacute phase. Intern Med. 2017;56(3):373–6. 35. Yan Q, Zhou M, Zou M, Liu WE. Hypervirulent Klebsiella pneumoniae induced ventilator-associated pneumonia in mechanically ventilated patients in China. Eur J Clin Microbiol Infect Dis. 2016;35(3):387–96. 36. Cubero M, Grau I, Tubau F, Pallares R, Dominguez MA, Linares J, Ardanuy C. Hypervirulent Klebsiella pneumoniae clones causing bacteraemia in adults in a teaching hospital in Barcelona, Spain (2007-2013). Clin Microbiol Infect. 2016;22(2):154–60. Liao CH, Huang YT, Chang CY, Hsu HS, Hsueh PR. Capsular serotypes and multilocus sequence types of bacteremic Klebsiella pneumoniae isolates associated with different types of infections. Eur J Clin Microbiol Infect Dis. 2014;33(3):365–9. 37. 38. Balm MN, La MV, Krishnan P, Jureen R, Lin RT, Teo JW. Emergence of Klebsiella pneumoniae co-producing NDM-type and OXA-181 carbapenemases. Clin Microbiol Infect. 2013;19(9):E421–3. 39. Castanheira M, Deshpande LM, Farrell SE, Shetye S, Shah N, Jones RN. Update on the prevalence and genetic characterization of NDM-1- producing Enterobacteriaceae in Indian hospitals during 2010. Diagn Microbiol Infect Dis. 2013;75(2):210–3. 40. Gharout-Sait A, Alsharapy SA, Brasme L, Touati A, Kermas R, Bakour S, Guillard T, de Champs C. Enterobacteriaceae isolates carrying the New Delhi metallo-beta-lactamase gene in Yemen. J Med Microbiol. 2014;63(Pt 10): 1316–23. 41. Barria-Loaiza C, Pincheira A, Quezada M, Vera A, Valenzuela P, Dominguez M, Lima CA, Araya I, Araya P, Prat S, et al. Molecular typing and genetic environment of the blaKPC gene in Chilean isolates of Klebsiella pneumoniae. J Glob Antimicrob Resist. 2016;4:28–34. 42. Uz Zaman T, Aldrees M, Al Johani SM, Alrodayyan M, Aldughashem FA, Balkhy HH. Multi-drug carbapenem-resistant Klebsiella pneumoniae infection carrying the OXA-48 gene and showing variations in outer membrane protein 36 causing an outbreak in a tertiary care hospital in Riyadh, Saudi Arabia. Int J Infect Dis. 2014;28:186–92. Luo Y, Wang Y, Ye L, Yang J. Molecular epidemiology and virulence factors of pyogenic liver abscess causing Klebsiella pneumoniae in China. Clin Microbiol Infect. 2014;20(11):O818–24. 43. 44. Gorrie CL, Mirceta M, Wick RR, Judd LM, Wyres KL, Thomson NR, Strugnell RA, Pratt NF, Garlick JS, Watson KM, et al. Antimicrobial- resistant Klebsiella pneumoniae carriage and infection in specialized geriatric care wards linked to acquisition in the referring hospital. Clin Infect Dis. 2018;67(2):161–70. 45. Qu TT, Zhou JC, Jiang Y, Shi KR, Li B, Shen P, Wei ZQ, Yu YS. Clinical and microbiological characteristics of Klebsiella pneumoniae liver abscess in east China. BMC Infect Dis. 2015;15:161. 46. Xiao SZ, Wang S, Wu WM, Zhao SY, Gu FF, Ni YX, Guo XK, Qu JM, Han LZ. The resistance phenotype and molecular epidemiology of Klebsiella pneumoniae in bloodstream infections in Shanghai, China, 2012-2015. Front Microbiol. 2017;8:250. 47. Moura Q, Esposito F, Fernandes MR, Espinoza-Munoz M, Souza TA, Santos SR, Cerdeira L, Cassettari V, Lincopan N. Genome sequence analysis of a hypermucoviscous/hypervirulent and MDR CTX-M-15/K19/ST29 Klebsiella pneumoniae isolated from human infection. Pathog Dis. 2017;75(9):ftx121. 48. Hsieh PF, Lin TL, Lee CZ, Tsai SF, Wang JT. Serum-induced iron-acquisition systems and TonB contribute to virulence in Klebsiella pneumoniae causing primary pyogenic liver abscess. J Infect Dis. 2008;197(12):1717–27. 49. Chen YT, Chang HY, Lai YC, Pan CC, Tsai SF, Peng HL. Sequencing and analysis of the large virulence plasmid pLVPK of Klebsiella pneumoniae CG43. Gene. 2004;337:189–98. Lu Y, Feng Y, McNally A, Zong ZZ. Occurrence of colistin-resistant hypervirulent Klebsiella variicola. J Antimicrob Chemother. 2018;73(11): 3001–4. Johnson TJ, Bielak EM, Fortini D, Hansen LH, Hasman H, Debroy C, Nolan LK, Carattoli A. Expansion of the IncX plasmid family for improved identification 50. 51.
10.1371_journal.pcbi.1011196	RESEARCH ARTICLE Dynamic recycling of extracellular ATP in human epithelial intestinal cells Nicolas Andres SaffiotiID Virginia Gentilini5,6, Gabriel Eduardo Gondolesi5,6, Pablo Julio Schwarzbaum1,2, Julieta SchachterID 1,2,3, Cora Lilia Alvarez1,4, Zaher Bazzi1,2, Marı´a 1,2 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Saffioti NA, Alvarez CL, Bazzi Z, Gentilini MV, Gondolesi GE, Schwarzbaum PJ, et al. (2023) Dynamic recycling of extracellular ATP in human epithelial intestinal cells. PLoS Comput Biol 19(6): e1011196. https://doi.org/10.1371/journal. pcbi.1011196 Editor: Melissa L. Kemp, Georgia Institute of Technology and Emory University, UNITED STATES Received: January 25, 2023 Accepted: May 17, 2023 Published: June 29, 2023 Copyright: © 2023 Saffioti et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: Data can be found in the following doi: https://doi.org/10.6084/m9. figshare.21938651.v1. Funding: Grants from Secretarı´a de Ciencia y Te´cnica, Universidad de Buenos Aires (PJS, UBACYT 20020170100152BA), Consejo Nacional de Investigaciones Cientı´ficas y Te´cnicas (PJS, CONICET PIP 1013) and Agencia Nacional de Promocio´n Cientı´fica y Tecnolo´gica (NAS, PICT- 2019 03218; JS, PICT 2019-0204; PJS, PICT 2021- 1 Instituto de Quı´mica y Fı´sico-Quı´mica Biolo´ gicas “Prof. Alejandro C. Paladini”, Universidad de Buenos Aires (UBA), Consejo Nacional de Investigaciones Cientı´ficas y Te´cnicas (CONICET), Facultad de Farmacia y Bioquı´mica, Buenos Aires, Argentina, 2 Universidad de Buenos Aires (UBA), Facultad de Farmacia y Bioquı´mica, Departamento de Quı´mica Biolo´gica, Ca´tedra de Quı´mica Biolo´ gica, Buenos Aires, Argentina, 3 Instituto de Nanosistemas, Universidad Nacional de General San Martin, Buenos Aires, Argentina, 4 Universidad de Buenos Aires (UBA), Facultad de Ciencias Exactas y Naturales, Departamento de Biodiversidad y Biologı´a Experimental, Buenos Aires, Argentina, 5 Fundacio´n Favaloro Hospital Universitario, Unidad de Insuficiencia, Rehabilitacio´n y Trasplante Intestinal, Buenos Aires, Argentina, 6 Instituto de Medicina Traslacional, Trasplante y Bioingenierı´a (IMETTyB, CONICET, Universidad Favaloro), Laboratorio de Inmunologı´a asociada al Trasplante, Buenos Aires, Argentina * pjs@qb.ffyb.uba.ar (PJS); jschachter@qb.ffyb.uba.ar (JS) Abstract Intestinal epithelial cells play important roles in the absorption of nutrients, secretion of elec- trolytes and food digestion. The function of these cells is strongly influenced by purinergic signalling activated by extracellular ATP (eATP) and other nucleotides. The activity of sev- eral ecto-enzymes determines the dynamic regulation of eATP. In pathological contexts, eATP may act as a danger signal controlling a variety of purinergic responses aimed at defending the organism from pathogens present in the intestinal lumen. In this study, we characterized the dynamics of eATP on polarized and non-polarized Caco-2 cells. eATP was quantified by luminometry using the luciferin-luciferase reaction. Results show that non-polarized Caco-2 cells triggered a strong but transient release of intracellular ATP after hypotonic stimuli, leading to low micromolar eATP accumulation. Subsequent eATP hydrolysis mainly determined eATP decay, though this effect could be counterbalanced by eATP synthesis by ecto-kinases kinetically characterized in this study. In polarized Caco-2 cells, eATP showed a faster turnover at the apical vs the basolateral side. To quantify the extent to which different processes contribute to eATP regulation, we cre- ated a data-driven mathematical model of the metabolism of extracellular nucleotides. Model simulations showed that eATP recycling by ecto-AK is more efficient a low micromo- lar eADP concentrations and is favored by the low eADPase activity of Caco-2 cells. Simula- tions also indicated that a transient eATP increase could be observed upon the addition of non-adenine nucleotides due the high ecto-NDPK activity in these cells. Model parameters showed that ecto-kinases are asymmetrically distributed upon polarization, with the apical side having activity levels generally greater in comparison with the basolateral side or the non-polarized cells. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 1 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells 00125). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. Finally, experiments using human intestinal epithelial cells confirmed the presence of functional ecto-kinases promoting eATP synthesis. The adaptive value of eATP regulation and purinergic signalling in the intestine is discussed. Author summary Intestinal epithelial cells play important roles in the absorption of nutrients, secretion of electrolytes and food digestion. When intracellular ATP is released into the intestinal milieu, either at the lumen or the internal side, the resulting extracellular ATP can act as an alert signal to engage cell surface purinergic receptors that activate the immune defence of the organism against pathogens. We worked with Caco-2 and primary human intestinal cell, and our results showed that extracellular ATP regulation is a complex network of reactions that simultaneously consume or generate ATP in whole viable intestinal epithelial cells. In particular, we cre- ated a mathematical model and fitted it to experimental data allowing to quantify the degree to which intracellular ATP release and the activity of a variety of ectoenzymes con- trol the concentration of extracellular ATP. 1. Introduction The surface of the intestine is covered by a layer of cells that form the intestinal epithelium. Intestinal epithelial cells play important roles in the absorption of nutrients, secretion of elec- trolytes, digestion of food and host defence mechanisms [1,2]. The function of intestinal epi- thelial cells is strongly influenced by extracellular nucleotides, supporting a complex signalling network that mediates short-term functions such as secretion and motility, and long-term functions like proliferation and apoptosis [3,4]. Among these nucleotides, extracellular ATP (eATP) was found to be an early danger signal response to infection with enteric pathogens that eventually promote inflammation of the gut [4,5]. An important source of eATP is the intracellular ATP (iATP) found in the cytosol and vesi- cles of many cell types [6]. Activation of iATP release was found in subepithelial intestinal fibroblasts, human epithelial cell lines and enteroendocrine cells in response to several stimuli, including agents that elevate cAMP, such as forskolin and cholera toxin [7], low medium phos- phate, hypoosmotic swelling and bacterial infection [7,8]. Currently, several mechanisms have been postulated to mediate regulated iATP release, and these mechanisms can vary according to the cell type and the stimuli [6]. For example, after a hypotonic shock, Schwann cells release ATP via the anionic channel pannexin-1 [9], while the treatment with lipopolysaccharide (LPS) induces ATP release via connexin-43 in macrophages [10]. Additionally, the vesicular release pathway for ATP was also described, as in the case of endothelial cells under hypoxia [11]. Extracellular ATP and other di- and tri-phosphonucleosides can activate purinergic recep- tors 2 (P2 receptors) unevenly distributed in the small and large intestine [12]. Purinergic sig- nalling is controlled by membrane bound ecto-nucleotidases and ecto-kinases capable of promoting the synthesis and/or hydrolysis of eATP, and/or its conversion into other extracel- lular nucleotides and nucleosides. For any cell type and metabolic context, a specific set of ecto-enzymes may control the rate, amount and timing of nucleotide turnover [13]. Ecto-nucleoside triphosphate diphosphohydrolases (Ecto-NTPDases) are a family of enzymes promoting the extracellular hydrolysis of eATP, eADP, eUTP and eUDP. One or PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 2 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells more members of this family are present in almost every cell. Ecto-NTPDase-1, -2, and -3, which differ regarding the specific preferences for nucleotides, are responsible for the hydroly- sis of nucleoside diphosphates (NDPs) and nucleoside triphosphates (NTPs) in various tissues of the gastrointestinal tract [1]. Regarding eATP and eADP hydrolysis, ecto-NTPDase-1 hydrolyses both nucleotides at similar rates, while ecto-NTPDase-2 has a high preference for eATP over eADP and ecto-NTPDase3 is a functional intermediate which preferably hydrolyses eATP [14]. The intestinal cell line HT29 cells expressed functional ecto-NTPDase-2 displaying high ecto-ATPase activity [15], while Caco-2 cells and their exosomes were reported to exhibit ecto- NTPDases-1 and -2 at the cell membrane [16,17]. Extracellular ATP can be also metabolized by ecto-kinases, with ecto-adenylate kinase (Ecto-AK) facilitating the reversible conversion of eADP to eATP and eAMP, and ecto-nucleo- side diphosphate kinase (Ecto-NDPK) promoting the exchange of terminal phosphate between extracellular NDPs and NTPs [13]. All these ecto-enzymes, if present and active, should be able to control the concentration of eATP. Up to now, although some ecto-enzymes have been identified in intestinal cells, no attempts have been made to characterize the dynamic interaction of these membrane proteins on eATP regulation of intestinal cells. In this study, we aimed to characterize iATP release and eATP recycling by ecto-enzymes, contributing to the regulation of eATP concentration ([eATP]) in Caco-2 cell line. The Caco-2 cells derive from colorectal adenocarcinoma and easily differenti- ate into cells exhibiting the morphology and function of enterocytes, the absorptive cells of the small intestine [18]. The experimental studies on eATP dynamics in polarized and non-polar- ized Caco-2 were complemented with a mathematical model quantifying the complex relation- ship among the different processes contributing to [eATP] regulation. Our results provide a quantitative description of the eATP dynamics of human intestinal epithelial cells. 2. Results In this section we show experimental results on eATP kinetics of non-polarized and polarized Caco-2 cells. To understand the dynamics of the different processes contributing to [eATP] regulation, a mathematical model was fitted to experimental data, and predictions were made. Finally, for a comparative purpose, we show results of a few experiments made on epithelial cells obtained from intestinal surgical pieces. 2.1. Non-polarized Caco-2 cells 2.1.1. eATP kinetics after hypotonic shock. The kinetics of eATP accumulation, i.e., eATP kinetics, results from the dynamic balance between iATP release mechanisms and the activities of ecto-enzymes capable of degrading and/or synthetizing eATP. As a first step towards the characterization of eATP kinetics, iATP release was triggered by exposing Caco-2 cells to hypotonic media (Fig 1A–1D). Hypotonic swelling is a stimulus that influences the uptake of nutrients by epithelial cells [19] and represents an inducer of iATP release in most cell types and tissues [8]. Under unstimulated conditions, [eATP] remained stable. Whereas addition of isotonic medium triggered a slight increase of [eATP], hypotonic media (100–180 mOsm) activated a stronger iATP release with different kinetics according to the osmotic gradient imposed (Fig 1A–1C). As shown in Fig 1D, [eATP] increased non-linearly with the magnitude of the hypo- tonic stimulus. The experimental [iATP] amounted to 1.81 mM. By comparing [iATP] with [eATP] along eATP kinetics, it was possible to estimate the energy cost of iATP release. Calculations were PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 3 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells Fig 1. eATP kinetics of hypotonically-stimulated Caco-2 cells. The time course of [eATP] from Caco-2 cells after a hypotonic shock was quantified by luminometry performed at room temperature. (A-C) Cells were maintained in isotonic medium and, at the times indicated by the arrow, were exposed to isotonic medium (grey) or to hypotonic media (blue) of 180 mOsm (A), 150 mOsm (B) and 100 mOsm (C). Results are expressed as means of [eATP] from 4, 3 and 2 independent experiments run in triplicate for the 180, 150 a 100 mOsm experiments, respectively. (D) Increases in [eATP] from data in A-C were evaluated as ΔATP, i.e., the difference between [eATP] at 30 minutes post-stimulus and basal [eATP]. Cells were exposed to 300 mOsm (light blue bars), 180 mOsm (blue bars), 150 mOsm (dark blue bars) and 100 mOsm (grey bars). Bars show mean values + standard error of the mean (s.e.m) from 2 to 5 independent experiments. Points represent the independents values for each condition. https://doi.org/10.1371/journal.pcbi.1011196.g001 made for cells exposed to isotonic or 180 mOsm media, two conditions where no lysis was detected [17]. During the isotonic shock, representing a mechanical stimulus in the absence of osmotic gradient, [eATP] amounted to 0.33% of [iATP], while under 180 mOsm this figure amounted to 3.6%. Thus, the energy cost of eATP production by iATP efflux was very small (see section 4.9 for further details). No iADP release was detected in the 180 mOsm stimulus (S1 Fig). In our previous work, we showed that ecto-nucleotidases present in Caco-2 cells catalyse significant rates of eATP hydrolysis, leading to eADP accumulation [17]. In principle, the resulting accumulated eADP could be used by the potential presence of ecto-kinases like ecto- AK and ecto-NDPK, present in several cell types, to synthetize eATP. Thus, in the following experiments the activities of ecto-AK and ecto-NDPK were assessed by quantifying eATP kinetics under different conditions. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 4 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells Fig 2. Synthesis of eATP from eADP in Caco-2 cells. (A) The time course of [eATP] synthetized from exogenous eADP (6–48 μM) in the extracellular medium of intact Caco-2 cells was quantified by luminometry. The cells were incubated with the luciferin-luciferase reaction mix and the [eADP] indicated in the figure were added at the time indicated by the arrow. Data are means of at least 3 independent experiments run in duplicate for each [eADP]. (B) Effect of treatment with Ap5A (adenylate kinase inhibitor) on eATP synthesis from eADP in Caco-2 cells. The cells were treated or not (w/o treatment) with 10 μM Ap5A and the [eATP] at 30 minutes was measured by luminometry under similar conditions as experiments in (A). The bars are means ± s.e.m. from at 3–5 independent experiments run in duplicate in the absence of Ap5A and 2 independent experiments in the presence of the inhibitor. ND means that there was non-detected [eATP]. https://doi.org/10.1371/journal.pcbi.1011196.g002 2.1.2. Ecto-AK activity in Caco-2 cells. AK catalyses the following reversible reaction: 2 eADP $ eATP + eAMP and is inhibited by Ap5A [20]. Ecto-AK activity was then assessed by following eATP synthesis when Caco-2 cells were incubated with exogenous eADP (6–48 μM). Non-linear [eATP] increases were proportional to [eADP] (Fig 2A). At 30 minutes post-stimu- lus, treatment with 10 μM Ap5A, which does not permeate cells, inhibited eATP synthesis by 100% (6–24 μM eADP) or by 90% (48 μM eADP) (Fig 2B), thus showing the presence of a functional ecto-AK in Caco-2 cells membrane. 2.1.3. Ecto-NDPK activity in Caco-2 cells. NDPK catalyses the transfer of a γ–phosphate from NTP to NDP. Thus, in the presence of eADP and a given eNTP, the following reaction: eADP + eNTP $ eATP + eNDP leads to eATP synthesis when eADP is phosphorylated by NDPK. Accordingly, incubation of cells with 100 μM eCTP at different [eADP] (3–12 μM) resulted in the rapid synthesis of eATP (Fig 3A). Maximal [eATP] values were obtained 30 minutes after the addition of substrates (Fig 3A). The experiments were conducted in the presence of 10 μM Ap5A to rule out any contribution of ecto-AK to the observed eATP kinetics. Addition of 5 mM eUDP, together with 100 μM eCTP and 12 μM eADP, decreased the eATP synthesis by 91% (Fig 3B), a result compatible with high [eUDP] favouring eUDP to eUTP conversion by ecto-NDPK, rather than eATP synthesis from eADP. In separate experiments, addition of increasing [eUTP] (1–100 μM) without the addition of exogenous eADP (only endogenous eADP present), resulted in a concentration-dependent increase of [eATP] (Fig 3C). Because this increase was abolished by 5 mM eUDP (S2 Fig), we hypothesized that eATP synthesis was due to ecto-NDPK activity using exogenous eUTP and endogenous eADP. This is because there is a basal eADP concentration in the extracellular media of 0.77 ± 0.47 μM eADP/mg protein (S3 Fig). A similar experiment using 100 μM eGTP, instead of eUTP, provided qualitatively similar results (Fig 3D). Overall results showed a functional ecto-NDPK activity capable of synthetizing eATP from different γ-phosphate donors (eCTP, eUTP and eGTP) in the presence of endogenous and exogenous eADP. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 5 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells Fig 3. Extracellular synthesis of eATP from eADP and eCTP, eUTP and eGTP in Caco-2 cells. (A) The time course of eATP synthesis from eCTP (100 μM) and eADP (light blue for 3 μM, blue for 6 μM and grey for 12 μM) in the extracellular medium of Caco-2 cells was quantified by luminometry. Cells were incubated with the reaction mix and eCTP and eADP were added at the time indicated by the arrow. Data are means of 3 to 5 independent experiments run in duplicate for each [ADP]. (B) Production of eATP after 30 minutes exposure of Caco-2 cells to 100 μM eCTP and 12 μM eADP. Experiments were run in the presence of 5 mM eUDP (grey bar and squares) or in its absence (blue bar and points). The [eATP] was measured under conditions similar to the experiments in (A). Results are expressed as [eATP] in μM/mg of protein, bars are means ± s.e.m from 4 to 7 independent experiments run in duplicate. * means p- value <0.01 in comparison with the condition without treatment. (C) and (D) The time course of eATP accumulation in the presence of eUTP (C; grey for 100 μM, dark blue for 10 μM, blue for 1 μM) or 100 μM eGTP (D). Data are the means from 4 independent experiments in the case of 100 μM eUTP, 3 in the case of 100 μM eGTP, and 2 independent experiments in the case of 10 μM or 1 μM eUTP. Nucleotides were added at the time indicated by the arrow. https://doi.org/10.1371/journal.pcbi.1011196.g003 2.1.4. Modelling eATP kinetics of non-polarized Caco-2 cells. Caco-2 cells regulate eATP kinetics by iATP release, eATP synthesis by the activities of ecto-AK and ecto-NDPK (as shown in this study), and hydrolysis by ecto-nucleotidases [17]. These processes are active simultaneously when measuring the eATP dynamics in Caco-2 cells, except when a specific inhibitor was added (like Ap5A in the experiments of Fig 3A and 3B). Thus, to quantify the individual contribution of these processes to eATP kinetics, we built a mathematical model that was then fitted to experimental data. Model parameters contain the kinetic information of each enzyme, allowing to assess the individual contribution of ecto-enzymes to eATP dynamics. A scheme of the model is depicted in Fig 4A. In the model, [eATP] can increase by iATP release, by lytic and by non-lytic mechanisms. In addition, [eATP] can be modulated by the activities of ecto-ATPases, ecto-AK and ecto-NDPK. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 6 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells Fig 4. A model of extracellular purinergic regulation in non-polarized Caco- 2 cells. (A) The scheme shows a representation of the model created to explain the experimental results in non-polarized cells. The yellow bolt indicates that the JATP depended on the application of a hypotonic shock. ADO means extracellular adenosine. The green star behind “eATP” indicates that this is the metabolite measured directly during experiments. (B) The plot shows, in red, the model fitting to eATP kinetics exposed to media of different osmolarities (experimental data correspond to those shown in Fig 1A and 1C). (C) iATP efflux (JATP) predicted by the model upon the hypotonic or isotonic shocks indicated in the figure. https://doi.org/10.1371/journal.pcbi.1011196.g004 The model provides functions describing each of the fluxes involved in transport and metabolism of extracellular nucleotides (see S1 Table and section 4.13.1). Fitting the model to the experimental eATP kinetics under the different conditions allowed to obtain the best-fit values for the parameters of these functions (S1 Table). In that way the contributions of each flux to eATP kinetics were quantified, and several predictions were made. 2.1.4.1. iATP release. For experiments under iso- and hypotonic media, the model found a good fit to experimental data (continuous lines in Fig 4B), thus allowing to predict the rate of iATP efflux (JATP) over time (Fig 4C). JATP was rapid and transient in nature, leading to a 12-fold increase of [eATP] to a maximum in less than 2 seconds under the 180 mOsm shock, followed by rapid inactivation. The magnitude of the JATP peak depended on the osmotic gra- dient imposed. Inactivation of JATP was observed under conditions where no lysis was detected (isotonic and 180 mOsm media). On the other hand, a lytic flux (JL) explains the continuous increase of [eATP] at 100 mOsm (Figs 1C and 4B). 2.1.4.2. Ecto-enzymes. Another factor shaping eATP kinetics is eATP hydrolysis by ecto- ATPase activity. We have previously observed that, in intact non-polarized Caco-2 cells, ecto- ATPase activity follows a linear function of micromolar [eATP] [17]. Thus, following a stimu- lus promoting iATP release, any increase of [eATP] should be at least partially counterbal- anced by an increase of ecto-ATPase activity. Model predictions made at 180 mOsm show that the initial peak of [eATP] increase due to JATP is about 8-fold higher than the rate of eATP hydrolysis, i.e., JATP was 1.2 μM iATP/min/ mg of protein (Fig 4C) and eATP hydrolysis was 0.15 μM eATP/min/mg of protein at 1.5 μM eATP (S1 Table). Thus, during the first seconds of [eATP] increase, eATP kinetics was mainly governed by iATP release. At later times, however, the JATP inactivated, and the ecto-ATPase activity progressively gained importance in controlling [eATP]. This is illustrated by modelling a change in the amount of ecto-ATPase over a wide range, showing that a 5-fold increase of ecto-ATPase activity could lead to rapid decay of [eATP], while a 5-fold decrease would pro- long high levels of [eATP] over the entire incubation period (Fig 5A). However, a similar pro- cedure, i.e, increasing or decreasing 5 times the activity of ecto-AK, had no influence on the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 7 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells Fig 5. Role of ecto-AK, ecto-ATPase and ecto-ADPase activity on eATP dynamics. (A) The simulation shows the [eATP] as a function of time upon a 180 mOsm hypotonic shock when the ecto-ATPase activity displayed its measured value (0.1, continuous line in blue), a 5-fold increase (0.5, dashed line in grey), and a 5-fold decrease (0.02, dashed line in light blue). The numbers in the plot indicate the kinetic constant of the activity in (μM eATP hydrolized)/(mg prot/ μM eATP/ min) units. (B) The simulation shows the [eATP] as a function of time upon a 180 mOsm shock at different initial [eADP] concentrations: calculated pre-stimulus eADP (0.22 μM continuous line in blue), a 3.5-fold increase (0.8 μM, dashed line in grey), and a 14-fold increase (3 μM, dashed line in light blue). (C) The simulation shows the [eATP] as a function of time upon addition of 6 μM eADP (the corresponding experimental results are shown in Fig 2A). The plot shows the eATP kinetics under various values of the kinetic constant for ecto-ADPase, i.e, the constant experimentally determined (0.008, continuous line in blue), a 5-fold increase (0.04, dashed line in grey), and a 12-fold increase (0.96, dashed line in light blue). The numbers in the plot indicate the kinetic constant of the activity in (μM eADP hydrolized)/(mg prot / μM eADP / min) units. (D) Ecto-ATPase, ecto-AK and ecto-ADPase activites as a function of their respective substrates, [eATP] for ecto-ATPase and [eADP] for ecto-AK and ecto-ADPase. The points show the initial velocities for eATP synthesis as a function of [eADP] by ecto-AK calculated from experimental data shown in Fig 2A. The points are means ± s.e.m. from 3 to 5 independent experiments run in duplicate. The continuous lines represent enzyme activities as a function of their respective substrates (see S1 Table for further details). Shadows behind the lines in panels A, B and C represent the uncertainty of the prediction calculated as indicated in section 4.13. Shadows behind the lines in panel D represent the interval ± the standard error of enzyme activity, calculated using the standard error of kinetic parameters of ecto-ATPase and ecto-ADPase activities. https://doi.org/10.1371/journal.pcbi.1011196.g005 [eATP] during the hypotonic shock (not shown). This can be attributed to the sigmoidal kinet- ics of ecto-AK, whose activity is very low below 3 μM eADP, but significantly higher above that concentration (Fig 5D). Thus, ecto-AK might influence eATP kinetics only when [eADP] is sufficiently high. Fig 5B shows a simulation where the initial [eADP] was raised up to 3 μM. At 3 μM eADP, eATP degradation was comparable to eATP synthesis by ecto-AK, indicating that ecto-AK can counterbalance ecto-ATPase activity. Note that a 180 mOsm hypoosmotic PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 8 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells medium does not change the activity of ecto-ATPase, ecto-ADPase, ecto-AMPase [17] or ecto- AK [21] in comparison with isosmotic medium. Another factor to consider is ecto-ADPase activity. We have previously shown that Caco-2 cells displays high ecto-ATPase but very low ecto-ADPase activity [17]. Nevertheless, a hypo- thetical increase of ecto-ADPase activity could negatively modulate ecto-AK activity. For example, an increase of 5- and 12-fold of ecto-ADPase activity would result in a 17% and 33% decrease in the [eATP] production respectively, at 6 μM eADP (Fig 5C). Model predictions showed above implied that the expression of ecto-AK in Caco-2 cells may have an important role in eATP kinetics. To assess this hypothesis, we compared ecto- ATPase, ecto-ADPase and ecto-AK activities as a function of their respective substrate’s con- centrations, that is, [eATP] for ecto-ATPase and [eADP] for ecto-ADPase and ecto-AK (Fig 5D). In Fig 5D, symbols of ecto-AK activities represent the initial velocities for eATP synthesis as a function of [eADP] calculated from experimental data shown in Fig 2A, and the continu- ous line represents the fit to data of the ecto-AK function included in the model (details in S1 Table and in the work of Sheng and collaborators [22]). The ecto-ATPase and ecto-ADPase activities are predictions made from data of our previous work [17]. Ecto-ATPase displayed the highest rate of the three reactions. On the other hand, although at low [eADP], ecto-AK and ecto-ADPase activities are similar and have relatively low values, the sigmoidal kinetics of ecto-AK allows a strong activity increase as [eADP] is raised, thus reaching activity levels well above those of ecto-ADPase activity (Fig 5D). Finally, in the presence of non-adenosine nucleotides, the influence of ecto-NDPK on eATP dynamics was assessed and analysed. Caco-2 cells synthetized eATP by ecto-NDPK activity in the presence of eCTP, eUTP and eGTP as NTP donors, and exogenous and endoge- nous eADP (Fig 3A–3D). The model found a good fit to the experimental eATP kinetics in the presence of 100 μM eCTP and different [eADP] (Fig 6A). Model predictions of ecto-NDPK activity at different [eADP] agreed well with initial velocities of experimental ecto-NDPK activities shown in Fig 3A (Fig 6B). We also studied the effect of eUTP addition without the addition of exogenous eADP (a condition where only endogenous eADP was present, S3 Fig) on the transient rise of [eATP] (Fig 3C, replicated in Fig 6C). To understand the role of eNTPs on ecto-NDPK activity, it is important to recall that ecto-NTPDases of Caco-2 cells can hydrolyse non-adenine nucleotides (S4 Fig). Model predictions show changes in ecto-NDPK and ecto-ATPase activities (Fig 6D), and the corresponding dynamics of [eATP] and [eADP] (Fig 6E), and of [eUTP] and [eUDP] (Fig 6F). Kinetics of eATP (Fig 6C and 6E) could be analysed in 3 stages. First, [eATP] increases due to a high an ecto-NDPK/ecto-ATPase activities ratio in the presence of high [eUTP] and basal [eADP] (stage 1 in Fig 6D, 6E and 6F). The resulting elevated [eATP] activates ecto- ATPase activity, while ecto-NDPK decreases deeply because its substrates (eUTP and eADP) are consumed by ecto-NTPase activity and by ecto-NDPK activity itself. A balance is then estab- lished between ecto-NDPK and ecto-ATPase activities in stage 2, where [eATP] is transiently stable. Finally in stage 3, [eUTP] continues decreasing, leading to a high ecto-ATPase/ecto- NDPK activities ratio, causing [eATP] to decrease and [eADP] to rise again (Fig 6E). 2.2. eATP regulation in polarized Caco-2 cells Because several reports showed differential activities of enzymes and transporters at each side of polarized epithelia [23,24], we speculated that [eATP] regulation might be different at the apical and basolateral sides of polarized Caco-2 monolayers. We then used polarized Caco-2 cells to test the effect of hypotonic shock on iATP release and resulting eATP kinetics at the apical and basolateral sides. Similarly to the procedure PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 9 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells Fig 6. Role of ecto-NDPK on eATP dynamics. (A) The plot shows the experimental results of eATP dynamics in the presence of 100 μM eCTP and various [eADP] (also shown in Fig 3A). Model fitting was applied to data and shown as continuous red lines. (B) The plot shows the ecto-NDPK activity expressed in μM of ATP synthetized per minute per mg of protein. The dots represent the initial velocity of ecto-NDPK (calculated from the experimental data shown in panel A) as a function of [eADP]. Points represent the means ± s.e.m. from 3 independent experiments run in duplicate. The continuous line represents the ecto-NDPK activity predicted by the model (details in S1 Table). (C) The plot shows the experimental eATP dynamics in the presence of various [eUTP] (also shown in Fig 3C) and the continuous red lines represent the model fitting. (D) The plot shows time changes of ecto-NDPK (blue line) and ecto- ATPase (grey line) activities predicted by the model. In the plot, the zones 1 (white background), 2 (pink background) and 3 (white background) represents the [eATP], increase, stabilization and decrease stages respectively. In (E) and (F) the plot shows the model predictions of [eATP] and [eADP], or [eUTP] and [eUDP] respectively as a function of time upon addition of 100 μM eUTP to non-polarized Caco-2 cells. Data is expressed in μM/mg protein, which was calculated by dividing the [eATP] at any time by the average protein mass in the experiments (0.25 mg in average). The shadows behind the lines in panels A, B and C represent the uncertainty of the prediction calculated as indicated in section 4.13. https://doi.org/10.1371/journal.pcbi.1011196.g006 employed for non-polarized cells, we fitted the model shown in Fig 4A to the experimental data to understand quantitatively the mechanisms involved in [eATP] regulation in differenti- ated monolayers of Caco-2 cells. Experimental results show that, following a 180 mOsm hypotonic shock, [eATP] increased at both sides of the monolayers, with qualitatively different kinetics. While at both sides the initial rate of [eATP] increase was fast, apical eATP kinetics achieved a maximum at 1.5 min- utes, followed by a rapid decay. This was not observed in the basolateral domain, where [eATP] continued increasing at a progressively lower rate, and a very slow [eATP] decay was observed only after 20 minutes (Fig 7A). The two different eATP kinetics suggested different activities of ecto-enzymes present at both sides of the monolayers. Therefore, we determined the activities of ecto-ATPase, ecto-AK and ecto-NDPK. For assessing ecto-ATPase activity, polarized Caco-2 cells were exposed to various [eATP] (0.2–7 μM) and eATP hydrolysis was estimated by quantifying [eATP] decay rates (S5 Fig). PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 10 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells Fig 7. Apical and basolateral eATP regulation in Caco-2 monolayers. (A) Effect of hypotonic shock on eATP kinetics. At the times indicated by the arrow, cells were exposed to 180 mOsm medium on the basolateral (grey) or the apical (blue) compartments. Data are the means from 5 independent experiments. (B) Ecto-ATPase activity measured from the eATP kinetics at different [eATP]. Data was obtained from eATP hydrolysis kinetics like the ones shown in S5A and S5B Fig. The points in the plot represent the mean ± s.e.m. of 3 independent experiments. The dashed lines represent a linear regression to the data allowing to obtain the ecto-ATPase kinetic constant which was 1.70 ± 0.08 and 0.36 ± 0.22 mM ATP hydrolized mM ATP mg prot min basolateral compartments respectively. (C) Ecto-AK activity. eATP kinetics in the presence of 12 μM eADP added to the basolateral (grey) or apical (blue) compartments. Data are the means from 2 independent experiments. (D) Ecto-NDPK activity. eATP kinetics in the presence of 100 μM eCTP + 12 μM eADP added to the basolateral (grey) or the apical (blue) compartments. Experiments were run in the presence of 10 μM Ap5A (adenylate kinase blocker). Data are the means from 3 independent experiments. (E) Ecto-AK initial velocities in polarized Caco-2 cells. Data are means + s.e.m. of 4 independent experiments. * indicates a p-value < 0.05 in comparison with the apical condition. (F) Ecto-NDPK initial velocities in polarized Caco-2 cells. Data are means ± s.e.m. of 3 independent experiments. (G) Scheme of the results interpretation showing that the increased activity of Ecto-AK, Ecto-NDPK and Ecto-ATPase leads to a faster eATP turnover. for the apical and https://doi.org/10.1371/journal.pcbi.1011196.g007 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 11 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells The initial rate values of [eATP] decay were used to calculate ecto-ATPase activity at each [eATP], so as to build a substrate curve (Fig 7B). Linear fitting to experimental data showed that ecto-ATPase activity was 4-fold higher in the apical than in the basolateral domain. To assess ecto-AK activity, Caco-2 cells were exposed to 12 μM eADP at the basolateral or apical domains. In the apical domain, [eATP] increased rapidly to a maximum, followed by a rapid decay towards pre-stimulated levels, while basolateral [eATP] increased steadily at a lower rate (Fig 7C). Initial velocity estimations showed that ecto-AK activity was significantly higher in the apical than in the basolateral compartment (Fig 7E). Ecto-NDPK activity was quantified using polarized cells exposed to 100 μM eCTP plus 12 μM eADP in the basal and apical domains. Experiments were run in the presence of 10 μM Ap5A to block ecto-AK activity. Production of eATP by ecto-NDPK was much higher than that observed under conditions used to measure ecto-AK activity, though the domain specific pattern of eATP kinetics was similar when assessing the two ecto-kinases, i.e., a biphasic pat- tern in the apical domain, and a steady [eATP] increase, at a lower rate, in the basal domain (Fig 7D). The initial velocity of ecto-NDPK was higher in the apical than in the basolateral domain although differences were not significant (Fig 7F, p value = 0.1). A good fitting of the model to all experimental data was achieved (red lines in Fig 7A, 7C and 7D). The model fitting allowed to obtain the ecto-NDPK and ecto-AK maximal velocity (Vmax) and compared them with the ones obtained from non-polarized cells (S6A and S6B Fig and S2 Table). Results indicated that the ecto-NDPK maximal activity in the apical compart- ment is a slightly higher than that of the non-polarized cells and significantly higher than that of the basolateral compartment. On the other hand, the ecto-AK maximal activity is signifi- cantly higher compared with the basolateral compartment or the non-polarized cells. Thus, the differences between the apical and basolateral eATP dynamics can be explained by an increase or decrease in ecto-enzymes activities. Altogether experimental results showed significantly higher activities of the ecto-enzymes (ecto-ATPase, ecto-AK and ecto-NPDK) in the apical, as compared to the basolateral domain (Fig 7G). Fig 8. Ecto-AK and ecto-NDPK activities in IECs. Time course of eATP synthetized from exogenous eADP (A) or eCTP and eADP (B) in the extracellular medium of IECs. (A) The cells were incubated with the luciferase-luciferin reaction mix and 12 μM eADP was added at the time indicated by the arrow in presence (grey) or absence (blue) of 10 μM Ap5A (B) 100 μM eCTP plus 12 μM eADP, in the presence of 10 μM Ap5A were added at the time indicated by the arrow. [eATP] was quantified by luminometry. Values are the means of 3 independent experiments run in duplicate. https://doi.org/10.1371/journal.pcbi.1011196.g008 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 12 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells 2.3. Ecto-AK and ecto-NDPK are active in human primary small intestinal epithelial cells Having characterized ecto-AK and ecto-NDPK activities of Caco-2 cells, we wondered whether these ecto-enzymes would be functional in IECs extracted from human small intes- tine. Accordingly, we used samples obtained from small intestine biopsies from healthy donors (Fig 8A and 8B). Results show that exposure of IECs to both 12 μM eADP (to assess ecto-AK) (Fig 8A) or to 12 μM eADP + 100 μM eCTP (to assess ecto-NDPK, Fig 8B) led to significant eATP produc- tion in the micromolar range. Furthermore, as expected, the presence of ApA5 totally inhibited the ecto-AK activity (Fig 8A). 3. Discussion Intestinal epithelial cells can release iATP and express several ecto-enzymes capable of regu- lating the amount and metabolism of eATP at the cell surface. The main goal of this study was to characterize quantitatively the dynamic interplay of iATP release, eATP hydrolysis and eATP synthesis contributing to the dynamic regulation of [eATP] in Caco-2 cells. Spe- cial emphasis was given to the role of ecto-kinases promoting eATP production under differ- ent conditions. Since Caco-2 cells undergo spontaneous enterocytic differentiation in culture, we decided to first approach the complexity of eATP regulation using the relatively simpler non-polarized cell model, and later extend the study to fully differentiated cells. These form apical and baso- lateral poles where morphological and biochemical features are segregated [23]. When exposed to hypotonicity, non-polarized Caco-2 cells triggered a strong iATP efflux that rapidly inactivated, leading to low μM [eATP] accumulation. A number of studies have confirmed that such micromolar [eATP] are capable of activating P2 receptors with high affin- ity for that nucleotide, such as P2Y2, P2Y11 and almost all P2X receptors [25]. In Caco-2 cells, eATP dose dependently activates P2Y receptors involved in the activation of MAPK cascades and transcription factors that promote cell proliferation [26,27], while higher [eATP] can induce apoptosis via P2X7 receptor [3]. In principle, purinergic activation by eATP should be transient, due to the presence of ecto- nucleotidases, the activities of which promotes strong eATP hydrolysis in Caco-2 cells [17]. Accordingly, our results show that hypotonicity induced iATP release and concomitant eATP accumulation, where [eATP] decay was accelerated by constitutive ecto-ATPase activity. This decay was even higher for a model predicted upregulation of eATP hydrolysis by one or more ecto-nucleotidases, as occurs in various cells and tissues during pathogen infection [28], cell differentiation [29] or tumorigenesis [30]. The above results imply that iATP release and eATP hydrolysis constitute two opposing fluxes shaping eATP kinetics of Caco-2 cells. However, the presence of ecto-kinases found in this study suggest that the dynamic regulation of [eATP] should also take the activities of these enzymes into account. In this respect, addition of exogenous eADP to Caco-2 cells dose dependently increased [eATP]. The fact that eATP synthesis was almost fully blunted by Ap5A, an AK blocker that does not permeate intact cells, suggested the presence of a functional ecto-AK. Results of the mathematical model allowed to envisage the contribution of ecto-AK to eATP kinetics. In the absence of exogenous eADP, the contribution of ecto-AK to eATP kinetics was negligible, so that [eATP] depended mainly on the balance between the rates of iATP release and eATP hydrolysis. This is due to the low endogenous [eADP] present under the experimental condi- tions. However, due to the sigmoidal nature of the AK reaction, model predictions show that PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 13 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells increasing [eADP] in the low micromolar range, suffices to promote significant eATP synthe- sis by ecto-AK, upregulating eATP kinetics. Thus, under certain conditions, e.g., when cell leak intracellular ADP (iADP) or eADP is supplied paracrinally by other cell types, eATP synthesis by ecto-AK of Caco-2 cells will transiently stabilize [eATP] levels, thereby favouring propaga- tion of eATP-dependent purinergic signalling. A similar stabilizing role of ecto-AK on [eATP] has been proposed for HT29 cells, lung epithelial cells and lymphocytes [14,15,31]. Modelling shows that ecto-ADPase activity, which facilitates eADP degradation, may com- pete with ecto-AK for the available eADP. However, Caco-2 cells -as HT29 cells [15]- displayed a relative low ecto-ADPase activity, in agreement with the presence of a functional ecto- NTPDase 2 in both cell types [15,17], and in addition the intrinsic sigmoidal nature of ecto- AK activity makes ecto-AK more sensitive to [eADP] than ecto-ADPase. Another consequence of ecto-AK activation relates to P1 signalling, since activity of this enzyme will provide eAMP from eADP for further hydrolysis to adenosine by ecto-5’NT present in Caco-2 cells [17], finally leading to extracellular adenosine accumulation. Our model predictions show how increasing [eADP] in the low μM range might lead to substantial adenosine accumulation, which may engage 4 different P1 receptors [32]. The con- sequences of P1 signalling on proliferation of Caco-2 cells and several other intestinal epithelial cell lines have been studied before [33]. In general, the balance between P1 and P2 receptors on epithelial cells regulate intestinal secretion [34–37] and absorption [38,39]; responses trig- gered by the P2 receptor stimulation by eATP and other nucleotides are sometimes counter- acted by P1 receptor stimulation by adenosine, though the potential role of ecto-AK was not considered in this context. Another factor affecting eATP kinetics is ecto-NDPK. Activity of this enzyme was detected in many cells and tissues such as astrocytoma cells [40], endothelial cells [41,42], lymphocytes [41], keratinocytes [43] and hepatocytes [44]. In general, ecto-NDPK will pri- marily serve to transfer phosphate groups between different extracellular nucleotides and thus potentially alter the pattern of P2 receptor activation. This is especially important since P2 receptor subtypes are differentially selective for adenine and uridine eNDPs and eNTPs [45,46]. Our results show that ecto-NDPK can use eCTP, eGTP and eUTP to phosphorylate eADP to eATP. As model predictions show, activities of ecto-NDPK (promoting eATP synthesis from eUTP and eADP) and ecto-nucleotidase (promoting eATP and eUTP hydrolysis) change in opposite directions to transiently stabilize [eATP]. Results analysed above show that, in non-polarized Caco-2 cells, [eATP] can increase by iATP release and ecto-kinase mediated eATP synthesis and decrease by ecto-nucleotidases mediated by eATP hydrolysis. Next, we studied eATP dynamics of polarized Caco-2 cells. These cells differentiate sponta- neously into polarized cells, with apical and basolateral domains exhibiting morphological and biochemical features of small intestine enterocytes [23,47]. In particular, the Caco-2 polarized phenotype is characterized by high levels of hydrolases typically associated with the brush bor- der membrane. The fact that in a variety of epithelia several ecto-nucleotidases and ecto-phos- phatases preferentially -but not exclusively- locate in the apical domain [48–50], anticipated a different eATP regulation at both poles of Caco-2 cells. Accordingly, hypotonically induced eATP kinetics had a faster resolution and was more effectively regulated at the apical, than at the basolateral side, a result in agreement with the observed higher apical (than basolateral) ecto-ATPase activity measured in this study. This is in agreement with several reports using intestinal epithelial cell from murine models and human intestinal cell lines, showing that various isoforms of ecto-NTPDases, ecto-phospha- tases and ecto-NPPases are preferentially located in the apical domain [50]. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 14 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells The mechanism mediating iATP release at both sides of the polarized Caco-2 cell mono- layer requires further investigation. We previously showed that iATP release in Caco-2 cells challenged by adrenergic stimulation or the presence of bacteria is reduced by treatment with carbenoxolone, a blocker of conductive iATP efflux [51]. Since the molecular mechanisms involved in iATP release may depend on the stimulus, further investigations will be conducted to clarify this topic. A qualitatively similar pattern was observed for ecto-AK and ecto-NDPK of Caco-2 cells, in that apical activities were much higher. Interestingly, the model describing eATP dynamics of non-polarized cells could be successfully fitted to eATP kinetics on each of the polarized domains, thus allowing to calculate the consequences of ectoenzymes sorting on eATP regulation. The results obtained with primary human IECs suggest that our model could be employed to explain the eATP kinetics in these cells, given the presence of both ecto-AK and ecto-NDPK. However, further studies would be required to adapt the Caco-2 model to IECs. These studies may need to account for bacterial periplasmic ATPases capable of hydrolyzing eATP, as we and others have shown that commensal bacteria can express periplasmic nucleotidases and release ATP into the intestinal lumen [51,52]. In this context, the model presented here may be taken as a starting point to progressively add other processes affecting [eATP] regulation in vivo. The fact that the apical domain exhibited a higher turnover of extracellular nucleotides, leading to higher eATP regulation may have adaptive value, considering that iATP release is a common response of epithelial intestinal cells to enteric pathogens [53]. Extracellular ATP may then act as a danger signal controlling a variety of purinergic responses aimed at defend- ing the organism from a variety of pathogens and their toxins present in the intestinal lumen. 4. Materials and methods 4.1. Ethics statement The protocol for handling human samples was approved by the Institutional Review Board of the Favaloro Foundation University Hospital (DDI [1587] 0621) and has been performed in accordance with the ethical standards laid down in the declarations of Helsinki and Istanbul. Informed written consent was obtained from donors. 4.2. Chemicals All reagents were of analytical grade. Bovine serum albumin (BSA), malachite green, adeno- sine 50 -triphosphate (ATP), adenosine 50 -diphosphate (ADP), cytidine 50-triphosphate diso- dium salt (CTP), adenosine 50 -monophosphate (AMP), uridine 5’-triphosphate (UTP), uridine 5’-diphosphate (UDP), guanosine-5’-triphosphate (GTP), phosphate-buffered saline (DPBS), 4-(2-hydroxyethyl)-1-piperazineetahnesulfonic acid (HEPES), ammonium molyb- date, Triton X-100, phenylmethylsulphonyl fluoride (PMSF), pyruvate kinase, phosphoenol- pyruvate (PEP), luciferase, coenzyme A and P1,P5-Di (adenosine-5´) pentaphosphate pentaso- dium salt (Ap5A) were purchased from Sigma-Aldrich (St Louis, MO, USA). D-luciferin was purchased from Molecular Probes Inc. (Eugene, OR, USA). 4.3. Solutions In the experiments to measure [eATP] by luminometry (section 4.5), cells were incubated with isotonic buffer called isosmotic DPBS (300 mOsm) containing: 137 mM NaCl, 2.7 mM KCl, 1 mM CaCl2, 2 mM MgCl2, 1.5 mM KH2PO4 and 8 mM Na2HPO4, pH 7.4 at 37˚C (assay medium). When applying a hypotonic shock to cells, the medium was changed for other con- taining the same components but with a lower NaCl concentration. Thus, DPBS with 100, 150 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 15 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells and 180 mOsm were prepared. The osmolarity of all media was measured with a vapor pres- sure osmometer (5100B, Lugan, USA). When measuring phosphate (section 4.8.4) the following medium without phosphate was employed instead of isotonic buffer: 145 mM NaCl, 5 mM KCl, 1 mM CaCl2, 10 mM HEPES, and 1 mM MgCl2, pH 7.4 at 37˚C. 4.4. Caco-2 cell culture Caco-2 cells (ATCC, Molsheim, France) were grown in Dulbecco’s modified Eagle’s medium (DMEM-F12, Gibco, Grad Island, NY, USA) containing 4.5 g/L glucose (Sigma-Aldrich, St Louis, MO, USA) supplemented with 10% v/v fetal bovine serum (Natocor, Co´rdoba, Argen- tina), 2 mM L-glutamine (Sigma-Aldrich, St Louis, MO, USA), 100 U/mL penicillin, 100 μg/ mL streptomycin and 0.25 μg/mL fungizone (Invitrogen, Carlsbad, CA, USA) in a humidified atmosphere of 5% CO2 at 37˚C. For eATP kinetics measurements cells were directly seeded on glass coverslips. For ecto-nucleotidase activity experiments using the malachite green method, cells were seeded in cell culture 24-well plates (Corning Costar, NY, USA). 4.4.1. Polarisation of Caco-2 cells. For preparation of polarized Caco-2 monolayers, cells were seeded in permeable supports (inserts) made of polyester (Transwell; 0.1 μm pore size, 1.12 cm2 cell growth area; Jet Biofil, China) in 12-well plates at a density of 3 × 104 cells/0.5 mL per insert. The medium was changed after 3 days, and then after every 3 or 4 days. The polar- ized Caco-2 monolayers were used for experiments after the transepithelial electrical resistance reached a plateau (approximately 21 days after seeding). In polarized and non-polarized cul- tures contamination (including Mycoplasma) was routinely tested. 4.5. Human Intestinal Epithelial Cells (IECs) isolation IECs were isolated from ileum biopsies collected from healthy volunteers who were endoscopi- cally evaluated for colon cancer (N = 3) at the Favaloro Foundation University Hospital. Sam- ples of non-tumoral, non-injured intestinal biopsies were collected and transported in ice-cold Hanks’s balanced salt solution (HBSS) for immediate processing. The biopsies were incubated in 5 mM ethylenediaminetetra-acetic acid (EDTA) and 1.5 mM dithiothreitol HBSS with agita- tion for 25–30 minutes at room temperature to obtain IECs. Cells were pelleted, re-suspended in DPBS and used immediately. 4.6. ATP measurements The [eATP] of non-polarized Caco-2, polarized Caco-2 monolayers or IECs was measured using the firefly luciferase reaction (EC 1.13.12.7, Sigma-Aldrich, St Louis, MO, USA), which catalyses the oxidation of D-luciferin in the presence of ATP to produce light [54]. As described below, using this method it was possible to determine eATP kinetics, the iATP con- tent and the activities of ecto-enzymes. Before the experiments, the cells were washed two times with the assay medium (isosmotic DPBS with or without Pi). In this work, the cells’ medium was substituted by the assay medium before any measure- ment, therefore exoenzymes (enzymes released to extracellular medium not bound to the membrane) were removed and only ecto-enzymes (membrane bound extracellular enzymes) were investigated. 4.7. eATP kinetics of non-polarized Caco-2 and IECs Non-polarized Caco-2 cells and IECs were seeded on glass coverslips. Under all conditions cells were mounted in the assay chamber of a custom-built luminometer, as previously PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 16 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells described [55]. Because luciferase activity at 37˚C is only 10% of that observed at 20˚C [56], to maintain full luciferase activity, [eATP] measurements were performed at room temperature. The setup allowed continuous measurements of [eATP] by the luciferin-luciferase reaction. A calibration curve was used to transform the time course of light emission into [eATP] versus time. Increasing [eATP] from 13 to 1000 nM were sequentially added to the assay medium from a stock solution of pure ATP dissolved in isosmotic or hypotonic medium, according to the experiment. Calibration curves displayed a linear relationship within the range tested. After each experiment, cells were lysed with a solution containing 1 mM PMSF and 0.1% of Triton X-100 and the protein contents of each sample were quantified [57]. Results were expressed as [eATP] at every time point of a kinetics curve denoted as “eATP kinetics”, with [eATP] expressed as μM of eATP/mg protein in a final assay volume of 100 μL. 4.8. eATP kinetics of polarized monolayers Polarized Caco-2 cells monolayers were placed in the insert physically separating an apical and a basolateral compartment. Detection of eATP was performed separately on either side, by adding the luciferin-luciferase mixture in one compartment (apical or basolateral) and adding isosmotic DPBS to the other side. In preliminary experiments, we observed that the luciferin- luciferase mix added in one compartment did not cross the monolayer into the other compart- ment. Thus, luminescence registered when measuring the [eATP] in one compartment was not contaminated by light from the other compartment due to luciferin-luciferase leakage. When an hypoosmotic shock was applied, a luciferin-luciferase mix in DPBS with an osmo- larity of 180 mOsm was added to the compartment of interest while, isosmotic DPBS was added to the other side. 4.9. Activities of ecto-enzymes Ecto-ATPase, ecto-AK and ecto-NDPK activities of intact cells were measured by luminome- try (section 4.5). Ecto-nucleotidase activities were measured by measuring the inorganic phos- phate (Pi) release. 4.9.1. Ecto-ATPase activity Cells were exposed to different [eATP] (0.2, 1.2, 4.2 or 7 μM). Following an acute increase of [eATP], ecto-ATPase activity was estimated from the initial velocity of eATP decay at each [eATP]. 4.9.2. Ecto-AK activity. Cells were exposed to different [eADP] (6, 12, 24 or 48 μM) and the eATP kinetics was quantified in the absence and presence of 10 μM Ap5A (an AK blocker). Ecto-AK initial velocity was estimated as indicated in section 4.12. 4.9.3. Ecto-NDPK activity. Cells were exposed to different [eADP] (3, 6 or 12 μM) in the presence of eCTP (100 μM), eGTP (100 μM) or eUTP (1, 10 or 100 μM). Then, the eATP kinet- ics was quantified in the presence of Ap5A to block the eADP to eATP conversion by ecto-AK activity. In some experiments 5 mM eUDP was added to inhibit ecto-NDPK activity. Ecto- NDPK initial velocity was estimated as indicated in section 4.12. 4.9.4. Ecto-NTPDase activities. Cells were incubated with 500 μM of eCTP, eUTP or eGTP at 37˚C. Samples were taken at 30, 60, 90 and 120 minutes after nucleotides addition and, the inorganic phosphate concentration was measured by the malachite green method [17,58]. Activities measured in section 4.8.1 were expressed as μM of eATP hydrolysed per minute, normalized by the cell protein mass in the experimental sample (μM of eATP /mg protein/ min). Results from experiments explained in sections 4.8.2 and 4.8.3 were expressed as μM of PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 17 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells eATP synthetized per minute, normalized by the cell protein mass in the experimental sample (μM of eATP /mg protein/min). Activities measured in section 4.8.4 were expressed as μM of inorganic phosphate released per minute, normalized by the cell protein mass in the experi- mental sample (μM of Pi /mg protein/min). 4.10. Intracellular ATP measurements Caco-2 (0–30,000 cells) were laid on coverslips, incubated with 45 μL of luciferin-luciferase reaction mix for 5 minutes and subsequently permeabilized with digitonin (1.6 mg/mL final concentration). Light emission was transformed into [eATP] as a function of time as indicated in section 4.6. After considering the total volume occupied by Caco-2 present in the chamber, and the relative solvent cell volume (3.66 μl per mg of protein) [59], [iATP] was calculated in mM. To calculate the % of iATP release, the following equation was employed: %iATP ¼ 100 x ATPcell ð1Þ where ATPcell represents the [ATP] obtained when iATP from all cells is released into the assay medium. The "x" denotes the [eATP] measured at any time. The value of ATPcell was 66 μM/mg protein and was calculated by multiplying the [iATP] (1.8 mM, section 2.1.1) by the Caco-2 cell volume (3.66 μl per mg of protein [59]) and diving by the assay volume (0.1 mL). 4.11. Extracellular ADP measurements For detection of eADP of intact Caco-2 cells, 3 U/100 μL of pyruvate kinase and 100 μM PEP were added to the luciferin-luciferase mix. Using PEP as a substrate, pyruvate kinase promotes the stoichiometric conversion of eADP into eATP [60]. The resulting eATP was then mea- sured by light emission using the luciferin-luciferase procedure described above. 4.12. Data analysis Statistical significance was determined using the non-parametric Mann-Whitney test. Data were analyzed and graphically represented using GraphPad Prism software v5.0 (Graph Pad Software, San Diego, CA, USA). Each independent experiment was carried out in an indepen- dent cell culture or tissue sample in a different day. 4.13. Initial velocity estimation To measure the initial velocity of Ecto-AK or Ecto-NDPK, the eATP dynamics were measured as indicated in section 4.8.2 and 4.8.3. Only the values of [eATP] obtained during the first 5 minutes after substrates addition were considered for further analysis. The following equation was fitted to experimental data: ½ eATP � ¼ Að1 (cid:0) e(cid:0) k timeÞ ð2Þ where A and k are parameters, whose value are optimized to achieve a good fitting of Eq 2 to experimental data. The initial velocity is the derivative of [eATP] as a function of time at time 0 (the time when substrates were added). Thus, the initial velocity was calculated by multiply- ing the value of A by the value of k. 4.14. Mathematical modelling Chemical models of extracellular nucleotides were built using COPASI (Complex Pathway Simulator) software in version 4.29 (source: https://copasi.org/) [61]. Parameter optimization PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 18 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells was performed using COPASI “parameter estimation function” with Hooke & Jeeves (50 itera- tion steps, 10−5 tolerance and 0.2 rho factor), Levenberg-Marquardt (2000 iteration steps, 10−6 tolerance), or Evolutionary programming (200 generations with a 20 population size) as opti- mization methods. An initial guess of the parameter value was proposed based on literature data for each kinetic step. A detailed description of the models employed in this work can be found in S1 and S2 Tables. Parameters obtained from the model fitting are expressed as the best value ± standard deviation. The COPASI and SBML files of the models described in sec- tion 4.13.1 and 4.13.2 can be found in the data repository (see data availability statement). When performing time course simulations, we used the deterministic LSODA method with default parameters in COPASI version 4.29. Specifically, we set the relative tolerance to 10−6, the absolute tolerance to 10−12, and allowed for a maximum of 105 internal steps without a limit on maximal step size. Model prediction uncertainties were calculated using the parameter scan task in COPASI, which explores the parameter space within the interval given by the parameter best value ± the standard error (S1 Table). Four parameter values were considered within the range of the parameter best value ± the standard error, one at the lowest value of the interval, one at the highest and two in the middle. Simulations were performed by testing all possible combina- tions of the selected parameter values. The shaded regions depicted in Figs 5 and 6 correspond to the area that includes all simulations results obtained by varying the parameters values. In Fig 5, the prediction uncertainty was calculated by simulating the eATP kinetics while varying the parameters KATP, KADP, Vmax Ecto-AMPase and FtrAK. These are the kinetic parameters that control eATP kinetics under the hypotonic stimulus. In Fig 6D, 6E and 6F, the prediction uncertainty was calculated by simulating the nucleotides kinetics by varying the parameters KATP, KmAD, KmUTP Vmax NDPK and KNTPase. 4.1.4.1. A model of purinergic homeostasis in non-polarized Caco-2 cells. To explain the experimental observations, a data driven mathematical model was created (depicted in Fig 4A). The model has 7 reactions to explain the chemical fluxes of transformations or transport of extracellular nucleotides in Caco-2 cells: JATP, JEcto-ATPase, JEcto-ADPase, JEcto-AMPase, JEcto-AK, JEcto-NDPK and JEcto-NTPDase. A detailed description of each flux, its mathematical description and parameters can be found in S1 Table. In the model, the concentration of each species as a function of time was calculated from the following differential equations: � ½ @ eATP @t ¼ JATP (cid:0) ð JEcto(cid:0) ATPase þ JEcto(cid:0) AK þ JEcto(cid:0) NDPK Þ � ½ @ eADP @t ¼ JEcto(cid:0) ATPase (cid:0) JEcto(cid:0) ADPase þ 2∗JEcto(cid:0) AK þ JEcto(cid:0) NDPK � ½ @ eAMP @t ¼ JEcto(cid:0) ADPase (cid:0) ð JEcto(cid:0) AK þ JEcto(cid:0) AMPase Þ � ½ @ eADO @t ¼ JEcto(cid:0) AMPase � ½ @ eCTP @t ¼ JEcto(cid:0) NDPK � ½ @ eCDP @t ¼ (cid:0) JEcto(cid:0) NDPK ð3Þ ð4Þ ð5Þ ð6Þ ð7Þ ð8Þ PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 19 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells � ½ @ eUTP @t ¼ JEcto(cid:0) NDPK (cid:0) JEcto(cid:0) NTPase � ½ @ eUDP @t ¼ (cid:0) JEcto(cid:0) NDPK þ JEcto(cid:0) NTPase ð9Þ ð10Þ Note that in the equations JEcto-AK and JEcto-NDPK were considered in the direction of eATP consumption, i.e., eAMP + eATP $ 2 eADP for JEcto-AK and eNDP + eATP $ eNTP + eADP for JEcto-NDPK. The model was written in COPASI 4.29 and was fitted simultaneously to all experimental data shown in Figs 1A, 1C, 2A, 3A and 3B. The fitting of the model to experi- mental data can be seen in Figs 4B, 5C, 6A and 6C as red lines. Some kinetic parameters of the enzymes catalyzing the reactions were obtained from the lit- erature. Parameters from the Jecto-ATPase and Jecto-ADPase were obtained from our previous work [17]. The Vmax of the Jecto-AMPase reaction was obtained from our previous work [17], while the Km was obtained from the work of Navarro et al. [62]. Kinetic parameters of the Jecto- AK activity were obtained from the work of Sheng et al [22]. The equilibrium constant (Keq) and the affinity for ATP (KmAT) of the Jecto-NDPK were obtained from the work of Garces and Cleland [63]. The affinity constants for product inhibition in Jecto-NDPK (KiNDP and KiADP) were estimated from the work from Lascu et Gonin [64]. The rest of the model parameters were obtained from model fitting to experimental data (see S1 and S2 Tables for more details). The shape of the JATP flux as a function of time was modeled based on findings of a previous work from our group [65]. 4.1.4.2. A model of purinergic homeostasis in polarized Caco-2 cells. The model fitted to experimental data from the apical and basolateral compartments data is the same model indicated in section 4.13.1, although the parameters of some reactions were fitted again (S2 Table). The JATP expression for the 180 mOsm hypotonic shock in the polarized cells was dif- ferent from the one employed on non-polarized cells (S2 Table). The mathematical expressions of the other 6 reactions were not modified. Four parameters were refitted to the data to account for differences in the ecto-ADPase, ecto-AK and ecto-NDPK activities after polariza- tion (values can be found in S2 Table). Moreover, in the case of ecto-NTPDase, the eCTP hydrolysis could not be neglected in the apical compartment and was necessary to achieve a good fit to experimental data. In contrast the eCTP hydrolysis could be avoided in the basolat- eral compartment without affecting model fitting. This suggest that the ecto-CTPase activity is greater in the apical than in the basolateral compartment, in agreement with the increased activity of other enzymes on the apical side. The differential equations for [eCTP] and [eCDP] are modified in the apical side model to account for the eCTP hydrolysis: � ½ @ eCTP @t ¼ JEcto(cid:0) NDPK (cid:0) JEcto(cid:0) NTPase � ½ @ eCDP @t ¼ (cid:0) JEcto(cid:0) NDPK þ JEcto(cid:0) NTPase ð11Þ ð12Þ The models for the apical and basolateral compartments were written in COPASI 4.29 and fit- ted to experimental data shown in Fig 7A, 7C and 7D. The COPASI files can be found at https://doi.org/10.6084/m9.figshare.21938651.v1. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 20 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells Supporting information S1 Fig. iADP release estimation. Increase in [eATP] after a 180 mOsm hypotonic shock in absence (blue) or presence (grey) of PK (3 U) and PEP (100 μM) were evaluated as ΔATP, i.e., the difference between [eATP] at 1 min post-stimulus and basal [eATP]. (TIF) S2 Fig. Inhibition by exogenous eUDP of ecto-NDPK activity in the presence of eCTP and eADP. Time course of eATP accumulation in the presence of 100 eUTP μM in the absence (blue) or in the presence of 5 mM eUDP (grey). The data showed are the means of 3 indepen- dent experiments. (TIF) S3 Fig. Measurement of eADP by the conversion to eATP. Caco-2 cells were incubated with luciferin-luciferase and, at the time indicated with the arrow, PK (3 U) and PEP (100 μM) were added. The value of the [eADP] in resting conditions was 0.77 ± 0.44 μM eADP/mg. Given a usual protein cell mass of 0.2 mg, the [eADP] in resting conditions is 0.15 ± 0.09 μM. The data showed are the means of 5 independent experiments. (TIF) S4 Fig. Ecto-nucleotidase activity of Caco-2 cells. Experiments were performed in assay medium without Pi at room temperature, and Pi production was measured by the malachite green method (section 4.9.4). The time course of Pi accumulation in the extracellular media of Caco-2 cells was measured and values of enzyme activity were derived from initial rates of nucleotides hydrolysis for 500 μM of eUTP (grey), eGTP (blue) and eCTP (light blue). The data are the means of ± s.e.m. from 3 to 5 independent experiments. (TIF) S5 Fig. Basolateral and apical ecto-ATPase activity of Caco-2 cells. eATP kinetics of cells exposed to [eATP] (0.2–7 μM). Levels of [eATP] were measured by luminometry at the baso- lateral (A) and apical (B) sides of the polarized Caco-2 monolayers. Data is the mean of 3 inde- pendent experiments run in duplicate. The initial velocity of the ecto-ATPase activity was calculated by linear regression to experimental data obtaining the slope and y-intercept of the line. The slope represented the eATP hydrolysis as a function of time, i.e. the ecto-ATPase activity at each [eATP] and in each compartment. (TIF) S6 Fig. Enzyme Vmax calculated from model fitting. The plot shows the enzymes’ Vmax in the apical and basolateral compartments, and in non-polarized cells. The ecto-NDPK Vmax (A) were obtained from model fitting to experimental data and are the same shown in S2 Table (for the apical and basolateral compartments) and in S1 Table (for the non-polarized cells). The ecto-AK Vmax (B) was calculated from the model parameters using the following formula: FtrAK k(cid:0) 2k1 , where the FtrAK was obtained from the model fitting (S2 Table for the apical and baso- k(cid:0) 2þk1 lateral compartments and S1 Table for the non-polarized cells). The k-2 and k1 parameters value can be found in S1 Table. (TIF) S1 Table. Mathematical model of eATP regulation in non-polarized Caco-2 cells. Numeri- cal values of constants were normalized by the protein cell mass in the experiments (Mcell), measured by the Bradford method (section 4.7 in the manuscript). Parameter fitting and simu- lations were performed by selecting the average cell mass in the experiments (Mcell = 0.2 mg). JL and JNL represent the lytic and non-lytic iATP release respectively upon an osmotic shock. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 21 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells The value of these terms was 0 before shock application. Jleakage represents a constant and small iATP release observed in the absence of any stimulus. The parameter values obtained from the model fitting are expressed as the best value ± standard deviation. (XLSX) S2 Table. A: JNL parameters obtained from fitting to experimental data at 180 mOsm shock in apical or basolateral compartments in polarized cells. The same value of kobs was considered for both compartments. Parameters obtained from the model fitting are expressed as the best value ± standard deviation. B: Parameters obtained from model fitting to experimental in apical or basolateral compartments in polarized cells. The model equations are the same shown in S1 Table, however, some parameters values were fitted again to experimental data from polarized cells. The parameters whose value has changed in comparison with the model of non-polarized cells are shown in this file. The rest of the parameters had the same value for non-polarized cells (shown in S1 Table). The KADPase and KNTPase (for eCTP) were considered 0 in the basolateral compartment. This does not mean that there is no ecto-ADPase or ecto- NTPase activity in the basolateral side but, they can be neglected in our experimental condi- tions. Parameters obtained from the model fitting are expressed as the best value ± standard deviation. (XLSX) Acknowledgments We are thankful to Dr. Cafferata for providing the Caco-2 cells. Author Contributions Conceptualization: Nicolas Andres Saffioti, Pablo Julio Schwarzbaum, Julieta Schachter. Data curation: Nicolas Andres Saffioti, Pablo Julio Schwarzbaum. Formal analysis: Nicolas Andres Saffioti, Julieta Schachter. Funding acquisition: Pablo Julio Schwarzbaum, Julieta Schachter. Investigation: Nicolas Andres Saffioti, Cora Lilia Alvarez, Zaher Bazzi, Marı´a Virginia Genti- lini, Julieta Schachter. Methodology: Nicolas Andres Saffioti, Cora Lilia Alvarez, Gabriel Eduardo Gondolesi, Pablo Julio Schwarzbaum, Julieta Schachter. Project administration: Pablo Julio Schwarzbaum, Julieta Schachter. Resources: Julieta Schachter. Supervision: Pablo Julio Schwarzbaum, Julieta Schachter. Writing – original draft: Nicolas Andres Saffioti, Pablo Julio Schwarzbaum, Julieta Schachter. Writing – review & editing: Nicolas Andres Saffioti, Cora Lilia Alvarez, Pablo Julio Schwarz- baum, Julieta Schachter. References 1. Lavoie EG, Gulbransen BD, Martı´n-Satue´ M, Aliagas E, Sharkey KA, Se´ vigny J. Ectonucleotidases in the digestive system: focus on NTPDase3 localization. Am J Physiol Liver Physiol. 2011; 300: G608– G620. https://doi.org/10.1152/ajpgi.00207.2010 PMID: 21233276 2. Stefan C, Jansen S, Bollen M. Modulation of purinergic signaling by NPP-type ectophosphodiesterases. Purinergic Signal. 2006; 2: 361–370. https://doi.org/10.1007/s11302-005-5303-4 PMID: 18404476 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 22 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells 3. Coutinho-Silva R, Stahl L, Cheung K-K, De Campos NE, de Oliveira Souza C, Ojcius DM, et al. P2X and P2Y purinergic receptors on human intestinal epithelial carcinoma cells: effects of extracellular nucleotides on apoptosis and cell proliferation. Am J Physiol Liver Physiol. 2005; 288: G1024–G1035. https://doi.org/10.1152/ajpgi.00211.2004 PMID: 15662049 4. Dal Ben D, Antonioli L, Lambertucci C, Spinaci A, Fornai M, D’Antongiovanni V, et al. Approaches for designing and discovering purinergic drugs for gastrointestinal diseases. Expert Opin Drug Discov. 2020; 15: 687–703. https://doi.org/10.1080/17460441.2020.1743673 PMID: 32228110 5. 6. Trautmann A. Extracellular ATP in the immune system: more than just a “danger signal.” Sci signal. 2009; 2: e6. Zimmermann H. Extracellular ATP and other nucleotides—ubiquitous triggers of intercellular messen- ger release. Purinergic Signal. 2016; 12: 25–57. https://doi.org/10.1007/s11302-015-9483-2 PMID: 26545760 7. Crane JK, Naeher TM, Choudhari SS, Giroux EM. Two pathways for ATP release from host cells in enteropathogenic Escherichia coli infection. Am J Physiol Liver Physiol. 2005; 289: G407–G417. https://doi.org/10.1152/ajpgi.00137.2005 PMID: 16093420 8. Ullrich N, Caplanusi A, Broˆ ne B, Hermans D, Larivière E, Nilius B, et al. Stimulation by caveolin-1 of the hypotonicity-induced release of taurine and ATP at basolateral, but not apical, membrane of Caco-2 cells. Am J Physiol Physiol. 2006; 290: C1287–C1296. 9. Wei Z-Y, Qu H-L, Dai Y-J, Wang Q, Ling Z-M, Su W-F, et al. Pannexin 1, a large-pore membrane chan- nel, contributes to hypotonicity-induced ATP release in Schwann cells. Neural Regen Res. 2021; 16: 899. https://doi.org/10.4103/1673-5374.290911 PMID: 33229726 10. Dosch M, Zindel J, Jebbawi F, Melin N, Sanchez-Taltavull D, Stroka D, et al. Connexin-43-dependent ATP release mediates macrophage activation during sepsis. Elife. 2019; 8: e42670. https://doi.org/10. 7554/eLife.42670 PMID: 30735126 11. To WKL, Kumar P, Marshall JM. Hypoxia is an effective stimulus for vesicular release of ATP from human umbilical vein endothelial cells. Placenta. 2015; 36: 759–766. 12. Burnstock G. Purinergic signalling in the gastrointestinal tract and related organs in health and disease. Purinergic Signal. 2014; 10: 3–50. https://doi.org/10.1007/s11302-013-9397-9 PMID: 24307520 13. Yegutkin GG. Enzymes involved in metabolism of extracellular nucleotides and nucleosides: functional implications and measurement of activities. Crit Rev Biochem Mol Biol. 2014; 49: 473–497. https://doi. org/10.3109/10409238.2014.953627 PMID: 25418535 Zimmermann H. Ectonucleoside triphosphate diphosphohydrolases and ecto-50-nucleotidase in puri- nergic signaling: how the field developed and where we are now. Purinergic Signal. 2021; 17: 117–125. https://doi.org/10.1007/s11302-020-09755-6 PMID: 33336318 14. 15. Bahrami F, Kukulski F, Lecka J, Tremblay A, Pelletier J, Rockenbach L, et al. Purine-metabolizing ectoenzymes control IL-8 production in human colon HT-29 cells. Mediators Inflamm. 2014; 2014. https://doi.org/10.1155/2014/879895 PMID: 25242873 16. Clayton A, Al-Taei S, Webber J, Mason MD, Tabi Z. Cancer exosomes express CD39 and CD73, which suppress T cells through adenosine production. J Immunol. 2011; 187: 676–683. https://doi.org/10. 4049/jimmunol.1003884 PMID: 21677139 17. Schachter J, Alvarez CL, Bazzi Z, Faillace MP, Corradi G, Hattab C, et al. Extracellular ATP hydrolysis in Caco-2 human intestinal cell line. Biochim Biophys Acta (BBA)-Biomembranes. 2021; 1863: 183679. https://doi.org/10.1016/j.bbamem.2021.183679 PMID: 34216588 18. Lu Y, Qi J, Wu W. Lipid nanoparticles: In vitro and in vivo approaches in drug delivery and targeting. Drug targeting and stimuli sensitive drug delivery systems. Elsevier; 2018. pp. 749–783. 19. Karlsson J, Ungell A-L, Gråsjo¨ J, Artursson P. Paracellular drug transport across intestinal epithelia: influence of charge and induced water flux. Eur J Pharm Sci. 1999; 9: 47–56. https://doi.org/10.1016/ s0928-0987(99)00041-x PMID: 10493996 20. Sinev MA, Sineva E V, Ittah V, Haas E. Domain closure in adenylate kinase. Biochemistry. 1996; 35: 6425–6437. https://doi.org/10.1021/bi952687j PMID: 8639589 21. Mayer RJ. Brain mitochondrial hexokinase and adenylate kinase: effect of osmotic conditions on enzyme activities. J Neurochem. 1972; 19: 2127–2138. https://doi.org/10.1111/j.1471-4159.1972. tb05122.x PMID: 5072389 22. Sheng XR, Li X, Pan XM. An iso-random Bi Bi mechanism for adenylate kinase. J Biol Chem. 1999; 274: 22238–22242. https://doi.org/10.1074/jbc.274.32.22238 PMID: 10428790 23. Harris DS, Slot JW, Geuze HJ, James DE. Polarized distribution of glucose transporter isoforms in Caco-2 cells. Proc Natl Acad Sci. 1992; 89: 7556–7560. https://doi.org/10.1073/pnas.89.16.7556 PMID: 1502167 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 23 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells 24. Maule´n NP, Henrı`quez EA, Kempe S, Ca´ rcamo JG, Schmid-Kotsas A, Bachem M, et al. Up-regulation and Polarized Expression of the Sodium-Ascorbic Acid Transporter SVCT1 in Post-confluent Differenti- ated CaCo-2 Cells *. J Biol Chem. 2003; 278: 9035–9041. https://doi.org/10.1074/jbc.M205119200 PMID: 12381735 25. Burnstock G, Kennedy C. P2X receptors in health and disease. Adv Pharmacol. 2011; 61: 333–372. https://doi.org/10.1016/B978-0-12-385526-8.00011-4 PMID: 21586364 26. Buzzi N, Boland R, de Boland AR. Signal transduction pathways associated with ATP-induced prolifera- tion of colon adenocarcinoma cells. Biochim Biophys Acta (BBA)-General Subj. 2010; 1800: 946–955. https://doi.org/10.1016/j.bbagen.2010.05.009 PMID: 20562007 27. Buzzi N, Bilbao PS, Boland R, de Boland AR. Extracellular ATP activates MAP kinase cascades through a P2Y purinergic receptor in the human intestinal Caco-2 cell line. Biochim Biophys Acta (BBA)-General Subj. 2009; 1790: 1651–1659. https://doi.org/10.1016/j.bbagen.2009.10.005 PMID: 19836435 28. Vuerich M, Robson SC, Longhi MS. Ectonucleotidases in intestinal and hepatic inflammation. Front Immunol. 2019; 10: 507. https://doi.org/10.3389/fimmu.2019.00507 PMID: 30941139 29. Matsumoto H, Erickson RH, Gum JR, Yoshioka M, Gum E, Kim YS. Biosynthesis of alkaline phospha- tase during differentiation of the human colon cancer cell line Caco-2. Gastroenterology. 1990; 98: 1199–1207. https://doi.org/10.1016/0016-5085(90)90334-w PMID: 2323513 30. Woods LT, Forti KM, Shanbhag VC, Camden JM, Weisman GA. P2Y receptors for extracellular nucleo- tides: Contributions to cancer progression and therapeutic implications. Biochem Pharmacol. 2021; 187: 114406. https://doi.org/10.1016/j.bcp.2021.114406 PMID: 33412103 31. Yegutkin GG. Nucleotide-and nucleoside-converting ectoenzymes: important modulators of purinergic signalling cascade. Biochim Biophys Acta (BBA)-Molecular Cell Res. 2008; 1783: 673–694. https://doi. org/10.1016/j.bbamcr.2008.01.024 PMID: 18302942 32. Kolachala VL, Bajaj R, Chalasani M, Sitaraman S V. Purinergic receptors in gastrointestinal inflamma- tion. Am J Physiol Liver Physiol. 2008; 294: G401–G410. https://doi.org/10.1152/ajpgi.00454.2007 PMID: 18063703 33. Lelièvre V, Muller J-M, Falcòn J. Adenosine modulates cell proliferation in human colonic carcinoma. II. Differential behavior of HT29, DLD-1, Caco-2 and SW403 cell lines. Eur J Pharmacol. 1998; 341: 299– 308. https://doi.org/10.1016/s0014-2999(97)01463-5 PMID: 9543252 34. Christofi FL. Purinergic receptors and gastrointestinal secretomotor function. Purinergic Signal. 2008; 4: 213–236. https://doi.org/10.1007/s11302-008-9104-4 PMID: 18604596 35. Grasl M, Turnheim K. Stimulation of electrolyte secretion in rabbit colon by adenosine. J Physiol. 1984; 346: 93–110. https://doi.org/10.1113/jphysiol.1984.sp015009 PMID: 6699790 36. Leipziger J, Kerstan D, Nitschke R, Greger R. ATP increases [Ca2+] i and ion secretion via a basolateral P2Y-receptor in rat distal colonic mucosa. Pflu¨gers Arch. 1997; 434: 77–83. https://doi.org/10.1007/ pl00008079 PMID: 9094258 37. Korman LY, Lemp GF, Jackson MJ, Gardner JD. Mechanism of action of ATP on intestinal epithelial cells: cyclic AMP-mediated stimulation of active ion transport. Biochim Biophys Acta (BBA)-Molecular Cell Res. 1982; 721: 47–54. 38. Yamamoto T, Suzuki Y. Role of luminal ATP in regulating electrogenic Na+ absorption in guinea pig dis- tal colon. Am J Physiol Liver Physiol. 2002; 283: G300–G308. https://doi.org/10.1152/ajpgi.00541.2001 PMID: 12121876 39. Kinoshita N, Takahashi T, Tada S, Shinozuka K, Mizuno N, Takahashi K. Activation of P2Y receptor enhances high-molecular compound absorption from rat ileum. J Pharm Pharmacol. 2006; 58: 195– 200. https://doi.org/10.1211/jpp.58.2.0006 PMID: 16451747 40. Lazarowski ER, Homolya L, Boucher RC, Harden TK. Identification of an ecto-nucleoside diphosphoki- nase and its contribution to interconversion of P2 receptor agonists. J Biol Chem. 1997; 272: 20402– 20407. https://doi.org/10.1074/jbc.272.33.20402 PMID: 9252347 41. Yegutkin GG, Henttinen T, Samburski SS, Spychala J, Jalkanen S. The evidence for two opposite, ATP-generating and ATP-consuming, extracellular pathways on endothelial and lymphoid cells. Bio- chem J. 2002; 367: 121–128. https://doi.org/10.1042/BJ20020439 PMID: 12099890 42. Yegutkin GG, Henttinen T, Jalkanen S. Extracellular ATP formation on vascular endothelial cells is mediated by ecto-nucleotide kinase activities via phosphotransfer reactions. FASEB J. 2001; 15: 251– 260. https://doi.org/10.1096/fj.00-0268com PMID: 11149913 43. Burrell HE, Wlodarski B, Foster BJ, Buckley KA, Sharpe GR, Quayle JM, et al. Human keratino- cytes release ATP and utilize three mechanisms for nucleotide interconversion at the cell surface. J Biol Chem. 2005; 280: 29667–29676. https://doi.org/10.1074/jbc.M505381200 PMID: 15958381 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 24 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells 44. Fabre ACS, Vantourout P, Champagne E, Terce´ F, Rolland C, Perret B, et al. Cell surface adenylate kinase activity regulates the F1-ATPase/P2Y13-mediated HDL endocytosis pathway on human hepato- cytes. Cell Mol Life Sci C. 2006; 63: 2829–2837. 45. Kennedy C. The P2Y/P2X divide: How it began. Biochem Pharmacol. 2021; 187: 114408. https://doi. org/10.1016/j.bcp.2021.114408 PMID: 33444568 46. Giuliani AL, Sarti AC, Di Virgilio F. Extracellular nucleotides and nucleosides as signalling molecules. Immunol Lett. 2019; 205: 16–24. https://doi.org/10.1016/j.imlet.2018.11.006 PMID: 30439478 47. Pieri M, Christian HC, Wilkins RJ, Boyd CAR, Meredith D. The apical (hPepT1) and basolateral peptide transport systems of Caco-2 cells are regulated by AMP-activated protein kinase. Am J Physiol Liver Physiol. 2010; 299: G136–G143. https://doi.org/10.1152/ajpgi.00014.2010 PMID: 20430871 48. Narumi K, Ohata T, Horiuchi Y, Satoh H, Furugen A, Kobayashi M, et al. Mutual role of ecto-5’-nucleo- tidase/CD73 and concentrative nucleoside transporter 3 in the intestinal uptake of dAMP. PLoS One. 2019; 14: e0223892. https://doi.org/10.1371/journal.pone.0223892 PMID: 31634358 49. Salem M, Lecka J, Pelletier J, Marconato DG, Dumas A, Vallières L, et al. NTPDase8 protects mice from intestinal inflammation by limiting P2Y6 receptor activation: identification of a new pathway of inflammation for the potential treatment of IBD. Gut. 2022; 71: 43–54. https://doi.org/10.1136/gutjnl- 2020-320937 PMID: 33452178 50. Zimmermann H, Zebisch M, Stra¨ter N. Cellular function and molecular structure of ecto-nucleotidases. Purinergic Signal. 2012; 8: 437–502. https://doi.org/10.1007/s11302-012-9309-4 PMID: 22555564 51. Alvarez CL, Corradi G, Lauri N, Marginedas-Freixa I, Leal Denis MF, Enrique N, et al. Dynamic regula- tion of extracellular ATP in Escherichia coli. Biochem J. 2017; 474: 1395–1416. https://doi.org/10.1042/ BCJ20160879 PMID: 28246335 52. Inami A, Kiyono H, Kurashima Y. ATP as a pathophysiologic mediator of bacteria-host crosstalk in the gastrointestinal tract. Int J Mol Sci. 2018; 19: 2371. https://doi.org/10.3390/ijms19082371 PMID: 30103545 53. Puhar A, Tronchère H, Payrastre B, Van Nhieu GT, Sansonetti PJ. A Shigella effector dampens inflam- mation by regulating epithelial release of danger signal ATP through production of the lipid mediator PtdIns5P. Immunity. 2013; 39: 1121–1131. https://doi.org/10.1016/j.immuni.2013.11.013 PMID: 24332032 54. Strehler BL. Bioluminescence assay: principles and practice. Methods Biochem Anal. 1968; 16: 99– 181. https://doi.org/10.1002/9780470110348.ch2 PMID: 4385967 55. Pafundo DE, Chara O, Faillace MP, Krumschnabel G, Schwarzbaum PJ. Kinetics of ATP release and cell volume regulation of hyposmotically challenged goldfish hepatocytes. Am J Physiol Integr Comp Physiol. 2008; 294: R220–R233. https://doi.org/10.1152/ajpregu.00522.2007 PMID: 17928510 56. Gorman MW, Marble DR, Ogimoto K, Feigl EO. Measurement of adenine nucleotides in plasma. Lumin J Biol Chem Lumin. 2003; 18: 173–181. https://doi.org/10.1002/bio.721 PMID: 12701093 57. Bradford MM. A rapid and sensitive method for the quantitation of microgram quantities of protein utiliz- ing the principle of protein-dye binding. Anal Biochem. 1976; 72: 248–254. https://doi.org/10.1006/abio. 1976.9999 PMID: 942051 58. Lanzetta PA, Alvarez LJ, Reinach PS, Candia OA. An improved assay for nanomole amounts of inor- ganic phosphate. Anal Biochem. 1979; 100: 95–97. https://doi.org/10.1016/0003-2697(79)90115-5 PMID: 161695 59. Blais A, Bissonnette P, Berteloot A. Common characteristics for Na+-dependent sugar transport in Caco-2 cells and human fetal colon. J Membr Biol. 1987; 99: 113–125. https://doi.org/10.1007/ BF01871231 PMID: 3123697 60. Mohammadi S, Nikkhah M, Nazari M, Hosseinkhani S. Design of a coupled bioluminescent assay for a recombinant pyruvate kinase from a thermophilic Geobacillus. Photochem Photobiol. 2011; 87: 1338– 1345. https://doi.org/10.1111/j.1751-1097.2011.00973.x PMID: 21790618 61. Hoops S, Sahle S, Gauges R, Lee C, Pahle J, Simus N, et al. COPASI—a COmplex PAthway SImula- tor. Bioinformatics. 2006; 22: 3067–3074. https://doi.org/10.1093/bioinformatics/btl485 PMID: 17032683 62. Navarro JM, Olmo N, Turnay J, Lo´ pez-Conejo MT, Lizarbe MA. Ecto-50-nucleotidase from a human colon adenocarcinoma cell line. Correlation between enzyme activity and levels in intact cells. Mol Cell Biochem. 1998; 187: 121–131. https://doi.org/10.1023/a:1006808232059 PMID: 9788749 63. Garces E, Cleland WW. Kinetic study of yeast nucleosidediphosphate kinase. Biochemistry. 1969; 8: 633–640. 64. Lascu I, Gonin P. The catalytic mechanism of nucleoside diphosphate kinases. J Bioenerg Biomembr. 2000; 32: 237–246. https://doi.org/10.1023/a:1005532912212 PMID: 11768307 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 25 / 26 PLOS COMPUTATIONAL BIOLOGYeATP recycling in human epithelial intestinal cells 65. Espelt M V, de Tezanos Pinto F, Alvarez CL, Alberti GS, Incicco J, Denis MFL, et al. On the role of ATP release, ectoATPase activity, and extracellular ADP in the regulatory volume decrease of Huh-7 human hepatoma cells. Am J Physiol Physiol. 2013; 304: C1013–C1026. https://doi.org/10.1152/ajpcell.00254. 2012 PMID: 23485713 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011196 June 29, 2023 26 / 26 PLOS COMPUTATIONAL BIOLOGY
10.1371_journal.pcbi.1011149	RESEARCH ARTICLE Disproportionate impacts of COVID-19 in a large US city Spencer J. FoxID Briana Betke4, Jose´ L. Herrera-Diestra4, Spencer Woody4, Kelly Pierce7, Kaitlyn E. Johnson8, Maureen Johnson-Leo´ n4, Michael Lachmann9, Lauren Ancel Meyers4,9 5, Graham C. Gibson6, 1,2,3, Emily JavanID 4, Remy PascoID 1 Department of Epidemiology & Biostatistics, University of Georgia, Athens, Georgia, United States of America, 2 Institute of Bioinformatics, University of Georgia, Athens, Georgia, United States of America, 3 Center for the Ecology of Infectious Diseases, University of Georgia, Athens, Georgia, United States of America, 4 Department of Integrative Biology, The University of Texas at Austin, Austin, Texas, United States of America, 5 Department of Industrial Engineering, The University of Texas at Austin, Austin, Texas, United States of America, 6 Los Alamos National Laboratory, Los Alamos, New Mexico, United States of America, 7 The Texas Advanced Computing Center, The University of Texas at Austin, Austin, Texas, United States of America, 8 The Rockefeller Foundation, New York, New York, United States of America, 9 The Santa Fe Institute, Santa Fe, New Mexico, United States of America sjfox@uga.edu Abstract COVID-19 has disproportionately impacted individuals depending on where they live and work, and based on their race, ethnicity, and socioeconomic status. Studies have docu- mented catastrophic disparities at critical points throughout the pandemic, but have not yet systematically tracked their severity through time. Using anonymized hospitalization data from March 11, 2020 to June 1, 2021 and fine-grain infection hospitalization rates, we esti- mate the time-varying burden of COVID-19 by age group and ZIP code in Austin, Texas. During this 15-month period, we estimate an overall 23.7% (95% CrI: 22.5–24.8%) infection rate and 29.4% (95% CrI: 28.0–31.0%) case reporting rate. Individuals over 65 were less likely to be infected than younger age groups (11.2% [95% CrI: 10.3–12.0%] vs 25.1% [95% CrI: 23.7–26.4%]), but more likely to be hospitalized (1,965 per 100,000 vs 376 per 100,000) and have their infections reported (53% [95% CrI: 49–57%] vs 28% [95% CrI: 27–30%]). We used a mixed effect poisson regression model to estimate disparities in infection and report- ing rates as a function of social vulnerability. We compared ZIP codes ranking in the 75th percentile of vulnerability to those in the 25th percentile, and found that the more vulnerable communities had 2.5 (95% CrI: 2.0–3.0) times the infection rate and only 70% (95% CrI: 60%-82%) the reporting rate compared to the less vulnerable communities. Inequality per- sisted but declined significantly over the 15-month study period. Our results suggest that fur- ther public health efforts are needed to mitigate local COVID-19 disparities and that the CDC’s social vulnerability index may serve as a reliable predictor of risk on a local scale when surveillance data are limited. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Fox SJ, Javan E, Pasco R, Gibson GC, Betke B, Herrera-Diestra JL, et al. (2023) Disproportionate impacts of COVID-19 in a large US city. PLoS Comput Biol 19(6): e1011149. https://doi.org/10.1371/journal.pcbi.1011149 Editor: Claudio Jose´ Struchiner, Fundac¸ão Getu´lio Vargas: Fundacao Getulio Vargas, BRAZIL Received: November 9, 2022 Accepted: May 2, 2023 Published: June 1, 2023 Copyright: This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication. Data Availability Statement: Code for replicating the analysis with synthetic data are available at https://github.com/sjfox/austin-disparities. Raw hospital admission and COVID-19 testing data are private due to their sensitive nature and will be made available upon reasonable request to Austin Public Health here; https://www.austintexas.gov/ email/health. Funding: RP, GCG, SJF, and LAM were supported by Grant U01IP001136 awarded to LAM from the CDC. EJ, SJF, BB, JLHD, and LAM were supported by Grant R01 AI151176 awarded to LAM from the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 1 / 22 PLOS COMPUTATIONAL BIOLOGYNIH. SW, KP, KEJ, MJL, ML, and LAM were supported by a generous donation from Love, Tito’s (the philanthropic arm of Tito’s Homemade Vodka, Austin, TX) all awarded to LAM. The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication. Competing interests: The authors have declared that no competing interests exist. Disproportionate impacts of COVID-19 in a large US city Author summary COVID-19 disproportionately impacted communities based on their socioeconomic and racial composition. Studies have documented catastrophic disparities at multiple geo- graphic scales, but have not yet tracked how they evolved over time. Here, we use fine- grain epidemiological data to estimate the time-varying disparate burden of COVID-19 by age group and ZIP code in Austin, Texas. During this 15-month period, we estimate that 23.7% (95% CrI: 22.5–24.8%) of the population was infected and 29.4% (95% CrI: 28.0–31.0%) of those infections were reported. Infections were not spread evenly across the region. Individuals over 65 were significantly less likely to be infected than younger age groups (11.2% [95% CrI: 10.3–12.0%] vs 25.1% [95% CrI: 23.7–26.4%]), suggesting that efforts to protect those populations may have been effective. We found that the most vulnerable ZIP codes in the region faced 2.5 times the infection risks compared with the least vulnerable ZIP codes, and that infections were only 70% as likely to be reported in vulnerable communities. Inequality persisted but declined significantly over the 15- month study period. Further public health efforts are needed to address local COVID-19 disparities. Introduction The WHO estimates that the COVID-19 pandemic caused nearly 15 million excess deaths worldwide between its emergence in 2019 and the end of 2021. The burden fell disproportion- ately on countries in South-East Asia, Europe, and the Americas, with 68% of the estimated excess deaths occurring in 10 countries containing 35% of the global population [1]. In the United States, pandemic burden was initially concentrated around New York City, New York, but spread geographically after the White House issued the Opening Up America Again guide- lines in spring of 2020 [2]. The pandemic disproportionately harmed essential workers and racial and ethnic minority groups [3–6] as well as US counties [7–10] and cities [11–13] with high social vulnerability indices [14]. In response to these glaring disparities, scientists and public health leaders advocated for programs to support marginalized communities, including accessible testing facilities, com- munity support programs to mitigate the socioeconomic, educational and healthcare harms resulting from lockdowns, proactive vaccination and antiviral campaigns, and effective public health communications [15–21]. Many US vaccination campaigns successfully prioritized vul- nerable regions [22–29], though others, such as in Texas, limited geographic prioritization efforts [30,31]. To prevent, detect, and reduce disparities in infectious disease burden, we need to increase the geographic and temporal resolution of our surveillance efforts, while reducing biases. Pub- lished estimates of COVID-19 burden in underserved populations are often derived directly from reported case or death counts, without correcting for ascertainment biases or disentangl- ing risks of infection from risks of severe outcomes [7–9,32–42]. When available, both serolog- ical [43,44] and hospitalization data [45] can be used to estimate reporting rates. Several studies have highlighted the disproportionate burden of COVID-19 infections within cities [46,47], but only at single time points during the pandemic. Here, we estimate the changing burden of COVID-19 at a local scale within a large US city throughout the first 15 months of the pandemic. Using ZIP-code and age-stratified hospitali- zation data, we track daily disparities in infection rates, hospitalization rates, and case report- ing rates. As the SARS-CoV-2 virus continues to evolve along with our arsenal of medical and PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 2 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city behavioral interventions, this method can help to ensure the reliability and equity of local risk assessments [48]. Results We analyzed spatial COVID-19 burdens in Austin, Texas using hospital admission data from March 11, 2020 to June 1, 2021. This period preceded the emergence of the Delta variant and included a small wave in April 2020, followed by larger waves in the summer and winter (Figs 1A and S1). As of June 1, 2021, there were 83,722 reported cases, 6,474 hospitalized patients, and 1,024 deaths of COVID-19 in Travis County, which has 1.3 million residents, covering 57% of the Austin metropolitan area population. We estimate that 23.7% (95% CrI: 22.5– 24.8%) of the population were infected in this time period and 29.4% (95% CrI: 28.0–31.0%) of all infections were reported. Statewide seroprevalence data suggest that Texans were 1.3 times as likely to be infected in this time period, with an estimated attack rate of 32% (95% CrI: 28– 36%) (Fig 1B) [49]. However, the estimated infection risks prior to September 23, 2020 are Fig 1. COVID-19 hospital admissions and estimated cumulative infections for Travis County (Austin, TX) from March 1, 2020 to June 1, 2021. (A) Daily reported COVID-19 hospital admissions per 1 million residents [50]. (B) Estimated cumulative infections with 95% credible intervals (black line and gray ribbon) compared to Texas statewide seroprevalence-based estimates (red points and error bars) [49]. https://doi.org/10.1371/journal.pcbi.1011149.g001 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 3 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city Fig 2. Estimated age-stratified COVID-19 burden in Travis country through June 1, 2021. (A) Reported COVID-19 hospital admissions by age group. (B) Reported COVID-19 cases by age group. (C) Estimated percent infected by age group. (D) Estimated COVID-19 case reporting rates by age group up to June 1, 2021. In (A)-(D), horizontal dashed lines indicate county-wide average rates. (E) Estimated daily infection rates (line) and 95% credible intervals (ribbons) by age group. (F) Distribution of infections across age groups for each period of the epidemic. The spring period refers to the two-month time period before the first major wave from March 1, 2020 to May 1, 2020, the summer period refers to the two-month period containing the first major wave from June 1, 2020 to August 1, 2020, and the winter period refers to the two-month period containing the second major wave from December 1, 2020 until February 1, 2021. Bars indicate the fraction of all infections during the time period in each age group, with the error bars indicating the 95% credible intervals. The horizontal colored lines in panel F indicate the proportion of the Travis county population in the specified age group. https://doi.org/10.1371/journal.pcbi.1011149.g002 higher for Travis county (10.6% [95% CrI: 10.0–11.1%] infected) than statewide (7.2% [95% CrI: 5.2–9.6%] infected). In Travis County, children aged 0–17 experienced the lowest hospitalization risk, with a cumulative count of 55.8 hospital admissions per 100,000, and adults over age 65 experienced the highest hospitalization risk of 1,965 per 100,000 (Fig 2A). In contrast, reported cases were relatively similar across age groups, ranging from 4,206 per 100,000 in children to 8,475 per 100,000 in young adults (Fig 2B). Using age-specific seroprevalence and hospital admissions data for the state of Texas, we estimate that one in 434 (95% CI: 243–625) infections in individ- uals aged 0–17 years and one in 4.7 (95% CI: 3.0–6.8) infections in individuals over age 65 led to hospitalization (Table 1). This is consistent with published estimates for the infection hospi- talization rate from China [51] and France [45] (S2 Fig). By June 1, 2021, we estimate that 28.5% (95% CrI: 26.6–30.5%) of 18–49 year olds were infected, while only 11.2% (95% CrI: 10.3–12.0%) of individuals over age 65 were infected Table 1. SARS-CoV-2 infection hospitalization rate (IHR) across Texas estimated from statewide seroprevalence and hospitalization data from July 29, 2020 through May 27, 2021. Age group 0–17 18–49 50–64 65+ Number of COVID-19 hospital admissions (95% CI) [52] Estimated infections based on seroprevalence data (95% CI) [53,54] Estimated IHR (95% CI) 4,870 (4,727–4,975) 53,913 (53,106–54,597) 58,466 (57,114–59,473) 100,881 (98,032–102,931) 2,081,849 (1,195,168–2,972,145) 3,795,068 (2,827,349–4,773,427) 1,084,729 (718,420–1,481,055) 0.25% (0.16%-0.41%) 1.45% (1.13%-1.90%) 5.57% (3.96%-8.16%) 499,962 (303,726–692,758) 21.12% (14.54%-33.04%) https://doi.org/10.1371/journal.pcbi.1011149.t001 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 4 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city (Fig 2C). The estimated percent of cases reported increases with age, ranging from 21.3% (95% CrI: 18.2–24.8%) in 0–17 year olds to 52.6% (95% CrI: 48.7–56.9%) in over 65 year olds (Fig 2D). All age groups experienced two large waves during the study period, though the summer 2020 was relatively mild for children (Fig 2E). Relative infection rates across age groups evened out over time (Fig 2F). For example, children, who account for 22% of the Travis county popu- lation, constituted 4.8% (95% CrI: 3.2–6.9%) of all infections between March 1, 2020 and May 1, 2020 and 19.6% (95% CrI: 16.5–22.9%) of all infections between December 1, 2020 and Feb- ruary 1, 2021. The proportion of infections occurring in 18–49 year olds, who make up 52.2% of the population, dropped from 73.9% (95% CrI: 71.3–76.5%) during the spring 2020 period to 56.1% (95% CrI: 53.3–59.0%) during the winter 2020–2021 wave (Figs 2F and S3). Reported case and hospitalization counts do not clearly exhibit this reversal in age-specific risks (S4 and S5 Figs). Infection rates for each age group were estimated to be lower in Travis county than statewide, by factors of 44% (95% CrI: 20–60%), 19.5% (95% CrI: 0.2–34%), 23.1% (0.2–40%), and 28.6% (95% CrI: 0.03–47%) in the 0–17, 18–49, 50–64, and 65+ year groups, respectively (S6 and S7 Fig). Estimated COVID-19 burden varies significantly across ZIP codes within Travis County, with Interstate 35 roughly partitioning the county into high risk ZIP codes in the East and low risk ZIP codes in the West (Fig 3A and 3B). High COVID-19 risk visibly aligns with high social vulnerability, as measured by ZIP-code level SVI (Fig 3C). Our estimates for ZIP-code level infection hospitalization rates exhibit the opposite geographic trend (Fig 3D) from the absolute hospitalization rates (S6 Fig). We estimate that a ZIP code in east Austin (78724) had the high- est infection rate of 53.7% (95% CrI: 42.7–67.1), while a Southwest Austin ZIP code (78739) had the lowest estimated infection rate of 4.8% (95% CrI: 2.6–8.5%) (Fig 3E) up to June 1, 2021. Downtown Austin (78701) had the lowest reporting rate of any ZIP code, with an esti- mated 15.2% (95% CrI: 11–20%) of infections reported, while a West Austin ZIP code (78732) had the highest reporting rate of 67% (95% CrI: 38–99%) (Fig 3F). Similar geographic patterns exist for each of the four age groups (S8–S10 Figs). Similar to previous work, we estimated the relationship between infections and reporting rate across ZIP codes using the function r= a�I−b and found a significant inverse relationship with a = 70.3 (95% CrI: 28–141) and b = 0.53 (95% CrI: 0.45–0.61) (S11 Fig) [55]. The cumulative infection rates, case rates, and hospitalization rates are positively correlated with social vulnerability across Travis County’s 46 ZIP codes (Figs 4A and S12). Of the 15 indi- vidual components of SVI, we find that minority population rates, educational attainment rates, and household makeup are the strongest predictors of both infection rates (S13 Fig) and reporting rates (S14 Fig). We compare the relative risks for individuals living in a ZIP code at Travis County’s 25th (SVI = 0.12) and 75th (SVI = 0.5) percentile by SVI, where higher SVI indicates higher social vulnerability. Controlling for random ZIP code-level effects, we esti- mate that ZIP codes in the 75th SVI percentile experienced 2.5 (95% CrI: 2.0–3.0) times the infection rate of those in the 25th percentile. Similar trends are observed for each age group (S15 Fig and S1 Table). COVID-19 burden is often estimated directly from reported case or hospitalization data, without correcting for geographic biases in testing and underlying risk factors. For Travis county, we find that the subset of case data from APH provides a reasonable approximation but hospitalization data tends to inflate the estimated disparities (S1 Table). We aggregate the estimated number of infections occurring in each ZIP code into four-week periods from March 1, 2020 to June 1, 2021, and measure the relationship between SVI and the relative infection risk during this period. Significant disparity (i.e., a relative risk greater than one) persisted throughout the period and was highest during the first three months of the pandemic (Fig 4B). In April 2020, individuals living in the 75th SVI percentile ZIP code had an expected 9.6 (95% CrI: 5.4–17.0) times greater infection risk than those living in the 25th PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 5 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city Fig 3. Reported and estimated COVID-19 burden by ZIP code for Travis County between March 1, 2020 and June 1, 2021. (A) Reported COVID-19 cases per 100,000. (B) Reported COVID-19 hospitalizations per 100,000. (C) Social Vulnerability Index [14] (D) Estimated infection hospitalization rate (IHR). (E) Estimated cumulative infections as of June 1, 2021. (F) Estimated percent of COVID-19 infections that were reported. Thin black curves indicate Interstate 35 and highway US 183. The ZIP code map was based on TIGER/Line shapefiles provided by the US Census Bureau [56] accessed through the tidycensus R package for the year 2019 [57]. https://doi.org/10.1371/journal.pcbi.1011149.g003 percentile SVI ZIP code. This ratio declined to 2.5 (95% CrI: 1.5–4.4) in August 2020 and hit a temporary minimum of 1.7 (95% CrI: 1.2–2.6) in November of 2020 before the large winter surge. COVID-19 case reporting rates are negatively correlated with social vulnerability. We esti- mate that infections occurring in the 75th SVI percentile ZIP code were only 70% (95% CrI: 60%-82%) as likely to have been reported than those occurring in the 25th SVI percentile ZIP code. We further stratified by age group using a small sample of age-specific case data reported by the Austin Public Health community testing programs, which targeted vulnerable PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 6 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city Fig 4. Infection and reporting rates correlate with social vulnerability in Travis County from March 1, 2020 to June 1, 2021. (A) Across the 46 ZIP codes, SVI is a significant predictor of estimated cumulative infections (p<0.001). The blue line and ribbon indicate the mean and 95% prediction interval from the fitted Poisson mixed-effects model. (B) Using the fitted model, we compare the expected infection rates among more and less vulnerable ZIP codes (specifically, ZIP codes at the 75th and 25th percentiles in the SVI distribution, respectively). The points indicate the expected ratio between these two values calculated using the estimated SVI regression coefficient from the 4-week time period; error bars indicate 95% CI’s. (C) Across the 46 ZIP codes, SVI is a significant predictor of estimated case reporting rates (p<0.001). The blue line and ribbon indicate the mean and 95% prediction interval from the fitted Poisson mixed- effects model. (D) Four week estimate for the inequality relationship between SVI and infection reporting rates across the 46 ZIP codes. Points and error bars show the mean and 95% CI for the relative reporting rate of individuals living in ZIP codes in the 75th SVI percentile compared with those living in the 25th SVI percentile. The red, horizontal dashed lines in B and D indicate if there were equitable infection risks or reporting rates across the 75th and 25th SVI percentile ZIP codes in the four week period. We overlay hospital admission time-series in B and D to showcase how inequality estimates compare with the progression of the local epidemic. For B and D we removed the ZIP codes reporting zero infections to stabilize the regression estimates. https://doi.org/10.1371/journal.pcbi.1011149.g004 populations in East Austin (S16 Fig) [58]. We found that the negative correlation between SVI and case reporting rates held for all age groups except those over 65 years, perhaps because of Austin’s efforts to improve testing access for high risk individuals (S17 Fig and S1 Table). Throughout the study period, the estimated ratio in reporting rates between the 75th and 25th SVI percentile ZIP codes fluctuated, often dropping to levels significantly less than one (Fig 4D). Discussion In the US, the first wave of the COVID-19 pandemic disproportionately harmed essential workers [5,6], residents of long-term care facilities [59], racial and ethnic minority populations [60], and socially vulnerable neighborhoods within cities [33,46,61,62]. Public health agencies and government officials have tried to address these disparities through targeted testing, vacci- nation, distribution of personal protective equipment, information campaigns, and paid sick leave [15–19]. Using a new method for inferring infection risks and reporting rates from COVID-19 hospital admissions data, we demonstrate that disparities persist on a granular scale within a large US city throughout the first year of the pandemic. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 7 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city Our estimates for the spatial burden of COVID-19 in Austin, Texas suggest that children were less likely to be infected than adults under age 65 during the first major wave of transmis- sion in the summer of 2020 but not during the subsequent winter wave. This is consistent with prior estimates [63–66] and may be attributable to early school closures, strict compliance with social distancing measures [67], or the emergence of variants that more efficiently infect children [68,69]. We also find that individuals over age 65 generally had the lowest risks of infection, despite suffering the highest per capita hospitalization rate, which may stem from heightened precautionary behavior and other protective measures such as COVID-19 screen- ing in long-term care facilities [70]. Reporting rates were lowest in children and highest in older adults (Fig 2), which may stem from the positive correlation between age and symptom severity [63]. Our results suggest that, by June of 2021, the cumulative risk of infection in Travis county was about 23% lower than the average risk across the state of Texas (Figs 1 and S6 and S7). The divergence is consistent with a potential shift of COVID-19 burden from urban to rural regions in the United States [71], and may have stemmed from stricter COVID-19 mitigation policies or higher levels of public adherence in Travis County compared to the rest of Texas. In October 2020, the Texas governor issued Executive Order GA 32, which standardized COVID-19 policies across the state and limited local authority to enact restrictions [72]. In late 2020, Travis County enacted strict policies in violation of the statewide order which effectively mitigated a large winter surge [73,74]. In contrast, El Paso, Texas experienced a catastrophic surge in the fall of 2020 [75] and several other major Texas cities reported higher COVID infection, hospitalization and death rates than Austin during the winter of 2020–2021. While prior studies have shown positive COVID-19 epidemiological outcomes are associated with more stringent policies [76–78], the interaction between local and state policies and their impact on COVID-19 burden is not yet understood. Historically marginalized populations in the “Eastern Crescent” of Austin were dispropor- tionately harmed throughout the first year of the pandemic [79,80], mirroring disparities reported for Santiago, Chile and New York City [46,47,61]. After controlling for the higher prevalence of underlying risk factors in more vulnerable communities, we find that the ZIP codes ranking in the 75th percentile of social vulnerability had a more than twofold higher infection rate and a roughly 70% the case reporting rate than those ranking in the 25th percen- tile. Our estimates for inequity differ depending on the metric of analysis. For example, we estimate larger disparities between ZIP codes using raw hospitalization rates rather than infec- tion rates, highlighting the importance of monitoring pandemic impacts across multiple indi- cators. In our analysis, higher hospitalization rate inequity may stem from the overlapping risks for infection and severe disease outcomes in vulnerable populations in Austin [60]. The estimated ratio in infection risk between more and less vulnerable regions decreased significantly during the first four months of the pandemic, perhaps because of local efforts to increase access to SARS-CoV-2 testing, isolation facilities, critical health information, and eventually vaccines [81]. The apparent decrease in disparity may also stem from higher infec- tion rates in vulnerable populations leading to a more rapid buildup of immunity or relatively higher infection rates in less vulnerable areas during later time periods [82,83]. Exploring these hypotheses will be important to preventing disparities and protecting vulnerable popula- tions during future infectious disease outbreaks. As of June of 2021, however, there remained a significant gap in COVID-19 risks and burden which informed targeted efforts by Austin Pub- lic Health to increase access to tests, vaccines, information and COVID-19 healthcare. We developed this estimation method because of data availability. At the time, we were not able to obtain sub-city level seroprevalence data or reliable case counts, but had access to line- list hospitalization data indicating patient age and residential ZIP code. Given similar data, PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 8 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city this approach can be applied to track longitudinal SARS-CoV-2 risks in cities and other geo- graphic scales worldwide [84], as well as to estimate infection rates for other pathogens with high proportions of subclinical infections, such as influenza or Zika Virus (ZIKV) [85]. We note several assumptions of our analysis. First, the hospital admission data are limited by the accuracy of patient ZIP codes. Fewer than 1% of patients had unknown addresses. How- ever, the missing data may correspond to vulnerable subgroups, such as people experiencing homelessness or undocumented residents, and thus obscure critical geographic or socioeco- nomic hotspots in our analysis. Second, we estimated each age- and ZIP-specific group inde- pendently rather than combining information across groups. This increases the uncertainty of our estimates but avoids the challenge of incorporating the changing contact and mobility pat- terns within the city throughout the pandemic [86–88]. Third, although we conducted analyses at a higher spatial resolution than most prior studies of COVID-19 burden, disparities in risk often occur at even more local scale [89]. Achieving COVID-19 health equity will require more granular surveillance and risk mitigation approaches. Finally, we made the simplifying assumptions that the IHR was constant and that individu- als could only be infected once. If reinfections were common during the analysis period, our model would underestimate the infection count. As such, we limited our analysis to the time period before the emergence of the Delta variant, after which reinfections and vaccine break- through infections were common [90–92]. Estimating local infection risks beyond June 2021 will require additional data and methods for accounting for reinfections and inferring the extent to which prior infection, prior vaccinations, and new variants modify the severity of infection. We estimate that less than 25% of the Austin, Texas population was infected by SARS-CoV- 2 prior to June 1, 2021 and that vulnerable communities in East Austin bore the brunt of the first two large waves of transmission. Our study introduces a framework for tracking infection and reporting rates on a granular scale using hospitalization data and provides evidence that the CDC’s social vulnerability index (SVI) is a strong predictor of risk that can inform targeted interventions. Materials and methods We estimate the daily age-stratified numbers of infections for each of the 46 ZIP codes in Tra- vis County, Texas from hospital linelist data provided by the three major local healthcare sys- tems to Austin Public Health [93]. As described below, we first estimate age-specific infection hospitalization rates (IHRs) from state-wide COVID-19 hospitalization data and SARS-CoV-2 seroprevalence data and then use the IHR estimates to infer the number and timing of infec- tions by age group and ZIP code. All code to recreate the analyses and figures can be accessed in the associated github repository (https://github.com/sjfox/austin-disparities). Estimating Texas statewide infection hospitalization rates (IHRs) The infection hospitalization rate is defined as the proportion of infected individuals that are hospitalized. We used age-stratified COVID-19 hospitalization data [52] and SARS-CoV-2 antibody seroprevalence data [54] to estimate the age-specific infection hospitalization rate in Texas. For each age group, we estimate the number of infections occurring between t0 (July 29, 2020) and t1 (May 27, 2021) according to a normal distribution as: NIk;state � normalðmean ¼ Ik;stateðt1Þ (cid:0) Ik;stateðt0Þ; var ¼ s2 k;stateðt0Þ þ s2 k;stateðt1ÞÞ PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 9 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city Where NIk,state corresponds to the estimated number of infections in the state in age group k, Ik,state corresponds to the mean CDC infection estimate at time, t, and s2 k;state tð Þ ¼ � maxfIk;state;97:5%ðtÞ (cid:0) Ik;stateðtÞ; Ik;stateðtÞ (cid:0) Ik;state;2:5%ðtÞg �2 1:96 Where Ik,state.2.5%(t) and Ik,state.97.5%(t) correspond to the lower and upper CDC infection confidence interval estimates respectively [53]. Texas seroprevalence samples were tested for SARS-CoV-2 anti-nucleocapsid antibodies during the time period of interest, so they only include individuals whose immunity derives from infection rather than vaccination [54]. We estimated statewide hospital admissions in the same time interval by accounting for the delay between infection and hospitalization. We fit a gamma distribution to the combined dis- tribution derived from the time to symptom onset estimated in [94] and the time between symptom onset and hospital admission estimated in [95]. We chose the gamma distribution because it was able to match the non-normal shape of the combined distribution. We esti- mated the delay distribution as: d � Gðshape ¼ 2:99; rate ¼ 0:27Þ We generated 1,000 samples for δ and generated a distribution of total hospital admissions for each age group as: NHk;state;i ¼ Xt1þdi t¼t0þdi Hk;stateðtÞ Where NHk,state,i is a single sample of the hospital admission distribution and Hk,state(t) is the raw hospital admission count for Texas at time, t. We aggregated hospital admission data, which are stratified into 0–17, 18–19, 20–29, 30–39, 40–49, 50–59, 60–69, 70–79, and 80+ year age groups, to match the stratification of the seroprevalence data (0–17, 18–49, 50–64, and 65 + years). For bins that do not align, we divided admissions evenly across years within a bin. Finally, we estimated the infection hospitalization rate for each of the 1,000 samples as: mk;state;i ¼ NHk;state;i NIk;state;i Where μk,state,i is a single sample of the infection hospitalization rate for age group k and NIk,state,i is a single sample from the estimated normal distribution. ZIP- and age-specific infection hospitalization rates (IHRs) Infection hospitalization rates depend on the underlying demographic makeup of a population [64]. To estimate age- and ZIP-specific IHRs from the statewide averages, we assumed that risk differences between ZIP codes could be captured by the proportion of the population esti- mated to be at high risk for severe COVID-19. We converted the statewide age-specific IHRs to ZIP-specific ones as: mk;z ¼ rk;z � mk;state;hr þ ð1 (cid:0) rk;zÞ � mk;state;lr where μk,z is the infection hospitalization rate for age group, k, and ZIP code z, μk,state,hr and μk, state,lr are the statewide estimated age-specific IHRs for those at high and low risk to severe COVID-19 outcomes respectively, and ρk,z is the proportion of the population at high risk to severe COVID-19 outcomes in that age and zip code. The estimates of ρk,z represent the pro- portion of the population having at least one chronic condition linked to increased risk of PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 10 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city severe COVID-19 disease such as cancer, obesity, diabetes, asthma, or HIV [5,96]. We assume a fixed hospitalization risk ratio between low and high risk individuals, μk,state,hr = ηk�μk,state,lr, where ηk is the age-specific hospitalization risk ratio estimated in [97]. For example, high risk individuals in the 20–24 and 75+ age groups are estimated to have 6.5 and 2.2 times the hospi- talization risk respectively compared with low risk individuals in the same age group (S2 Table). We then estimate μk,state,lr and μk,state,hr as: mk;state;lr ¼ mk;state ðZk (cid:0) 1Þ � rk;state þ 1 mk;state;hr ¼ Zk � mk;state;lr where μk,state is the estimated statewide age-specific IHR, and ρk,state is the statewide age-spe- cific estimate for the proportion of the population at high risk for severe COVID-19 [5,96]. Confidence intervals for μk,z are derived by converting the lower and upper bound estimates for μk,state in the same fashion. Age- and ZIP code- specific infection estimates We estimate the number of infections (Ik,z) in a specific age group (k) and ZIP code (z) using the reported hospital admissions (Hk,z) and the infection hospitalization rate (mk,z). We assume that infections are independent from one another and that every infected individual within an age group and ZIP code has the same chance of being hospitalized (mk,z). We describe their relationship with a binomial distribution as: pðHk;zjIk;z; mk;zÞ � binomðIk;z; mk;zÞ We use a discrete uniform prior for Ik,z that ensures there are at least as many infections as hospital admissions (Hk,z) and no more infections than the total population size (Nk,z), and we assume an informative prior beta distribution for mk,z as it is a flexible distribution that ensures the rate remains between zero and one: pðmk;zÞ � betaðak;z; bk;zÞ pðIk;zÞ � unifðHk;z; Nk;zÞ We estimate the parameters for the informative beta prior distribution, ak,z and bk,z, using the ZIP and age-specific IHR estimates estimated from seroprevalence data in the previous sec- tion. Specifically we use the equation: ak;z ¼ bk;z � mk;z;mean 1 (cid:0) mk;z;mean and identify the value of bk,z that minimizes the difference between μk,z,2.5% and the 2.5th per- centile of the resulting beta distribution, beta(ak,z, bk,z). In essence, we estimate the shape parameters of a beta distribution that match the mean and lower bound estimate of the IHR. Our posterior distribution for Ik,z and mk,z can then be defined as: pðIk;z; mk;zjHk;zÞ / pðHk;zjmk;z; Ik;zÞ � pðIk;zÞ � pðmk;zÞ We used Markov chain Monte Carlo (MCMC) sampling of the posterior distribution using the rjags package in the R programming language [98,99]. Specifically we sample 1,000 draws from the posterior distribution across four chains thinning every two samples and with a 200 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 11 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city sample burn-in period. Throughout the paper we summarize the posterior distributions using their mean and 95% credible intervals. Estimating the timing of infections We created a distribution of the infection timing using the hospital admission timing and the previously estimated delay distribution between infection and hospital admission, δ~Γ(shape =2.99,rate =0.27). We draw 1,000 infection time-series samples using the 1,000 posterior infection samples for each age- and ZIP-group. For each sample and for infection i2{1,. . .,Ik,z} we: 1. Draw a single hospital admission date, dH,i, randomly from the dates of all hospital admis- sions recorded in the age- and ZIP-group 2. Draw a single infection to hospital admission delay, δi, from the delay distribution 3. Assign the date of the infection, dI,i, as dI,i = dH,i−δi Each of the 1,000 estimated dI vectors capture a single infection time-series. For age and ZIP code groups that had estimated infections but zero reported hospitalizations, we drew the hospital admission date, dH,i, from the dates of all hospital admissions within the ZIP code. We present an example outlining the infection estimation procedure in the supplement (S18 Fig). Reporting rate estimates We assumed that every infection, Ij, had the same chance of being reported, rj, and that reported cases, Cj, were independent from one another, so that reported cases were distributed binomially as: pðCjÞ � binomðIj; rjÞ Where j describes the specific subgroup of interest (age group, k, and/or ZIP code, z). Assuming a uniform conjugate beta prior distribution on the reporting rate, the posterior for rj can be calculated as: pðrjjIj; CjÞ � betað1 þ Cj; 1 þ Ij (cid:0) CjÞ For subgroups where Ij<Cj, we increase Ij so that Ij = Cj. We estimated overall and subgroup reporting rates for the full study period through June 1, 2021 using cumulative age-specific case counts for Travis County [100] as well as ZIP code specific counts provided by Austin Public Health (APH) [101]. Separately, we estimated the age- and ZIP-specific reporting rates from a subset of testing data provided to us directly from Austin Public Health, which included 60% of reported cases in Travis County during the time period. Social Vulnerability Index (SVI) as a predictor The CDC’s Social Vulnerability Index (SVI) is an indicator that estimates a community’s abil- ity to withstand a hazardous emergency event such as a hurricane or disease outbreak [14]. SVI values are based on 15 different American Community Survey (ACS) variables and are given at the level of census tract as percentile ranks (range 0.0–1.0) within each state based on the 2014–2018 5-year ACS. For example, an SVI of 0.6 indicates that a census tract is more vul- nerable than 60% of other census tracts in the state. We aggregated SVI to ZIP codes using weighted averages based on the percent of residential addresses in a ZIP code that fall in each census tract [102,103]. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 12 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city We estimated the impact of SVI on infection and reporting rates using a mixed effect pois- son regression model using the lme4 R package [104]. For estimating the impact of SVI on infection rates the model can be described as: Iz;i � PoisðexpðbI;1 � SVIz þ bI;0 þ pI;zÞÞ Where IZ,i is the ith infection estimate sample estimated for ZIP code, z, SVIz is the ZIP codes’ SVI, πI,z is the ZIP code level random effect, βI,1 is the fixed effect of SVI on infections, and βI,0 is an intercept term. We use the ZIP code population as an offset in the model to stan- dardize infection rates. For estimating the impact of SVI on reporting rates the model can be described as: Cz � PoisðexpðbC;1 � SVIz þ bC;0 þ pC;zÞÞ where Cz is the reported case count for ZIP code, z, πC,z is the ZIP code level random effect, βC,1 is the fixed effect of SVI on cases, and βC,0 is an intercept. We use the 1,000 ZIP code infec- tion estimate samples as an offset in the model to standardize reporting rates. For the age and ZIP code-stratified analysis we also include an interaction term between age and SVI, so the equations become: Ik;z;i � PoisðexpðbI0;k;1 � SVIz þ bI0;0 þ pI0;zÞÞ Ck;z � PoisðexpðbC0;k;1 � SVIz þ bC0;0 þ pC0;zÞÞ where Ik,z,i is the ith infection estimate sample in age group, k for ZIP code, z, Ck,z is the reported cases in age group, k for ZIP code, z, βI0,k,1 and βC0,k,1 are the SVI regression coefficient for age group, k for the infection and case interaction terms respectively, βI0,0 and βC0,0 are the intercepts, and πI0,z and πC0,z are the ZIP code level random effects. For the infection and reporting rate models we use the age- and ZIP- population and infection estimates as offsets respectively. The SVI regression coefficients (βI,1, βC,1, βI0,k,1, and βC0,k,1) can be interpreted as inequality metrics, quantifying the relative infection and reporting risks as a function of SVI. Because there are no ZIP codes with a value of 0 or 1 for SVI in our sample, we report the relative infec- tion and reporting rates between ZIP codes in the 25th and 75th percentile in Travis County throughout the manuscript. Supporting information S1 Fig. Daily COVID-19 burden estimates for Travis County, Texas from March 1, 2020 until June 1, 2021. Daily new reported case (A) and mortality (B) counts as reported by the New York Times for Travis County, Texas [100]. (TIFF) S2 Fig. Comparison of age-dependent estimates for infection-hospitalization rates. Age- stratified estimates of the risk of severe COVID-19 (defined as risk for hospitalization) from China [51], France [45]. (TIFF) S3 Fig. Weekly estimated relative infection rates from March 1, 2020 until June 1, 2021 across age groups. Points and error bars indicate the median and 95% confidence interval for the weekly infection rate with the size of the population. Values of 1 (horizontal dashed line) indicate that the fraction of the infections occurring that week equals the population fraction for the specific age group, while values below or above one indicate the age group faced PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 13 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city disproportionately low or high infection risk respectively during that week. Only the 65+ age group consistently experienced disproportionately low infection rates compared with their population size over the whole pandemic. (TIFF) S4 Fig. Reported 7-day average of case counts by age group from April 22, 2020 until May 28, 2021. Daily reported cases counts for each age group provided by Austin Public Health [101]. (TIFF) S5 Fig. Reported hospital admissions by age group from March 1, 2020 until June 1, 2021. Age-specific admission data provided by Austin Public Health. (TIFF) S6 Fig. COVID-19 estimated cumulative infections for Travis County (Austin, TX) and the state of Texas from March 1, 2020 to June 1, 2021. Estimated cumulative infections in Travis County with 95% credible intervals (black line and gray ribbon) compared to Texas statewide seroprevalence-based estimates (red points and error bars) for each of the four age groups [49]. (TIFF) S7 Fig. Relative COVID-19 infection risk by age group in Travis County as compared to the state of Texas from March 1, 2020 to June 1, 2021. For each age group, we compare the mean model-estimated infection rates for Travis County with the mean statewide seropreva- lence estimates in Texas to estimate the mean relative infection risks between the two (points and smoothed lines). Values above the horizontal dashed line indicate that Travis County resi- dents faced higher infection risks than residents of Texas, while values below the line indicate higher statewide infection risks. As of June 1, 2021, infection rates were 45% (95% CrI: 20– 61%), 19.5% (95% CrI: 0.1–33%), 22.7 (0.1–40%), and 29.8% (95% CrI: 2–48%) lower for indi- viduals 0–17, 18–49, 50–64, and 65+ respectively in Travis compared with Texas as whole. (TIFF) S8 Fig. Estimated ZIP code and age-specific IHR for each ZIP code in Travis County. Infec- tion hospitalization rates derived from Texas-specific estimates (Table 1) using population risk estimation methodology for each age group as detailed in [5,96,105]. (TIFF) S9 Fig. Cumulative infection estimates for each ZIP code and age group in Travis County using hospitalization data up to June 1, 2021. (TIFF) S10 Fig. Cumulative estimated reporting rate for each ZIP code and age group in Travis County using reported case data up to June 1, 2021. Testing data used for reporting rates are only a subset of all tests performed, as age and ZIP code stratified data were only available for Austin Public Health administered tests. (TIFF) S11 Fig. Relationship between the infections per 100,000 and the reporting rate for ZIP codes in Travis County (Austin, TX), between March 1, 2020 and June 1, 2021. Points and error bars indicate the mean and 95% confidence intervals for each ZIP code. The dashed line indicates the mean estimated relationship across 1,000 posterior samples, with a = 70.4 (95% CrI: 28–141) and b = 0.533 (95% CrI: 0.45–0.61). (TIFF) PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 14 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city S12 Fig. Reported case and hospitalization counts correlate with social vulnerability in Travis County from March 1, 2020 to June 1, 2021. (A) Across the 46 ZIP codes, SVI is a sig- nificant predictor of reported case counts (p<0.001). The blue line and ribbon indicate the mean and 95% prediction interval from the fitted Poisson mixed-effects model. (B) Across the 46 ZIP codes, SVI is a significant predictor of reported hospitalization counts (p<0.001). The blue line and ribbon indicate the mean and 95% prediction interval from the fitted Poisson mixed-effects model. (TIFF) S13 Fig. Comparison between the estimated SARS-CoV-2 infection rates in each Travis County ZIP code from March 1, 2020 to June 1, 2021 with the 15 individual components of the Social Vulnerability Index (SVI). Points and error bars indicate the mean and 95% credi- ble intervals for estimated infection rates in a ZIP code. (TIFF) S14 Fig. Comparison between the estimated SARS-CoV-2 reporting rates in each Travis County ZIP code from March 1, 2020 to June 1, 2021 with the 15 individual components of the Social Vulnerability Index (SVI). Points and error bars indicate the mean and 95% credi- ble intervals for estimated reporting rates in a ZIP code. (TIFF) S15 Fig. Estimated infection rates correlate with social vulnerability in Travis County from March 1, 2020 to June 1, 2021 across all age groups. Across the 46 ZIP codes, SVI has a positive relationship with cumulative infection rates as a proportion of the population for every age group (S1 Table). Estimated age-specific SVI relationships from the poisson mixed effects regression model are shown in the blue line (mean) and blue ribbon (95% confidence interval). (TIFF) S16 Fig. Observed biases in the subset of reported case data stratified by age and ZIP code. (A) Fraction of all reported cases included in the subset of age- and ZIP-code stratified data collected through Austin Public Health’s community testing programs by ZIP code. Overall, the data set covers 60% of all reported cases, but the data set, which does not include all cases identified by private testing sites, has high levels of coverage in the vulnerable ZIP codes of East Austin. (B) Reported case coverage from the dataset correlates positively with SVI. Blue line indicates the mean of a fitted linear regression model. (TIFF) S17 Fig. Estimated reporting rates correlate with social vulnerability in Travis County from March 1, 2020 to June 1, 2021. Across the 46 ZIP codes, SVI has a flat or slightly nega- tive relationship with cumulative infection reporting rates for every age group except for those aged 65+ (S1 Table). Estimated age-specific SVI relationships from the poisson mixed effects regression model are shown in the blue line (mean) and blue ribbon (95% confidence inter- val). (TIFF) S18 Fig. Infection estimation methodology for a hypothetical region with 150 hospital admissions and a mean infection hospitalization rate of 0.2. (A) Prior distribution of the infection hospitalization rate for the example region specified by α = 25 and β = 100. (B) Hos- pital admission counts by day in the example region. (C) Estimated cumulative infection dis- tribution for the region based on the hospital admission count and IHR distribution. IHR PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 15 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city distribution is made up of 1,000 draws from the posterior distribution. (D) Cumulative esti- mated infections over time for each of the 1,000 posterior infection draws. Timing is based on the hospital admission timing and the delay distribution between infection and hospitaliza- tion. (TIFF) S1 Table. Comparison of age-stratified risk ratio between more vulnerable (75th SVI per- centile) and less vulnerable (25th SVI percentile) ZIP codes for Travis county, from March 1, 2020 to June 1, 2021. Estimates based on reported COVID-19 hospitalizations are consis- tently higher than those based on model-derived estimates of ZIP-code level infection rates and observed COVID-19 case rates. (XLSX) S2 Table. Relative hospitalization rates for high risk individuals compared with low risk individuals from [97]. (XLSX) Acknowledgments The authors thank Austin Public Health and the City of Austin for providing the raw data and useful feedback on early drafts of the manuscript. Author Contributions Conceptualization: Spencer J. Fox, Emily Javan, Michael Lachmann, Lauren Ancel Meyers. Data curation: Briana Betke. Formal analysis: Spencer J. Fox, Emily Javan, Remy Pasco, Spencer Woody, Michael Lach- mann, Lauren Ancel Meyers. Funding acquisition: Lauren Ancel Meyers. Investigation: Maureen Johnson-Leo´n. Methodology: Spencer J. Fox, Remy Pasco, Graham C. Gibson, Briana Betke, Jose´ L. Herrera- Diestra, Spencer Woody, Kelly Pierce, Kaitlyn E. Johnson, Maureen Johnson-Leo´n, Michael Lachmann, Lauren Ancel Meyers. Resources: Lauren Ancel Meyers. Software: Emily Javan, Remy Pasco, Graham C. Gibson, Kelly Pierce, Michael Lachmann. Supervision: Lauren Ancel Meyers. Validation: Spencer J. Fox. Visualization: Spencer J. Fox, Emily Javan, Briana Betke, Spencer Woody, Kelly Pierce, Kai- tlyn E. Johnson, Michael Lachmann. Writing – original draft: Spencer J. Fox. Writing – review & editing: Spencer J. Fox, Emily Javan, Remy Pasco, Graham C. Gibson, Bri- ana Betke, Jose´ L. Herrera-Diestra, Spencer Woody, Kelly Pierce, Kaitlyn E. Johnson, Mau- reen Johnson-Leo´n, Michael Lachmann, Lauren Ancel Meyers. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 16 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city References 1. 14.9 million excess deaths associated with the COVID-19 pandemic in 2020 and 2021. [cited 17 Jun 2022]. Available: https://www.who.int/news/item/05-05-2022-14.9-million-excess-deaths-were- associated-with-the-covid-19-pandemic-in-2020-and-2021 2. Bajema KL, Wiegand RE, Cuffe K, Patel SV, Iachan R, Lim T, et al. Estimated SARS-CoV-2 Seroprev- alence in the US as of September 2020. JAMA Intern Med. 2021; 181: 450–460. https://doi.org/10. 1001/jamainternmed.2020.7976 PMID: 33231628 3. Webb Hooper M, Na´poles AM, Pe´rez-Stable EJ. COVID-19 and Racial/Ethnic Disparities. JAMA. 2020; 323: 2466–2467. https://doi.org/10.1001/jama.2020.8598 PMID: 32391864 4. Romano SD, Blackstock AJ, Taylor EV, El Burai Felix S, Adjei S, Singleton C-M, et al. Trends in Racial and Ethnic Disparities in COVID-19 Hospitalizations, by Region—United States, March-December 2020. MMWR Morb Mortal Wkly Rep. 2021; 70: 560–565. https://doi.org/10.15585/mmwr.mm7015e2 PMID: 33857068 5. Pasco RF, Fox SJ, Johnston SC, Pignone M, Meyers LA. Estimated Association of Construction Work With Risks of COVID-19 Infection and Hospitalization in Texas. JAMA Netw Open. 2020; 3: e2026373. https://doi.org/10.1001/jamanetworkopen.2020.26373 PMID: 33119111 6. Lancet The. The plight of essential workers during the COVID-19 pandemic. Lancet. 2020; 395: 1587. https://doi.org/10.1016/S0140-6736(20)31200-9 PMID: 32446399 7. Karmakar M, Lantz PM, Tipirneni R. Association of Social and Demographic Factors With COVID-19 Incidence and Death Rates in the US. JAMA Netw Open. 2021; 4: e2036462. https://doi.org/10.1001/ jamanetworkopen.2020.36462 PMID: 33512520 8. Khazanchi R, Beiter ER, Gondi S, Beckman AL, Bilinski A, Ganguli I. County-Level Association of Social Vulnerability with COVID-19 Cases and Deaths in the USA. J Gen Intern Med. 2020; 35: 2784– 2787. https://doi.org/10.1007/s11606-020-05882-3 PMID: 32578018 9. Dasgupta S. Association Between Social Vulnerability and a County’s Risk for Becoming a COVID-19 Hotspot—United States, June 1–July 25, 2020. MMWR Morb Mortal Wkly Rep. 2020;69. https://doi. org/10.15585/mmwr.mm6942a3 PMID: 33090977 10. Hughes MM, Wang A, Grossman MK, Pun E, Whiteman A, Deng L, et al. County-Level COVID-19 Vaccination Coverage and Social Vulnerability—United States, December 14, 2020-March 1, 2021. MMWR Morb Mortal Wkly Rep. 2021; 70: 431–436. https://doi.org/10.15585/mmwr.mm7012e1 PMID: 33764963 11. Sy KTL, Martinez ME, Rader B, White LF. Socioeconomic disparities in subway use and COVID-19 outcomes in New York City. medRxiv. 2020. https://doi.org/10.1101/2020.05.28.20115949 PMID: 32511568 12. Carrio´n D, Colicino E, Pedretti NF, Arfer KB, Rush J, DeFelice N, et al. Neighborhood-level disparities and subway utilization during the COVID-19 pandemic in New York City. Nat Commun. 2021; 12: 3692. https://doi.org/10.1038/s41467-021-24088-7 PMID: 34140520 13. Kim SJ, Bostwick W. Social Vulnerability and Racial Inequality in COVID-19 Deaths in Chicago. Health Educ Behav. 2020; 47: 509–513. https://doi.org/10.1177/1090198120929677 PMID: 32436405 14. CDC’s Social Vulnerability Index (SVI). 15 Oct 2020 [cited 21 Dec 2020]. Available: https://www.atsdr. cdc.gov/placeandhealth/svi/index.html 15. Chiriboga D, Garay J, Buss P, Madrigal RS, Rispel LC. Health inequity during the COVID-19 pan- demic: a cry for ethical global leadership. Lancet. 2020; 395: 1690–1691. https://doi.org/10.1016/ S0140-6736(20)31145-4 PMID: 32419711 16. Johnson-Agbakwu CE, Ali NS, Oxford CM, Wingo S, Manin E, Coonrod DV. Racism, COVID-19, and Health Inequity in the USA: a Call to Action. J Racial Ethn Health Disparities. 2020. https://doi.org/10. 1007/s40615-020-00928-y PMID: 33197038 17. Asundi AO’Leary C, Bhadelia N. Global COVID-19 vaccine inequity: The scope, the impact, and the challenges. Cell Host Microbe. 2021; 29: 1036–1039. https://doi.org/10.1016/j.chom.2021.06.007 PMID: 34265241 18. Wang Z, Tang K. Combating COVID-19: health equity matters. Nat Med. 2020; 26: 458. https://doi. org/10.1038/s41591-020-0823-6 PMID: 32284617 19. Berkowitz SA, Cene´ CW, Chatterjee A. Covid-19 and Health Equity—Time to Think Big. N Engl J Med. 2020; 383: e76. https://doi.org/10.1056/NEJMp2021209 PMID: 32706955 20. Largent EA, Persad G, Mello MM, Wenner DM, Kramer DB, Edmonds BT, et al. Incorporating Health Equity Into COVID-19 Reopening Plans: Policy Experimentation in California. Am J Public Health. 2021; 111: 1481–1488. https://doi.org/10.2105/AJPH.2021.306263 PMID: 34111945 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 17 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city 21. Wrigley-Field E, Kiang MV, Riley AR, Barbieri M, Chen Y-H, Duchowny KA, et al. Geographically tar- geted COVID-19 vaccination is more equitable and averts more deaths than age-based thresholds alone. Sci Adv. 2021; 7: eabj2099. https://doi.org/10.1126/sciadv.abj2099 PMID: 34586843 22. Dooling K, McClung N, Chamberland M, Marin M, Wallace M, Bell BP, et al. The Advisory Committee on Immunization Practices’ Interim Recommendation for Allocating Initial Supplies of COVID-19 Vac- cine—United States, 2020. MMWR Morb Mortal Wkly Rep. 2020; 69: 1857–1859. https://doi.org/10. 15585/mmwr.mm6949e1 PMID: 33301429 23. Dooling K, Marin M, Wallace M, McClung N, Chamberland M, Lee GM, et al. The Advisory Committee on Immunization Practices’ Updated Interim Recommendation for Allocation of COVID-19 Vaccine— United States, December 2020. MMWR Morb Mortal Wkly Rep. 2021; 69: 1657–1660. https://doi.org/ 10.15585/mmwr.mm695152e2 PMID: 33382671 24. National Academies of Sciences, Engineering, and Medicine, Health and Medicine Division, Board on Population Health and Public Health Practice, Board on Health Sciences Policy, Committee on Equita- ble Allocation of Vaccine for the Novel Coronavirus. Framework for Equitable Allocation of COVID-19 Vaccine. National Academies Press; 2020. 25. Schmidt H, Weintraub R, Williams MA, Miller K, Buttenheim A, Sadecki E, et al. Equitable allocation of COVID-19 vaccines in the United States. Nat Med. 2021; 27: 1298–1307. https://doi.org/10.1038/ s41591-021-01379-6 PMID: 34007071 26. Woody S, Javan E, Johnson K, Pasco R, Johnson-Leo´n M, Lachmann M, et al. Spatial distribution of COVID-19 infections and vaccinations in Austin, Texas. 2021. Available: https://covid-19.tacc.utexas. edu/media/filer_public/fe/f2/fef289f8-800c-4390-ab27-6eff3b229f59/covid_infections_and_ vaccinations_-_austin_zip_codes_-_ut_-_041221.pdf 27. Sacarny A, Daw JR. Inequities in COVID-19 Vaccination Rates in the 9 Largest US Cities. JAMA Health Forum. 2021; 2: e212415–e212415. https://doi.org/10.1001/jamahealthforum.2021.2415 PMID: 35977181 28. Diesel J, Sterrett N, Dasgupta S, Kriss JL, Barry V, Vanden Esschert K, et al. COVID-19 Vaccination Coverage Among Adults—United States, December 14, 2020-May 22, 2021. MMWR Morb Mortal Wkly Rep. 2021; 70: 922–927. https://doi.org/10.15585/mmwr.mm7025e1 PMID: 34166331 29. Ogilvie GS, Gordon S, Smith LW, Albert A, Racey CS, Booth A, et al. Intention to receive a COVID-19 vaccine: results from a population-based survey in Canada. BMC Public Health. 2021; 21: 1017. https://doi.org/10.1186/s12889-021-11098-9 PMID: 34051770 30. Garnham JP, Platoff E. Dallas County axes plan to prioritize vaccinating communities of color after state threatens to slash allocation. The Texas Tribune. 21 Jan 2021. Available: https://www. texastribune.org/2021/01/20/dallas-vaccine-plan-communities-of-color/. Accessed 23 Mar 2023. 31. Governor Abbott DSHS announce COVID-19 vaccine distribution plan. [cited 23 Mar 2023]. Available: https://gov.texas.gov/news/post/governor-abbott-dshs-announce-covid-19-vaccine-distribution-plan 32. Adhikari S, Pantaleo NP, Feldman JM, Ogedegbe O, Thorpe L, Troxel AB. Assessment of Commu- nity-Level Disparities in Coronavirus Disease 2019 (COVID-19) Infections and Deaths in Large US Metropolitan Areas. JAMA Netw Open. 2020; 3: e2016938. https://doi.org/10.1001/jamanetworkopen. 2020.16938 PMID: 32721027 33. Tortolero GA, Otto M de O, Ramphul R, Yamal J-M, Rector A, Brown M, et al. Examining Social Vul- nerability and the Association With COVID-19 Incidence in Harris County, Texas. Front Public Health. 2021; 9: 798085. https://doi.org/10.3389/fpubh.2021.798085 PMID: 35071172 34. Rossen LM, Branum AM, Ahmad FB, Sutton P, Anderson RN. Excess Deaths Associated with COVID-19, by Age and Race and Ethnicity—United States, January 26-October 3, 2020. MMWR Morb Mortal Wkly Rep. 2020; 69: 1522–1527. https://doi.org/10.15585/mmwr.mm6942e2 PMID: 33090978 35. Gold JAW, Rossen LM, Ahmad FB, Sutton P, Li Z, Salvatore PP, et al. Race, Ethnicity, and Age Trends in Persons Who Died from COVID-19—United States, May-August 2020. MMWR Morb Mortal Wkly Rep. 2020; 69: 1517–1521. https://doi.org/10.15585/mmwr.mm6942e1 PMID: 33090984 36. Price-Haywood EG, Burton J, Fort D, Seoane L. Hospitalization and Mortality among Black Patients and White Patients with Covid-19. N Engl J Med. 2020; 382: 2534–2543. https://doi.org/10.1056/ NEJMsa2011686 PMID: 32459916 37. 38. Jones JM, Stone M, Sulaeman H, Fink RV, Dave H, Levy ME, et al. Estimated US Infection- and Vac- cine-Induced SARS-CoV-2 Seroprevalence Based on Blood Donations, July 2020-May 2021. JAMA. 2021; 326: 1400–1409. https://doi.org/10.1001/jama.2021.15161 PMID: 34473201 Fitzpatrick V, Rivelli A, Blair C, Copeland K, Richards J. Disparities in Seroprevalence of SARS-CoV-2 Immunoglobulin Antibodies in a Large Midwestern Health Care System. Public Health Rep. 2021; 136: 361–367. https://doi.org/10.1177/0033354921999168 PMID: 33673779 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 18 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city 39. Anand S, Montez-Rath M, Han J, Cadden L, Hunsader P, Kerschmann R, et al. Estimated SARS- CoV-2 Seroprevalence in US Patients Receiving Dialysis 1 Year After the Beginning of the COVID-19 Pandemic. JAMA Netw Open. 2021; 4: e2116572. https://doi.org/10.1001/jamanetworkopen.2021. 16572 PMID: 34251441 40. Asch DA, Islam MN, Sheils NE, Chen Y, Doshi JA, Buresh J, et al. Patient and Hospital Factors Asso- ciated With Differences in Mortality Rates Among Black and White US Medicare Beneficiaries Hospi- talized With COVID-19 Infection. JAMA Netw Open. 2021; 4: e2112842. https://doi.org/10.1001/ jamanetworkopen.2021.12842 PMID: 34137829 41. Fu J, Reid SA, French B, Hennessy C, Hwang C, Gatson NT, et al. Racial Disparities in COVID-19 Outcomes Among Black and White Patients With Cancer. JAMA Netw Open. 2022; 5: e224304. https://doi.org/10.1001/jamanetworkopen.2022.4304 PMID: 35344045 42. Schultes O, Lind ML, Brockmeyer J, Sosensky P, Cummings DAT, Ko AI. Closing the health inequity gap during the pandemic: COVID-19 mortality among racial and ethnic groups in Connecticut, March 2020 to December 2021. J Epidemiol Community Health. 2022; 76: 695–696. https://doi.org/10.1136/ jech-2022-218975 PMID: 35428672 43. Pathela Crawley, Weiss Maldin. Seroprevalence of SARS-CoV-2 following the largest initial epidemic wave in the United States: Findings from New York City, May 13-July 21, 2020. The Journal of. Avail- able: https://stacks.cdc.gov/view/cdc/107346 44. Mustanski B, Saber R, Ryan DT, Benbow N, Madkins K, Hayford C, et al. Geographic disparities in COVID-19 case rates are not reflected in seropositivity rates using a neighborhood survey in Chicago. Ann Epidemiol. 2022; 66: 44–51. https://doi.org/10.1016/j.annepidem.2021.10.006 PMID: 34728335 45. Salje H, Tran Kiem C, Lefrancq N, Courtejoie N, Bosetti P, Paireau J, et al. Estimating the burden of SARS-CoV-2 in France. Science. 2020; 369: 208–211. https://doi.org/10.1126/science.abc3517 PMID: 32404476 46. Mena GE, Martinez PP, Mahmud AS, Marquet PA, Buckee CO, Santillana M. Socioeconomic status determines COVID-19 incidence and related mortality in Santiago, Chile. Science. 2021; 372. https:// doi.org/10.1126/science.abg5298 PMID: 33906968 47. Yang W, Kandula S, Huynh M, Greene SK, Van Wye G, Li W, et al. Estimating the infection-fatality risk of SARS-CoV-2 in New York City during the spring 2020 pandemic wave: a model-based analysis. Lancet Infect Dis. 2021; 21: 203–212. https://doi.org/10.1016/S1473-3099(20)30769-6 PMID: 33091374 48. Hunter DJ, Abdool Karim SS, Baden LR, Farrar JJ, Hamel MB, Longo DL, et al. Addressing Vaccine Inequity—Covid-19 Vaccines as a Global Public Good. The New England journal of medicine. 2022. pp. 1176–1179. https://doi.org/10.1056/NEJMe2202547 PMID: 35196425 49. National Center for Immunization, Division of Viral Diseases. Nationwide Commercial Laboratory Seroprevalence Survey. 2020. Available: https://data.cdc.gov/Laboratory-Surveillance/Nationwide- Commercial-Laboratory-Seroprevalence-Su/d2tw-32xv 50. Experience. [cited 3 Oct 2022]. Available: https://experience.arcgis.com/experience/ 3793562ab87e4299b106e0c282bb1fc5 51. Verity R, Okell LC, Dorigatti I, Winskill P, Whittaker C, Imai N, et al. Estimates of the severity of corona- virus disease 2019: a model-based analysis. Lancet Infect Dis. 2020; 20: 669–677. https://doi.org/10. 1016/S1473-3099(20)30243-7 PMID: 32240634 52. U.S. Department of Health & Human Services. COVID-19 Reported Patient Impact and Hospital Capacity by State Timeseries. 2020. Available: https://healthdata.gov/Hospital/COVID-19-Reported- Patient-Impact-and-Hospital-Capa/g62h-syeh 53. Havers FP, Reed C, Lim T, Montgomery JM, Klena JD, Hall AJ, et al. Seroprevalence of Antibodies to SARS-CoV-2 in 10 Sites in the United States, March 23-May 12, 2020. JAMA Intern Med. 2020. https://doi.org/10.1001/jamainternmed.2020.4130 PMID: 32692365 54. CDC. COVID data tracker. In: Centers for Disease Control and Prevention [Internet]. 28 Mar 2020 [cited 23 Dec 2021]. Available: https://covid.cdc.gov/covid-data-tracker/ 55. Pullano G, Di Domenico L, Sabbatini CE, Valdano E, Turbelin C, Debin M, et al. Underdetection of cases of COVID-19 in France threatens epidemic control. Nature. 2021; 590: 134–139. https://doi.org/ 10.1038/s41586-020-03095-6 PMID: 33348340 56. US Census Bureau. TIGER/Line Shapefiles and TIGER/line files technical documentation. 2023 [cited 12 Apr 2023]. Available: https://www.census.gov/programs-surveys/geography/technical- documentation/complete-technical-documentation/tiger-geo-line.2019.html 57. Walker K, Eberwein K, Herman M. Tidycensus: Load us census boundary and attribute data as’ tidy- verse’and’sf’-ready data frames. URL https://CRAN.R-project.org/package=tidycensus R package version 0 4. 2018;1. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 19 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city 58. COVID-19 Testing Options Expand in Austin-Travis County. [cited 6 Oct 2022]. Available: https:// www.austintexas.gov/news/covid-19-testing-options-expand-austin-travis-county 59. Grabowski DC, Mor V. Nursing Home Care in Crisis in the Wake of COVID-19. JAMA. 2020; 324: 23– 24. https://doi.org/10.1001/jama.2020.8524 PMID: 32442303 60. Zelner J, Trangucci R, Naraharisetti R, Cao A, Malosh R, Broen K, et al. Racial Disparities in Coronavi- rus Disease 2019 (COVID-19) Mortality Are Driven by Unequal Infection Risks. Clin Infect Dis. 2021; 72: e88–e95. https://doi.org/10.1093/cid/ciaa1723 PMID: 33221832 61. Gozzi N, Tizzoni M, Chinazzi M, Ferres L, Vespignani A, Perra N. Estimating the effect of social inequalities on the mitigation of COVID-19 across communities in Santiago de Chile. Nat Commun. 2021; 12: 2429. https://doi.org/10.1038/s41467-021-22601-6 PMID: 33893279 62. Wadhera RK, Wadhera P, Gaba P, Figueroa JF, Joynt Maddox KE, Yeh RW, et al. Variation in COVID-19 Hospitalizations and Deaths Across New York City Boroughs. JAMA. 2020; 323: 2192– 2195. https://doi.org/10.1001/jama.2020.7197 PMID: 32347898 63. Davies NG, Klepac P, Liu Y, Prem K, Jit M, CMMID COVID-19 working group, et al. Age-dependent effects in the transmission and control of COVID-19 epidemics. Nat Med. 2020; 26: 1205–1211. https://doi.org/10.1038/s41591-020-0962-9 PMID: 32546824 64. Levin AT, Hanage WP, Owusu-Boaitey N, Cochran KB, Walsh SP, Meyerowitz-Katz G. Assessing the age specificity of infection fatality rates for COVID-19: systematic review, meta-analysis, and public policy implications. Eur J Epidemiol. 2020; 35: 1123–1138. https://doi.org/10.1007/s10654-020- 00698-1 PMID: 33289900 65. Viner RM, Mytton OT, Bonell C, Melendez-Torres GJ, Ward J, Hudson L, et al. Susceptibility to SARS- CoV-2 Infection Among Children and Adolescents Compared With Adults: A Systematic Review and Meta-analysis. JAMA Pediatr. 2021; 175: 143–156. https://doi.org/10.1001/jamapediatrics.2020.4573 PMID: 32975552 66. Dattner I, Goldberg Y, Katriel G, Yaari R, Gal N, Miron Y, et al. The role of children in the spread of COVID-19: Using household data from Bnei Brak, Israel, to estimate the relative susceptibility and infectivity of children. PLoS Comput Biol. 2021; 17: e1008559. https://doi.org/10.1371/journal.pcbi. 1008559 PMID: 33571188 67. AISD Open for Learning Newsletter: First Week, Afterschool Care and More! / Boletı´n informativo Abierto para el aprendizaje del AISD: La primera semana, cuidados despue´ s de clases ¡y ma´ s! [cited 19 Apr 2021]. Available: https://www.austinisd.org/announcements/2020/10/09/aisd-open-learning- newsletter-first-week-afterschool-care-and-more-boletin 68. Laxminarayan R, Wahl B, Dudala SR, Gopal K, Mohan BC, Neelima S, et al. Epidemiology and trans- mission dynamics of COVID-19 in two Indian states. Science. 2020; 370: 691–697. https://doi.org/10. 1126/science.abd7672 PMID: 33154136 69. Clarke KEN, Jones JM, Deng Y, Nycz E, Lee A, Iachan R, et al. Seroprevalence of Infection-Induced SARS-CoV-2 Antibodies—United States, September 2021-February 2022. MMWR Morb Mortal Wkly Rep. 2022; 71: 606–608. https://doi.org/10.15585/mmwr.mm7117e3 PMID: 35482574 70. Stratil JM, Biallas RL, Burns J, Arnold L, Geffert K, Kunzler AM, et al. Non-pharmacological measures implemented in the setting of long-term care facilities to prevent SARS-CoV-2 infections and their con- sequences: a rapid review. Cochrane Database Syst Rev. 2021 [cited 2 May 2022]. https://doi.org/10. 1002/14651858.CD015085.pub2 PMID: 34523727 71. Cuadros DF, Branscum AJ, Mukandavire Z, Miller FD, MacKinnon N. Dynamics of the COVID-19 epi- demic in urban and rural areas in the United States. Ann Epidemiol. 2021; 59: 16–20. https://doi.org/ 10.1016/j.annepidem.2021.04.007 PMID: 33894385 72. Governor’s Order GA 32 Activated for Trauma Service Area that Includes Austin-Travis County. [cited 19 Apr 2021]. Available: https://www.austintexas.gov/news/governors-order-ga-32-activated-trauma- service-area-includes-austin-travis-county 73. AG Paxton sues city of Austin, mayor of Austin, Travis county, and county judge for imposing unlawful shutdown orders. In: Texas Attorney General [Internet]. [cited 24 Mar 2023]. Available: https://www. texasattorneygeneral.gov/news/releases/ag-paxton-sues-city-austin-mayor-austin-travis-county-and- county-judge-imposing-unlawful-shutdown 74. Bradshaw K. Austin-Travis County moves to Stage 5 of coronavirus threat; tightest restrictions urged. Austin American Statesman. 23 Dec 2020. Available: https://www.statesman.com/story/news/2020/ 12/23/austin-travis-county-moves-stage-5-coronavirus-restrictions/4024400001/. Accessed 11 May 2023. 75. Garcia T. “We”re in a crisis stage’: Texas border city reels from coronavirus surge. The Guardian. 27 Oct 2020. Available: https://www.theguardian.com/us-news/2020/oct/27/covid-coronavirus-el-paso- texas-surge. Accessed 11 May 2023. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 20 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city 76. Auger KA, Shah SS, Richardson T, Hartley D, Hall M, Warniment A, et al. Association Between State- wide School Closure and COVID-19 Incidence and Mortality in the US. JAMA. 2020; 324: 859–870. https://doi.org/10.1001/jama.2020.14348 PMID: 32745200 77. Hale T, Angrist N, Hale AJ, Kira B, Majumdar S, Petherick A, et al. Government responses and COVID-19 deaths: Global evidence across multiple pandemic waves. PLoS One. 2021; 16: e0253116. https://doi.org/10.1371/journal.pone.0253116 PMID: 34242239 78. Hale T, Angrist N, Goldszmidt R, Kira B, Petherick A, Phillips T, et al. A global panel database of pan- demic policies (Oxford COVID-19 Government Response Tracker). Nat Hum Behav. 2021; 5: 529– 538. https://doi.org/10.1038/s41562-021-01079-8 PMID: 33686204 79. Pace KL. Forgetting Waller Creek: An Environmental History of Race, Parks, and Planning in Down- town Austin, Texas. J South Hist. 2021; 87: 603–644. 80. Bixler RP, Yang E, Richter SM, Coudert M. Boundary crossing for urban community resilience: A social vulnerability and multi-hazard approach in Austin, Texas, USA. International Journal of Disaster Risk Reduction. 2021; 66: 102613. 81. Austin Public Health COVID-19, and equity. [cited 2 May 2022]. Available: https://www.austintexas. gov/aphequity 82. Ma KC, Menkir TF, Kissler S, Grad YH, Lipsitch M. Modeling the impact of racial and ethnic disparities on COVID-19 epidemic dynamics. Elife. 2021;10. https://doi.org/10.7554/eLife.66601 PMID: 34003112 83. Zelner J, Naraharisetti R. Declining COVID-19 mortality risk ratios must be interpreted with caution. Journal of epidemiology and community health. 2022. p. 840. https://doi.org/10.1136/jech-2022- 219176 PMID: 35568395 84. Barber RM, Sorensen RJD, Pigott DM, Bisignano C, Carter A, Amlag JO, et al. Estimating global, regional, and national daily and cumulative infections with SARS-CoV-2 through Nov 14, 2021: a sta- tistical analysis. Lancet. 2022;0. https://doi.org/10.1016/S0140-6736(22)00484-6 PMID: 35405084 85. Haby MM, Pinart M, Elias V, Reveiz L. Prevalence of asymptomatic Zika virus infection: a systematic review. Bull World Health Organ. 2018; 96: 402–413D. https://doi.org/10.2471/BLT.17.201541 PMID: 29904223 86. 87. Feehan DM, Mahmud AS. Quantifying population contact patterns in the United States during the COVID-19 pandemic. Nat Commun. 2021; 12: 893. https://doi.org/10.1038/s41467-021-20990-2 PMID: 33563992 Zhang J, Litvinova M, Liang Y, Wang Y, Wang W, Zhao S, et al. Changes in contact patterns shape the dynamics of the COVID-19 outbreak in China. Science. 2020; 368: 1481–1486. https://doi.org/10. 1126/science.abb8001 PMID: 32350060 88. Nouvellet P, Bhatia S, Cori A, Ainslie KEC, Baguelin M, Bhatt S, et al. Reduction in mobility and COVID-19 transmission. Nat Commun. 2021; 12: 1090. https://doi.org/10.1038/s41467-021-21358-2 PMID: 33597546 89. Snow J California’s “Equity” algorithm could leave 2 million struggling californians without additional vaccine supply. ACLU NorCal. 2021. Available: https://www.aclunc.org/blog/californias-equity- algorithm-could-leave-2-million-struggling-californians-without-additional 90. Fowlkes A, Gaglani M, Groover K, Thiese MS, Tyner H, Ellingson K, et al. Effectiveness of COVID-19 Vaccines in Preventing SARS-CoV-2 Infection Among Frontline Workers Before and During B.1.617.2 (Delta) Variant Predominance—Eight U.S. Locations, December 2020-August 2021. MMWR Morb Mortal Wkly Rep. 2021; 70: 1167–1169. 91. Mlcochova P, Kemp SA, Dhar MS, Papa G, Meng B, Ferreira IATM, et al. SARS-CoV-2 B.1.617.2 Delta variant replication and immune evasion. Nature. 2021; 599: 114–119. https://doi.org/10.1038/ s41586-021-03944-y PMID: 34488225 92. Farinholt T, Doddapaneni H, Qin X, Menon V, Meng Q, Metcalf G, et al. Transmission event of SARS- CoV-2 delta variant reveals multiple vaccine breakthrough infections. BMC Med. 2021; 19: 255. https://doi.org/10.1186/s12916-021-02103-4 PMID: 34593004 93. Health. [cited 22 Jun 2022]. Available: https://www.austintexas.gov/department/health 94. 95. Lauer SA, Grantz KH, Bi Q, Jones FK, Zheng Q, Meredith HR, et al. The Incubation Period of Corona- virus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Applica- tion. Ann Intern Med. 2020; 172: 577–582. https://doi.org/10.7326/M20-0504 PMID: 32150748 Faes C, Abrams S, Van Beckhoven D, Meyfroidt G, Vlieghe E, Hens N, et al. Time between Symptom Onset, Hospitalisation and Recovery or Death: Statistical Analysis of Belgian COVID-19 Patients. Int J Environ Res Public Health. 2020;17. https://doi.org/10.3390/ijerph17207560 PMID: 33080869 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 21 / 22 PLOS COMPUTATIONAL BIOLOGYDisproportionate impacts of COVID-19 in a large US city 96. Wang X, Pasco RF, Du Z, Petty M, Fox SJ, Galvani AP, et al. Impact of Social Distancing Measures on Coronavirus Disease Healthcare Demand, Central Texas, USA. Emerg Infect Dis. 2020; 26: 2361– 2369. https://doi.org/10.3201/eid2610.201702 PMID: 32692648 97. Stokes EK, Zambrano LD, Anderson KN, Marder EP, Raz KM, El Burai Felix S, et al. Coronavirus Dis- ease 2019 Case Surveillance—United States, January 22-May 30, 2020. MMWR Morb Mortal Wkly Rep. 2020; 69: 759–765. https://doi.org/10.15585/mmwr.mm6924e2 PMID: 32555134 98. Plummer M.< rjags: Bayesian graphical models using MCMC. R package version 3–10. Statistical software. 99. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Founda- tion for Statistical Computing; 2020. Available: https://www.R-project.org/ 100. covid-19-data: An ongoing repository of data on coronavirus cases and deaths in the U.S. Github; Available: https://github.com/nytimes/covid-19-data 101. City of Austin. [cited 3 Jan 2022]. Available: https://atc-covid19data-austin.hub.arcgis.com/ 102. HUD USPS ZIP code crosswalk files. [cited 22 Jun 2022]. Available: https://www.huduser.gov/portal/ datasets/usps_crosswalk.html 103. 104. 105. Zottarelli LK, Sharif HO, Xu X, Sunil TS. Effects of social vulnerability and heat index on emergency medical service incidents in San Antonio, Texas, in 2018. J Epidemiol Community Health. 2021; 75: 271–276. https://doi.org/10.1136/jech-2019-213256 PMID: 33055179 Linear Mixed-Effects Models using “Eigen” and S4 [R package lme4 version 1.1–29]. 2022 [cited 3 May 2022]. Available: https://cran.r-project.org/web/packages/lme4/index.html Fox SJ, Lachmann M, Tec M, Pasco R, Woody S, Du Z, et al. Real-time pandemic surveillance using hospital admissions and mobility data. Proc Natl Acad Sci U S A. 2022;119. https://doi.org/10.1073/ pnas.2111870119 PMID: 35105729 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011149 June 1, 2023 22 / 22 PLOS COMPUTATIONAL BIOLOGY
10.1186_s13567-019-0627-1	Li et al. Vet Res (2019) 50:9 https://doi.org/10.1186/s13567-019-0627-1 RESEARCH ARTICLE Open Access Role of p53 in pseudorabies virus replication, pathogenicity, and host immune responses Xun Li1†, Wei Zhang2†, Yunjia Liu1, Jiaxun Xie1, Chuanhuo Hu1 and Xiaoye Wang1* Abstract As a key cellular transcription factor that plays a central role in cellular responses to a broad range of stress factors, p53 has generally been considered as a host cell restriction factor for various viral infections. However, the defined roles of p53 in pseudorabies virus (PRV) replication, pathogenesis, and host responses remain unclear. In the present study, we initially constructed a p53 overexpressing a porcine kidney epithelial cell line (PK-15) to detect the effect of p53 on PRV replication in vitro. The results show that viral glycoprotein B (gB) gene copies and the titers of virus were sig- nificantly higher in p53 overexpressing PK-15 cells than in PK-15 and p53 inhibitor treated p53 overexpressing PK-15 cells. A similar result was also found in the p53 inhibitor PFT-α-treated PK-15 cells. We then examined the effects of p53 on PRV infection in vivo by using p53-knockout (p53− −) mice. The results show that p53 knockout not only led to significantly reduced rates of mortality but also to reduced viral replication and development of viral encephalitis in the brains of mice following intracranial inoculation. Furthermore, we examined the effect of p53 knockout on the expression of the reported host cell regulators of PRV replication in the brains of mice by using RNA sequencing. The results show that p53 knockout downregulated the interferon (IFN) regulator genes, chemokine genes, and antiviral genes after PRV infection. This finding suggests that p53 positively regulates viral replication and pathogenesis both in vitro and in vivo. These findings offer novel targets of intrinsic host cell immunity for PRV infection. / Introduction Pseudorabies virus (PRV) belongs to the genus Varicell- ovirus in the subfamily Alphaherpesvirinae and it is the pathogen that causes porcine Aujeszky’s disease (AD) [1]. PRV causes nervous and respiratory system disorders in newborn piglets and reproductive failure in sows [2]. The virus has a broad host range and can infect most mam- mals; however, pigs are the natural reservoir and the only animal that can survive PRV infection [1]. The clinical manifestations of other animals infected by PRV are fatal and acute, and accompanied by extreme itching. The tumor suppressor protein p53 is a major host cel- lular response protein to a broad range of stress factors such as viral infection through its modulation of cel- lular pathways, including innate immune control, host cell cycling, proliferation, DNA repair, and apoptosis Correspondence: xywang@gxu.edu.cn †Xun Li and Wei Zhang contributed equally to this work 1 College of Animal Science and Technology, Guangxi University, Nanning 530004, Guangxi, People’s Republic of China Full list of author information is available at the end of the article [3–5]. Viral infection is a type of cellular stress that acti- vates p53 response that triggers apoptosis of the infected cells, leading to the suppression of viral replication [6–8]. Thus, p53 is considered as a host restriction factor in a range of viral infections. However, p53 appears to have both positive and negative effects on various viral infec- tions. The replication of various viruses is enhanced by the knockout or knockdown of p53 and inhibited by the overexpression of p53. Examples of such viruses include hepatitis C virus (HCV), influenza A virus (IAV), Japa- nese encephalitis virus (JEV), and vesicular stomatitis virus (VSV) [6, 8, 9]. Furthermore, many viruses have acquired a variety of distinct mechanisms to counter- act the negative effects of p53 in infected cells [6]. Con- versely, p53 is required for efficient viral replication of other viruses. p53 knockdown impairs the replication of herpes simplex virus 1 (HSV-1) and the associated viral pathogenesis of the central nervous system (CNS) of mice [10, 11]. It was also reported that the replications of human cytomegalovirus (HCMV) and porcine circovirus type 2 were impaired by p53 knockdown [12, 13]. © The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/ publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Li et al. Vet Res (2019) 50:9 Page 2 of 12 Collectively, the studies described above indicate that p53 is a critical host restriction factor for a variety of viruses. However, there are no studies that have exam- ined the effects of p53 on PRV infection. The biological significance of p53 in PRV replication, pathogenicity, and host immune responses remains to be elucidated. In the present study, we investigated the role(s) of p53 in PRV replication by using p53 overexpressing PK-15 cells and the p53 inhibitor PFT-α-treated PK-15 cells in vitro. Fur- thermore, we investigated the effects of p53 on the rep- lication and pathogenesis of PRV in vivo by using p53 knockout mice. The primary aim of this study was to elucidate the underlying mechanisms responsible for the role of p53 involvement in the replication and pathogen- esis of PRV and to offer novel targets of intrinsic host cell immunity for PRV infection. Materials and methods Cells, mice, and virus The PK-15 cell line was purchased from the American Type Culture Collection (ATCC, Rockville, MD, USA). The cells were grown in Dulbecco’s Modified Eagle’s medium (DMEM, Wisent) supplemented with 10% fetal bovine serum (FBS, Invitrogen, Carlsbad, CA, USA) and 80 μg of gentamycin/mL at 37 °C in a humidified atmos- phere of 5% CO2. and cloned (pCDH-MCS-GFP-Puro) A p53 overexpressing PK-15 cell line (PK-15 pCDH- p53) was constructed and maintained in our labora- tory. The p53 gene (GenBank No.AF098607.1) was into pCDH-CMV-MCS- synthesized EF1-CopGFP-T2A-Puro to generate the recombinant lentiviral vector pCDH-CMV- p53-EF1-CopGFP-T2A-Puro (pCDH-p53-GFP-Puro). The insertion fragment was identified by polymerase chain reaction (PCR), restriction endonuclease analysis, and DNA sequencing. The plasmid lentiviral vector sys- tem was transfected into PK-15 cells with Lipofectin 2000 reagent for packaging into mature lentivirus. Polybrena was used to screen stably expressing p53 PK-15 cells. An empty lentiviral vector was also transfected into PK-15 cells as the negative control (PK-15 pCDH). After 48 h of transfection, the cells were collected and stored at −80 °C until the expression level of p53 was measured by real- time PCR and Western blot. p53 knockout (p53−/−) mice and wild-type mice litter- mates were obtained by interbreeding heterozygous mice (stock number B-EM-020), provided by the Beijing Bio- cytogen Co., Ltd. The PRV strain Bartha K61 was isolated and purified from vaccine (Jiangsu Nannong Gaoke Animal Phar- maceutical Co., Ltd.) and maintained in our laboratory. PK-15 was used for the propagation of PRV. Cell studies PK-15 p53+/+ cells were infected with 0.1 multiplicity of infection (MOI) PRV. At 8, 12, 24, and 48 h post- infection, the cells were harvested to detect the viral titer, and glycoprotein B (gB) mRNA expression of PRV utilized real-time PCR and 50% tissue culture infective dose (TCID50), respectively. PK-15 cells were treated with PFT-α (p53 inhibi- tor) and then infected with 0.1 MOI of PRV. The cells were also harvested to detect the gB mRNA expres- sion of PRV by real-time PCR at 8, 12, 24, and 48 h post-infection. Animal studies All the animal experiments and the protocols used in this study were approved by the Research Ethics Com- mittee of the College of Animal Science, Guangxi Uni- versity, Guangxi, China. The animals were maintained under constant conditions of light (12 h of light) and temperature (22 °C) and housed in groups of four mice per cage until the beginning of the tests. The mice, had free access to pelleted food and tap water. For intrac- ranial infection, 4- to 8-week-old male p53−/− mice (n = 40) and wild-type mice (n = 40) were injected intracranially with 105 TCID50 of PRV (Bartha). Fif- teen mice among the total of either p53−/− or wild- type mice were monitored daily, and the mortality was recorded from 1 to 14 days post-infection. Twenty-five mice among the total p53−/− or wild-type mice were killed at 6 days after infection by decapitation under a mild dose of anesthetic ether within 30 min. The brains were excised and cleaned. Twelve brains of p53−/− or wild-type mice were stored at −80 °C to detect the host cellular gene expression by RNA sequencing. Six brains of p53−/− or wild-type mice were stored at −80 °C to detect the viral titer as well as the expression of the gB gene and inflammatory factors by using TCID50 and real-time PCR, respectively. Blood serum was also collected and kept at 4 °C to detect the inflammatory factor through specific antibodies by enzyme-linked immunosorbent assay (ELISA). The other 7 brains of p53−/− or wild-type mice were fixed in neutral-buffered formalin for histological analysis. Histopathology For histological analysis, the brains were dehydrated in an ethanol series and embedded in paraffin wax, followed by histologic sectioning (5 μm) and routine hematoxylin–eosin staining. To describe the histopa- thology in the brain in the p53−/− and p53+/+ mice, the pathological changes were determined through observation of the morphologic characteristics. For Li et al. Vet Res (2019) 50:9 Page 3 of 12 this purpose, 5 slides were selected from 1 sample, and 2 sections from each slide were examined microscopi- cally (magnification: ×400). Table 1 Primers and annealing temperature for real-time PCR. Genes Primer sequence (5′‑3′) Annealing (°C) RNA extraction and real‑time reverse transcription PCR (RT‑PCR) Total RNA from the brains or cells was extracted using the TRIzol extraction method (TRIzol reagent, TaKaRa Japan). The procedures for RNA isolation and purifi- cation, as well as the on-column deoxyribonuclease treatment (Qiagen), were performed according to the manufacturer’s instructions. For each sample, equal amounts of all the RNA samples were reverse transcribed simultaneously using an oligo (deoxythymidine) 15 primer and M-MLV reverse transcriptase (TaKaRa Japan) according to the manufacturer’s instructions. All RT reactions were performed at 42 °C and included a nega- tive control, which contained nuclease-free water instead of RNA. The SYBR Green Quantitative Real-Time PCR Master Mix (Roche, Mannheim, Germany) was used to quantify or relatively quantify the abundance of the tar- get mRNA according to the manufacturer’s instructions, and the accumulated fluorescence was detected using a real-time PCR detection system (Prism 7300, Applied Biosystems Inc., Foster City, USA). The primer sequences are shown in Table 1. β-Actin served as the endogenous control. The real-time PCR amplification conditions were as follows: initial denaturation at 95 °C for 10 min, fol- lowed by 40 cycles of 95 °C for 20 s, annealing for 30 s at 59–61 °C, and then extension at 72 °C for 30 s. To quantify the viral load, the PRV viral gB mRNA level was determined by absolute quantification real-time PCR. The gB gene was constructed from the pMD-19T vec- tor (Promega, Madison, WI, USA) and termed pMD- gB. A standard graph of the CT values was constructed from a tenfold serial dilution of the pMD-gB. The CT val- ues from the test samples were plotted on the standard curve, and the copy number was calculated automatically by Sequence Detector version 1.6 (PE Applied Biosys- tems), a software package for data analysis. Each sample was tested in duplicate, and the mean of the two values was considered as the copy number of the sample. The expression levels of the other target genes were meas- ured by relative quantitative real-time PCR. The data for each sample were calculated using the 2−ΔΔCT method as described previously. RNA sequencing For RNA sequencing analysis, equal amounts of the total RNA from either the p53−/− or wild-type mice (n = 12) were pooled into one sample. The procedure used for RNA sequencing has been described previ- ously. Briefly, sequencing libraries were constructed gB IL-6 TNF-α Oasl2 Ifi44 Usp18 Ifit1 Ly6a Gbp10 Gbp4 Gbp3 Gbp7 Stat1 Ddx58 Ifih1 Ifitm3 Mx1 Ifi44 Ccl2 Cxcl10 β-actin (F)CGG CAT CGC CAA CTT CTT C (R)GTC CTC CTT GAG CGT CTT CGT (F)AGT CCG GAG AGG AGA CTT CA (R)ATT TCC ACG ATT TCC CAG AG (F)GGG ACA GTG ACC TGG ACT GT (R)GCT CCA GTG AAT TCG GAA AG (F)ACA ATT TCC AAA ACG AGG TC (R)TTC CCA TCC CTT TCT TCT TC (F)GAC AGA TAC CAG TTC GAT TC (R)TTT TCT TGA TCT TTG CCA CC (F)CAA GGA ACA GTC TGA AAT ACAC (R)CAC AGT AAT GAC CAA AGT CAG (F)AGA ACA GCT ACC ACC TTT AC (R)TTC TTG ATG TCA AGG AAC TG (F)GAG AGG AAG TTT TAT CTG TGC (R)TCT CAA ATG GGA CTC CAT AG (F)CTA ACC GGA AGT GTT TTG TC (R)CAG AAT CCC TAG TTT ATT CCC (F)AGC TAA CGA AGG AAC AAA AG (R)GAT GTT ATG TCC CAG TTG ATG (F)CTG TTC GAG ATT TTG CTC TG (R)TGG ACT TTG AGA TTG TCT CC (F)GAG TGA AGG CAA ATC ATG TC (R)CTG TTT CTG TCT TAG TAG CTC (F)TTT GAC AGT ATG ATG AGC AC (R)AGC AAA TGT GAT GCT CTT TC (F)GAG AGT CAC GGG ACC CAC T (R)CGG TCT TAG CAT CTC CAA CG (F)TGA TGC ACT ATT CAA GAA CTA ACA (R)TCT GTG AGA CGA GTT AGC CAAG (F)TCA TCA TTG TTC TTA ACG CTCA (R)CGG AAG TCG GAA TCC TCT AT (F)GAA GGC AAG GTC TTG GAT G (R)GCT GAC CTC TGC ACT TGA CT (F)TTC AAC TCA GTG GAA GTC TGCT (R)GGA GTG TTT CCC CGC TTT TTC (F)TCT GTG CTG ACC CCA AGA AGG (R)TGG TTG TGG AAA AGG TAG TGGAT (F)TCC CTC TCG CAA GGAC (R)TTG GCT AAA CGC TTT CAT (F)AGG TGA CAG CAT TGC TTC TG (R)GCT GCC TCA ACA CCT CAA C 61 60 61 59 60 60 61 60 59 60 60 60 61 61 59 60 60 61 59 59 60 with the SureSelect Strand-Specific RNA library (Agi- lent); 100-bp paired-end sequencing was performed using an Illumina HiSeq 2500 sequencer according to the manufacturer’s instructions. The raw sequence reads were mapped to the mouse genome by using the TopHat program. The normalized transcription pro- files were estimated on the basis of the mapping results using the Cufflinks program. The number of reads per kilobase of exon per mil- lion mapped reads (RPKM) was converted from the Li et al. Vet Res (2019) 50:9 Page 4 of 12 row read counts of each transcript using the program Cuffdiff. Elisa Levels of the cytokines interleukin (IL)-6 and tumor necrosis factor (TNF)-α were measured by ELISA using the mouse IL-6 and TNF-α kits (Biosource). Blood samples were collected, and serum was separated. The serum was then added to wells coated with monoclonal antibodies against IL-6 and TNF-α. After 3 washes with washing buffer (0.05% Tween-20 in phosphate-buffered saline, PBS), peroxidase-conjugated avidin, biotinylated antibodies against IL-6 and TNF-α, and chromogenic substrates were added to each well. The absorbance was read at 450 nm in an ELISA plate reader. Virus infection and titration PRV was propagated in PK-15 cells in DMEM supple- mented with 10% FBS until harvested. Then, viral stock titers were measured using the plaque-forming unit (PFU) assay, and TCID50 was calculated using the Reed- Muench method. Statistical analysis All the data are shown as mean ± S.E.M. The differences were considered to be significant when P < 0.05. Statisti- cal analysis was performed using one-way analysis of var- iance (ANOVA) with SPSS 17.0. Results p53 facilitates PRV viral replication in vitro gB is the most conserved envelope glycoprotein across the herpesvirus family. Because of its indispensable role in fusion during virus entry and cell-to-cell viral spread, it is required for viral replication. To investi- gate the effect of p53 on PRV replication in vitro, we estimated the viral load and virus yield by quantifica- tion real-time PCR and TCID50 assays in PFT-α (p53 inhibitor) treated PK-15 pCDH-p53 and PK-15 pCDH- p53 cells compared with PK-15 pCDH. As shown in Figure 1A, the viral gB gene copies gradually increased from 4 h to 24 h in PFT-α treated PK-15 pCDH-p53, PK-15 pCDH-p53 and PK-15 pCDH cells, but the viral load in PK-15 pCDH-p53 cells exceeded that in PK-15 pCDH and PFT-α treated PK-15 pCDH-p53 cells in 8 h, 12 h, and 24 h. The viral gB gene copies show no significant difference between PFT-α treated PK-15 pCDH-p53 and PK-15 pCDH cells. The results of viral titers were consistent with the viral load test (Figure 1B). To further confirm the promotive effect of p53 on PRV viral replication, PK-15 cells treated with PFT-α were also used to estimate the virus yield after Figure 1 Effect of p53 on the PRV replication in vitro. A The viral gB gene copies were detected in PFT-α (p53 inhibitor) treated PK-15 pCDH-p53, PK-15 pCDH-p53 and PK-15 pCDH cells after 0.1 MO PRV infection in 4, 8, 12, and 24 h by quantitative real-time PCR. B Titers of PRV were determined in PFT-α treated PK-15 pCDH-p53, PK-15 pCDH-p53 and PK-15 pCDH cells after 0.1 MO PRV infection in 4, 8, 12, and 24 h by TCID50. C The viral gB gene copies were detected in PFT-α treated PK-15 and mock PK-15 cells after 0.1 MO PRV infection in 4, 8, 12, and 24 h by quantitative real-time PCR. D Titers of PRV were determined in PFT-α treated PK-15 and mock PK-15 cells after 0.1 MO PRV infection in 4, 8, 12, and 24 h by TCID50. All the results were confirmed by three independent experiments. Error bars represent the standard deviations of triplicate experiments. P < 0.05; *P < 0.01. PRV infection compared with mock PK-15 cells. The results show that PFT-α-treated PK-15 cells exhibited significantly lower viral gB gene copies (Figure 1C) and viral titers (Figure 1D) than mock PK-15 cells. These results indicate that p53 facilitates PRV viral Li et al. Vet Res (2019) 50:9 Page 5 of 12 replication by increasing gB expression and PRV prog- eny yields in vitro. p53 promoted the PRV replication in mice brain in vivo Because PRV infection mainly causes neurological symp- toms, in order to investigate the role of p53 in PRV rep- lication in vivo, we detected the mortality, viral load, and viral yield in the brains of wild-type and p53−/− mice. The survival of the infected mice was monitored for 14 days post-infection. As shown in Figure 2A, the survival rate of the p53−/− mice was 100%, while the survival rate of wild-type mice was only 53%, suggesting that p53−/− mice are susceptible to PRV infection. Another group of wild-type and p53−/− mice were infected as described above. The virus titers and viral gB gene in the brains were assayed at 6 days post-infection. The viral titers in the brains of the wild-type mice were significantly higher than those in the brains of the p53−/− mice (Figure 2B). The PRV gB gene copy numbers of the wild-type mice were nearly tenfold that of the p53−/− mice, which was consistent with the viral titers (Figure 2C). The above results indicate that knockout of p53 inhibited PRV rep- lication in mouse brains; this implies that p53 promoted PRV replication in vivo. In summary, these findings fur- ther reveal that p53 facilitates PRV viral replication both in vitro and in vivo. p53 promoted PRV infection pathogenicity in mice Encephalitis caused by PRV infection in the CNS is a key factor contributing to animal death. To detect the degree of encephalitis in wild-type and p53−/− mice, we histopathologically analyzed the brains at 6 days post- infection. As shown in Figure 3, the brains of the wild- type mice show a significant increase in the number of necrotic neurons (Figure 3I), and more glial cells were accumulated around the degenerate neurons (glial nod- ules) than those in p53−/− mice (Figure 3D). Additionally, neuronophagia was found in the brains of the wild-type mice (Figure 3I). In the cerebellum of the wild-type mice, a large number of Purkinje cells show shrinkage or necro- sis (Figure 3K) and more obvious nuclear disintegration (Figure 3K) than that in the cerebellum of p53−/− mice (Figure 3E). Furthermore, perivascular cuffing (Fig- ure 3O) and hyperemia (Figure 3M) were more obvious in the brains of wild-type mice. We further determined IL-6 and TNF-α mRNA and protein levels in the brains of the wild-type and p53−/− mice by relative quantitative real-time PCR and ELISA. The results show that both IL-6 and TNF-α mRNA expression levels in the brains of the wild-type mice were significantly higher (80-fold and eightfold, respectively) than those in the p53−/− mice (Figures 4A and B). A / / = = 15) and wild-type mice − and wild-type mice was − mice (n 15) were inoculated with 105 TCID50 of PRV (Bartha) intracranially. Figure 2 Effect of p53 on PRV replication in mice brains in vivo. Four- to 8-week-old male p53− (n A The survival of the infected p53− monitored for 14 days post-infection. Statistical significance was determined by the log-rank test. B, C At 5 days post-infection, the brains of the infected mice were harvested. Viral gB gene copies and virus titers in the brains of the p53− − and wild-type mice were assayed by quantitative real-time PCR and TCID50, respectively. Each data point is the viral titer or viral gB gene copies in the brain of one mouse. The horizontal bars indicate the mean for each group. P < 0.05; *P < 0.01. / Li et al. Vet Res (2019) 50:9 Page 6 of 12 ▸ / / − = = − (n 7) and wild-type (n Figure 3 Histopathological features of the brains of the p53− and wild‑type mice following intracranial inoculation. Four- to 8-week-old male p53− 7) mice were inoculated with 105 TCID50 of PRV (Bartha) intracranially. At 6 days post-infection, the brains of the infected mice were harvested, sectioned, and stained with hematoxylin and eosin. B, C Magnified images of the regions indicated with black rectangles in A. G, H Magnified images of the regions indicated with black rectangles in F. D, E, I, K, M, and O Magnified images of the regions indicated with black rectangles in B, C, G, H, L, and N, respectively. Representative images are shown. N: neurons; G; glial cells; NN: necrotic neurons; NP: neuronophagia; P: Purkinje cells; PS: shrinkage of Purkinje cells; PN: necrotic Purkinje cells; ND: nuclear disintegration of Purkinje cells; H: hyperemia; and PC: perivascular cuffing. similar result was also found in protein levels. The pro- tein levels of IL-6 and TNF-α were significantly higher in the brains of the wild-type mice than those in the p53−/− mice (Figures 4C and D); this finding suggests that the wild-type mice experienced a more severe inflammatory response than the p53−/− mice. These results indicate that p53 promotes the pathogenicity of PRV infection in vivo. p53 knockout suppresses the inflammatory response in the brains of infected mice To investigate the effects of p53 knockout on the host cell gene expression in the brains of the infected mice, we performed a whole-transcriptome shotgun sequencing analysis of the brains from the infected p53−/− and wild- type mice following intracranial infection. In the expres- sion profiles, 122 mRNA were differentially expressed, comprising 104 downregulated and 18 upregulated mRNA (Figure 5A). A functional annotation analysis of the target genes of the differentially expressed mRNA was performed to identify the gene ontology (GO) terms and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. As shown in Figure 5B, most of the enriched GO terms of the differentially expressed mRNA were involved in GTPase activity, GTP binding, and guanyl ribonucleotide binding molecular functions as well as the immune response and immune system process. KEGG pathway enrichment analysis indicates that the differ- entially expressed genes (DEG) tended to be involved in herpes simplex infection, ribosome and antigen process- ing and presentation as well as TNF signaling and Toll- like receptor signaling pathways (Figure 5C). Furthermore, the downregulated genes of the DEG were further grouped according to their functions. The results show that compared to the wild-type mice, the major downregulated genes included the IFN-related genes, chemokines, and antiviral genes. As shown in Li et al. Vet Res (2019) 50:9 Page 7 of 12 − and wild‑type mice following intracranial Figure 4 IL‑6 and TNF‑α mRNA and antibody levels were detected in the brains of the p53− 6) mice were inoculated with 105 TCID50 of PRV (Bartha) intracranially. At inoculation. Four- to 8-week-old male p53− = 6 days post-infection, the brains of the infected mice were harvested (A–D). IL-6 and TNF-α mRNA and antibody levels were detected by relative quantitative real-time PCR and ELISA. Each data point is the level of IL-6 and TNF-α mRNA or antibodies in the brain of one mouse. The horizontal bars indicate the mean for each group. P < 0.05; *P < 0.01. 6) and wild-type (n − (n = / / Table 2, the levels of IFN signal pathway-related fac- tors 1 and 7 (IRF1 and IRF7) in the brains of the p53−/− mice were reduced 8.0- and 25.4-fold, respectively, compared to those in the brains of wild-type mice. Further, compared with those in the wild-type mice, the IFN-induced protein family genes (Ifit1, Ifit2, Ifit3, Ifit3b, Ifitm3, Ifip1, Ifih1, Ifi44, and Ifi2712a) in the p53−/− mice were significantly downregulated, with a fold change ranging from 5.7 to 38.2. Similarly, the guanylate-binding protein family genes, which belong to another IFN-induced family, including Gbp 2, Gbp 3, Gbp 4, Gbp 5, Gbp 6, Gbp 7, Gbp 8, Gbp 9, and Gbp 10, were decreased, with a fold change ranging from 6.4 to 38.2. Chemokines are considered to be pro-inflam- matory agents and can be induced during an immune response to recruit cells of the immune system to the site of infection. As shown in Table 3, the chemokine gene family, including Ccl2, Ccl7, Cxcl9, and Cxcl10, Li et al. Vet Res (2019) 50:9 Page 8 of 12 / Figure 5 GO terms and KEGG pathway enrichment analysis of the DEG mRNA. A The volcano plot for the DEG between the brains of p53− − and wild-type mice; the x-axis indicates the log fold change and the y-axis indicate the -log (P value). B The most enriched terms for the differentially expressed mRNA are displayed here. The enrichment score of the GO term equals the -log10 (P-value). The listed GO items are categorized as biological processes or molecular functions. C The rich factor plot of the KEGG pathway enrichment analysis results. The degree of the color stands for the P-value; the size of the node stands for the gene count in this item. Li et al. Vet Res (2019) 50:9 Table 2 Differential expressed of IFN-related genes in brains of PRV-infected p53− / − and wild-type mice. Gene ID Gene name Gene description 16362 54123 15957 15958 15959 667370 66141 60440 71586 99899 76933 14469 55932 17472 229898 100702 229900 236573 626578 Irf1 Irf7 Ifit1 Ifit2 Ifit3 Ifit3b Ifitm3 Iigp1 Ifih1 Ifi44 Ifi27l2a Gbp2 Gbp3 Gbp4 Gbp5 Gbp6 Gbp7 Gbp9 Gbp10 Interferon regulatory factor 1 Interferon regulatory factor 7 Interferon-induced protein with tetratricopeptide repeats 1 Interferon-induced protein with tetratricopeptide repeats 2 Interferon-induced protein with tetratricopeptide repeats 3 Interferon-induced protein with tetratricopeptide repeats 3b Interferon induced transmembrane protein 3 Interferon inducible GTPase 1 Interferon induced with helicase C domain 1 Interferon-induced protein 44 Interferon alpha-inducible protein 27 like 2A Guanylate binding protein 2 Guanylate binding protein 3 Guanylate binding protein 4 Guanylate binding protein 5 Guanylate binding protein 6 Guanylate binding protein 7 Guanylate binding protein 9 Guanylate binding protein 10 a Fold activation represents the fold increase in the level of activation in p53− / − mice compared with the level of activation in wild-type mice. Page 9 of 12 Foldchangea / (p53− − vs wild‑type) 8.0 − 25.4 − − − − − 19.6 12.4 12.9 10.4 − − 5.7 − 37.7 7.5 − 17.6 9.5 − 38.2 − − − − − 13.3 31.7 26.8 12.0 6.4 − 8.0 − 24.6 − was found to be downregulated, with a fold change ranging from 44.2 to 133.9. As shown in Table 4, the antiviral genes of the p53−/− mice related to the IFN pathways were downregulated, with fold changes rang- ing from 3.0- to 76.9-fold; for example, the expression of genes encoding virus entry inhibitors such as Mx1 reduced 76.9-fold and that of genes encoding virus translation and replication inhibitors such as Oasl1 and Oasl2 was reduced 68.0- and 9.3-fold, respectively. To validate the data obtained by RNA sequencing analysis, 16 downregulated genes were selected and their expression was detected by relative quantita- tive real-time PCR. As shown in Figure 6, the expres- sion levels of the 16 genes (Oasl2, Ddx58, Mx1, Stat1, Gbp3, Gbp4, Gbp7,Gbp10, Ifih1, lfitm3, Ifi44, Ifit1, Ccl2, Cxcl10, Ly6a, Usp18) in the p53−/− mice were sig- nificantly decreased compared to those in the wild-type mice. These results were consistent with the findings of RNA sequencing analysis. Discussion PRV causes encephalitis that can result in severe neuro- logical defects and death in swine. Many of the host cell factors involved in the regulation of PRV infection have been investigated. However, most of these factors are immunological regulators and function through immu- nological pathways to restrict PRV infection. These Table 3 Differential expressed chemokines in brains of PRV-infected p53− − and wild-type mice. / Gene ID Gene name Gene description Foldchangea / (p53− − and wild‑ type) 20296 20306 17329 Ccl2 Ccl7 Cxcl9 Chemokine (C–C motif ) ligand 2 Chemokine (C–C motif ) ligand 7 Chemokine (C-X-C motif ) ligand 9 65.7 42.2 75.1 − − − 15945 Cxcl10 Chemokine (C-X-C motif ) ligand 10 133.9 − a Fold activation represents the fold increase in the level of activation in p53− mice compared with the level of activation in wild-type mice. / − factors therefore provide limited information on the intrinsic host cell regulators that may be involved in the facilitation of PRV infection. Here, we demonstrate that a host cell protein, p53, which has generally been con- sidered as a host cell restriction factor for various viral infections, is required for efficient PRV replication and pathogenesis in mice. This is the first report showing that p53 positively regulates PRV replication and patho- genesis in vitro and in vivo and provides insights into the molecular mechanism of p53. Li et al. Vet Res (2019) 50:9 Table 4 Differential expressed anti-viral genes in brains of PRV-infected p53− / − and wild-type mice. Gene ID Gene name Gene description 20846 53817 230073 20210 110454 100041546 231655 23962 17857 58185 Stat1 Ddx39b Ddx58 Saa3 Ly6a Ly6c2 Oasl1 Oasl2 Mx1 Rsad2 Signal transducer and activator of transcription 1 DEAD (Asp-Glu-Ala-Asp) box polypeptide 39B DEAD (Asp-Glu-Ala-Asp) box polypeptide 58 Serum amyloid A 3 Lymphocyte antigen 6 complex locus A Lymphocyte antigen 6 complex locus C2 2′-5′ oligoadenylate synthetase-like 1 2′-5′ oligoadenylate synthetase-like 2 MX dynamin-like GTPase 1 Radical S-adenosyl methionine domain containing 2 a Fold activation represents the fold increase in the level of activation in p53− / − mice compared with the level of activation in wild-type mice. Page 10 of 12 Foldchangea / (p53− − and wild‑ type) − − − − 3.0 − 14.8 5.6 − 44.0 3.8 − 23.0 68.0 9.3 − 76.9 37.1 − − However, in DNA virus infection, viral genome replica- tion induces host DNA damage responses (DDRs), which activate apoptotic p53 responses [6, 10]. In agreement with these observations, we found that p53 overexpres- sion promoted PRV replication, whereas inhibiting p53 by a specific inhibitor reduced PRV replication in PK-15 cells in vitro. p53−/− mice were also not susceptible to PRV infection compared to wild-type mice in vivo. Fol- lowing intracranial inoculation, p53 knockout reduced viral replication in the brains of mice and led to signifi- cantly reduced rates of mortality. However, PRV titers and viral gene copy numbers in the brain as well as mor- tality after intracranial inoculation appear to be surpris- ingly low. The possible reason is that the attenuated PRV vaccine strain (PRV-Bartha) which was used for intracra- nial inoculation in our studies. Brittle et al. [17] reported that mice infected with an attenuated PRV vaccine strain (PRV-Bartha) survive approximately three times longer than virulent (e.g., PRV-Becker, PRVKaplan, or PRV- NIA3) strain-infected mice. Furthermore, they indicated that the absence of the genes encoding US9, gE and gI in PRV-Bartha accounts for much of its attenuation. The absence of US9, gE and gI genes led to the time delay of PRV-Bartha entering various tissue types. In addition, the amount of infectious PRV-Bartha in the brainstem was lower relative to virulent PRV at the equivalent time point. Taken together, these results suggest that p53 has a positive effect on PRV replication both in vitro and in vivo. It has previously been reported that PRV invades and spreads within the trigeminal pathway (the nasal mucosa, the trigeminal ganglion, the pons/medulla, and the cerebellum/thalamus) of neonatal pigs [18, 19]. PRV induces encephalitis in both pigs and mice with similar Figure 6 The expression levels of the 16 downregulated genes of the p53‑/‑ mice were detected by relative quantitative real‑time PCR. Oasl2, Ddx58, Mx1, Stat1, Gbp3, Gbp4, Gbp7, Gbp10, Ifih1, lfitm3, Ifi44, Ifit1, Ccl2, Cxcl10, Ly6a and Usp18 were detected by relative quantitative real-time PCR to validate the data obtained by RNA sequencing analysis. P < 0.05; **P < 0.01. Animals usually have many defense strategies against viral infection; however, viruses have evolved complex tactics to override these lines of defense, for instance, by using the host defensive proteins to promote viral replication [14–16]. p53 is a critical host restriction fac- tor because its regulation of the various cellular life pro- cesses, such as cell cycle arrest, apoptosis, and autophagy [6], has both positive and negative effects on various viral infections [10, 12]. On the basis of previous stud- ies, double-stranded RNA are produced in RNA virus infections, and these double-stranded RNA trigger anti- viral responses mediated by type I interferon (IFN-I) signaling, in which p53 appears to reduce the replication of some viruses; therefore, these viruses have acquired mechanisms to counteract p53 in infected cells [7, 8]. Li et al. Vet Res (2019) 50:9 Page 11 of 12 pathological signs [18, 20]. Although it was reported that the attenuated Bartha strain does not cause severe brain pathology despite viral replication and spread through- out the brain in chicken embryos [21], our histopathol- ogy results show much more serious encephalitis in the brains of wild-type mice than in those of the p53−/− mice. Compared to the p53−/− mice, wild-type mice show neu- ronal degeneration and necrosis in the brain. Notably, Purkinje cells in the cerebellum of wild-type mice show more obvious shrinkage or necrosis and nuclear disin- tegration than those in p53−/− mice. These histolpatho- logical changes provided morphological data to support our observation of serious neurological signs, including convulsion, ataxia, and abnormal behavior, in wild-type mice after PRV infection. Our findings coincided with the conclusion of previous studies that PRV infection impairs cerebellar development and differentiation [18, 20]. Furthermore, two cytokines associated with inflam- mation, namely IL-6 and TNF-α, in the brains of the p53−/− mice were also significantly lower than those in the brains of wild-type mice; this finding coincided with the development of encephalitis. These results indicate that p53 is required for efficient virulence and replication of the virus and for the consequent development of viral encephalitis in the brains of mice following intracranial inoculation. This requirement of p53 for efficient viral replication in the brains of mice is in agreement with the findings of our in vitro study using cell culture, as described above. To our knowledge, this is the first report showing that p53 plays a positive role in PRV replication and pathogenesis both in vitro and in vivo. We further investigated mRNA profiles in the PRV- infected p53−/− and wild-type mice by RNA sequenc- ing. The gene function analysis shows that differentially expressed RNA enrichment was observed in a number of the pathways in the host antiviral immune response and the inflammatory response processes. The DEG primarily enriched the interferon-related pathways, including interferon-regulated genes (IRG), interferon- inducible protein family genes, interferon-inducible guanylate-binding protein family genes (Gbp), and inter- feron-stimulated genes (ISG). Interferon is an important antiviral factor of the body [22, 23]. It has been reported that alpha/beta interferon receptor deficiency in mice significantly enhanced their susceptibility to PRV infec- tion [24]. In molecular biology, interferon-inducible pro- tein family genes, namely Gbps and ISG, are key factors for protective immunity against microbial and viral path- ogens [25–27]. The expression of all the above mentioned genes was significantly downregulated in the p53−/− mice compared with that in wild-type mice, suggesting that the PRV replication progress in the p53−/− mice was inhibited. Moreover, some of the cytokine ligands and anti-viral genes show significantly different expression between the p53−/− and wild-type mice. Notably, Cxcl10, Oasl1, and Ccl7 in the brain of p53−/− mice were reduced by 133-, 67-, and 42-folds, respectively, compared with those in wild-type mice. Previous studies have demon- strated that Cxcl10 and Ccl7 can offer protective immu- nity against PRV, while OASL1 deficiency promotes antiviral immunity against local mucosal viral infection [28–30]. These data corroborate our suggestion that p53 promotes PRV viral replication and pathogenesis in mice. In conclusion, this study provides evidence suggesting that p53 positively regulates viral replication and patho- genesis both in vitro and in vivo. We demonstrate that the host cell protein p53 can be a host cell restriction fac- tor for PRV infection. Our study offers novel therapeutic targets of intrinsic host cell immunity for PRV infection. Competing interests The authors declare that they have no competing interests. Authors’ contributions XW and YL designed and performed animal experiments for the study. JX performed cells experiments. WZ and YL analyzed the data and aided in data interpretation. XL and WZ discussed the results. XL, WZ and XW wrote the manuscript. CH provided laboratory materials. XW and WZ edited the manuscript and provided funding. All authors read and approved the final manuscript. Acknowledgements This work was supported by a grant from the National Natural Science Foun- dation of China (NSFC) (31502079), Natural Science Foundation of Guangxi Province (2017GXNSFAA198071), and Natural Science Foundation of the Jiangsu Higher Education Institutions of China (15KJB31004). Author details 1 College of Animal Science and Technology, Guangxi University, Nan- ning 530004, Guangxi, People’s Republic of China. 2 Department of Biochem- istry and Molecular Biology, School of Basic Medical Sciences, Nanjing Medical University, Nangjing 211166, People’s Republic of China. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. Received: 6 November 2018 Accepted: 3 January 2019 References 1. Pomeranz LE, Reynolds AE, Hengartner CJ (2005) Molecular biololgy of pseudorabies virus: impact on neurovirology and veterinary medicine. Microbiol Mol Biol Rev 69:462–500 2. Klupp BG, Hengartner CJ, Mettenleiter TC, Enquist LW (2004) Com- 3. 4. plete, annotated sequence of the pseudorabies virus genome. J Virol 78:424–440 Laptenko O, Prives C (2006) Transcriptional regulation by p53: one pro- tein, many possibilities. Cell Death Differ 13:951–961 Farnebo M, Bykov VJ, Wiman KG (2010) The p53 tumor suppressor: a master regulator of diverse cellular processes and therapeutic target in cancer. Biochem Biophys Res Commun 396:85–89 5. Kruse JP, Gu W (2009) Modes of p53 regulation. Cell 137:609–622 6. Sato Y, Tsurumi T (2013) Genome guardian p53 and viral infections. Rev Med Virol 23:213–220 Li et al. Vet Res (2019) 50:9 Page 12 of 12 7. Takaoka A, Hayakawa S, Yanai H, Stoiber D, Negishi H, Kikuchi H, Sasaki S, Imai K, Shibue T, Honda K (2003) Integration of interferon-alpha/beta signalling to p53 responses in tumour suppression and antiviral defence. Nature 424:516–523 8. Munoz-Fontela C, Macip S, Martinez-Sobrido L, Brown L, Ashour J, Garcia-Sastre A, Lee SW, Aaronson SA (2008) Transcriptional role of p53 in interferon-mediated antiviral immunity. J Exp Med 205:1929–1938 9. Pampin M, Simonin Y, Blondel B, Percherancier Y, Chelbi-Alix MK (2006) Cross talk between PML and p53 during poliovirus infection: implications for antiviral defense. J Virol 80:8582–8592 10. Maruzuru Y, Fujii H, Oyama M, Kozuka-Hata H, Kato A, Kawaguchi Y (2013) Roles of p53 in herpes simplex virus 1 replication. J Virol 87:9323–9332 11. Maruzuru Y, Koyanagi N, Takemura N, Uematsu S, Matsubara D, Suzuki Y, Arii J, Kato A, Kawaguchi Y (2016) p53 is a host cell regulator during herpes simplex encephalitis. J Virol 90:6738–6745 12. Casavant NC, Luo MH, Rosenke K, Winegardner T, Zurawska A, Fortunato EA (2006) Potential role for p53 in the permissive life cycle of human cytomegalovirus. J Virol 80:8390–8401 13. Xu D, Du Q, Han C, Wang Z, Zhang X, Wang T, Zhao X, Huang Y, Tong D (2016) p53 signaling modulation of cell cycle arrest and viral replication in porcine circovirus type 2 infection cells. Vet Res 47:120 14. Kumar R, Khandelwal N, Thachamvally R, Tripathi BN, Barua S, Kashyap SK, Maherchandani S, Kumar N (2018) Role of MAPK/MNK1 signaling in virus replication. Virus Res 253:48–61 15. Skalsky RL, Cullen BR (2010) Viruses, microRNAs, and host interactions. Annu Rev Microbiol 64:123–141 20. Yang QY, Sun Z, Tan FF, Guo LH, Wang YZ, Wang J, Wang ZY, Wang LL, Li XD, Xiao Y, Tian KG (2016) Pathogenicity of a currently circulating Chinese variant pseudorabies virus in pigs. World J Virol 5:23–30 21. Clase AC, Banfield BW (2003) Corticosteroids are unable to protect against pseudorabies virus-induced tissue damage in the developing brain. J Virol 77:4979–4984 22. Schneider WM, Chevillotte MD, Rice CM (2014) Interferon-stimulated genes: a complex web of host defenses. Annu Rev Immunol 32:513–545 23. Schoggins JW (2014) Interferon-stimulated genes: roles in viral pathogen- esis. Curr Opin Virol 6:40–46 24. Wei J, Ma Y, Wang L, Chi X, Yan R, Wang S, Li X, Chen X, Shao W, Chen JL (2017) Alpha/beta interferon receptor deficiency in mice significantly enhances susceptibility of the animals to pseudorabies virus infection. Vet Microbiol 203:234–244 25. Robertsen B (2008) Expression of interferon and interferon-induced genes in salmonids in response to virus infection, interferon-inducing compounds and vaccination. Fish Shellfish Immunol 25:351–357 26. Man SM, Place DE, Kuriakose T, Kanneganti T-D (2017) Interferon-induc- ible guanylate-binding proteins at the interface of cell-autonomous immunity and inflammasome activation. J Leukoc Biol 101:143–150 27. MacMicking JD (2012) Interferon-inducible effector mechanisms in cell- autonomous immunity. Nat Rev Immunol 12:367–382 28. Han YW, Aleyas AG, George JA, Kim SJ, Kim HK, Yoo DJ, Kang SH, Eo SK (2009) Genetic co-transfer of CCR7 ligands enhances immunity and pro- longs survival against virulent challenge of pseudorabies virus. Immunol Cell Biol 87:91–99 16. Walsh D, Mohr I (2011) Viral subversion of the host protein synthesis 29. Huang J, Ma G, Fu L, Jia H, Zhu M, Li X, Zhao S (2014) Pseudorabies machinery. Nat Rev Microbiol 9:860–875 17. Brittle EE, Reynolds AE, Enquist LW (2004) Two modes of pseudorabies virus neuroinvasion and lethality in mice. J Virol 78:12951–12963 18. Taharaguchi S, Kon Y, Yoshino S, Ono E (2003) Impaired development of the cerebellum in transgenic mice expressing the immediate-early protein IE180 of pseudorabies virus. Virology 307:243–254 19. Kritas SK, Pensaert MB (1994) Role of gp63 and gIII of Aujeszky’s disease virus in the invasion of the olfactory nervous pathway in neonatal pigs. Acta Vet Hung 42:309–316 viral replication is inhibited by a novel target of miR-21. Virology 456–457:319–328 30. Oh JE, Lee MS, Kim YJ, Lee HK (2016) OASL1 deficiency promotes antiviral protection against genital herpes simplex virus type 2 infection by enhancing type I interferon production. Sci Rep 6:19089
10.1371_journal.pcbi.1009023	RESEARCH ARTICLE Mechanistic model of nutrient uptake explains dichotomy between marine oligotrophic and copiotrophic bacteria Noele NorrisID 1,2,3, Naomi M. LevineID 2, Vicente I. Fernandez3, Roman Stocker3 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Norris N, Levine NM, Fernandez VI, Stocker R (2021) Mechanistic model of nutrient uptake explains dichotomy between marine oligotrophic and copiotrophic bacteria. PLoS Comput Biol 17(5): e1009023. https://doi.org/ 10.1371/journal.pcbi.1009023 Editor: Pedro Mendes, University of Connecticut School of Medicine, UNITED STATES Received: November 1, 2020 Accepted: April 28, 2021 Published: May 19, 2021 Copyright: This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication. Data Availability Statement: All relevant data are within the manuscript and its supporting information. The code is now available on GitHub: https://github.com/noelenorris/ABC_proteome_ allocation. Funding: This work was supported by a grant from the Simons Foundation (https://www. simonsfoundation.org; 542395 to R.S. and 542389 to N.L.), as part of the Principles of Microbial Ecosystems (PriME) Collaborative, and by a Gordon and Betty Moore Foundation Investigator 1 Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, United States of America, 2 Department of Biological Sciences, University of Southern California, Los Angeles, United States of America, 3 Institute of Environmental Engineering, Department of Civil, Environmental and Geomatic Engineering, ETH Zu¨ rich, Zu¨rich, Switzerland * noelen@alum.mit.edu (NN); romanstocker@ethz.ch (RS) Abstract Marine bacterial diversity is immense and believed to be driven in part by trade-offs in meta- bolic strategies. Here we consider heterotrophs that rely on organic carbon as an energy source and present a molecular-level model of cell metabolism that explains the dichotomy between copiotrophs—which dominate in carbon-rich environments—and oligotrophs— which dominate in carbon-poor environments—as the consequence of trade-offs between nutrient transport systems. While prototypical copiotrophs, like Vibrios, possess numerous phosphotransferase systems (PTS), prototypical oligotrophs, such as SAR11, lack PTS and rely on ATP-binding cassette (ABC) transporters, which use binding proteins. We develop models of both transport systems and use them in proteome allocation problems to predict the optimal nutrient uptake and metabolic strategy as a function of carbon availability. We derive a Michaelis–Menten approximation of ABC transport, analytically demonstrating how the half-saturation concentration is a function of binding protein abundance. We predict that oligotrophs can attain nanomolar half-saturation concentrations using binding proteins with only micromolar dissociation constants and while closely matching transport and metabolic capacities. However, our model predicts that this requires large periplasms and that the slow diffusion of the binding proteins limits uptake. Thus, binding proteins are critical for oli- gotrophic survival yet severely constrain growth rates. We propose that this trade-off funda- mentally shaped the divergent evolution of oligotrophs and copiotrophs. Author summary Marine bacteria utilize carbon as a building block and an energy source and thus exert an important control on the amount of carbon that is sequestered in the ocean versus respired into the atmosphere. They use a spectrum of strategies to consume carbon: while copiotrophic bacteria dominate in nutrient-rich environments, oligotrophic bacteria dominate in nutrient-poor environments and are typically smaller, nonmotile, and slower growing. Yet the paragon oligotroph SAR11 is the planet’s most abundant organism. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 1 / 21 Award (https://www.moore.org; GBMF3783 to R. S.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. Marine oligotrophs versus copiotrophs Despite this, most of our understanding of bacteria derives from research on copiotrophs. Here we use molecular-level models to understand how an oligotroph’s physiology enables it to outperform copiotrophs in nutrient-poor but not in nutrient-rich environ- ments. We contrast copiotrophs’ prevalent method of sugar transport with oligotrophs’ reliance on binding proteins, which trap nutrients in the periplasm. Binding proteins allow cells to attain affinities that are much higher than the transport proteins’ intrinsic affinities. However, our model predicts that attaining such high affinities requires large periplasms with high abundances of the slowly diffusing binding proteins, which pre- cludes high growth rates. By quantifying the benefits and costs of binding proteins, we provide a mechanistic explanation for the divergent evolution of oligotrophs and copiotrophs. Introduction Approximately half of global carbon fixation occurs in the ocean [1]. The fate of that carbon is governed by diverse species of heterotrophic bacteria [2–4] that differ in their carbon prefer- ences and uptake rates [5–7]. Yet we lack a fundamental understanding of how and why spe- cies’ metabolic strategies differ, an understanding needed to predict how a changing climate will affect rates of carbon flux in the ocean [8]. An important driver of species’ differentiation is nutrient availability, leading to a spectrum of microbial lifestyles: at opposite ends, copiotrophs dominate in nutrient-rich environments, whereas oligotrophs dominate in nutrient-poor environments [9–11]. Prototypical copio- trophs, like Vibrios, exhibit a feast-and-famine lifestyle and swim to colonize sporadic, nutri- ent-rich patches and particles [12,13]. They reach volumes greater than 1 μm3 and doubling times less than one hour [14]. Conversely, the abundant oligotrophs of the SAR11 clade are nonmotile and free-living [15] and have volumes smaller than 0.1 μm3 and doubling times greater than 5 hours [14]. Although copiotrophs typically attain higher doubling rates and have larger per cell biomass, the slow-growing oligotrophs comprise the majority of marine bacterial biomass [16,17]. Despite this, most of our understanding of bacterial metabolism derives from research on copiotrophic-like bacteria, which are easier to culture [18]. Genomic analyses suggest that the divergent phenotypic traits of copiotrophs and oligo- trophs are correlated with their suite of genes for nutrient transport [14,19–21]. Prototypical copiotrophs have many genes for phosphotransferase systems (PTS) used to uptake specific sugars [14,22]. In contrast, prototypical oligotrophs, like SAR11 and Sphingopyxis alaskensis, lack PTS [14,23] and instead rely heavily on ATP-binding cassette (ABC) transport systems, which are comprised of a transmembrane transport unit and a periplasmic substrate-binding protein. ABC transport systems have higher affinities than PTS [24,25]. Although it has long been held that the high affinity of ABC transport is a consequence of high-affinity binding pro- teins [26,27], Bosdriesz and others recently suggested that the affinity of ABC transport is a function of binding protein abundance and, specifically, that ABC transport confers high affin- ity only when the abundance of binding proteins exceeds that of transport units [28]. Thus, oli- gotrophs’ high abundances of binding proteins may explain their ability to grow in low nutrient conditions [19,29]. However, it is not understood why oligotrophs such as SAR11 cannot achieve higher grower rates in nutrient-rich conditions or why typical copiotrophs— which do, in fact, possess many ABC transport systems—cannot achieve higher affinities in nutrient-limited conditions [9]. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 2 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs To understand the metabolic constraints governing the dichotomy between the oligotro- phic and copiotrophic lifestyles, we develop molecular-level transport and cellular proteome allocation models to compare the performance of ABC transport and PTS. We derive a Michaelis–Menten approximation of ABC transport kinetics that predicts that the specific affinity of transport is proportional to binding protein abundance when the binding protein to transport unit ratio is sufficiently high. We thus find that ABC transport allows independent tuning of affinity and maximal uptake rate so that cells can achieve high affinities while closely matching transport and metabolic capacities. We thus predict that an oligotroph can attain a half-saturation concentration over a thousand-fold smaller than its binding protein’s dissocia- tion constant. However, attaining this high affinity requires a great abundance of binding pro- teins, which diffuse slowly and require large periplasms. Consequently, the reliance on binding proteins to achieve high affinity precludes high growth rates. Moreover, the ability of ABC transport to achieve high affinities while matching metabolic capacity makes metabolic imbal- ances unlikely and thus mechanisms for handling sudden nutrient up-shifts typically unneces- sary, which may explain the toxicity of high-nutrient conditions to SAR11. Together, these findings provide a mechanistic explanation for the divergence of the copiotrophic and oligo- trophic lifestyles, as the consequence of trade-offs between PTS and ABC transport. Results The specific affinity of ABC transport is a function of both transport and binding protein abundance To contrast the nutrient acquisition strategies of PTS and ABC transport systems, we present models of both, which show that, whereas the half-saturation concentration of PTS is an intrinsic property of the transporter, the half-saturation concentration of ABC transport is a function of binding protein abundance [28]. A PTS is used for the cytoplasmic uptake of a spe- cific sugar and modifies the sugar once it enters the cytoplasm by binding the sugar to a phos- phate group. PTS uptake kinetics can be described by the canonical model for transport [30]. It describes transport as a two-step reaction, in which (i) the periplasmic substrate (Sp) binds to the membrane-bound transport unit (T) with rate constant k1 to form a bound complex (T:S), and (ii) the substrate is translocated irreversibly into the cytoplasm with rate k2 (Sc) (Fig 1 and Section A in S1 Appendix). Using mass-action kinetics, we find that the cytoplasmic uptake rate (the rate at which Sp is converted to Sc) for PTS at steady-state is vc;PTS ¼ k2 T:S½ � ¼ k2½T�total ½S�p KT þ ½S�p ; ð1Þ where KT = k2/k1 is the transport unit dissociation constant and [T]total is the abundance of membrane-bound transport units divided by the volume of the periplasm. (Note that we here express all transport rates in terms of change in periplasmic concentration per time. We use the conversion factor fp/(1−fp) to obtain the uptake rate in terms of change in cytoplasmic con- centration, where fp is the fraction of the cell’s volume comprised of the periplasm. See Section D.3 in S1 Appendix.) The solution in Eq 1 has the classic Michaelis–Menten form of nutrient transport [31], with maximal uptake rate Vmax proportional to [T]total and half-saturation con- stant KM equal to KT (Fig 2). In contrast to PTS, the kinetics of ABC transport does not follow the classic Michaelis– Menten form [28]. ABC transport uses binding proteins (BP) in the periplasm that scavenge for incoming nutrients. These binding proteins, when in complex with the substrate, bind to membrane-bound transport units that require ATP to translocate the substrate from the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 3 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs Fig 1. Schematic of transport systems. For a nutrient to enter the cytoplasm, a transport unit bound to the inner membrane must expend energy to modify the substrate or translocate the substrate against a concentration gradient. For transport of a sugar by a phosphotransferase system (PTS), the sugar binds directly to the transport unit, and a cascade of specific proteins phosphorylate that particular sugar. For transport of a substrate by an ATP-binding cassette (ABC) transport system, binding proteins in the periplasm first scavenge for and store the substrate in the periplasm. When bound to substrate, a binding protein can then bind to a membrane-bound transport unit, which uses ATP to translocate the substrate. While a single type of binding protein may be able to bind to different substrates, it can bind to only a single, corresponding type of transport unit. To limit the number of free parameters when modeling these two transport systems, we use a simple model of PTS that assumes that binding of the substrate to the transport unit is irreversible. We extend the model for ABC transport to account for the reversible binding of the substrate to the binding protein and the dissociation of the binding protein from the transport unit after translocation. https://doi.org/10.1371/journal.pcbi.1009023.g001 0f and dissociation rate k0 periplasm into the cytoplasm [32–34]. Similar to previous models of transport by binding pro- teins [27,28], we describe ABC uptake by extending the PTS model to account for a four-step reaction: (i) the substrate–binding protein complex (S:BP) is formed by a reversible reaction with association rate k0 binding protein (S:BP) binds with rate constant k0 1 to the membrane-bound transport unit (T) to form a bound complex (T:S:BP), (iii) the substrate is translocated irreversibly into the cyto- plasm (Sc) with rate k0 2, and (iv) the transport unit and binding protein dissociate with rate k0 (Fig 1). At steady state, we obtain a system of four equations that can be solved exactly for the cytoplasmic uptake rate for ABC transport, vc,ABC, in terms of change in periplasmic concen- tration per time as a function of the concentration of free substrate in the periplasm, [S]p (Sec- tions A.2 and D.3.2 in S1 Appendix): 0r, (ii) the bound complex of substrate and 3 vc;ABC ¼ k0 2 � k0 3 2 þ k0 k0 3 � ½T�total ½S:BP� T þ ½S:BP� K0 ; K 0 T ¼ k0 2k0 3Þ 2 þ k0 1ðk0 3 ; k0 ½ S:BP � ¼ ½S�p½BP� KD þ k0 1½T�=k0 0f ; KD ¼ k0 k0 0r 0f ; ½BP� ¼ ½BP�total (cid:0) PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 ½S:BP� (cid:0) ð1 þ k0 2=k0 3Þ½T:S:BP�; ð2Þ ð3Þ ð4Þ 4 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs Fig 2. Maximal uptake rates, half-saturation concentrations, and specific affinities of PTS and ABC transport systems. We can approximate cytoplasmic uptake rates using the Michaelis–Menten equation: vc = Vmax[S]p/ (KM+[S]p), where Vmax is the maximal uptake rate and KM the half-saturation concentration. While the exact solution of the cytoplasmic uptake rate for our model of PTS is in the form of a Michaelis–Menten equation, the exact solution of the uptake rate for ABC transport is not. Because our simulations suggest that the abundance of binding proteins should exceed the abundance of transport units in the oligotrophic conditions where ABC transport is optimal, we make the approximations that (i) [T:S:BP]+[T:BP]�[BP]total and (ii) k0 the above estimates for the effective maximal rate and half-saturation concentration. For PTS, the half-saturation concentration is a constant equal to the dissociation constant KT = k2/k1. For ABC transport, the half-saturation concentration depends on both the transport dissociation constant K0 dissociation constant KD = k0 approximation, the specific affinity a0 = V0 abundances of transport units and of binding proteins. T ¼ k0 0f and is additionally a function of the abundance of binding proteins. Under this M of ABC transport is thus proportional to the product of the 1[T]�k0r (Section B in S1 Appendix) to obtain 3ÞÞ and the binding protein max/K0 2 þ k0 3=ðk0 0r/k0 1ðk0 2k0 https://doi.org/10.1371/journal.pcbi.1009023.g002 ½T� ¼ ½T�total (cid:0) ð1 þ k0 2=k0 3Þ½T:S:BP�: ð5Þ This model is a simplification of the ABC transport model developed by Bosdriesz and oth- ers [28]; in contrast to their model, our model assumes that translocation as well as the associa- tion and dissociation of binding protein and transport unit proceed irreversibly and thus has three fewer free parameters. Yet we find that our model provides a good fit for the well-charac- terized maltose ABC transport system in Escherichia coli (Section C in S1 Appendix). The model accurately predicts the observed KM as well as the shape of the uptake rate curves as functions of both extracellular maltose concentration and binding protein abundance (Figs A-D in S1 Appendix). To obtain a compact analytical expression describing how transport protein abundances affect uptake rate, we used our model to derive an approximation of ABC transport kinetics in Michaelis–Menten form. By assuming that binding proteins are much more abundant than active transport units [28,35] ([BP]total�[T:S:BP]+[T:BP]) and that the abundance of unbound transport units is low (so that [T]�k0 from Eqs 2–5 the following approximation for the cytoplasmic uptake rate: 1) (Fig 1 and Section B in S1 Appendix), we obtain 0r/k0 vABC � 3 k0 2k0 2 þ k0 k0 3 � ½T�total � ½BP�total T þ ½BP�total K0 ½S�p K0 KD T þ½BP�total K0 T þ ½S�p : ð6Þ This Michaelis–Menten equation well approximates ABC transport when the binding pro- tein to transport unit ratio sufficiently exceeds one and thus captures the dynamics of the full ABC transport model (Eqs 2–5) over a wide range of parameter values (S1 Fig). This formulation shows analytically how the half-saturation “constant” KM is, in fact, a function of the concentration of binding proteins in the periplasm ([BP]total, Fig 2). For PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 5 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs ½BP�total � ½T�total and ½BP�total � K0 T, as is the case for E. coli’s ABC maltose transport system, the approximation predicts that the half-saturation concentration KM is proportional to both the transport dissociation constant K0 T and the binding protein dissociation constant KD and is inversely proportional to the total abundance of binding proteins [BP]total. Therefore, express- ing high abundances of binding proteins enables oligotrophs to attain small KM values and thus high affinities. At low nutrient concentrations, the classic Michaelis–Menten uptake rate is proportional to the specific affinity [36], a = Vmax/KM. Whereas a bacterium using PTS has constant KM and thus can increase its specific affinity in oligotrophic conditions only by tun- ing Vmax (via expression of the transport unit; Eq 1), a bacterium using ABC transport can increase its specific affinity by tuning either Vmax or KM, by tuning the expression levels of the transport units and binding proteins, respectively (Eq 6). A rate–affinity trade-off drives the differentiation of oligotrophs and copiotrophs The derived Michaelis–Menten kinetics (Eqs 1 and 6) show how ABC transport systems allow bacteria to achieve higher substrate affinities than PTS by expressing high abundances of bind- ing proteins. To understand the costs associated with achieving these high affinities and thus to determine how the optimal expression levels of transport units and binding proteins differ in low-nutrient and high-nutrient environments, we integrate our solutions for the cyto- plasmic uptake rates of PTS (Eq 1) and ABC transport (Eqs 2–5) into a mechanistic, single-cell metabolic model (Fig 3, Methods, and Section D in S1 Appendix). Similar to the self-replicator model of Molenaar and others [37], our highly idealized metabolic model accounts for only four protein groups–transport proteins, metabolic proteins, ribosomes, and membrane bio- synthesis proteins–and is used to solve a proteome allocation problem that determines the optimal amount of each protein group that the cell should express in order to maximize its growth rate for a given extracellular nutrient concentration. Our metabolic model tracks the transport of a nutrient into the cytoplasm and the subsequent transformation of that nutrient into the proteins and metabolites required for replication. The abundances of proteins and metabolites are constrained by the cell’s surface-area-to-volume ratio. Because the cellular components occupy volume, they are limited by maximum cytoplasmic and periplasmic densities to prevent molecular overcrowding [37], and this favors smaller sur- face-area-to-volume ratios. On the other hand, the surface of the inner membrane must be suffi- ciently large because the membrane-bound transport units carry “real estate costs” [37,38]. Larger surface-area-to-volume ratios also support higher specific uptake rates by diffusion at low-nutri- ent conditions [39–41]. Thus, taken together, the surface-area-to-volume ratio creates a trade-off between the cell’s capacity for uptake and its capacity for synthesis. Therefore, in addition to determining the optimal proteome allocations, our model also determines the optimal surface- area-to-volume ratio, the protein and metabolite concentrations that are constrained by this ratio, and the fraction of the volume devoted to the periplasm (Methods, Section D in S1 Appendix). Central to this optimization problem are the costs and benefits of expressing more of a par- ticular protein group. While expressing more transport units or binding proteins increases the uptake rate, it incurs a proteomic cost [35,42–44]. This cost is an opportunity cost. For exam- ple, because growth rate depends on the proteome fraction allocated to ribosomes [45], expressing greater abundances of transport proteins may limit growth, as it limits the propor- tion of the proteome devoted to ribosomes. We assume that the transport units of PTS and ABC transport systems have the same proteomic cost and that the proteomic cost of an ABC binding protein is four times less than the cost of a transport unit (see Section D.3 in S1 Appendix for justifications). PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 6 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs Fig 3. A simple metabolic model tracks the utilization of a generic nutrient by the cell. The nutrient diffuses into the periplasm via a porous outer membrane and is then transported into the cytoplasm by membrane-bound transport units. The cell uses either transport by PTS, in which the substrate directly binds to the transport unit, or ABC transport, in which the substrate must first bind to a binding protein and then this complex binds to the transport unit. The intracellular substrate is next metabolized by a protein group that transforms the substrate into a precursor (a generic amino acid) that is needed to build the cell. The precursors are used (i) by a membrane biosynthesis protein group to build both the outer and inner membranes and (ii) by ribosomes to make proteins comprising the six protein groups. This model is subject to a number of constraints to determine the proteome allocation that maximizes the steady-state exponential growth rate. While this model does not consider the utilization of carbon for energy, we expanded the model to consider energy to show that differences in the energetic requirements of PTS and ABC transport do not change our results (Section E in S1 Appendix). https://doi.org/10.1371/journal.pcbi.1009023.g003 The effect that these transport proteomic costs have on the optimal proteome allocation strongly depends on the uptake rate per transport unit. This uptake rate is often limited by rates of diffusion within the periplasm [46,47]. Hence, we argue that differences in substrate diffusion drive a trade-off between PTS and ABC transport. Because ABC binding proteins are much larger than the substrates they bind, the diffusivity of the binding proteins is lower than the diffusivity of the substrate, limiting the achievable rates of ABC transport rel- ative to PTS, as suggested by [28]. For example, a typical binding protein (MalE) has a molecular weight of approximately 40 kDa and thus an estimated cytoplasmic diffusivity of 2 μm2/s, whereas glucose has a molecular weight of 0.18 kDa and thus an estimated cyto- plasmic diffusivity of 200 μm2/s [48]. Our model therefore assumes that the association rate k0 because it depends on the slow diffusion of the binding protein toward the membrane- bound transport units. 1(Fig 1) is one-hundred times smaller than the equivalent rate for PTS (i.e., k0 1 = 0.01k1) We choose the other rate values to enable a fair comparison between PTS and ABC trans- port. We assume that both their translocation rates and transport unit dissociation constants are equal (k0 3)), we are thus assuming that the rate of the dissociation of the binding protein from the transport T = KT). Therefore, since KT = k2/k1 and K0 2 = k2, K0 T = k0 3/(k0 2+k0 1(k0 2k0 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 7 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs Fig 4. Optimal proteome allocation for PTS and ABC transport systems. Proteome fractions shown are fractions of the proteome available for the four specified protein groups. (A) While it is optimal for cells relying on either PTS or ABC transport systems to devote nearly all of their proteome to transport at low nutrient concentrations, for ABC transport systems, it is the proteome fraction of the binding proteins that increases as nutrient concentration decreases and not the fraction allocated to the membrane-bound transport units. (B) As the nutrient concentration decreases, the optimal maximal uptake of transport increases for PTS but remains constant for ABC transport systems. This results in an increasing ratio of optimal maximal uptake and maximal metabolic rates for transport by PTS as nutrient concentration decreases, while it is optimal for ABC transport systems to maintain this ratio closer to one. https://doi.org/10.1371/journal.pcbi.1009023.g004 2). 3 � 0.01k0 2 ¼ k2 ¼ 200 sec-1 and K0 unit is approximately one-hundred times smaller than its translocation rate (k0 This is a reasonable assumption because the dissociation of the binding protein from the trans- port unit and its movement away from the inner membrane is also limited by the slow diffu- sion of the binding protein [47]. Indeed, these parameter assumptions well match differences between E. coli’s maltose ABC transport system and glucose PTS (Sections C and D.3 in S1 Appendix). We thus set the translocation rate and dissociation constant to those measured for E. coli’s glucose PTS, k0 T ¼ KT ¼ 10 μM. In addition, we assume that the kinetics of the binding of substate to binding protein matches that of maltose to MalE so that KD = 1 μM, with association rate k0 0r = 100 sec-1. Because we assume that the diffusive rates of binding proteins limit ABC uptake rates, our model shows that PTS can achieve higher maximal uptake rates Vmax per proteomic cost than ABC transport (Fig 2). Specifically, at saturating extracellular nutrient concentrations, the opti- mal cell using PTS devotes 80 times less proteome to transport than the optimal cell using ABC transport (Fig 4A) yet achieves a slightly (3%) higher Vmax (S2 Fig). Therefore, cells using PTS achieve higher growth rates than cells using ABC transport when nutrient concentrations are high (Fig 5A). 0f ¼ 105 mM-1sec-1 and dissociation rate k0 Conversely, our model shows that ABC transport systems have higher specific affinities (a) per proteomic cost than PTS (Figs 2 and 5A). As the nutrient concentration decreases to 1 nM, the cytoplasm of the optimal ABC cell shrinks to concentrate the limiting metabolites so that the optimal cytoplasmic concentrations remain nearly constant over all extracellular nutrient concentrations (S3 Fig). Yet the optimal periplasmic volume increases so that the optimal ABC cell at 1 nM has a periplasmic volume that is, in fact, larger than its cytoplasmic volume (S4 Fig). This increase in periplasmic volume prevents molecular overcrowding while permitting an increase in the abundance of binding proteins, which is limited by the periplasmic density constraint (Section D.1 in S1 Appendix and Figs O and P in S2 Appendix). Thus, although the optimal periplasmic binding protein concentration remains constant as nutrient levels decrease (S5 Fig), the binding protein to transport unit ratio ([BP]/[T]total) increases to seven PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 8 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs Fig 5. A rate–affinity trade-off. Plots show the results of proteome allocation problems using either PTS or ABC transport and solved for different extracellular nutrient concentrations (x-axis). We assume that the transport association rate is 100 times lower for ABC transport than for PTS (k0 1 = 0.01k1) but that the translocation rate as well as the transport unit dissociation constant are equal (k0 2 ¼ k2; K 0 T ¼ KT). We additionally limit the radius of the cell to a minimum of 60 nm, corresponding to a maximum surface-area-to-volume ratio of 50 μm-1. (A) shows the maximal growth rates achieved using the optimal proteome allocation, and (B) shows the optimal surface-area-to-volume ratio used to achieve those maximal growth rates. ABC transport achieves higher growth rates at low nutrient concentrations because it supports higher substrate affinities per transport proteomic cost, whereas PTS achieves higher growth rates at high nutrient concentrations because it supports higher maximal uptake rates per transport proteomic cost. https://doi.org/10.1371/journal.pcbi.1009023.g005 (Fig 6). This increase in binding protein to transport unit ratio increases the probability that a transport unit will be bound, thus increasing uptake affinity (S6 Fig). In this way, using a bind- ing protein with KD = 1 μM and a transport unit with dissociation constant K0 T = 10 μM, the optimal ABC cell achieves an effective half-saturation concentration of K0 M � 3 nM (Fig 6). Thus, although the optimal ABC cell devotes 16% less of its proteome to transport than the optimal PTS cell, it achieves a half-saturation concentration that is over three thousand times lower than the half-saturation concentration of the optimal PTS cell, KM = KT = 10 μM. There- fore, cells using ABC transport achieve higher growth rates than cells using PTS when nutrient concentrations are low (Fig 5A). Many bacterial species have both PTS and ABC transport systems for the same nutrient, using PTS when the nutrient is plentiful and ABC transport when the nutrient is scarce [24,49]. Because of this redundancy, it has long been hypothesized that there exists a rate-affin- ity trade-off between PTS and ABC transport [24,28]. Our results provide a mechanistic expla- nation for this trade-off and furthermore demonstrate that this trade-off, in particular, drives the differences in performance between the optimal ABC and PTS cell. Alternative hypotheses on the mechanisms creating a trade-off between the two transport mechanisms are not sup- ported by our metabolic model. We find that the advantage of PTS in high-nutrient conditions does not stem from either lower energetic or lower proteomic costs because these costs are minimal in high-nutrient conditions. When we expanded our model to include the energetic costs of transport and furthermore incorrectly assumed that the association rate, k1, was the same for both PTS and ABC transport systems and also that the dissociation rate k0 gible (i.e., k0 of ABC transport, the maximal growth rate achieved by the optimal ABC cell was always greater than or equal to the maximal growth rate achieved by the optimal PTS cell (Section E in S1 Appendix). 3 was negli- 2), we observed no trade-off: despite the higher proteomic and energetic costs 3 � k0 We therefore conclude that the only trade-off that can explain the redundancy of species that utilize both PTS and ABC transport systems for the same nutrient is a rate–affinity trade- off that is a consequence of the high affinity achieved by using binding proteins and the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 9 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs Fig 6. The effective half-saturation concentration of ABC transport. ABC transport systems achieve low optimal half-saturation concentrations (Keff, magenta curve and axis)—and thus high specific affinities—as nutrient concentrations decrease by maintaining a high surface-area-to-volume ratio and increasing the ratio of the abundance of binding proteins to the abundance of membrane-bound transport units (turquoise curve and axis). For high binding protein to transport unit ratios, the Michaelis-Menten approximation of ABC transport (Eq 6) holds (S1 Fig). At 1 nM, where the binding protein to transport unit ratio is approximately seven, the approximation gives Keff � 1.5 nM, while the calculated Keff = 2.9 nM. (For a plot showing how we calculate the effective half-saturation concentration, see S7 Fig). https://doi.org/10.1371/journal.pcbi.1009023.g006 binding proteins’ limiting rates of diffusion. Specifically, our model predicts that it is the disso- ciation rate of the binding protein and transport unit, k0 3, that limits ABC uptake rate. (See sen- sitivity analyses in S2 Appendix.) After translocation, the bulky, unbound binding protein must dissociate from the membrane-bound transport unit and diffuse away from the inner membrane to allow a bound binding protein to associate with the transport unit. Therefore, while the translocation rate k2 governs the uptake rate by PTS, we predict that it is the diffusiv- ity of the binding protein that governs the ABC uptake rate. The heavy reliance of heterotro- phic bacteria on ABC transport systems in the oligotrophic ocean suggests that this trade-off is central to the dichotomy between the copiotrophic and oligotrophic lifestyles and that it may explain the fundamental difference in their achievable growth rates. Our conclusion that a rate–affinity trade-off between PTS and ABC transport underpins the differentiation of oligotrophs and copiotrophs is further supported by our model’s predictions on the optimal surface-area-to-volume ratio. We find that the optimal surface-area-to-volume ratio (which is inversely proportional to the cell radius) is smaller for PTS cells in nutrient-rich condi- tions than it is for ABC cells in all nutrient conditions (Fig 5B). This result is in accordance with observations that typical copiotrophic marine bacteria, like Vibrio, are over ten times larger than typical oligotrophic ones, like SAR11 [14,40]. The model further reveals that increasing transloca- tion rates for PTS decreases the optimal surface-area-to-volume ratio (S8 Fig). This indicates that, whereas larger surface-area-to-volume ratios allow the cell to achieve higher cytoplasmic concen- trations of metabolites and proteins—and hence higher processing rates—in low-nutrient condi- tions, smaller surface-area-to-volume ratios are optimal at high nutrient uptake rates because they provide the cell with more space in which to process the substrate and transform it into bio- mass. Thus, the higher achievable uptake rates of PTS support larger optimal cell volumes. A suite of sensitivity analyses confirmed that the fundamental trends observed from the rate-affinity trade-off depend primarily on our assumption that it is the association and PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 10 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs 2 to k0 3 ¼ k0 3) that limit ABC trans- 3, we find that a rate-affinity trade-off in the relative achievable growth rates of 1 and k0 2) nor the association rate of the substrate to binding pro- dissociation rates of the binding protein and transport unit (k0 port and not the translocation rate (k0 tein (k0 0f ). The trends do not depend on the precise magnitude of the rates (S2 Appendix). Although our results are most sensitive to the dissociation rate of the binding protein and transport unit k0 PTS versus ABC transport still exists when we increase the dissociation constant from k0 0.01 k0 2 (S2 Appendix). While modifications to the surface-area-to-volume ratio and density constraints modulate the magnitude of the rate-affinity trade-off, we find that the mag- nitude of the trade-off is most sensitive to the relative proteomic costs of transport versus pro- tein synthesis (S3 Appendix). If we increase the proteomic cost of transport relative to protein synthesis, then the optimal ABC cell has less of an advantage in oligotrophic conditions (that is, there is a smaller difference in the achieved growth rates of the ABC and PTS cells), whereas the optimal PTS cell has a greater advantage in high-nutrient conditions (Fig A in S3 Appen- dix). Conversely, if we increase the protein synthesis proteomic cost relative to the transport proteomic cost, the optimal ABC cell now has a greater advantage in oligotrophic conditions, whereas the optimal PTS cell has a smaller advantage in high-nutrient conditions (Fig B in S3 Appendix). 3 = ABC cells achieve high affinities while closely matching metabolic and transport capacities Unlike a cell using PTS, a cell using ABC transport can increase its specific affinity without increasing its maximal uptake rate (Eq 6). Our model predicts that, for both the optimal PTS cell and the optimal ABC cell, as the extracellular nutrient concentration decreases, the frac- tion of the proteome devoted to transport increases, while the fraction of the proteome devoted to metabolic enzymes decreases (Fig 4A). For PTS, increasing the transport proteome fraction increases the maximal uptake rate Vmax (Fig 2). As a result, our model demonstrates a mis- match between metabolic and transport capacities for a PTS cell optimized for growth in low- nutrient conditions: for the optimal PTS cell, the ratio of the transport capacity Vmax and the metabolic capacity—which is proportional to the abundance of metabolic enzymes—exceeds one for all nutrient concentrations below 1 mM. Indeed, at 1 nM, this ratio approaches 10,000 (Fig 4B). Therefore, a PTS cell optimized for growth in nutrient-poor conditions that suddenly encounters a higher nutrient concentration would uptake more nutrient than it can process and could quickly accumulate toxic levels of metabolites if it cannot excrete them. In contrast, as the extracellular nutrient concentration decreases, the optimal ABC cell does not allocate any additional proteome to membrane-bound transport units but only to binding proteins to increase its affinity (Fig 4A). As a result, our model shows that the optimal ABC cell maintains a transport to metabolic capacity ratio of one for all nutrient concentrations above 0.1 μM. At 1 nM, the maximum ratio is below ten (Fig 4B). Therefore, our model sug- gests that a cell using ABC transport is much less prone to mismatches between its proteome and the environment that may cause toxic accumulations of metabolites within the cell. Hence, cells may rarely if ever need to excrete metabolites that are consumed only using ABC transport systems. Discussion We used a simple metabolic model to quantify the costs and benefits of using PTS versus ABC transport systems and thus understand the divergence of the copiotrophic and oligotrophic lifestyles of heterotrophic marine bacteria that rely on carbon as an energy source. By deriving an approximation of ABC transport in Michaelis–Menten form, we predict that, when the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 11 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs abundance of binding proteins sufficiently exceeds the abundance of transport units, the spe- cific affinity of ABC transport is directly proportional not only to the transport unit abundance but also to the binding protein abundance, corroborating previous theoretical work that found that KM is a function of binding protein abundance [28]. Our analysis thus suggests that cells should maintain high binding protein to transport unit ratios to achieve high specific affinities. We predict that, for oligotrophs such as members of the SAR11 clade, the KM value may be over a thousand-fold smaller than the dissociation constant of the binding protein KD. Although we are aware of only two experimental studies that considered the effects of varying binding protein abundance on uptake, both provide support to our model. We used one of the experimental studies—on E. coli’s ABC maltose transport system—to directly verify our mod- el’s predictions on the dependence of uptake on binding protein abundance (Section C in S1 Appendix). The second study found that a Salmonella typhimurium mutant that expresses five- fold higher levels of binding protein for histidine uptake has a fourfold lower KM—thirtyfold lower than the estimated KD of the binding protein for histidine [50]. Our Michaelis-Menten approximation of ABC transport is consistent with these observed values, predicting that the Salmonella mutant’s binding protein concentration is approximately thirty times greater than the transport unit dissociation constant KT. As ABC transport systems are ubiquitous in gram-negative bacteria, the fact that KM may be drastically different from KD has important implications for our ability to predict microbial dynamics. Because of the difficulty of measuring the value of KM for uptake directly, much pre- vious work has estimated the performance of ABC transport systems using binding assays that measure KD instead [25,34]. Our work suggests that this estimate could differ from KM by orders of magnitude for oligotrophs that use high abundances of binding proteins, thus poten- tially leading to substantial underestimates of oligotrophs’ nutrient uptake rates. In addition, a variety of microbial ecosystem models assume a constant value of KM for uptake [51,52], but this assumption may be flawed because bacteria may vary their binding protein abundance and thus their KM value as a function of environmental conditions. Alternatively, it is also pos- sible that cells have evolved to express a constant binding protein abundance to maintain a constant, ecologically relevant KM value. Experiments are needed to determine the extent to which binding protein abundance and the value of KM vary within a species, as well as the impacts of the variability in KM on ecosystem dynamics. Our model provides a mechanistic explanation for the differences in performance observed between the glycine betaine transport systems of E. coli and of a SAR11 strain that is prevalent in the vast nutrient-poor expanses of the ocean [29]. The SAR11 strain can achieve nanomolar values for the half-saturation concentration of glycine betaine uptake, whereas E. coli’s geneti- cally similar glycine betaine transport system uses a binding protein with only a micromolar dissociation constant [21]. It was posited that SAR11 achieves these higher affinities by achiev- ing higher binding protein concentrations in a large periplasm [21]. Our model corroborates this hypothesis and furthermore demonstrates how SAR11 can achieve a nanomolar half-satu- ration concentration using a binding protein with the same micromolar binding affinity as E. coli’s glycine betaine binding protein. To achieve such a high specific affinity using a binding protein with only a micromolar dissociation constant, our model predicts that SAR11 main- tains a high binding protein to transport unit ratio. Although our model cannot rule out the alternative possibility that oligotrophs evolved binding proteins with lower KD values to achieve very low KM values, it does demonstrate that this is not required. To determine the rel- ative roles of low KD values versus high binding protein abundances for achieving high affini- ties in SAR11, their binding proteins must be purified and used in binding assays to directly measure KD values and contrast them with the KM values attained from uptake rate measurements. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 12 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs It was previously hypothesized that typical oligotrophs achieve higher specific affinities than copiotrophs by having higher ratios of transport units to metabolic enzymes—and thus extremely high ratios of transport to catabolic capacity [53], and this theory is still often used to explain the nutrient acquisition strategy of SAR11 [29]. Here we propose an alternative the- ory: unlike cells using PTS, cells using ABC transport are able to increase their specific affinity without increasing their maximal uptake rate. As a result, oligotrophs may be able to closely match their transport and metabolic capacities for a number of important compounds. We hypothesize that it is for this reason that members of the SAR11 clade experience large nutrient upshifts as toxic [9,54,55]: since transport capacity rarely exceeds metabolic capacity, they may not be able to excrete substrates consumed via ABC transport as they would not need to do so in the nutrient-poor ocean in which they evolved. Consequently, an atypical, large nutrient upshift would overwhelm the cytoplasm with substrate that the cell can neither process nor excrete. Our metabolic model indicates that ABC transport systems are more efficient than PTS at low nutrient concentrations because expressing an additional binding protein has a lower proteomic cost than expressing an additional transport unit and, furthermore, does not incur real-estate costs on the inner membrane. Instead, the binding protein abundance is subject only to a constraint on the periplasmic density, a constraint that a cell can miti- gate by modifying the fraction of its volume devoted to the periplasm. Our model predicts that the optimal periplasmic volume fraction increases as extracellular nutrient concentra- tion decreases (S4 Fig): observations suggesting that the periplasm occupies up to 70% of the volume of a SAR11 Pelagibacter cell [56] are in line with this prediction. Therefore, our model predicts that a majority of an oligotroph’s proteome is comprised of binding pro- teins (Fig 4A). This prediction is corroborated by metaproteomic analyses showing that binding proteins are among the most prevalent bacterial proteins found in the oligotrophic ocean [19]. Our results provide a mechanistic explanation for the long-standing hypothesis of a rate– affinity trade-off for nutrient uptake by marine bacteria [57,58]. An oligotroph’s reliance on binding proteins to achieve high affinities precludes its ability to attain high growth rates because our model assumes that the rate of ABC transport is diffusion-limited due to the bulki- ness of the binding proteins. In particular, our model predicts that it is the dissociation of the transport unit and binding protein that is the limiting step of ABC transport and, specifically, that this dissociation step is much slower than translocation because of the size of the binding protein. To test this hypothesis, we must measure the association and dissociation rates of binding protein and transport unit for different ABC transport systems and determine whether these rates are functions of the size of the binding protein. We also find that the mechanism of this rate-affinity trade-off explains observations that the surface-area-to-volume ratio of a typical oligotroph, like a SAR11 cell, is at least fivefold greater than that of a typical copiotroph, like a Vibrio [59]. We thus propose that the high translocation rates of PTS in copiotrophs are advantageous not only because they support greater uptake rates at high nutrient concentrations but also because these higher uptake rates support larger optimal cell volumes. This is of particular importance to motile copiotrophs, which must be large enough to overcome rotational diffusion in order to swim effectively toward nutrient hotspots [60]. In addition, motile cells may not be able to attain values of KM as low as those of oligotrophs because of the large periplasmic volume fractions that this requires. Because the distance between the outer and inner membranes dictates the length of the flagellar rotor, periplasmic volume is carefully regulated in motile cells [61] and typically does not exceed 20% of the cell volume [62]. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 13 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs Although this work considered optimal cell physiologies in different homogeneous, unchanging environments, it also suggests how cells may optimally regulate their proteomes and morphologies in response to changes in nutrient levels. Our analysis suggests that, in response to a decrease in nutrient concentration, a cell should shrink its cytoplasm and inflate its periplasm to increase the ratio of ABC binding proteins to transport units. It has been observed that Vibrios, at the onset of starvation, divide—thus shrinking in size and shedding their flagella [63,64]. It would be interesting to determine if the shedding of the flagella enables Vibrio to increase periplasmic volume to thus increase binding protein abundance. Similarly, although previous work suggests that SAR11 cells remodel very little of their proteome in response to environmental fluctuations [29], our model suggests that oligotrophs should regu- late periplasmic volume and binding protein abundance due to the high costs of growing the outer membrane and expressing high ratios of binding proteins to transport units. Future experiments should investigate the extent to which SAR11 may vary binding protein abun- dance in response to nutrient levels. In summary, our work suggests that the constraints imposed by a rate–affinity trade-off between PTS and ABC transport systems shaped the divergent evolution of copiotrophic and oligotrophic bacteria in the ocean. By quantifying this trade-off, our model helps predict the achievable nutrient uptake rates and affinities of marine heterotrophic bacteria. These mecha- nistic predictions could be used to constrain the parametrizations of marine microbial ecosys- tem models used to understand how bacterial population dynamics may affect carbon flux rates in a changing ocean. Methods To compare the performance of PTS and ABC transport, we incorporated models of each (Eqs 1 and 2–5) into a single-cell metabolic model that is a modification of the self-replicator model proposed by Molenaar and others [37]. We used this model to solve the following proteome allocation problem: maximize x m subject to : equality constraints EqC: 1 (cid:0) 7; inequality constraints Ineq:C 1 (cid:0) 3; xi � 0 8i; and cell radius r > 60 nm; where μ is the steady-state exponential growth rate; the independent variables to be optimized are x = (xm, ϕ, r, fp, μ); the vector of intracellular metabolite concentrations xm = ([S]p, [S]c, [A], [W], [P]) is comprised of the periplasmic concentration of the generic carbon substrate [S]p, the cytoplasmic concentration of the carbon substrate [S]c, the cytoplasmic concentration of amino acids [A], the number of generic cell membrane units divided by the cytoplasmic vol- ume of the cell [W], and the number of amino acids incorporated into protein divided by the cytoplasmic volume of the cell [P]; the vector ϕ = (ϕBP, ϕT, ϕE, ϕM, ϕR) denotes the fraction of the proteome devoted to ABC binding proteins, transport units, metabolic enzymes, mem- brane biosynthesis enzymes, and protein synthesis enzymes respectively; and fp is the fraction of the cell’s volume devoted to the periplasm. The equality constraints 1–5 are ordinary differential equations that assume balanced, steady-state exponential growth of each of the five cellular components: EqC: 1 (cid:0) 5 : dxm dt ¼ Nvr (cid:0) mxm ¼ 0; PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 14 / 21 PLOS COMPUTATIONAL BIOLOGYwhere N is a stoichiometry matrix; and vr is a vector of Michaelis-Menten reaction rates, Marine oligotrophs versus copiotrophs 2 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 vr ¼ vdiff vc � kE E½ � kW M½ kE R½ � ½S�c KM;E þ ½S�c ½A� KM;W þ ½A� ½A� KM;R þ ½A� ðSext ! SpÞ ðSp ! ScÞ ð5Sc ! 6AÞ ðA ! WÞ ðA ! PÞ 3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 5 ; where enzyme concentration [X] = ϕXαX[P]. We here assume that the periplasmic concentra- tion of substrate is limited by diffusion and not by porin abundance so that the periplasmic uptake rate is vdiff ¼ 3D ½S�ext (cid:0) fpr2 ½S�p ; where D is the diffusivity of the substrate and [S]ext is the specified concentration of substrate in the external environment. The cytoplasmic uptake rate vc is either that of PTS (Eq 1) or of ABC transport (Eqs 2–5). Equality constraint EqC. 6 ensures that the proteome fractions sum to one: EqC: 6 : 1 ¼ �O;cyto þ �O;peri þ X i2P �i; where ϕO,cyto (ϕO,peri) is a required constant fraction of the proteome devoted to “other” pro- tein components in the cytoplasm (periplasm). Equality constraint EqC. 7 ensures that the concentration of cell membrane units, [W], is sufficient to cover both the inner and outer membranes of the cell: � EqC: 7 : 4p 1 þ ð1 (cid:0) � 2 3 fpÞ r2 ¼ W½ � � � 4pr3 3 aw; where aw is the surface area of a single membrane unit. Inequality constraints IneqC. 1&2 are density constraints on the cytoplasm and periplasm: X IneqC: 1 : mjxmðjÞ � rcyto; j2Mcyto IneqC: 2 : X j2Mperi mjxmðjÞ � rperi; where mj is the molecular weight of metabolite j and ρcyto (ρperi) is the maximal allowed density of the cytoplasm (periplasm). Inequality constraint IneqC. 3 ensures that the surface area of the inner membrane in suffi- ciently large to contain all inner membrane-bound transport units: � � � IneqC: 3 : fSA 4pð1 (cid:0) � 2 3r2 fpÞ � T½ � 4pr3 3 aT; where fSA is the fraction of the surface area available for transport units and aT is the surface area of a single transport unit. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 15 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs Values for all of the parameters specified in this model are given and justified in Section D in S1 Appendix. To solve the optimization problem, we used MATLAB’s constrained nonlin- ear multivariable function solver, fmincon. To ensure that the solver found globally optimal solutions, we transformed the units of the constraints and variables so that their predicted magnitudes were all approximately 1 and ran the solver 50 times for each optimization prob- lem, each time using a different initial guess for the variables x. The code is available at: https:// github.com/noelenorris/ABC_proteome_allocation. Supporting information S1 Appendix. Transport and proteome allocation models. This supplemental appendix con- tains derivations of the PTS and ABC transport models and ABC Michaelis-Menten approxi- mation; analysis of E. coli’s ABC maltose transport system; full exposition of metabolic model with parameter value justifications; and discussion of the energetic costs of transport. Figs A-E in the S1 Appendix support the analysis of E. coli’s ABC maltose transport system. Fig A: Effects of maltoporin abundance on uptake. Fig B: Effects of binding protein abundance on uptake. Fig C: Effects of binding protein abundance on maximal uptake rate. Fig D: Effects of binding protein abundance on the half-saturation concentration of uptake. Fig E: The perme- ability of the outer membrane limits half-saturation concentration of uptake. (PDF) 1 and k0 3. Fig F: Sensitivity analysis, 2. Fig B: Sensitivity 1. Fig D: Sensitivity anal- 3: optimal proteome fractions. Fig I: Sensitivity analysis, k0 1: optimal proteome fractions. Fig E: Sensitivity analysis, k0 3: optimal proteome fractions. Fig G: Sensitivity analysis, k0 2: optimal proteome fractions. Fig C: Sensitivity analysis, k0 S2 Appendix. Sensitivity analyses of ABC transport system. This supplemental appendix presents Figs A-V, showing the optimal solutions of the ABC cell when specified parameter values are modified against the baseline value. Fig A: Sensitivity analysis, k0 analysis, k0 ysis, k0 k0 sis, k0 analysis, k0 analysis, ;O,cyto: optimal proteome fractions. Fig M: Sensitivity analysis, ρcyto. Fig N: Sensitivity analysis, ρcyto: optimal proteome fractions. Fig O: Sensitivity analysis, ρperi. Fig P: Sensitivity analysis, ρperi: optimal proteome fractions. Fig Q: Sensitivity analysis, fSA. Fig R: Sensitivity analysis, fSA: optimal proteome fractions. Fig S: Sensitivity analysis, number of amino acids comprising binding protein. Fig T: Sensitivity analysis, number of amino acids comprising binding protein: optimal proteome fractions. Fig U: Sensitivity analysis, D. Fig V: Sensitivity analysis, D: optimal proteome fractions. (PDF) 0f: optimal proteome fractions. Fig K: Sensitivity analysis, ;O,cyto. Fig L: Sensitivity 3. Fig H: Sensitivity analy- 0f. Fig J: Sensitivity 1 and k0 S3 Appendix. Sensitivity analyses of rate-affinity trade-off. This supplemental appendix presents Figs A-G, which assesses the sensitivity of the rate-affinity trade-off by contrasting the optimal ABC and PTS cells when particular parameters are modified. Fig A: PTS versus ABC, transport proteomic costs x10. Fig B: PTS versus ABC, protein synthesis proteomic cost x10. Fig C: PTS versus ABC, ;O,cyto = 0.5. Fig D: PTS versus ABC, ρcyto x0.01. Fig E: PTS versus ABC, ρcyto x100. Fig F: PTS versus ABC, ρperi x0.01. Fig G: PTS versus ABC, ρperi x100 (PDF) S1 Fig. Comparison of approximate and exact ABC transport half-saturation concentra- tion values. Here we compare our Michaelis-Menten approximation of the half-saturation concentration for ABC transport with the exact half-saturation concentration obtained by solving the set of four equations for ABC transport rates using baseline values for the kinetics rates and modifying the periplasmic concentration of transport units and binding proteins. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 16 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs Note that, for a periplasmic transport unit concentration of 1.16 mM, the half-saturation con- centration does not asymptote to the approximation because, in this case, [T]total>k0 1f. Yet the exact solution follows the same trend as the approximation. (TIF) 0r/k0 S2 Fig. Optimal maximal uptake rates and specific affinities. Here are plots showing the optimal maximal uptake rates, Vmax, and corresponding optimal specific affinities, Vmax/KM. (TIF) S3 Fig. Optimal cytoplasmic concentrations for ABC cell. Optimal cytoplasmic concentra- tions of intracellular nutrient (Sc), amino acids, and total protein (in units of amino acids) over extracellular nutrient condition for cell with ABC transport. Although the extracellular nutri- ent concentration varies over many magnitudes, the optimal intracellular concentrations vary by less than a factor of three. (TIF) S4 Fig. Optimal periplasmic volume fraction for ABC transport. The optimal periplasmic volume fraction increases as nutrient concentration decreases to allow for greater abundances of binding proteins, which are subject to a density constraint on the periplasm. (TIF) S5 Fig. Optimal periplasmic concentrations for ABC cell. Although the optimal binding protein concentration remains nearly constant over all extracellular nutrient concentrations, the periplasmic transport unit concentration ([T]total) decreases as nutrient concentration decreases due to the inflation of the periplasm. While the periplasm inflates, the cytoplasm shrinks so that, for an extracellular nutrient concentration of 1 nM, the optimal periplasmic concentration of transport units is less than the abundance of transport units divided by the cytoplasmic volume ([T]totalVperi/Vcyto). (TIF) S6 Fig. Impact of modifications to periplasmic volume around optimal solution of ABC cell at nutrient concentration of 1 nM. To understand why the periplasm inflates as the nutri- ent concentration decreases to 1 nM, we plot the proportion of bound transport units (A) and effective half-saturation constant, KM, (B) as we modify the periplasmic volume of the optimal solution for a nutrient concentration of 1 nM. We assumed that both the concentration of binding proteins in the periplasm and the abundance of transport units on the inner mem- brane remain constant. Therefore, as the periplasm grows, the periplasmic concentration of transport units decreases and the ratio of binding proteins to transport units increases. (A) shows how the increase in abundance of binding proteins due to the inflation of the periplasm leads to an increase in the proportion of bound transport units, where we here assume that the concentration of free substrate in the periplasm ([S]p) is equal to 1 nM. (B) shows the calcu- lated half-saturation constant by fitting the Michaelis-Menten equation to the exact solutions of ABC transport uptake (Eqs 2 to 5), as well as our Michaelis-Menten approximation of the half-saturation constant (Eq 6), which holds only when the binding protein concentration suf- ficiently exceeds the transport unit concentration. (TIF) S7 Fig. Calculating the half-saturation concentration of ABC transport. To calculate the effective half-saturation concentration of an optimal solution to a particular proteome alloca- tion problem, we used the system of equations describing ABC transport to determine the uptake rate over a range of nutrient concentrations (x-axis). Here we show the calculated uptake rates over various nutrient concentrations for the proteome allocation obtained when PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 17 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs optimized the cell for growth at an extracellular concentration of [S]ext = 1 nM. (TIF) S8 Fig. Sensitivity analysis of proteome allocation for PTS transport to changes in translo- cation rate k2 at extracellular carbon concentration [S]ext = 100 mM. Increases in the trans- location rate result in (A) higher achievable growth rates and (B) larger optimal cell radii (that is, smaller surface-area-to-volume ratios). (TIF) S9 Fig. Active constraints on cell radius. Both the surface area “real estate" constraints and the density constraints are active for the PTS transport proteome allocation problem. Increases in maximal allowed density result in smaller optimal cell radii (red and yellow). Increases in the fraction of the surface area available to the membrane-bound transport units result in larger optimal cell radii (purple and green). (TIF) Acknowledgments Contributions to the editing of this paper by Dr. Russell Naisbit are gratefully acknowledged. The authors thank Terry Hwa and Cameron Thrash for discussions and feedback. Author Contributions Conceptualization: Noele Norris, Vicente I. Fernandez, Roman Stocker. Data curation: Noele Norris. Formal analysis: Noele Norris. Funding acquisition: Naomi M. Levine, Roman Stocker. Methodology: Noele Norris, Naomi M. Levine. Software: Noele Norris. Supervision: Roman Stocker. Visualization: Noele Norris. Writing – original draft: Noele Norris. Writing – review & editing: Noele Norris, Naomi M. Levine, Roman Stocker. References 1. Falkowski PG, Barber RT, Smetacek V. Biogeochemical Controls and Feedbacks on Ocean Primary Production. Science. 1998; 281: 200–206. https://doi.org/10.1126/science.281.5374.200 PMID: 9660741 2. Giovannoni SJ, Stingl U. Molecular diversity and ecology of microbial plankton. Nature. 2005; 437: 343– 348. https://doi.org/10.1038/nature04158 PMID: 16163344 3. Azam F, Malfatti F. Microbial structuring of marine ecosystems. Nature Reviews Microbiology. 2007; 5: 782–791. https://doi.org/10.1038/nrmicro1747 PMID: 17853906 4. Burd AB, Frey S, Cabre A, Ito T, Levine NM, Lønborg C, et al. Terrestrial and marine perspectives on modeling organic matter degradation pathways. Glob Change Biol. 2016; 22: 121–136. https://doi.org/ 10.1111/gcb.12987 PMID: 26015089 5. Poretsky RS, Sun S, Mou X, Moran MA. Transporter genes expressed by coastal bacterioplankton in response to dissolved organic carbon. Environ Microbiol. 2010; 12: 616–627. https://doi.org/10.1111/j. 1462-2920.2009.02102.x PMID: 19930445 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 18 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs 6. Gifford SM, Sharma S, Booth M, Moran MA. Expression patterns reveal niche diversification in a marine microbial assemblage. ISME J. 2013; 7: 281–298. https://doi.org/10.1038/ismej.2012.96 PMID: 22931830 7. Hellweger FL. Heterotrophic substrate specificity in the aquatic environment: The role of microscale patchiness investigated using modelling: Substrate specificity and microscale patchiness. Environmen- tal Microbiology. 2018; 20: 3825–3835. https://doi.org/10.1111/1462-2920.14397 PMID: 30175444 8. Widder S, Allen RJ, Pfeiffer T, Curtis TP, Wiuf C, Sloan WT, et al. Challenges in microbial ecology: building predictive understanding of community function and dynamics. ISME J. 2016; 10: 2557–2568. https://doi.org/10.1038/ismej.2016.45 PMID: 27022995 9. Koch AL. Oligotrophs versus copiotrophs. Bioessays. 2001; 23: 657–661. https://doi.org/10.1002/bies. 1091 PMID: 11462219 10. Giovannoni SJ, Cameron Thrash J, Temperton B. Implications of streamlining theory for microbial ecol- ogy. ISME J. 2014; 8: 1553–1565. https://doi.org/10.1038/ismej.2014.60 PMID: 24739623 11. Kirchman DL. Growth Rates of Microbes in the Oceans. Annual Review of Marine Science. 2016; 8: 285–309. https://doi.org/10.1146/annurev-marine-122414-033938 PMID: 26195108 12. Koch AL. The Adaptive Responses of Escherichia coli to a Feast and Famine Existence. In: Rose AH, Wilkinson JF, editors. Advances in Microbial Physiology. Academic Press; 1971. pp. 147–217. https:// doi.org/10.1016/s0065-2911(08)60069-7 PMID: 4950180 13. Srinivasan S, Kjelleberg S. Cycles of famine and feast: the starvation and outgrowth strategies of a mar- ineVibrio. J Biosci. 1998; 23: 501–511. https://doi.org/10.1007/BF02936144 14. 15. Lauro FM, McDougald D, Thomas T, Williams TJ, Egan S, Rice S, et al. The genomic basis of trophic strategy in marine bacteria. PNAS. 2009; 106: 15527–15533. https://doi.org/10.1073/pnas. 0903507106 PMID: 19805210 Luo H, Moran MA. How do divergent ecological strategies emerge among marine bacterioplankton line- ages? Trends in Microbiology. 2015; 23: 577–584. https://doi.org/10.1016/j.tim.2015.05.004 PMID: 26051014 16. Malmstrom RR, Cottrell MT, Elifantz H, Kirchman DL. Biomass production and assimilation of dissolved organic matter by SAR11 bacteria in the Northwest Atlantic Ocean. Appl Environ Microbiol. 2005; 71: 2979–2986. https://doi.org/10.1128/AEM.71.6.2979-2986.2005 PMID: 15932993 17. Yooseph S, Nealson KH, Rusch DB, McCrow JP, Dupont CL, Kim M, et al. Genomic and functional adaptation in surface ocean planktonic prokaryotes. Nature. 2010; 468: 60–66. https://doi.org/10.1038/ nature09530 PMID: 21048761 18. Gest H. The Modern Myth of “Unculturable” Bacteria/ Scotoma of contemporary microbiology. 2008. Available: https://scholarworks.iu.edu/dspace/bitstream/handle/2022/3149/Microbial%20Growth.pdf; sequence=1 19. Sowell SM, Wilhelm LJ, Norbeck AD, Lipton MS, Nicora CD, Barofsky DF, et al. Transport functions dominate the SAR11 metaproteome at low-nutrient extremes in the Sargasso Sea. ISME J. 2009; 3: 93–105. https://doi.org/10.1038/ismej.2008.83 PMID: 18769456 20. Cottrell MT, Kirchman DL. Transcriptional Control in Marine Copiotrophic and Oligotrophic Bacteria with Streamlined Genomes. Appl Environ Microbiol. 2016; 82: 6010–6018. https://doi.org/10.1128/AEM. 01299-16 PMID: 27474718 21. Noell SE, Giovannoni SJ. SAR11 bacteria have a high affinity and multifunctional glycine betaine trans- porter. Environmental Microbiology. 2019;21. https://doi.org/10.1111/1462-2920.14649 PMID: 31090982 22. Kotrba P, Inui M, Yukawa H. Bacterial phosphotransferase system (PTS) in carbohydrate uptake and control of carbon metabolism. Journal of Bioscience and Bioengineering. 2001; 92: 502–517. https:// doi.org/10.1263/jbb.92.502 PMID: 16233138 23. Schwalbach MS, Tripp HJ, Steindler L, Smith DP, Giovannoni SJ. The presence of the glycolysis operon in SAR11 genomes is positively correlated with ocean productivity. Environmental Microbiology. 2010; 12: 490–500. https://doi.org/10.1111/j.1462-2920.2009.02092.x PMID: 19889000 24. Neidhardt FC, Ingraham JL, Schaechter M. Physiology of the Bacterial Cell: A Molecular Approach. Sinauer Associates; 1990. 25. Tam R, Saier MH. Structural, functional, and evolutionary relationships among extracellular solute-bind- ing receptors of bacteria. Microbiology and Molecular Biology Reviews. 1993; 57: 320–346. 26. Krupka RM. Kinetics of transport systems dependent on periplasmic binding proteins. Biochimica et Biophysica Acta (BBA)—Biomembranes. 1992; 1110: 1–10. https://doi.org/10.1016/0005-2736(92) 90287-V 27. Bohl E, Shuman HA, Boos W. Mathematical treatment of the kinetics of binding protein dependent transport systems reveals that both the substrate loaded and unloaded binding proteins interact with the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 19 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs membrane components. Journal of Theoretical Biology. 1995; 172: 83–94. https://doi.org/10.1006/jtbi. 1995.0006 PMID: 7891451 28. Bosdriesz E, Magnu´sdo´ ttir S, Bruggeman FJ, Teusink B, Molenaar D. Binding proteins enhance spe- cific uptake rate by increasing the substrate-transporter encounter rate. FEBS J. 2015; 282: 2394– 2407. https://doi.org/10.1111/febs.13289 PMID: 25846030 29. Giovannoni SJ. SAR11 Bacteria: The Most Abundant Plankton in the Oceans. Annu Rev Mar Sci. 2017; 9: 231–255. https://doi.org/10.1146/annurev-marine-010814-015934 PMID: 27687974 30. Lehninger AL, Nelson DL, Cox MM. Lehninger Principles of Biochemistry. Macmillan; 2005. 31. Michaelis L, Menten ML, Johnson KA, Goody RS. The original Michaelis constant: translation of the 1913 Michaelis-Menten paper. Biochemistry. 2011; 50: 8264–8269. https://doi.org/10.1021/bi201284u PMID: 21888353 32. Hengge R, Boos W. Maltose and lactose transport in Escherichia coli: Examples of two different types of concentrative transport systems. Biochimica et Biophysica Acta (BBA)—Reviews on Biomembranes. 1983; 737: 443–478. https://doi.org/10.1016/0304-4157(83)90009-6 PMID: 6349688 33. Hosie AHF, Poole PS. Bacterial ABC transporters of amino acids. Research in Microbiology. 2001; 152: 259–270. https://doi.org/10.1016/s0923-2508(01)01197-4 PMID: 11421273 34. Davidson AL, Dassa E, Orelle C, Chen J. Structure, Function, and Evolution of Bacterial ATP-Binding Cassette Systems. Microbiol Mol Biol Rev. 2008; 72: 317–364. https://doi.org/10.1128/MMBR.00031- 07 PMID: 18535149 35. Peebo K, Valgepea K, Maser A, Nahku R, Adamberg K, Vilu R. Proteome reallocation in Escherichia coli with increasing specific growth rate. Molecular BioSystems. 2015; 11: 1184–1193. https://doi.org/ 10.1039/c4mb00721b PMID: 25712329 36. Button DK, Robertson B, Gustafson E, Zhao X. Experimental and Theoretical Bases of Specific Affinity, a Cytoarchitecture-Based Formulation of Nutrient Collection Proposed To Supercede the Michaelis- Menten Paradigm of Microbial Kinetics. Applied and Environmental Microbiology. 2004; 70: 5511– 5521. https://doi.org/10.1128/AEM.70.9.5511-5521.2004 PMID: 15345439 37. Molenaar D, Berlo R van, Ridder D de, Teusink B. Shifts in growth strategies reflect tradeoffs in cellular economics. Molecular Systems Biology. 2009; 5: 323. https://doi.org/10.1038/msb.2009.82 PMID: 19888218 38. Szenk M, Dill KA, de Graff AMR. Why Do Fast-Growing Bacteria Enter Overflow Metabolism? Testing the Membrane Real Estate Hypothesis. Cell Systems. 2017; 5: 95–104. https://doi.org/10.1016/j.cels. 2017.06.005 PMID: 28755958 39. Koch AL. WHAT SIZE SHOULD A BACTERIUM BE? A Question of Scale. Annual Review of Microbiol- ogy. 1996; 50: 317–348. https://doi.org/10.1146/annurev.micro.50.1.317 PMID: 8905083 40. Young KD. The Selective Value of Bacterial Shape. Microbiol Mol Biol Rev. 2006; 70: 660–703. https:// doi.org/10.1128/MMBR.00001-06 PMID: 16959965 41. Armstrong RA. Nutrient uptake rate as a function of cell size and surface transporter density: A Michae- lis-like approximation to the model of Pasciak and Gavis. Deep Sea Research Part I: Oceanographic Research Papers. 2008; 55: 1311–1317. https://doi.org/10.1016/j.dsr.2008.05.004 42. Kaleta C, Scha¨ uble S, Rinas U, Schuster S. Metabolic costs of amino acid and protein production in Escherichia coli. Biotechnology Journal. 2013; 8: 1105–1114. https://doi.org/10.1002/biot.201200267 PMID: 23744758 43. Noor E, Flamholz A, Bar-Even A, Davidi D, Milo R, Liebermeister W. The Protein Cost of Metabolic Fluxes: Prediction from Enzymatic Rate Laws and Cost Minimization. PLOS Computational Biology. 2016; 12: e1005167. https://doi.org/10.1371/journal.pcbi.1005167 PMID: 27812109 44. Basan M. Resource allocation and metabolism: the search for governing principles. Current Opinion in Microbiology. 2018; 45: 77–83. https://doi.org/10.1016/j.mib.2018.02.008 PMID: 29544124 45. Scott M, Klumpp S, Mateescu EM, Hwa T. Emergence of robust growth laws from optimal regulation of ribosome synthesis. Molecular Systems Biology. 2014; 10: 747. https://doi.org/10.15252/msb. 20145379 PMID: 25149558 46. Brass JM, Higgins CF, Foley M, Rugman PA, Birmingham J, Garland PB. Lateral diffusion of proteins in the periplasm of Escherichia coli. J Bacteriol. 1986; 165: 787–795. https://doi.org/10.1128/jb.165.3.787- 795.1986 PMID: 3005237 47. Vijaykumar A, Bolhuis PG, ten Wolde PR. The intrinsic rate constants in diffusion-influenced reactions. Faraday Discussions. 2016; 195: 421–441. https://doi.org/10.1039/c6fd00104a PMID: 27727353 48. Trovato F, Tozzini V. Diffusion within the Cytoplasm: A Mesoscale Model of Interacting Macromole- cules. Biophys J. 2014; 107: 2579–2591. https://doi.org/10.1016/j.bpj.2014.09.043 PMID: 25468337 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 20 / 21 PLOS COMPUTATIONAL BIOLOGYMarine oligotrophs versus copiotrophs 49. Schleif R. AraC protein, regulation of the l-arabinose operon in Escherichia coli, and the light switch mechanism of AraC action. FEMS Microbiology Reviews. 2010; 34: 779–796. https://doi.org/10.1111/j. 1574-6976.2010.00226.x PMID: 20491933 50. Ames GF-L, Lever J. COMPONENTS OF HISTIDINE TRANSPORT: HISTIDINE-BINDING PROTEINS AND hisP PROTEIN. PNAS. 1970; 66: 1096–1103. https://doi.org/10.1073/pnas.66.4.1096 PMID: 4920090 51. 52. Tilman D. Resource Competition between Plankton Algae: An Experimental and Theoretical Approach. Ecology. 1977; 58: 338–348. https://doi.org/10.2307/1935608 Zakem EJ, Levine NM. Systematic Variation in Marine Dissolved Organic Matter Stoichiometry and Remineralization Ratios as a Function of Lability. Global Biogeochemical Cycles. 2019; 33: 1389–1407. https://doi.org/10.1029/2019GB006375 53. Button DK. Biochemical Basis for Whole-Cell Uptake Kinetics: Specific Affinity, Oligotrophic Capacity, and the Meaning of the Michaelis Constant. Appl Environ Microbiol. 1991; 57: 2033–2038. https://doi. org/10.1128/AEM.57.7.2033-2038.1991 PMID: 16348524 54. Schut F, Vries EJ de, Gottschal JC, Robertson BR, Harder W, Prins RA, et al. Isolation of Typical Marine Bacteria by Dilution Culture: Growth, Maintenance, and Characteristics of Isolates under Laboratory Conditions. Appl Environ Microbiol. 1993; 59: 2150–2160. https://doi.org/10.1128/AEM.59.7.2150- 2160.1993 PMID: 16348992 55. Cho J-C, Giovannoni SJ. Cultivation and Growth Characteristics of a Diverse Group of Oligotrophic Marine Gammaproteobacteria. Appl Environ Microbiol. 2004; 70: 432–440. https://doi.org/10.1128/ aem.70.1.432-440.2004 PMID: 14711672 56. Zhao X, Schwartz CL, Pierson J, Giovannoni SJ, McIntosh JR, Nicastro D. Three-Dimensional Struc- ture of the Ultraoligotrophic Marine Bacterium “Candidatus Pelagibacter ubique.” Appl Environ Micro- biol. 2017;83. https://doi.org/10.1128/AEM.02807-16 PMID: 27836840 57. Gudelj I, Beardmore RE, Arkin SS, Maclean RC. Constraints on microbial metabolism drive evolutionary diversification in homogeneous environments. Journal of Evolutionary Biology. 2007; 20: 1882–1889. https://doi.org/10.1111/j.1420-9101.2007.01376.x PMID: 17714305 58. Bosdriesz E, Wortel MT, Haanstra JR, Wagner MJ, Corte´ s P de la T, Teusink B. Low affinity uniporter carrier proteins can increase net substrate uptake rate by reducing efflux. Scientific Reports. 2018; 8: 5576. https://doi.org/10.1038/s41598-018-23528-7 PMID: 29615663 59. Ghuneim L-AJ, Jones DL, Golyshin PN, Golyshina OV. Nano-Sized and Filterable Bacteria and Archaea: Biodiversity and Function. Front Microbiol. 2018;9. https://doi.org/10.3389/fmicb.2018.00009 PMID: 29387050 60. Mitchell JG. The influence of cell size on marine bacterial motility and energetics. Microb Ecol. 1991; 22: 227–238. https://doi.org/10.1007/BF02540225 PMID: 24194338 61. Miller SI, Salama NR. The gram-negative bacterial periplasm: Size matters. PLOS Biology. 2018; 16: e2004935. https://doi.org/10.1371/journal.pbio.2004935 PMID: 29342145 62. Oliver DB. Periplasm. Escherichia coli and Salmonella: Cellular and Molecular Biology. ASM Press; 1996. 63. Morita RY. Starvation-Survival of Heterotrophs in the Marine Environment. Advances in Microbial Ecol- ogy. Springer, Boston, MA; 1982. pp. 171–198. https://doi.org/10.1007/978-1-4615-8318-9_5 64. Stretton S, Danon SJ, Kjelleberg S, Goodman AE. Changes in cell morphology and motility in the marine Vibrio sp. strain S14 during conditions of starvation and recovery. FEMS Microbiol Lett. 1997; 146: 23–29. https://doi.org/10.1111/j.1574-6968.1997.tb10166.x PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009023 May 19, 2021 21 / 21 PLOS COMPUTATIONAL BIOLOGY
10.1371_journal.pcbi.1007770	RESEARCH ARTICLE A framework for integrating directed and undirected annotations to build explanatory models of cis-eQTL data David LamparterID Alice ZhangID, Victor Hanson-SmithID* ¤a, Rajat Bhatnagar, Katja HebestreitID, T. Grant BelgardID ¤b, Verge Genomics, South San Francisco, CA 94080, USA ¤a Current address: Health 2030 Genome Center, 1202 Geneva, Switzerland ¤b Current address: The Bioinformatics CRO, Niceville, FL 32578, USA * victor@vergegenomics.com Abstract A longstanding goal of regulatory genetics is to understand how variants in genome sequences lead to changes in gene expression. Here we present a method named Bayesian Annotation Guided eQTL Analysis (BAGEA), a variational Bayes framework to model cis- eQTLs using directed and undirected genomic annotations. We used BAGEA to integrate directed genomic annotations with eQTL summary statistics from tissues of various origins. This analysis revealed epigenetic marks that are relevant for gene expression in different tis- sues and cell types. We estimated the predictive power of the models that were fitted based on directed genomic annotations. This analysis showed that, depending on the underlying eQTL data used, the directed genomic annotations could predict up to 1.5% of the variance observed in the expression of genes with top nominal eQTL association p-values < 10−7. For genes with estimated effect sizes in the top 25% quantile, up to 5% of the expression vari- ance could be predicted. Based on our results, we recommend the use of BAGEA for the analysis of cis-eQTL data to reveal annotations relevant to expression biology. Author summary Many geneticists wish to map changes in DNA sequences to changes in human traits and to understand these processes mechanistically. Here we present BAGEA, a framework to study this question for gene expression. Specifically, BAGEA models a genome variant’s impact on gene expression based on established genome annotations. BAGEA predicts not only whether a variant has an impact on gene expression, but also the sign of the effect. We applied BAGEA to datasets from different tissues and cell types and found that annota- tions most predictive of gene expression in a given tissue were typically derived from simi- lar tissues. Based on our results, we recommend the use of BAGEA to reveal annotations relevant to expression biology and to build predictive models of gene expression. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Lamparter D, Bhatnagar R, Hebestreit K, Belgard TG, Zhang A, Hanson-Smith V (2020) A framework for integrating directed and undirected annotations to build explanatory models of cis- eQTL data. PLoS Comput Biol 16(6): e1007770. https://doi.org/10.1371/journal.pcbi.1007770 Editor: Roger Pique-Regi, Wayne State University, UNITED STATES Received: June 28, 2019 Accepted: March 3, 2020 Published: June 9, 2020 Peer Review History: PLOS recognizes the benefits of transparency in the peer review process; therefore, we enable the publication of all of the content of peer review and author responses alongside final, published articles. The editorial history of this article is available here: https://doi.org/10.1371/journal.pcbi.1007770 Copyright: © 2020 Lamparter et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: Software package is available on github (https://github.com/dlampart/ bagea/). PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 1 / 27 PLOS COMPUTATIONAL BIOLOGYFunding: This research was supported by Verge Genomics, a venture funded drug discovery company (https://www.vergegenomics.com/). The funders of Verge Genomics had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors of this paper are current and former employees of Verge Genomics, a venture-backed startup company. The authors have declared that no competing interests exist. This does not alter our adherence to all PLOS Computational Biology policies on sharing data and materials. Integrating directed and undirected annotations to build explanatory models of cis-eQTLs Introduction A longstanding goal in the field of genetics is to accurately predict the phenotypic conse- quences of any given variant from the genome sequence alone, i.e. to ‘read the genome’ [1]. This would help to reveal the phenotypic effects of very rare variants even if their effect is weak. The effects of such variants are typically studied via whole genome sequencing studies. However these studies often have limited statistical power because, by definition, there are few carriers in any sampled population [2]. Recently, progress has been made in predicting epigenetic marks and transcription factor (TF) binding from genome sequence alone; these sequence-based models predict the effect of any given sequence variant on epigenetic marks (and TF binding) [3–7]. The question now is how to extend these models to predict effects on genetically complex phenotypes, such as com- mon diseases. A mechanistic stepping stone between the regulation of epigenetic marks and the regulation of complex phenotypes is the regulation of gene expression, as suggested by the previous observation that disease-causing sequence variants are enriched in gene expression quantitative trait loci (eQTLs) [8, 9]. Thus, there is a need for sequence-based models to pre- dict gene expression. One strategy to build sequence-based models of gene expression is to leverage sequence- based models of epigenetic marks. Results of these sequence-based models can be interpreted as directed genome annotations. A genome annotation is defined as a collection of genome regions that have a shared property such as coverage by a particular epigenetic mark, or evolu- tionary conservation across species. Each region can potentially carry an intensity value to denote the annotation strength, such as the strength of conservation. We call such an annota- tion undirecected if its value is independent of the alleles its covering in a given individual. For directed annotations, the sign of its intensity value depends on characteristics of the sequence in the region, such as the presence of a specific allele. A simple motivating example is that of a SNP in a TF binding site. In this situation, the TF can have higher binding affinity for one allele versus the other allele. This can cause consistent directional transcriptional effects: the allele inhibiting binding of an activating TF for instance should lead to decreased expression of the target gene. Conversely, an allele inhibiting binding of a repressive TF would lead to increase in expression, allowing us to discern activators and repressors de novo. One strategy to express this effect as a directional annotation would be to use TF position weight matrices that calcu- late TF affinity for a given sequence, while computationally more sophisticated methods express the same relationship using deep neural networks [3–7]. Methods to evaluate the effect of directed genome annotations on gene expression have recently been proposed [7, 10]. Specifically, Zhou et al. predicted variant impact without exploiting eQTL data using models that predict expression from chromatin patterns directly [7]. Reshef et al. presented a fast method to determine which directed annotations are enriched in variants causal for a given phenotype. However, the method from Reshef et al. is geared towards screening and hypothesis testing rather than towards detailed predictive modeling. For instance, the Reshef model does not account for interactions between the effect of an annotation and the distance to the transcription start site (TSS). Approaches using directed annotations to predict gene expression have been developed rel- atively recently. Methods integrating undirected annotations with eQTL data have a longer history. These methods allow the prior probability distribution of a SNP’s effect size to vary based on the genome annotations with which it overlaps. This is achieved via bayesian hierar- chical models [11–15]. This in turn allows to fine-map causal SNPs, find annotations that are either enriched or depleted in causal SNPs, and increase power to call eQTLs. Methods differ in the modeling assumption they make. For instance, assuming only one causal SNP in a locus PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 2 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs makes the model independent of linkage disequilibrium (LD), thereby simplifying modeling approaches and lower computational burden [11, 12, 15]. Allowing for multiple causal SNPs per locus can improve fine mapping but necessitates the modeling of LD [13, 14]. These types of models have also been employed for integrating functional annotations with GWAS signal [14, 16]. Here we present a new predictive model of gene expression, named Bayesian Annotation Guided eQTL Analysis (BAGEA). BAGEA is a variational Bayes modeling framework to ana- lyze eQTLs using both directed and undirected annotations in a multivariate fashion. BAGEA can model interactions between these annotations by weighting the impact of the directed annotation based on the undirected annotations. Consequently, BAGEA can directly model phenomena relevant to genetic architecture, such as the relatively larger impact of SNPs close to the TSS on directed annotations compared to that of distal SNPs, making BAGEA mores useful for predictive modeling. BAGEA’s results are interpretable and highlight genome anno- tations that are particularly predictive for gene expression. Further, BAGEA can model multi- ple causal SNPs per region. Our software implementation of BAGEA can be run on summary statistics using external LD information as well as on individual level genotype data directly. Optionally, using a low rank approximation of the LD information improves run-time and decreases BAGEA’s memory requirements. We used BAGEA to analyze results from a cis-eQTL meta-analysis in human monocytes and from cis-eQTL summary statistics derived from tissues of various origins [9, 17]. As addi- tional input, we gave the method regulatory impact predictions of common variants on epige- netic marks from a recent deep neural network model [7]. We specified these predictions as directed annotations in the method. We show that BAGEA highlighted biologically sensible annotations as particularly predictive of eQTLs. Further we estimated the predictive power of the directed annotations for various eQTL datasets. Overall, our results suggest that BAGEA is a useful framework to build predictive models of gene expression based on directed annota- tions, find biologically relevant annotations, and benchmark methods that produce such directed annotations. Model overview BAGEA models gene expression as dependent on SNP genotypes in cis. In general, SNP effects on gene expression depend on both directed and undirected annotations (Fig 1A). BAGEA builds predictors of gene expression and ranks annotations by their impact on gene expres- sion. For every gene j, BAGEA takes as input a genotype matrix Xj, an expression vector yj, annotation matrices Vj, F j and Cj. Xj has dimensions (n × mj), where n is the number of indi- viduals assayed, and mj is the number of SNPs in cis of gene j’s TSS. The matrices V j F j and Cj are of dimensions (mj × s), (mj × q), and (mj × t) respectively, where s, q and t are the number of annotations used. BAGEA models gene expression as a linear combination of SNP geno- types: yj ¼ Xjbj þ ϵj; ð1Þ where ϵj is an i.i.d normal noise vector and bj is a vector of SNP effects. The effect of SNP i on gene j bij is modeled as: bij � NððωTvj iÞðνTf j iÞ; a(cid:0) 1 ji Þ; Detailed descriptions of the terms are as follows: PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 ð2Þ 3 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs Fig 1. Illustration of BAGEA model components. (A) The core components of the BAGEA model in the summary statistics formulation. Observed variables are in squares while estimated variables are circled. Given are zj, the eQTL z- scores for gene j, as well as the LD matrix Sj, defining the correlation between summary statistics. Further, z-scores are influenced by the true eQTL effects bj. These effects in turn depend on directed and undirected annotations, Vj and Fj respectively. While undirected annotations can cover regions of any size, directed annotation have the same size as the genomic variants themselves. The impact of annotations on bj is estimated from the data via ω and ν. (B) An example of the modeling of different priors of elements of ω using meta-annotations via υ variable vectors. We assume that directed annotations are available for nine annotations, which were derived from tissues Liver, Blood and Brain via 3 assay types DHS, H3K27ac and H3K4me3. It is reasonable to assume that for a given eQTL study, particular tissues or cell types are more relevant than others. We model this by introducing a variable υ for each tissue (or cell type) that affects the prior distribution of only those elements of ω that are derived from this tissue, e.g. υLiver only affects elements of ω tied to experiments performed in liver. We model different priors for various for assay types analogously. Shown is the resulting network of influences of the variable υtissue, υassay on ω. (We used the actual group names as indices, while in the main text, elements of υ’s and ω are indexed by natural numbers). https://doi.org/10.1371/journal.pcbi.1007770.g001 • vj i encodes the directed annotaton for SNP i in the region of gene j. It is the ith row of Vj. In our applications of BAGEA, Vj is previously computed from sequence-based models, where column of Vj represents an epigenetic mark and each row Vj represents a SNP. Each entry in Vj expresses the predicted effect of a genotype change on the epigenetic mark in question. • f j i encodes the undirected annotations for SNP i in the region of gene j. It is the ith row of Fj. Each element in Fj expresses the presence or absence of the annotation at a SNP’s location. In our applications of BAGEA, Fj is derived from the relative positions of a SNP and gene j’s TSS, where each column represents a particular region around the TSS. For example, if a col- umn Fj encodes a region of 20 kilobases (KB) upstream from the TSS, all entries for rows corresponding to SNPs within 20 KB upstream of that TSS will be set to 1 and entries for all other rows will be set to 0. By default, the first column of Fj is an intercept column consisting only of ones. • ω and ν are vectors of lengths s and q respectively that are estimated by BAGEA. Specifically, ω and ν are the effects of annotations in Fj and Vj on the SNP effects bj. By default, the effect of the interecept weight ν1 is fixed at close to 1 via constraining priors. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 4 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs • αij is a scalar estimated by BAGEA and models the variance of bij. Allowing different vari- ances for the elements of bj typically produces sparse estimates for bj with many elements close to zero as integrating out the Gamma distributed prior αij yields a t-distribution for the different bj [18]. Further, αj, the vector of length mj with αij its ith element, is modeled as dependent on the undirected annotation matrix Cj. Cj can potentially be identical to Fj but can model different undirected annotations as well (see Method Details). Typically, directed annotations are grouped by their cell type or assay type. BAGEA can use this grouping structure in order to select groups of annotations that are useful for predicting gene expression (Fig 1B). BAGEA selects annotation groups via a modeling strategy that yields sparsity on the annotation group level similar to the group lasso [19]. In BAGEA, this grouping strategy is implemented by partitioning annotations into multiple meta-annotations (such as different cell types, assay types etc.). When using this partitioning mechanism, BAGEA includes an extra random variable vector υ of the same length as the number of elements in the partition structure (e.g. the number of cell types, or the number of assay types) (See Methods as well as Fig 1B for an illustrative example). The kth element of υ, υk, controls the variance of the effect sizes for annotations that fall into partitioning group k. Specifically, υk is proportional to the inverse of the variance of the respective elements of ω. u(cid:0) 1 is therefore called the variance k modifier of annotation partition element k (see Methods). Importantly, the model can be reformulated in terms of the summary statistics j yj= and LD matrices Σj ¼ XT zj ¼ XT BAGEA to studies for which only summary statistics are available, by estimating Sj from exter- nal sources (see Methods). j Xj=n. The reformulation enables the application of p ffiffiffi n Evaluation strategy for model fit We developed an approach to evaluate the performance of BAGEA when fitting directed anno- tations to genotype and gene expression data. An important feature of BAGEA is that its results can be used to predict gene expression for a gene without using any expression data for that gene, but rather using genotypes and genome annotations whose weights are fitted from other genes. We can therefore validate BAGEA by training it on gene expression data for one set of genes, and then calculating the extent to which the trained model predicts gene expression for other genes. We propose a so-called directed predictor ^μj, which predicts gene expression for gene j based on knowledge of directed annotations and genotype for gene j. Set ^ηj as the expected mean shift of bj due to the annotations. Using the same notation as in Eqs (1) and (7), we have ^Zij ¼ E½bijjν ¼ ^ν; ω ¼ ^ω� ¼ ð ^ωTvj iÞð^νTf j iÞ; i.e. ^Zij is the ith element of ^ηj . the predictor ^μj is then computed by ^μj ¼ E½yjjν ¼ ^ν; ω ¼ ^ω� ¼ Xj ^ηj: ð3Þ ð4Þ j ^μj measures how much gene expression variance the model The squared magnitude Sj ¼ ^μT attempts to explain via the predictor ^μj. To evaluate the predictor’s accuracy and degree of j ¼ ðyj (cid:0) ^μjÞTðyj (cid:0) ^μjÞ=n. The evalu- overfitting, we use the directed mean squared error MSEdir ation of the predictor is performed on a set of genes independent of the ones used to estimate ω and ν. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 5 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs We can reformulate MSEdir j Σj ¼ XT j Xj=n, and ^ηj: in terms of summary statistics zj ¼ XT j yj= p ffiffiffi n , LD matrices MSEdir j ¼ 1 (cid:0) 2^ηjzj= p ffiffiffi n þ ^ηT j Σj ^ηj; ð5Þ if we assume that yT y = n. In principle, the reformulation allows us to calculate a predictor’s directed mean squared error, even if only summary statistics are available, by approximating Sj from external sources. Directed annotations derived from blood can partially explain cis-eQTLs in monocytes We used BAGEA to determine the extent to which annotations can predict gene expression in CD14 positive monocytes. To this end, we aggregated data from two eQTL studies on expres- sion genetics in CD14 positive monocytes [20, 21]. For directed annotations, we used predic- tions of genetic variant effects on epigenetic marks (12 different histone mark assays and DNase1 Hypersensitivity Site (DHS) calls with 4 different peak calling strategies) in various blood-derived cell types from the pre-trained ExPecto model. ExPecto is a deep learning frame- work that predicts epigenetic marks based on sequence context and performs in silico muta- genesis to evaluate the consequences of sequence variants [7]. ExPecto yielded 2002 directed annotations of which 253 were from blood related cell types. These are referred to as the Blood annotation subset in this paper. We partitioned these directed annotations by cell type and assay type, respectively, and modeled separate prior variance terms for each partition (Fig 1A). To train BAGEA, we used gene expression data from human chromosomes 1 through 15. Only 2410 genes that had a SNP in cis showing a signficant association with a p-value lower than 10−10 were included. To test model fit, we predicted expression for 917 genes on chromo- somes 16 through 22 with a top nominal cis-eQTL p-value below 10−10. Specifically, we used the model fit on the training set to derive the estimates ^ω and ^ν (see Eq 7). We then used these estimates to calculate the directed predictors ^μj for genes on the test set (see Eq 4). To assess the predictive power of ^μj, we calculated MSEdir for every gene in the test set. j We observed that directed genome annotations can partially explain gene expression vari- j j ^μj), such that for the top quartile of ance (Fig 2). The average MSEdir across all genes was 99.5%, which was significantly smaller than 100% (as evaluated by bootstrap sampling genes; p-value smaller than 10−4). MSEdir showed a dependence on predictor size Sj (where Sj ¼ ^μT genes when ranked by Sj, the directed component was estimated to predict 1% to 3% of expres- sion variance (Fig 2A). For each gene, the variance explained is bounded by the additive genetic variance component in cis which is typically much lower than 100%. We estimated the variance of expression explained for each gene in cis in an unbiased way via Haseman-Elston (HE) regression [22]. This approach suggested that around 6.6% of the total genetic variance in cis was explained by the externally fitted directed component ^μj for genes in the top quartile w.r.t Sj (Fig 2B). Across all strong cis-eQTLs, we estimated that the directed component explained 2.5% of total additive genetic variance in cis. We further tested the impact of the dis- tance modifier by constraining all elements of ^ν (except the intercept element) to zero, show- ing that the modeling the distance modifier leads to higher predictive power (S1 Fig). These results show that BAGEA can be used to model how sequence changes affect gene expression. Note that this evaluation metric relies on global parameter (i.e. ω, ν) estimates only. This allows to form predictors for a gene’s expression even if the gene was not included in the training set, but has lower predictive power than approaches that use genewise local parameter estimates (i.e. bj). These approaches can predict expression potentially in an PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 6 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs , the mean squared error (MSE) when predicting gene expression yj from ^μj. To Fig 2. Gene expression variance can be partially explained by directed genome annotations. The BAGEA model was fitted on genes in the training set (all genes on chromosomes 1 through 15) using monocyte eQTL data on genes with a top nominal p-value below 10−10, and with ExPecto-derived directed annotations. ExPecto includes 2002 annotations in total, of which one of two subsets were used: 253 annotations derived from histone and DHS assays in a blood related cell types (Blood), or, alternatively, 690 annotations derived from TF ChIP-Seq (TF). For each gene j in the test set (all genes on chromosomes 16 through 22 with a top nominal p-value below 10−10), we calculated the directed predictor of expression ^μj. As a measure of a predictor’s size, we use its squared magnitude Sj ¼ ^μT performance, we calculated MSEdir j estimate what the smallest attainable MSEdir , the additive genetic variance in cis via Haseman- Elston regression per gene. (A) The relationship between the MSE of the predictor and its squared magnitude. We sorted results by predictor Size Sj and averaged MSEdir j within a sliding window containing 25% of genes and step size of 5% of data. Averaged Directed Predictor Size �S: The mean value of Sj per window on the horizontal axis; Averaged Directed MSE (MSEdir ): The averaged MSEdir j of genes falling into the window on the vertical axis. The 95% confidence interval for each window was derived by bootstrapping. Most variance is explained by genes in the top quartile when ranked by Sj. for genes in the top quartile when ranked by Sj. Genetic Variance (s2 (B) The relationship between MSEdir ): gcis j ) on the vertical axis. 95% The estimated additive genetic variance in cis on the horizontal axis. Directed MSE (MSEdir confidence intervals for the mean of both the MSEdir and s2 gcis confidence interval for the average MSEdir is given by the upper and lower corner, whereas the confidence interval for the average s2 gcis confidence interval shown in grey. is given by the right and left corner respectively). A linear regression is plotted as the blue line, with 95% are represented as the corners of the red diamond (i.e. the j would be, we estimated s2 gcis j ^μj. To evaluate the predictor’s j and s2 gcis https://doi.org/10.1371/journal.pcbi.1007770.g002 out-of-sample fashion, but only for genes in the training set. To illustrate this, we fit BAGEA on a subset of the monocyte samples (134 samples from the Fairfax et al. study) and extracted the local parameter estimates ^bj [20]. We then used these estimates to predict gene expression in the other available monocyte samples [21] [20]. As expected, a substantial fraction of the genetic variance in cis could be predicted using the local parameter estimates (S2 Fig). Further, using BAEGA in an annotation uninformed manner lowered the variance explained. Joint modeling of cis-eQTLs and directed annotations highlights biologically relevant epigenetic marks We next evaluated if BAGEA can effectively be used to discover which annotations, or groups of annotations, are most predictive of gene expression. We grouped the directed annotations by cell type and assay type, and for each set of annotation groups, we modeled separate prior variance modifiers υ−1 (Fig 1B). For each annotation group k we measured its contribution to gene expression as its estimated variance modifier u(cid:0) 1 k (See Model Overview). For the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 7 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs monocyte data, BAGEA estimated the largest variance modifiers for annotations from DHS as well as H3K27ac and H3K4me3 assays (Fig 3A). This observation is consistent with results from a previous method, using undirected annotations, suggesting that SNPs with an effect on gene expression are enriched in open chromatin (DHS), activated enhancers and promoters (H3K27Ac, H3K4me3) [11]. Across cell type annotations, BAGEA estimated the largest vari- ance modifiers for annotations from two blood cell types that were both CD14 positive (Fig 3B). This observation matches our expectations because the cells in the underlying expression data were derived from CD14 positive cells [20, 21]. Across all tested pairs of assays and cell types, BAGEA estimated the largest positive effect sizes for annotations from DHS, H3K27ac, H3K4me3 assays in CD14 positive cells (Fig 3C). Additionally, we saw a negative effect size for DHS assayed in CD3 positive cells, albeit with lower absolute effect size than CD14 positive cells. One explanation could be that DHS that occur in CD14 but not CD3 positive cells have larger predictive value than DHS that occur in both cell types. It is well known that eQTLs occur more likely and increase in effect size closer to the TSS. This suggests that the effects of directed annotations might also be bigger for SNPs close to the TSS than for SNPs that are distal. BAGEA models SNP distance dependence of directed anno- tation effects by weighting the directed annotation effect term Vjω across SNPs, with a distance modifier Fjν (see Model Overview). We next tested whether BAGEA estimated directed anno- tation effect sizes to be dependent on a SNP’s distance to the TSS. We examined the value of a SNP’s estimated distance modifier Fj^ν against its position relative to the TSS. We observed a characteristic peak around the TSS (Fig 3D), suggesting that BAGEA can indeed produce a similar pattern of distance dependence for the effect sizes derived from directed annotations as for the eQTL effect sizes themselves. We repeated this analysis with a different set of directed annotations, namely 690 ExPecto annotations derived from transcription factor (TF) ChIP-Seq in any cell type. We estimated the TF annotation subset to be similarly predictive of gene expression as the Blood annotation subset (Fig 2A). When looking at the estimates of ω, MYC assayed in the cell line NB4 had the largest effect size among all tested 690 annotations (S3 Fig). Additionally, SPI1, MAX CTCF had effects larger than 10% of the maximal effect size. For SPI1 and CTCF, effects from multi- ple cell lines reached this threshold with consistent effect size directions. NB4 is a promyelocytic leukemia cell line that can be differentiated into neutrophils or monocytes [23]. NB4 is therefore expected to have similar expression genetics as CD14 positive monocytes, and, given that no TF ChIP-Seq experiment was performed in monocyte cell lines directly, the large ω values for NB4 data are consistent with our expectations. However, inter- pretations of cell type selection for the TF subset are complicated by the fact that the underly- ing TF ChIP-seq experiments did not sample each TF comprehensively across all cell lines which might lead to biases. When checking expression of the highlighted TFs in monocytes via Protein Atlas, we found all were TFs classified as expressed but not elevated in monocytes [24]. Effect size directions of CTCF and MAX were negative, which naively interpreted would sug- gest that their binding have a suppressive effect on gene expression. CTCF can act as repressor, activator and insulator [25]. Our data suggests that, globally and in the studied context, repres- sive effects outweigh. MAX and MYC are part of a family of TFs that form heterodimers [26]. The MYC/MAX dimer is usually regarded as an activator, which might seem to be at odds with our results as the MAX annotation had a negative effect size. However, another important fam- ily member MXD1 (also known as MAD) was not assayed. The MAD/MAX heterodimer is thought to act as a repressor. A positive effect size for MYC and a smaller negative effect size for MAX could just imply that MAX binding in absence of concomitant MYC binding has a negative effect on expression because it tracks with MAD/MAX heterodimer binding. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 8 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs Fig 3. BAGEA, fitted on monocyte eQTL data, selects relevant epigenetic marks and increases directional effect sizes for SNPs close to a TSS. Parameter estimates when applying BAGEA to monocyte eQTL data using as directed annotations histone and DHS ExPecto predictions derived from blood-related cell types (i.e. Blood from Fig 2). (A) For each chromatin assay type, BAGEA models an assay variance modifier ^u (cid:0) 1 assay that captures the extent to which that assay type is predictive of gene expression. Shown are the square roots for the assay types with the ten highest variance modifiers (from 17 assay types total). In the BAGEA model, DHS, H3K27Ac and H3K4me3 assays have the largest modifiers. (B) For each cell type, BAGEA models a cell type variance modifier ^u(cid:0) 1 in panel A. Shown are the square roots for the cell types with the ten highest variance modifiers (out of 61 cell types). In the BAGEA model, CD14 positive cells have the largest modifiers. (C) BAGEA reveals experiments underlying the directed annotations that were most predictive of gene expression. Assay Type x Cell Type: Each experiment is a particular assay type performed in a particular cell type. Effect Size ( ^o i, for experiment i): The BAGEA-estimated effect on gene expression. Shown are the ten largest directed annotation effect sizes. In the BAGEA model, the experiments using DHS, H3k27Ac and H3Kme4 with CD14 positive cells have the largest effect sizes. We also see that most of the 253 annotations are estimated to have a close to zero effect. (D) Shown is the estimated distance modifier of the directed component, F^ν. We see a characteristic peak around the TSS, implying that the directed annotations are upweighted close to the TSS. cell, similar to the assay variance modifier https://doi.org/10.1371/journal.pcbi.1007770.g003 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 9 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs Developmentally, during monocyte/macrophage maturation, a switch from high levels of MYC/MAX to MAD/MAX is observed [27]. A further highlighted transcription factor is SPI1 which is regarded as one of the most important transcription factor in monocyte and macro- phage development [28]. Modeling directional components is robust to the use of summary statistics In many cases it is not feasible to compute LD for the population from which the summary sta- tistics were derived (i.e., the study population), and LD has to be derived from other sources (i.e., external genotypes) [29, 30]. The use of external genotypes allows publicly available sum- mary statistics to be analyzed without access to restricted individual level genotype data [9]. However, LD computed on external genotypes can only approximate LD patterns of the study population. We therefore need to test the accuracy of methods when using external genotypes. We evaluated if directed annotation effects were robust to the genetic source of LD infor- mation. We used 1000 Genomes data to compute LD [31]. We re-fit the BAGEA model to the monocyte data with the Blood annotation subset, using LD matrices derived from European 1000 Genomes data. We then compared ω estimates when using LD from 1000 Genomes to ω estimates when using LD from the monocyte data itself, for every annotation in the monocyte Blood data. We observed that the two approaches produced similar effect sizes with a linear regression R2 of 97.5% and regression slope of 0.96 (S4A Fig). This suggests that directed anno- tation effect estimates are robust to the source of LD information. We then explored if the source of LD information affected our estimates of directed mean squared error (MSEdir). To this end, we estimated MSEdir on chromosomes 16 through 22 from summary statistics and external LD matrices derived from 1000 Genomes alone, and then compared these MSEdir val- ues to the original MSEdir values computed with LD derived from monocyte data. We ensured that the same SNPs were included, by removing SNPs with low minor allele frequency (MAF) in either of the sets. We observed that the two sources of LD produced MSEdir values that agree with each other, with a linear regression R2 of 99.9% and regression slope of 1.002 (S4B Fig). Exploring BAGEA’s ability to identify causal marks through simulation To explore BAGEA’s ability to select the causal annotations among the set of annotations, we used simulation (see S1 Appendix). Naturally, this depends on the correlation structure of the directed annotations, as highly correlated annotations will make it difficult to isolate the causal one. We therefore used empirically observed directed annotations in our simulation experi- ment. We assumed a model were the truly causal annotations were sparse (with the non-zero effects varying from 3 to 12). We tested cases where the non-zero effects clustered in terms of meta-annotations (e.g. clustered in terms of cell types and assay type) (structured) and cases where non-zero effects did not cluster (unstructured). While fitting BAGEA we also used two parameter settings, either making use of the meta-annotations available for cell type and assay type, (group-lasso) or ignoring the meta-annotation information and letting each ωi parameter be controlled by a separate υi parameter (lasso). While we saw high recovery of the causal annotations for lower number of causal variables, precision and recall tended to drop as more but individually smaller non-zero effects were added (S5 Fig). Drop-off was fastest when pair- ing unstructured data generation, with the group-lasso fitting procedure, presumably, because this parameter setting tried to enforce a structure that was not actually present. Conversely, structured data generation paired with group-lasso fitting procedure showed the highest perfor- mance of all settings. When evaluating the gene expression prediction power of the model fits, we saw that in all cases the results were close to optimal, suggesting that even in higher PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 10 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs complexity settings when incorrect variables get picked, the chosen variables are highly corre- lated with the correct ones (S6 Fig). Building predictive expression models from GTEx summary statistics Having established that BAGEA performs well when using summary statistics, we next deter- mined if BAGEA can identify relevant directed annotations for empirical data for which sum- mary statistics are available but genotypes are not. Specifically, we fit BAGEA on summary statistics for eQTL studies of 13 tissues produced by the GTEx consortium with a sample size of at least 300 for each study [9]. We additionally supplemented this set with results for Lym- phoblastoid cell lines (LCL) derived from a meta-analysis of GTEx and GEAUVADIS [17]. Because GTEx gathered eQTLs in complex tissues and sampled fewer individuals than were sampled in the monocyte studies, we expected lower power to produce robust parameter esti- mates. We therefore used different parameter values than in our monocyte analysis, including genes with top nominal cis-eQTL p-value lower than 10−7. We fitted models either using ExPecto derived annotation for all 1187 histone or DHS annotations derived from Roadmap consortium data or the derived annotations for non-histone ChIP-seq data from ENCODE [32, 33]. When using Roadmap annotations we used BAGEA in the group-lasso setting, whereas for ENCODE annotations we used the lasso setting, the rationale being, that the Road- map consortium performed most assays for a given cell type, whereas ENCODE ChIP-seq was less complete, i.e. many TFs were assayed in only few cell lines, leading to potentially strong biases. We again split genes into training and test set, fitting BAGEA on the training set and build- ing directed expression predictors ^μj for all genes in the test set. We observed that the average MSEdir per dataset was variable across GTEx datasets ranging from 100% to below 98.5% (Fig 4A). When looking at only the highest quartile of genes in terms of squared magnitude Sj, we saw the lowest average MSEdir go to approximately 0.95 (Fig 4B). Furthermore the gains in average MSEdir were in line with squared magnitude Sj values, suggesting that BAGEA does not substantially overfit. We saw that the ENCODE TF annotation set tended to outperform the histone and DNase1 Roadmap set and that in the Roadmap group lasso setting, BAGEA would not always select any annotations, potentially due to poor overlap between annotation and GTEx cell types and stringent regularization. We next compared the predictive power achieved by BAGEA on the GTEx data to results derived via ExPecto directly. To predict expression from genomic variants, the authors of ExPecto propose a strategy, where results from two statistical models are combined. The first model is a deep neural network that predicts the impact of genomic sequence variants on chro- matin marks (the results of which are also used by BAGEA in this analysis). The second model is a l2-boosting model that predicts gene expression from (spatially transformed) chromatin marks directly. as part of the ExPecto release, results of the second model were already publicly available for 13 relevant GTEx datasets [4]. Combining these results with the directed annota- tions and the z-scores from GTEx, allowed us to estimate the scalar product between the gene epression vector yj and the corresponding directed predictor ^μj (see S1 Appendix). This allowed us to compare model quality in terms of the fraction of genes with a scalar product larger than zero. When comparing results from BAGEA (using TF annotation subset and lasso setting) and ExPecto (using all annotations) in terms of this metric, we saw that, while perfor- mance was comparable across all genes, BAEGA substantially outperformed ExPecto for genes in the highest quantiles in terms of effect size (S7 Fig). We further compared BAGEA to Torus, a tool which allows to model SNP effect priors in terms of undirected annotations [14, 15]. We therefore made our annotations undirected by PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 11 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs Fig 4. Directed annotations partially explain gene expression variance in GTEx. The BAGEA model was fit using various GTEx eQTL data (supplemented with GEAUVADIS eQTL data) and with ExPecto-derived directed annotations on genes in the trainig set (chr1,‥,chr15) with a top nominal p-value<10−7. ExPecto includes 2002 total annotations, of which either 1187 histone and DHS annotations from Roadmap (Roadmap) or 690 non-histone ChIP-Seq from ENCODE (TF) were used. For the Roadmap annotation set we enforced structure on the priors of ω by using the meta- annotations available for cell type and assay type, (group-lasso), while for the (TF annotation set, each ωi parameter be controlled by its individual υi parameter (lasso). For each gene j in the test set (chr16,‥,chr22 and top nominal p-value< 10−7), we calculated an approximate version of Sj, the squared magnitude of the directed predictor ^μj, where the approximation uses external LD information. Further, we calculated an approximate version of MSEdir j error (MSE) when predicting gene expression yj from ^μj. (A) Displayed is the average (approximated) MSEdir genes for each GTEx experiment, and annotation subset. 95% Confidence intervals are computed by bootstrap sampling. (B) For each GTEx experiment and annotation subset, we sorted results by predictor size Sj and and averaged MSEdir j within the top quartile. Displayed is the relationship between the MSE of the predictor and its mean squared magnitude Sj. Averaged Sj, top quartile �Sj jSj > F(cid:0) 1 ð0:75Þ: The mean value of the directed predictor size Sj in the top quartile on the horizontal axis; Averaged Directed MSE (MSEdir ): The averaged MSEdir j of genes falling into the top quartile in terms of Sj on the vertical axis. The 95% confidence interval for each window was derived by bootstrap sampling. We see that the average squared magnitude Sj is of similar size as the gains in directed MSE suggesting that the BAGEA does not substantially overfit. , the mean squared j across all Sj https://doi.org/10.1371/journal.pcbi.1007770.g004 taking absolute values and adding the undirected annotations used in BAGEA to generate annotations for Torus (see S1 Appendix). We fit data from 13 GTEx experiment on the TF annotation subset filtering SNPs and genes as for BAGEA. Varying amounts of l2 regulariza- tion were applied and an additional overfit strategy was added to have an upper bound on results achieved with an optimal regularization strategy (see S1 Appendix). As an evaluation strategy we recorded for each gene in the test set the SNP with the highest prior. As an evalua- tion metric, we used the fraction of genes for which that SNP had an absolute z-score close to the largest absolute z-score for that gene. We saw that, while BAGEA had a slightly higher value than even the overfit strategy, the increases were modest (S8 Fig). As Torus was run in ridge mode, few effects were very close to zero. BAGEA in its default parameter resembles lasso, in that it only selects a limited number of effect sizes substantially different from zero. To compare the variables selected, we split the distribution torus effect sizes estimates into 3 groups based on whether BAGEA estimated them as (close to) zero, negative or positive. We saw that the distribution of torus estimated effect sizes was substantially shifted for the non- zero BAGEA effect groups compared to the zero BAGEA effect group (S9 Fig). PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 12 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs Fig 5. Model fit for GTEx summary statistics selects directional annotations mainly from biologically consistent cell types. Shown here are various parameter estimates from fitting 13 different GTEx eQTL summary statistics data (supplemented with GEAUVADIS eQTL data) using histone and DHS ExPecto predictions derived from Roadmap (1187 annotations). (A) BAGEA reveals the experiments underlying the directed annotations that are most predictive of gene expression. GTEx x Roadmap(Rm): Each GTEx eQTL dataset highlights particular Roadmap annotations. Shown here are the 10 largest positive effect sizes across all eQTL and annotation pairings. Effect Size: The estimate of ^o i for experiment i. (B) For each chromatin assay type, BAGEA models an assay variance modifier ^u (cid:0) 1 to which that assay type is predictive of gene expression. Shown here is the distribution of the square roots of the assay variance modifier for any given assay type across all 13 GTEx eQTL datasets. Results are sorted by the maximal value achieved for each assay type and only the 10 highest scoring assay types are shown. We see that DNase.all.peaks H3K27ac annotations dominate. The DNase.fdr0.01.peaks was prioritized in Lung tissue, which had the lowest value for DNase.all. peaks among all experiments. The highest value in DNase.all.peaks was achieved in Fibroblast, an experiment that also showed low average MSEdir. assay that expresses the extent https://doi.org/10.1371/journal.pcbi.1007770.g005 GTEx expression models make use of biologically relevant annotations To mitigate the impact of limited power during variable selection, we additionally fit models without splitting chromosomes into test and validation sets. Exploring the results from the Roadmap annotation subset first, we saw that the distribution of effect sizes of the directional annotations revealed a bias towards positive values (S10 Fig). Focusing on the largest positive effect sizes(top ten or ^oi > 0:06), we saw many biologically consistent pairings between the tissue assayed by GTEx via eQTL and the tissue assayed by Roadmap for epigenetic marks (Fig 5A). While some of the pairings are obvious from the annotation names themselves (such as correct pairings for lymphoblastoid cells, lung and adipose tissues) others were suprising yet on closer inspection, turned out to be consistent with the biological literature. For instance bone marrow derived mesenchymal stem cells (BMD MSC) are paired with fibroblast. A recent study found no functional differences between the two cell types leading the authors to support a longstanding opinion in the field that these two cell types should be classified as the same [34, 35]. The pairing between Esophagus Mucosa and keratinocytes can be explained by the fact that the Esophagus Mucosa is mainly composed of squamous cells, i.e. keratinocytes [36, 37]. The pairing between tibial artery and BMD MSC can be explained by the fact that fibro- blasts are the main component of vascular adventitia [38]. Our model also paired tibial nerve and muscle, which seems physiologically the least biologically consistent among the ten pair- ings. When looking at the largest negative values, we saw some of the same tissue pairings repeated, with only one pairing with effect size ^oi smaller than -0.06 for the pairing between PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 13 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs fibroblasts and BMD MSC) (S11 Fig). When looking at the variance modifier estimates for the different assay types, we saw that DHS and H3K27ac epigenetic marks were ranked consis- tently highly (Fig 5B). Interestingly, among various annotations derived from the same DHS experiments, some performed consistently better than others: DNase1 peak call annotations outperformed DNase1 hotspots calls. These two annotation types make use of the same under- lying DNAse-seq data, but use different data processing pipelines [39]. Hotspots calls have var- iable length and are typically wider than peak calls which have a fixed length of 150 bp. BAGEA makes a clear choice which should be preferred to model sequence impacts on gene expression. Turning our attention to the results for the non-histone ChIP-Seq ENCODE anno- tation subset, we saw for most GTEx experiments, that the largest effect size was associated with a Pol2 assay (S12 Fig). As all annotations are fit together, it is possible, that large Pol2 effects could obscure transcription factor action as Pol2 binding could be a result of the bind- ing of other factors. We therefore removed all Pol2 assay annotations, and refit the model again. We then counted the number GTEx experiment for which a given TF showed either a substantial positive or negative effect (S13A Fig). When looking for TFs That showed negative effect sizes in at least half of the assayed experiments, we found EZH2, SIN3A and ZEB1, which all have substantial literature support for being transcriptional repressors [40–42]. TFs that showed substantial positive effects in at least half of the experiments, we saw PHF8 and ELF1. PHF8 is known to demethylate H3K9me2, a mark strongly associated with transcriptional repression [43]. The fact that this assay shows significant positive effects in most analysed GTEx experiments, might highlight an underappreciated importance of this mode of tran- scriptional control. ELF1 has been cited in the literature as having both activator and repressor properties [44, 45]. Our results suggest that activator properties outweigh. To systematically evaluate whether our results are in line with the literature, we compared them to the most recent human version of TRRUST, literature database of regulatory interactions between TFs and their targets [46]. Interaction in TRRUST are annotated as repressive or activating in nature. We derived a TFs repressor activity score based on the fraction of annotated interac- tions defined as repressive. We derived a second TF repressor actvity score as the number of positive effects minus the number of negative effects (S13B Fig). We saw a strong correlation between these scores (p-value below 0.0001, R2 = 0.43). When removing the 3 bona-fide repressors, results remained significant at the 0.05, level albeit less so (one-tail p-value below 0.015, R2 = 0.15). Additionally, we saw that while distance modifier did vary between fits, the characteristic peak around the TSS was replicated in all cases (S14 Fig). Discussion Here we introduced a new method, named Bayesian Annotation Guided eQTL Analysis (BAGEA). BAGEA integrates directed and undirected genome annotations in a multivariate fashion with eQTL data in a variational Bayesian framework to build predictive models of gene expression. We applied this method to eQTL results from CD14 positive monocytes as follows: First, we derived directed annotations by predicting functional impacts on epigenetic marks for all common SNPs using the pre-trained ExPecto deep neural net [7]. Second, from these ExPecto results, we extracted two annotation subsets of particular interest: histone ChIP-Seq and DHS in blood-derived cell types (the Blood annotation subset), and TF ChIP- Seq in any cell type (the TF annotation subset). We then ran BAGEA on both annotation sub- sets separately, while allowing the effect of the directed annotations to depend on the distance to the TSS. We tested whether the model had explanatory power with a training and test proto- col (i.e. explanatory power was estimated on genes that were excluded from training). We saw that the directed component μ of the model explained part of the gene expression variance in a PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 14 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs statistically significant manner (Fig 2A). For genes with a strong cis-eQTL (p-value<10−10) and in the top quartile for μTμ, we estimated that the Blood derived directed component explained 6.6% of total additive genetic variance in cis (Fig 2B). Importantly, BAGEA priori- tized annotations that cohere with widely accepted biological knowledge and are supported by existing literature (Fig 3). Next, we investigated to which extent the model fit was affected when the LD information was approximated via reference genomes. We observed agreement between the results in terms of the directed component, suggesting that the use of eQTL summary statistics together with external LD data is justified (S4 Fig). We next used simulation to investigate, whether BAGEA is able to reliably detect causal annotations and found that to be the case if the generat- ing model was sufficiently sparse S5 Fig. We then used BAGEA to analyze eQTL summary sta- tistics results from GTEx. To accommodate the wide range of tissues explored in GTEx, we expanded the number of directed annotations used in the fitting process to over a thousand. While for some tissues, the analysis strategy was underpowered to derive a predictive model of gene expression from directed annotations, others had a significant fraction of gene expression explained by directed annotations (Fig 4). We compared results from BAGEA to both ExPecto for the ability predicting gene expression from annotations alone and Torus for predicting causal SNPs from annotations alone. In terms of predicting gene expression BAGEA showed favourable performance for genes with large effect sizes in terms of Sj. While BAGEA did show improvement over Torus w.r.t. prediction of causal SNPs, the improvement was marginal. One plausible explanation is that distance to TSS is already a very strong predictor of causal SNP location. Many of the directed annotations BAGEA selected were derived from tissues that were biologically related to the original tissue of the eQTL studies (Fig 5A). Additionally, we observed that DNAse1 and H3K27ac epigenetic marks were selected across many different eQTL studies (Fig 5B). Furthermore, we used the results to classify TFs de novo into transcrip- tional activators and repressors, showcasing an application that relies crucially on directed annotations S3 Fig. BAGEA belongs to a class of models that allow the prior probability distribution of a SNP’s effect size to vary based on the genome annotations with which it overlaps [11–13]. These prior models explored the impact of undirected annotations. While BAGEA can model undi- rected annotations, the main novelty comes from the concomitant modeling of directed and undirected annotations as well as interactions thereof. Using directed annotations to explain natural variation in phenotypes was also recently proposed by both Zou et al. and Reshef et al., albeit with different modeling philosophies [7, 10]. Zou et al. use a model that predicts expres- sion from chromatin patterns directly. This has the advantage that genotype data is not needed. However, this method does not model the causal impact of epigenetic marks on expression levels but rather correlations between them. This modeling approach therefore assumes a priori that causality flows from epigenetic marks to gene expression. However, recent integrative analysis modeling causality between expression and chromatin marks sug- gest that this is not always the case as expression can itself reorganize proximal epigenetic pat- terns [47]. Reshef et al.’s LD profile regression method has more similarities to BAGEA as it can also be used to analyze directed annotations and eQTL summary statistics. However, the method is geared towards multiple hypothesis testing rather than high predictive accuracy. Compared to BAGEA, the fitted model is simpler allowing for fast analysis of large collections of data. The increased speed comes at the cost of not being able to model certain features like interactions of directed and undirected annotations (such as distance to TSS). BAGEA uses a modeling approach that has both prediction and interpretability in mind. It allows for more complex model features while still being useful for revealing relevant biology. Indeed, when using PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 15 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs BAGEA on various eQTL datasets, BAGEA highlighted many relevant cell types. Further, allowing the directed component to depend on the distance to the TSS improved the model fit S1 Fig. There are at least two drawbacks to BAGEA’s model complexity. First, there is a substantial computational cost to fit the model. To mitigate this issue, we used various computational strategies such as fast matrix inversion of approximated LD matrices and parallelization (see S15 Fig for an overview of the duration of various analyses). Second, variational model fitting approach does not provide confidence intervals. While it does provide credibility intervals, the approximative nature of mean field variational inference makes these credibility intervals often unreliable [48]. In our analysis, we opted for evaluating statistical significance of the model results by using a training and test protocol. Future research could investigate whether using a different variational approximation rather than the mean field approximation provides better estimates of the true credibility inter- vals. An avenue also not explored here, is to learn across datasets by adjusting the priors. The bayesian nature of our framework offers a simple iterative strategy here: After fitting various datasets once, adjust the priors according to the results seen across the different datasets and refit. Whether this strategy can substantially improve predictive power remains to be seen. Further, while the method can be extended to predict expression effects of rare variants, we focused here on prediction of relatively common SNPs. With the cost of whole genome sequencing (WGS) dropping, WGS eQTL studies suited for this purpose should become widely available. We estimated the extent to which epigenetic marks are able to predict the genetic compo- nent of gene expression in cis. Our results show that while the current generation of directed annotations can partially explain the genetic cis component of gene expression, most of the genetic cis component remains unexplained, indicating that there is still room for improve- ment. Future gains in this space will likely come from both improved directed annotations as well as improved modeling. Methods Model details We assume individual level genotype and expression data for n individuals. For gene j, we model its n × 1 expression vector yj as yj ¼ Xjbj þ ϵj; ð6Þ where Xj is the n × mj genotype matrix for the mj SNPs surrounding gene j’s TSS. bj is the mj × 1 vector of SNP effect sizes and ϵj the expression noise unexplained by the genotype. The noise term precision λj is modeled in a hierarchical fashion: ϵj � Nnð0; ðljÞ(cid:0) 1InÞ: lj � Gðl1; l2Þ; l2 � Gðr1; r2Þ: with hyperparameters λ1, ρ1 and ρ2 (while this notation is overloaded, we expect it is clear from context which parameter is meant). We model the vector of effect sizes bj as a multivari- ate normal, whose mean and covariance is affected by annotation matrices. For gene j we assume annotation matrix Fj and a directed continuous annotation matrix Vj, with PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 16 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs dimensions mj × q and mj × s respectively, with the current implementation of BAGEA expect- ing Fj to be 0 − 1 coded due to performance reasons. Then the ith element of bj is modeled as bij � NððωTvj iÞðνTf j iÞ; a(cid:0) 1 ji Þ; ð7Þ i and f j i being the ith row of Vj and Fj respectively. αij being an element of a vector of with vj independently drawn gamma distributed random variables (the independence is conditional on its parental hyperparameters, the modeling of which is described further down). ω and ν are s and q dimensional multivariate normal distributed random variables respectively. ω denotes the vector of activities of directed annotations, whereas ν allows the overall weight that the directed annotations contribute to the effect size vary based on undirected annotations. This allows, for instance, the impact of the directed annotations to vary dependent on the dis- tance to the TSS. ω is modeled in a hierarchical fashion ω � Nsð0; diagðδ(cid:0) 1ÞÞ; where δ is again modeled as a random variable. The choice of model for δ enables the imple- mentation of a grouping structure on the directional annotations (in our application, these groupings are the assay used to derive the annotation and the cell type in which the assay was performed). We allow the model to fit differences in prior variances based on group member- ship. Thereby, entire groups of directional annotation effects are shrunk to zero (akin to the group lasso [19]). Let dl be a positive integer vector of length s taking hl different values, i.e dl partitions the vector of directed annotations into hl groups (l = 1, ‥, w runs over the meta- annotations, e.g. if the modeled meta-annotations are cell type and assay type, l can either take the value one or two). Let υl be a random vector of length hl (i.e. these are the group specific weights). Then, di ¼ Yw l¼1 ; ul dl i ul k ¼ Gðw1l; w2lÞ; with hyperparameter χ1l. χ2l is modeled as with hyperparameters z1 and z2. ν is modeled as w2j � Gðz1; z2Þ; ν � Nqðc; diagðpÞ(cid:0) 1Þ; where p and c are hyperparameter vectors of length q. The vector of precisions of the effect size vector αj is modeled as aij � Gðg1; kjgijÞ; where γ1 is a hyperparameter. Note that letting the precision for each SNP vary leads to sparse estimates for bj; this is akin to automatic relevance determination (ARD) regression [18]. κj is a genewise parameter modeled in a hierarchical fashion kj � Gðt1; t2Þ; t2 � Gðx1; x2Þ; PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 17 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs where τ1, ξ1 and ξ2 are hyperparameters. To model γij, we again make use of annotation matri- ces. For gene j, assume undirected 0 − 1 coded annotation matrix Cj of dimension m × t (BAGEA currently only 0 − 1 coded matrices for Cj coded due to performance reasons). Then the SNP-wise precision modifier γij is modeled as Y gij ¼ ak k:C j ik ¼1 where Cj ik ¼ 1 if annotation k is active at index i in gene region j. Further, ak ¼ Gð�1; �2Þ; where ϕ1 and ϕ2 are hyperparameters. Summary statistics adaptation Instead of using individual level genotype and expression data, we can reformulate the model for the use of summary statistics. Multiplying Eq 6 with 1ffiffi p XT gives n 1 p XTϵj: ffiffiffi n 1 p XTXjbj þ ffiffiffi n 1 p XTyj ¼ ffiffiffi n A natural model to use with summary statistics is therefore, p ffiffiffi n zj ¼ Σjbj þ ϵ0 j; where zj is the vector of summary statistics, Sj is the LD matrix and ϵ0 be approximated from external sources such as 1KG [31]. Alternatively, we can use an approxi- mate and regularized version of the empirical LD matrix (see below). j ΣjÞ. Sj can j � Nmð0; l(cid:0) 1 Model fitting The model was fit using a variational bayes approach [48]. As the model is in the conjugate exponential family, we can use the variational message passing strategy [49]. For detailed updating steps see S1 Appendix. Naive updates can be prohibitively expensive due to the requirement to invert many large matrices of the form (cXT X + Dα), where c is a constant and Dα is a diagonal matrix. To speed up computation, we can approximate the LD matrix cXT X with a low rank approximation AT t At, where At is a t × m matrix with t < m. This allows us to speed up a time critical matrix inversion step. ðcXTX þ DαÞ(cid:0) 1 � D(cid:0) 1 α (cid:0) D(cid:0) 1 α AtðIt þ AT t D(cid:0) 1 α AtÞ(cid:0) 1AT t D(cid:0) 1 α : If X is already low rank, it is computationally advantageous to use an At s.t. cXTX ¼ AT t At. If t At deviates from cXTX, we need to use the summary statistics formulation to avoid conver- AT gence issues. For more detail, see S1 Appendix. Deriving annotations For common SNPs (minor allele frequency (MAF) above 2.5% in the 1000 Genomes European population [31]), we ran the ExPecto model to predict the effect of the variant on epigenetic marks [7]. For each SNP we predicted the epigenetic effects within the 200 bp region encom- passing it. For most SNPs the effects are very close to zero, allowing us to sparsify the results. Absolute effects smaller than 0.008 were set to zero and all other effects were shrunk towards zero by 0.008 via xnew = x − 0.008 � sgn(x). Next, results for both strands were averaged and the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 18 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs shrinking procedure repeated with a threshold of 0.008. This yielded a matrix with 98.4% of entries zero. The directed annotations were then scaled to have all the same 2-norm. The mag- nitude of the 2-norm was set to the average of the unscaled 2-norms. These were the directed annotations used in BAGEA. For undirected annotations, we used upstream and downstream distances to the TSS. Dis- tance to TSS annotations as well as SNP positional annotations were downloaded from the UCSC genome annotation database with SNP and gene annotations taken from the refGene and snp147Common tables respectively (see link below) [50]. cis-eQTL datasets For monocyte eQTL data, we used two preprocessed monocyte datasets with a combined sam- ple size of 1176 (418 from Fairfax et al. and 758 from Rotival et al. respectively) [20, 21]. Expression matrices were quantile normalized and 10 PEER factors as well as 5 genotype PCs removed [51]. Genotype data was quality control filtered (4% SNP level missingness; 5% indi- vidual level missingness; Hardy-Weinberg p-value above 10−13 relatedness below 0.1875) and imputed using the human genome reference panel [52]. We further downloaded eQTL summary statistics for various tissues produced by the GTEx project if the number of samples was above 300 individuals [9]. Additionally, for LCL, we meta-analyzed eQTL summary statistics released for 117 samples by GTEx with summary sta- tistics derived from 358 European PEER-controlled samples collected as part of the GEUVA- DIS study [17]. Running BAGEA For the monocyte eQTL analysis, BAGEA was run with default hyperparameter settings (see S1 Appendix). Genotypes within a window of 150KB around a gene’s TSS were used to con- struct a genewise LD matrix. Each genewise LD matrix was approximated via singular value decompostion with a low rank symmetric matrix of equal top eigenvalues and eigenvectors, such that the trace of the approximation matrix was at least 99% of the trace of the original LD matrix. Then, a scaled identity matrix was added such that the trace of the resulting matrix was equal to the trace of the original LD matrix. As undirected annotations, distance windows around the TSS (50KB, 20KB, 10KB, 5KB, 2KB, 1KB, 0.5KB, 0.25KB) split into upstream and downstream windows were used. To analyse summary statistics with BAGEA, LD was approxi- mated via 1KG European samples. Variants where the reference allele in 1KG did not agree with the reference allele in the UCSC SNP annotation table, were removed. Reference 1KG LD matrices were calculated and replaced with low rank approximations with 95% of the matrix trace kept, anlagously to the above procedure. For all GTEx summary statistics analysis, default hyperparameter settings where used except for c which was set to 0.3 instead of 0 to yield con- sistently positive signs for ν estimates. BAGEA was run for 300 iterations in each analysis. URLs • Code to run BAGEA can be found at https://github.com/dlampart/bagea • Auxiliary preprocessed data automatically installed by BAGEA is downloaded from https:// s3-us-west-1.amazonaws.com/bagea-data/bagea_data_freeze/ Links to publicly available external data sources are as follows: • UCSC: http://hgdownload.cse.ucsc.edu/goldenpath/hg19/database/ PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 19 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs • ExPecto: https://github.com/FunctionLab/ExPecto/ • 1KG: ftp://ftp-trace.ncbi.nih.gov/1000genomes/ftp/release/20130502/ • GTEX: https://gtexportal.org/home/datasets • GEUVADIS: ftp://ftp.ebi.ac.uk/pub/databases/microarray/data/experiment/GEUV/ E-GEUV-3/analysis_results/ • TRRUST-db: https://www.grnpedia.org/trrust/ • Protein atlas: https://www.proteinatlas.org • First monocyte study: https://www.ebi.ac.uk/ega/studies/EGAS00001000411 • Second monocyte study: https://www.ebi.ac.uk/ega/studies/EGAS00000000109 Supporting information S1 Appendix. Supporting methods. (PDF) S1 Fig. Removing distance modifier leads to lower predictive power. We used BAGEA to predict gene expresion for CD14 positive monocytes using the Blood annotation set, analo- gously to Fig 2 but removing the distance modifier by constraining all elements of ^ν except the intercept element to zero (Without Distance Modifier). For comparison, we additionally plot- ted results achieved with the same data and settings except using the default prior for ν (With Distance Modifier). We see substantial decrease in power when the distance dependence of the effect sizes is not modeled. (TIFF) S2 Fig. BAGEA SNP effect size estimates predict gene expression out of sample. SNP effect size estimates ^bj were derived from a subsample (134 individuals) of one dataset [20]. These estimates were used to predict gene expression in the other available monocyte samples [20] [21]. Fits were performed for genes on chromosomes 1 to 22 that had at least a marginal eQTL p-value of 10−10 or below in the combined data. Shown is the average estimated expression var- iance explained in the test data using the Blood annotation subset and default distance annota- tions. BAGEA was run either with the default parameter setting (BAGEA With Annotations), or with annotation uniformed setting where a was constrained close to 1 and ω was con- strained close to 0 (BAGEA Without Annotations). Additionally, we compared those estimates to estimates of average genetic variance explained in cis as estimated by Haseman-Elston regression on the test data. 95% confidence intervals were derived by bootstrap sampling genes. We see that 60% of estimated genetic variance in cis is explained by BAGEA out-of-sam- ple estimates of ^bj . Further, running BAGEA in annotation uninformed mode dropped this fraction to 0.567%. Overall, we saw that 62.5% of assayed genes had a lower MSE in the anno- tation informed model than in the annotation uninformed model. (TIFF) S3 Fig. Parameter estimates for the directed annotation TF subset when using BAGEA on monocyte eQTL data. Shown are parameter estimates from fitting monocyte eQTL data using TF ExPecto predictions in all cell types. (A) BAGEA reveals the experiments underlying the directed annotations that are most predictive of gene expression. Assay Type x Cell Type: Each experiment is a particular assay type performed in a particular cell type. Effect Size ( ^oi, PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 20 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs for experiment i): The BAGEA-estimated effect on gene expression. Shown here the ten largest directed annotation effect sizes. We see c-Myc annotation in NB4 dominates. (B) Shown is the estimated distance modifier of the directed component, F^ν. We see a characteristic peak around the TSS, implying that the directed annotations are upweighted close to the TSS. (TIFF) S4 Fig. Directed annotation effect estimates and modeling error are robust to source of LD information. (A) Shown is a comparison of estimates of the directed annotation effect vector ω when using external reference LD information or individual level genotypes. We retrained BAGEA with the blood monocyte summary statistics using reference LD matrices from the 1000 Genomes Project (1KG). ^oi (1KG): Directed annotation effect, measured as ω estimates from BAGEA using 1KG reference LD information. ^oi (Monocyte LD): Directed annotation effect, measured as ω estimates from BAGEA using individual-level genotypes from the mono- cyte data itself (i.e. using the same genotypes as for the deriving the summary statistics). (B) To investigate the extent to which MSEdir j can be approximated using summary statistics and refer- ence 1KG LD matrices, we calculated MSEdir tics of monocyte cis-eQTLs (see formula in main text). We then compared these to the original j values that were computed using genotypes of the monocyte datasets. The same SNPs MSEdir were used in both calculations. R2: The coefficient of determination, measuring goodness-of- fit, from a linear regression of the data shown. (TIFF) j on chromosomes 16 to 22 from summary statis- S5 Fig. Simulation results confirm BAGEA’s ability to recover relevant annotations. Shown are precision and recall for various simulation settings (see S1 Appendix) and two parameter settings. For each simulation setting we fitted BAGEA either making use of the meta-annota- tions available for cell type and assay type, (group-lasso) or ignoring the meta-annotation infor- mation and letting each ωi parameter be controlled individual υi parameter (lasso). Upper panels: shown are example results when fitting BAGEA either in group-lasso setting (A) or lasso setting (B), in the structured simulation setting with 9 variables (see S1 Appendix for sim- ulation details). True effect sizes for ω are indicated via red dots. Scaled BAGEA estimates of ω are given as black lines (We scaled ω to account for differences in estimates of ^ν versus ν. We multiplied each element of ν by the coverage of its associated annotations and summed the resulting vector. We treated the estimate ν analogously and divided the two to get the scaling factor for ω. These scaling factors where 0.83 and 0.90 for the group-lasso (A) and lasso (B) set- ting respectively). When defining all scaled effect size estimates above 0.01 as positives and below 0.01 as negatives, we see that both settings yield a precision of one, whereas group-lasso also had a precision of 1.0 and recall of 0.88 and lasso had a precision of 0.83 and recall of 0.55 (five out of nine annotations recovered, one false positive). When looking at precision (C) and recall (D) across all simulation settings, we see that precision and recall drop as more variable are added. As expected, in an unstructured simulation setting, it is disadvantageous to enforce a structure on the estimates via the group-lasso setting. On the other hand, group-lasso main- tains good precision and recall in a structured setting with up to 12 variables. (TIFF) S6 Fig. Predictive power of BAGEA for various simulation settings show little deviations form predictive power achieved for true parameter settings. Shown are average MSEdir j val- ues for all genes in the test set (chromosomes 3). The Upper panel shows average MSEdir j across all test genes, whereas the lower panel shows average MSEdir j terms of Sj. We see that the performance is very close to optimal even for settings where for genes in the top quartile in PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 21 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs BAGEA did not select the correct variables, suggesting that the selected variables were highly correlated to the correct variables. (TIFF) S7 Fig. Comparing BAGEA (TF) results with ExPecto result for GTEx data shows favour- able performance for genes with large predicted effect sizes. Shown is the comparison between BAGEA and ExPecto for 13 the GTEx experiments w.r.t. agreement between gene expression and the estimated directed predictor. (Sj > quantile x (ðSj > F(cid:0) 1 Sj experiments, genes were sorted by the squared magnitude Sj (Sj were computed for both BAGEA and ExPecto separately, i.e. for ExPecto SExP j was used). For each GTEx experiment, the top n-th percent of genes w.r.t Sj were then used to calculate the proportion of genes with j ^μj): The proportion of genes for which the scalar product positive yT between the gene expression vector yj and the directed predictor ^μj was larger than 0. We see that for genes with large relative effect sizes, BAGEA leads to higher concordance between yj and ^μj. 95% confidence band is derived by bootstrap sampling. (TIFF) j ^μj. (Proportion with yT ðxÞÞ): Per GTEx S8 Fig. Comparing BAGEA (TF) results with Torus result for GTEx data shows comparable performance. Shown is the comparison between BAGEA and Torus for 13 the GTEx experi- ments. To evaluate a method, we determined for each gene in the test set the SNP with the highest prior of being causal. For Torus, this amounted to ranking SNPs based on the scalar product between the SNP’s annotations and their estimated effect sizes. For BAGEA, we ranked SNPs based on E½b2 ijjG�, where G refers to all global BAGEA parameter estimates (see S1 Appen- dix). (fitting method): The various methods used in the fitting and evaluation. For Torus we used various l2 parameter settings as well as an overfitting strategy as upper bound (see S1 Appendix for details) [14, 15]. (Proportion with \|ztop\| − \|zannot\| < 0.2): To evaluate a given method, we picked the SNP for each gene in the test set for which the method predicted the largest absolute effect sizes based on the annotations alone and recorded its z-score (denoted \|zannot\|). We then compared this value to the overall largest absolute z-score for this gene (denoted \|ztop\|). We evaluated the power by the proportion of genes for which \|ztop\| − \|zannot\| was below 0.2. 95% confidence interval is derived by bootstrap sampling. (TIFF) S9 Fig. Comparison of effect size estimates between BAGEA and Torus. Shown are the dis- tribution of effect size estimates of Torus when fitted on 13 GTEx datasets for the TF annota- tion subset (l2 = 100). As Torus was run in ridge mode, few effects were very close to zero. BAGEA in its default parameter resembles lasso, in that it only selects a limited number of effect sizes substantially different from zero. When fitting BAGEA using the same datasets in lasso mode, we saw 257 annotations overall larger than 0.001 (of which 192 where also larger than 0.01). When color-coding those 257 effect sizes based on direction and comparing them against the rest, we saw that the Torus effect sizes were markedly shifted away from zero for both the positive and the negative effect size BAGEA group. (TIFF) S10 Fig. Histogram of directed effect sizes ^ω across all 14 GTEx datasets. Displayed are esti- mated directed annotation effect sizes ^o for all GTEx (and GEAUVADIS) datasets, with values with absolute value below 10−3 removed. Shown are results when fitting on data from all auto- somes. (TIFF) PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 22 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs S11 Fig. Largest negative directed annotation effect sizes for GTEx summary statistics repeat some of the same tissue pairings as large positive effect sizes. Shown are the largest negative directed annotation effect from fitting 14 different GTEx (and GEAUVADIS) eQTL summary statistics datasets using Histone and DHS ExPecto predictions derived from Road- map. (TIFF) S12 Fig. BAGEA fits for GTEx summary statistics with non-histone ChIP-seq annotations show largest effect sizes for Pol2 assays. Shown is the largest directed annotation effect for each of the fitted 14 different GTEx (and GEAUVADIS) eQTL summary statistics datasets using ExPecto predictions derived from ENCODE non-histone ChIP-seq experiments [33]. (TIFF) S13 Fig. BAGEA fits with TF annotations predict TF activators and repressors. After removal of Pol2 from the TF annotations subset, we fit BAGEA to the GTEx summary statistics using lasso mode. (A) Shown are the number of GTEx experiments for which a given TF-ChIP-Seq assay shows a postively or negatively signed effect with absolute value above 0.01 (If multiple annotations mapped to the same TF we summed the effects, this step only affected a few TFs because of the regularization strategy employed). (B) Shown is the comparison between BAGEA’s prediction of repressor/activator activity of a TF’s with predictions derived from the trrust-db v2 [46]. (sign bias of effect size direction (#)): For each TF we take the dif- ference between the number of postive effect directions (blue bar in panel (A)) and the number of negative effect directions (red bar in panel (A)) to get a prediction of whether a TF acts as activator (>0) or repressor (<0). (Fraction of Repressor Annotation (TRRUST-db)): The fraction of annotations in the human TTRUST db (human) for a given TF which claimed repressor activity among all annotations with a clear assigned direction (i.e. after removal of all annotations with unknown direction from TRRUST-db). We see a clear dependence between results from TRRUST (unidirectional p-value lower than 0.0001, R2 = 0.43), suggesting that results from BAGEA are in broad agreement with the literature in terms of determining activa- tor and repressor TFs. (TIFF) S14 Fig. Comparison of distance modifier estimates for BAGEA fits on GTEx data. Shown is the estimated distance modifier of the directed component, F^ν for all GTEx experiments when fit with the TF annotation subset in lasso mode. Individual results are plotted in grey and averages are plotted in black. While there is some fluctuation for individual results around the mean, the general peak shape is respected in all cases. (TIFF) S15 Fig. Speed of variable update varies across annotation subset. Shown is the speed with wich each updating iteration of the variational algorithm takes for the main analyses per- formed. For Blood we used the setting of the monocyte analysis (see for instance Fig 2). For TF and Histone/DHS we used the settings used in the respective GTEx analyses (see for instance Fig 4). All analyses were performed on an AWS r4 × 4 instance using 15 cores. As we ran the algorithm for 300 iterations, we see that in this setting, the algorithm took between 50 minutes and 5 and a half hours to complete. (TIFF) S1 Table. BAGEA effect size estimates for GTEx experiments. Given are the ω effect size estimates for various BAGEA fits to GTEx data. Only effect sizes with absolute value above PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 23 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs 0.001 are included. (TXT) S2 Table. BAGEA effect size estimates for monocyte experiment. Given are the ω effect size estimates for various BAGEA fits to the monocyte data. Only effect sizes with absolute value above 0.001 are included. (TXT) S3 Table. Directed Annotations to tissue/cell type mapping. Given are the mappings between the ExPecto annotations and the Roadmap tissues, as well as ENCODE cell lines. (TXT) S4 Table. MSEdir estimates for GTEx experiments. Given are the MSEdir on the test set for various BAGEA fits to the GTEx data. (TXT) S5 Table. MSEdir estimates for monocyte experiments. Given are the MSEdir on the test set for various BAGEA fits to the monocyte data. (TXT) Acknowledgments Special thanks to Prof. Zoltan Kutalik for helpful discussions. Author Contributions Conceptualization: David Lamparter, Rajat Bhatnagar, Katja Hebestreit, T. Grant Belgard, Victor Hanson-Smith. Formal analysis: David Lamparter. Funding acquisition: Alice Zhang. Methodology: David Lamparter. Project administration: Victor Hanson-Smith. Software: David Lamparter. Writing – original draft: David Lamparter, Rajat Bhatnagar, Katja Hebestreit, T. Grant Bel- gard, Alice Zhang, Victor Hanson-Smith. References 1. Pai AA, Pritchard JK, Gilad Y. The genetic and mechanistic basis for variation in gene regulation. PLoS Genet. 2015; 11(1):e1004857. https://doi.org/10.1371/journal.pgen.1004857 PMID: 25569255 2. Timpson NJ, Greenwood CMT, Soranzo N, Lawson DJ, Richards JB. Genetic architecture: the shape of the genetic contribution to human traits and disease. Nat Rev Genet. 2018; 19(2):110–124. https:// doi.org/10.1038/nrg.2017.101 PMID: 29225335 3. Moyerbrailean GA, Kalita CA, Harvey CT, Wen X, Luca F, Pique-Regi R. Which Genetics Variants in DNase-Seq Footprints Are More Likely to Alter Binding? PLoS Genet. 2016; 12(2):e1005875. https:// doi.org/10.1371/journal.pgen.1005875 PMID: 26901046 4. 5. Zhou J, Troyanskaya OG. Predicting effects of noncoding variants with deep learning-based sequence model. Nat Methods. 2015; 12(10):931–4. https://doi.org/10.1038/nmeth.3547 PMID: 26301843 Lee D, Gorkin DU, Baker M, Strober BJ, Asoni AL, McCallion AS, et al. A method to predict the impact of regulatory variants from DNA sequence. Nat Genet. 2015; 47(8):955–61. https://doi.org/10.1038/ng. 3331 PMID: 26075791 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 24 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs 6. Kelley DR, Snoek J, Rinn JL. Basset: learning the regulatory code of the accessible genome with deep convolutional neural networks. Genome Res. 2016; 26(7):990–9. https://doi.org/10.1101/gr.200535. 115 PMID: 27197224 7. Zhou J, Theesfeld CL, Yao K, Chen KM, Wong AK, Troyanskaya OG. Deep learning sequence-based ab initio prediction of variant effects on expression and disease risk. Nat Genet. 2018; 50(8):1171– 1179. https://doi.org/10.1038/s41588-018-0160-6 PMID: 30013180 8. Gusev A, Mancuso N, Won H, Kousi M, Finucane HK, Reshef Y, et al. Transcriptome-wide association study of schizophrenia and chromatin activity yields mechanistic disease insights. Nat Genet. 2018; 50(4):538–548. https://doi.org/10.1038/s41588-018-0092-1 PMID: 29632383 9. Gamazon ER, Segrè AV, van de Bunt M, Wen X, Xi HS, Hormozdiari F, et al. Using an atlas of gene reg- ulation across 44 human tissues to inform complex disease- and trait-associated variation. Nat Genet. 2018; 50(7):956–967. https://doi.org/10.1038/s41588-018-0154-4 PMID: 29955180 10. Reshef YA, Finucane HK, Kelley DR, Gusev A, Kotliar D, Ulirsch JC, et al. Detecting genome-wide directional effects of transcription factor binding on polygenic disease risk. Nat Genet. 2018; 50 (10):1483–1493. https://doi.org/10.1038/s41588-018-0196-7 PMID: 30177862 11. Veyrieras JB, Kudaravalli S, Kim SY, Dermitzakis ET, Gilad Y, Stephens M, et al. High-resolution map- ping of expression-QTLs yields insight into human gene regulation. PLoS Genet. 2008; 4(10): e1000214. https://doi.org/10.1371/journal.pgen.1000214 PMID: 18846210 12. Gaffney DJ, Veyrieras JB, Degner JF, Pique-Regi R, Pai AA, Crawford GE, et al. Dissecting the regula- tory architecture of gene expression QTLs. Genome Biol. 2012; 13(1):R7. https://doi.org/10.1186/gb- 2012-13-1-r7 PMID: 22293038 13. Das A, Morley M, Moravec CS, Tang WHW, Hakonarson H, MAGNet Consortium, et al. Bayesian inte- gration of genetics and epigenetics detects causal regulatory SNPs underlying expression variability. Nat Commun. 2015; 6:8555. https://doi.org/10.1038/ncomms9555 PMID: 26456756 14. Wen X, Lee Y, Luca F, Pique-Regi R. Efficient Integrative Multi-SNP Association Analysis via Determin- istic Approximation of Posteriors. Am J Hum Genet. 2016; 98(6):1114–1129. https://doi.org/10.1016/j. ajhg.2016.03.029 PMID: 27236919 15. Wen X, et al. Molecular QTL discovery incorporating genomic annotations using Bayesian false discov- ery rate control. The Annals of Applied Statistics. 2016; 10(3):1619–1638. https://doi.org/10.1214/16- AOAS952 16. Pickrell JK, Marioni JC, Pai AA, Degner JF, Engelhardt BE, Nkadori E, et al. Understanding mecha- nisms underlying human gene expression variation with RNA sequencing. Nature. 2010; 464 (7289):768–72. https://doi.org/10.1038/nature08872 PMID: 20220758 17. Lappalainen T, Sammeth M, Friedla¨ nder MR, ’t Hoen PAC, Monlong J, Rivas MA, et al. Transcriptome and genome sequencing uncovers functional variation in humans. Nature. 2013; 501(7468):506–11. https://doi.org/10.1038/nature12531 PMID: 24037378 18. Bishop CM. Pattern Recognition and Machine Learning. Springer; 2006. 19. Yuan M, Lin Y. Model selection and estimation in regression with grouped variables. Journal of the Royal Statistical Society: Series B (Statistical Methodology). 2006; 68(1):49–67. https://doi.org/10. 1111/j.1467-9868.2005.00532.x 20. Fairfax BP, Humburg P, Makino S, Naranbhai V, Wong D, Lau E, et al. Innate immune activity condi- tions the effect of regulatory variants upon monocyte gene expression. Science. 2014; 343 (6175):1246949. https://doi.org/10.1126/science.1246949 PMID: 24604202 21. Rotival M, Zeller T, Wild PS, Maouche S, Szymczak S, Schillert A, et al. Integrating genome-wide genetic variations and monocyte expression data reveals trans-regulated gene modules in humans. PLoS Genet. 2011; 7(12):e1002367. https://doi.org/10.1371/journal.pgen.1002367 PMID: 22144904 22. Haseman JK, Elston RC. The investigation of linkage between a quantitative trait and a marker locus. Behav Genet. 1972; 2(1):3–19. https://doi.org/10.1007/BF01066731 PMID: 4157472 23. Clark CS, Konyer JE, Meckling KA. 1alpha,25-dihydroxyvitamin D3 and bryostatin-1 synergize to induce monocytic differentiation of NB4 acute promyelocytic leukemia cells by modulating cell cycle pro- gression. Exp Cell Res. 2004; 294(1):301–11. https://doi.org/10.1016/j.yexcr.2003.11.013 PMID: 14980523 24. Uhle´n M, Fagerberg L, Hallstro¨ m BM, Lindskog C, Oksvold P, Mardinoglu A, et al. Proteomics. Tissue- based map of the human proteome. Science. 2015; 347(6220):1260419. https://doi.org/10.1126/ science.1260419 PMID: 25613900 25. Holwerda SJB, de Laat W. CTCF: the protein, the binding partners, the binding sites and their chromatin loops. Philos Trans R Soc Lond B Biol Sci. 2013; 368(1620):20120369. https://doi.org/10.1098/rstb. 2012.0369 PMID: 23650640 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 25 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs 26. Grandori C, Cowley SM, James LP, Eisenman RN. The Myc/Max/Mad network and the transcriptional control of cell behavior. Annu Rev Cell Dev Biol. 2000; 16:653–99. https://doi.org/10.1146/annurev. cellbio.16.1.653 PMID: 11031250 27. Ayer DE, Eisenman RN. A switch from Myc:Max to Mad:Max heterocomplexes accompanies mono- cyte/macrophage differentiation. Genes Dev. 1993; 7(11):2110–9. https://doi.org/10.1101/gad.7.11. 2110 PMID: 8224841 28. Kurotaki D, Sasaki H, Tamura T. Transcriptional control of monocyte and macrophage development. Int Immunol. 2017; 29(3):97–107. https://doi.org/10.1093/intimm/dxx016 PMID: 28379391 29. Yang J, Ferreira T, Morris AP, Medland SE, Genetic Investigation of ANthropometric Traits (GIANT) Consortium, DIAbetes Genetics Replication And Meta-analysis (DIAGRAM) Consortium, et al. Condi- tional and joint multiple-SNP analysis of GWAS summary statistics identifies additional variants influ- encing complex traits. Nat Genet. 2012; 44(4):369–75, S1–3. https://doi.org/10.1038/ng.2213 PMID: 22426310 30. Ehret GB, Lamparter D, Hoggart CJ, Genetic Investigation of Anthropometric Traits Consortium, Whitta- ker JC, Beckmann JS, et al. A multi-SNP locus-association method reveals a substantial fraction of the missing heritability. Am J Hum Genet. 2012; 91(5):863–71. https://doi.org/10.1016/j.ajhg.2012.09.013 PMID: 23122585 31. 1000 Genomes Project Consortium, Auton A, Brooks LD, Durbin RM, Garrison EP, Kang HM, et al. A global reference for human genetic variation. Nature. 2015; 526(7571):68–74. https://doi.org/10.1038/ nature15393 PMID: 26432245 32. Roadmap Epigenomics Consortium, Kundaje A, Meuleman W, Ernst J, Bilenky M, Yen A, et al. Integra- tive analysis of 111 reference human epigenomes. Nature. 2015; 518(7539):317–30. https://doi.org/10. 1038/nature14248 PMID: 25693563 33. ENCODE Project Consortium. An integrated encyclopedia of DNA elements in the human genome. Nature. 2012; 489(7414):57–74. https://doi.org/10.1038/nature11247 34. Hematti P. Mesenchymal stromal cells and fibroblasts: a case of mistaken identity? Cytotherapy. 2012; 14(5):516–21. https://doi.org/10.3109/14653249.2012.677822 PMID: 22458957 35. Denu RA, Nemcek S, Bloom DD, Goodrich AD, Kim J, Mosher DF, et al. Fibroblasts and Mesenchymal Stromal/Stem Cells Are Phenotypically Indistinguishable. Acta Haematol. 2016; 136(2):85–97. https:// doi.org/10.1159/000445096 PMID: 27188909 36. Whelan KA, Muir AB, Nakagawa H. Esophageal 3D Culture Systems as Modeling Tools in Esophageal Epithelial Pathobiology and Personalized Medicine. Cell Mol Gastroenterol Hepatol. 2018; 5(4):461– 478. https://doi.org/10.1016/j.jcmgh.2018.01.011 PMID: 29713660 37. Doupe´ DP, Alcolea MP, Roshan A, Zhang G, Klein AM, Simons BD, et al. A single progenitor population switches behavior to maintain and repair esophageal epithelium. Science. 2012; 337(6098):1091–3. https://doi.org/10.1126/science.1218835 PMID: 22821983 38. Di Wang H, Ra¨ tsep MT, Chapman A, Boyd R. Adventitial fibroblasts in vascular structure and function: the role of oxidative stress and beyond. Can J Physiol Pharmacol. 2010; 88(3):177–86. https://doi.org/ 10.1139/Y10-015 PMID: 20393583 39. Thurman RE, Rynes E, Humbert R, Vierstra J, Maurano MT, Haugen E, et al. The accessible chromatin landscape of the human genome. Nature. 2012; 489(7414):75–82. https://doi.org/10.1038/nature11232 PMID: 22955617 40. Ding X, Wang X, Sontag S, Qin J, Wanek P, Lin Q, et al. The polycomb protein Ezh2 impacts on induced pluripotent stem cell generation. Stem Cells Dev. 2014; 23(9):931–40. https://doi.org/10.1089/scd. 2013.0267 PMID: 24325319 41. Yang X, Zhang F, Kudlow JE. Recruitment of O-GlcNAc transferase to promoters by corepressor mSin3A: coupling protein O-GlcNAcylation to transcriptional repression. Cell. 2002; 110(1):69–80 PMID: 12150998 42. Papadopoulou V, Postigo A, Sa´nchez-Tillo´ E, Porter ACG, Wagner SD. ZEB1 and CtBP form a repres- sive complex at a distal promoter element of the BCL6 locus. Biochem J. 2010; 427(3):541–50. https:// doi.org/10.1042/BJ20091578 PMID: 20175752 43. Zhu Z, Wang Y, Li X, Wang Y, Xu L, Wang X, et al. PHF8 is a histone H3K9me2 demethylase regulating rRNA synthesis. Cell Res. 2010; 20(7):794–801. https://doi.org/10.1038/cr.2010.75 PMID: 20531378 44. Song H, Suehiro Ji, Kanki Y, Kawai Y, Inoue K, Daida H, et al. Critical role for GATA3 in mediating Tie2 expression and function in large vessel endothelial cells. J Biol Chem. 2009; 284(42):29109–24. https:// doi.org/10.1074/jbc.M109.041145 PMID: 19674970 45. Sugi Y, Takahashi K, Nakano K, Hosono A, Kaminogawa S. Transcription of the Tollip gene is elevated in intestinal epithelial cells through impaired O-GlcNAcylation-dependent nuclear translocation of the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 26 / 27 PLOS COMPUTATIONAL BIOLOGYIntegrating directed and undirected annotations to build explanatory models of cis-eQTLs negative regulator Elf-1. Biochem Biophys Res Commun. 2011; 412(4):704–9. https://doi.org/10.1016/ j.bbrc.2011.08.035 PMID: 21867680 46. Han H, Cho JW, Lee S, Yun A, Kim H, Bae D, et al. TRRUST v2: an expanded reference database of human and mouse transcriptional regulatory interactions. Nucleic Acids Res. 2018; 46(D1):D380– D386. https://doi.org/10.1093/nar/gkx1013 PMID: 29087512 47. Delaneau O, Zazhytska M, Borel C, Giannuzzi G, Rey G, Howald C, et al. Chromatin three-dimensional interactions mediate genetic effects on gene expression. Science. 2019; 364 (6439). https://doi.org/10. 1126/science.aat8266 PMID: 31048460 48. Blei DM, Kucukelbir A, McAuliffe JD. Variational inference: A review for statisticians. Journal of the American Statistical Association. 2017; 112(518):859–877. https://doi.org/10.1080/01621459.2017. 1285773 49. Winn J, Bishop CM. Variational message passing. Journal of Machine Learning Research. 2005; 6 (Apr):661–694. 50. Haeussler M, Zweig AS, Tyner C, Speir ML, Rosenbloom KR, Raney BJ, et al. The UCSC Genome Browser database: 2019 update. Nucleic Acids Res. 2019; 47(D1):D853–D858. https://doi.org/10. 1093/nar/gky1095 PMID: 30407534 51. Stegle O, Parts L, Piipari M, Winn J, Durbin R. Using probabilistic estimation of expression residuals (PEER) to obtain increased power and interpretability of gene expression analyses. Nat Protoc. 2012; 7 (3):500–7. https://doi.org/10.1038/nprot.2011.457 PMID: 22343431 52. McCarthy S, Das S, Kretzschmar W, Delaneau O, Wood AR, Teumer A, et al. A reference panel of 64,976 haplotypes for genotype imputation. Nat Genet. 2016; 48(10):1279–83. https://doi.org/10.1038/ ng.3643 PMID: 27548312 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1007770 June 9, 2020 27 / 27 PLOS COMPUTATIONAL BIOLOGY
10.1186_s40359-021-00605-7	Yung et al. BMC Psychol (2021) 9:104 https://doi.org/10.1186/s40359-021-00605-7 RESEARCH ARTICLE Open Access Understanding the experiences of hikikomori through the lens of the CHIME framework: connectedness, hope and optimism, identity, meaning in life, and empowerment; systematic review Jolene Y. K. Yung1* , Victor Wong2, Grace W. K. Ho3 and Alex Molassiotis4 Abstract Background: Hikikomori is a phenomenon describing people who exhibit behaviors of self-secluding themselves at home for long durations of time and usually only having face-to-face social interactions with none other than family. Existing interventions for hikikomori are inconclusive and the majority are absent in using a theoretical framework to guide its components. Therefore, applicability of the psychosocial recovery framework of Connectedness, Hope and Optimism, Identity, Meaning in Life, and Empowerment (CHIME) towards hikikomori care was reviewed. Method: Five databases were searched in April 2020 with the search formula from a published systematic review on hikikomori combined with search terms specific to domains of the CHIME framework. Articles included in the review were of the English language, of all publication years, peer-reviewed, quantitative or qualitative research studies and case studies, included study designs that were observational or interventional in nature, and involved populations of socially withdrawn youth. Results: CHIME’s comprehensive structure and organized approach could guide researchers or service providers in determining areas needing assessments, measurement, and areas of focus. It is suggested that the CHIME framework is applicable after modifying a specific dimension—‘meaning of mental illness experiences’ into ‘meaning of the hikikomori experience’. Thematic overlap occurred between the domains of connectedness, identity, and meaning. Yet, additional dimensions or domains such as trust building, non-linearity, and spatiality can be included for address- ing specific limitations in this application, which would help towards catering services to help hikikomori in recovery or in increasing quality-of-life of those individuals’ while entrapped in this withdrawn lifestyle. Conclusion: CHIME framework could be applicable towards hikikomori care after applying the suggested modifica- tions. Additionally, many knowledge gaps were found in literature during this review that warrants further investiga- tion to improve hikikomori care. Keywords: Hikikomori, Youth in social withdrawal, CHIME, Recovery, Connectedness, Hope and optimism, Identity, Meaning in life, Empowerment *Correspondence: jolene-yk.yung@connect.polyu.hk 1 A130, School of Nursing, The Hong Kong Polytechnic University, Hung Hom, Kowloon, HKSAR, China Full list of author information is available at the end of the article © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Yung et al. BMC Psychol (2021) 9:104 Page 2 of 30 Background Hikikomori is a term that originated in Japan and was used as early as the 1990s [1] to describe people who socially withdraw from society or the phenomenon of their doing so [2]. Other terms, such as socially with- drawn youth [3] or hidden youth [4], have been used to describe this phenomenon in other places. Individuals with this condition seclude themselves at home for six months or longer, refrain from going to work or school, and do not maintain friendships [5]. These individu- als live reclusive lifestyles and usually have face-to-face interactions only with family members [6]. Although the common definition of hikikomori refers to self-seclu- sion for the mentioned extended period of time, some researchers and organizations have suggested a lower threshold of three months of self-seclusion to aid in early detection and treatment [3, 6, 7]. While Hamasaki et al. [8] have suggested hikikomori as a spectrum continuum of social withdrawal (hikikomori) severity. Five reviews of research on hikikomori were identified through a literature search. One was a systematic review and the other four were narrative reviews. The aim of the systematic review was to consolidate available research evidence on hikikomori, not to raise research questions or hypotheses [9]. Similar topics discussed in the reviews were the definition, etiology, and diagnosis of hikikomori, and interventions to treat the condition. Three definitions were mentioned in the reviews, with the major differ- ences between them arising from the inclusion of indi- viduals with psychiatric conditions [10, 11] or of those who might leave their home but avoid social interactions [10]. There was also further categorical differentiation into primary and secondary hikikomori, namely, those without psychiatric comorbidity and those with psychi- atric comorbidities, respectively [9]. In the reviews, the following similar etiologies were described: adverse or traumatic childhood experiences, bullying, peer rejec- tion, dysfunctional family dynamics, changes in the labor market [11, 12], and overprotective parenting styles [9, 11, 12]. The differences were: psychiatric condition [12], introverted personality, shyness, parental attachment issues, dysfunctional family dynamics, parental psychi- atric conditions, poor academic performance and high expectations, technology, globalization, the Internet, the breakdown of social cohesion [13], and the overdepend- ence of children [9]. Common issues of diagnosis men- tioned in the reviews were the difficulty of differentiating hikikomori from psychiatric disorders, because those exhibiting socially isolating behaviors could potentially suffer from any one of a spectrum of psychiatric illnesses [12, 13]. There is also uncertainty over whether a psychi- atric condition is the cause of hikikomori symptoms or if hikikomori leads to a psychiatric condition [10]. The following Interventions were commonly reported in the reviews: psychotherapy, pharmacological treatment, family therapy, nidotherapy, milieu therapy with the provision of a safe environment for hikikomori, support groups with the avoidance of labeling, rigid schedules, or categorization of role identity [9, 12, 13]. Less commonly reported were the following interventions: group therapy, horse-assisted therapy, communal cooking, online plat- forms [13], Chinese medicine, narrative therapy, naikan therapy, and engagement with social workers [9]. The following are additional interventions not mentioned in the reviews: animal-assisted therapy [14], jogging therapy [15], and the online mobile game Pokémon Go [16]. Many current interventions in hikikomori care brought up in case reports seem to lack a focal factor to target in order to achieve a recovery. The majority also fail to use a theoretical framework to guide the components of the intervention. It would be beneficial to apply a psychoso- cial recovery model to the task of developing methods of caring for hikikomori because such a model provides a comprehensive structure and an organized approach to guide a researcher or service provider in determining what areas need to be assessed or measured, or in identi- fying areas that could be focused on for care [17]. To the best of our knowledge, in only one study [18] has a psy- chological recovery model or framework for hikikomori care been applied. In their study, Yokoyama et al. [18] combined concepts from dialectic behavior therapy and from the mental health recovery model used by Mental Health America to design online modules for hikikomori, which uses elements of self-realization, caring for one- self, acquiring change, and future planning in the inter- vention. Other psychosocial recovery frameworks have not been used in hikikomori interventions. Saito [19] has proposed conceptual models on the power operates in the “hikikomori system” and the vicious circles pre- venting treatment for hikikomori; however, they were not focusing on psychosocial recovery. After reviewing different psychosocial recovery frameworks such as the Recovery Model [20], Psychosocial Rehabilitation Model [21], Strength Based Model [22], Coach-based Model [23], and the CHIME framework for personal recovery [24], it was concluded that all of the frameworks had something beneficial to offer for hikikomori care, such as a non-linear approach, a focus on the positive attrib- utes of an individual or holistic care. However, it seems most fitting to apply the CHIME framework for per- sonal recovery to hikikomori care because of the follow- ing two reasons. First, the framework was synthesized for psychosocial recovery and hikikomori are in need of psychosocial recovery from a behavioral and etiologi- cal perspective. Second, some domains and dimensions of the framework shed light on what hikikomori lack, Yung et al. BMC Psychol (2021) 9:104 Page 3 of 30 such as connectedness, identity, or meaningful social roles, which provide accuracy in targeting specific areas requiring re-establishing for hikikomori. There is aware- ness that the CHIME framework has been designed for personal recovery in the area of mental health; therefore, some may hypothesize that it may be more applicable to people who have experienced mental health challenges; whereas young people with experience of primary social withdrawal may not have been diagnosed with any mental health issues or exhibited any mental health syndromes. However, the domains of the CHIME frame- work seem broad and encompassing which may possibly extend its application to individuals without psychiatric disorders but in need of psychosocial recovery. The CHIME framework The CHIME framework for personal recovery was first synthesized by Leamy et al. [24] from a systematic review of 87 articles on frameworks used for personal recovery in mental health. It is the most comprehensive depic- tion of the recovery process to date [25]. This framework is versatile and has been used in studies as wide-ranging as those on cultural diversity and depression [25], and art therapy for mental health recovery [26]. The CHIME framework consists of five domains, i.e., connected- ness, hope and optimism, identity, meaning in life, and empowerment. In addition, each domain contains spe- cific dimensions. Connectedness refers to the connection with peers, relationships, being part of the community, and receiving support from peers and others [24]. Hope and optimism refer to a belief in recovery, the motivation to change, having hope-inspiring relationships, thinking positively and valuing success, and having dreams and aspirations [24]. Identity encompasses the dimension of identity, rebuilding or redefining a positive sense of identity, and overcoming stigma [24]. Dimensions of identity have been further explained by the research team as the view that an individual can have multiple identities pertaining not only to their medical diagnosis but also including the aspects of culture, ethnicity, and sexual identity [27]. For hikikomori, much focus is placed on a status-driven or non-status driven identity. The former refers to the sta- tus of a student, worker, or trainee, while the latter may refer to, but is not limited to, the status of social activ- ist, serious leisure devotee, volunteer, carer, and oth- ers. Hikikomori may have the identity of not being in education, training, or employment (NEET). However, it should be noted that not all NEET can be considered hikikomori, as some NEET have an active social life [6]. In the third domain of the CHIME model of meaning in life, the dimensions are: meaning of mental illness experi- ences, spirituality, quality of life, meaningful social roles and goals, and rebuilding life [24]. The meaning of mental illness experiences is described as finding understanding or meaning from the illness experience itself [25, 27]. If this domain is to be applied to the hikikomori popula- tion, this dimension could be replaced by the meaning of the hikikomori experience, as not all hikikomori have a comorbidity of mental illness. The last domain, empow- erment, includes the dimensions of personal responsibil- ity, control over life, and focusing on strengths [24]. As mentioned previously, the CHIME framework was described as the most comprehensive of recovery pro- cesses [25] and has domains or dimensions that match deficits in hikikomori; hence, the CHIME framework can be applied to hikikomori care. Therefore, the aim of this review was to investigate the applicability of the CHIME framework to the hikikomori population to understand their life experiences and the phenomenon. The objectives of this review were: (1) To identify stud- ies on the hikikomori population in relation to each domain of the CHIME Framework; (2) To synthesize the identified studies and apply them to each domain of the CHIME framework; (3) To determine if the identified studies would fit into the domains of the CHIME Frame- work; and (4) To identify whether there are any dimen- sions in the CHIME Framework in which studies on the hikikomori population are lacking, which may indicate a knowledge gap. The authors hypothesize that the CHIME framework would provide an encompassing understand- ing of the hikikomori life experience or hikikomori phe- nomenon, which would be evidenced by the presence of hikikomori literature being found related to the domains of the framework and the literature would give a depic- tion of the life experiences of individuals with a hikiko- mori lifestyle or the hikikomori phenomenon. If the framework were not applicable to hikikomori, then there would been an absence of literature or a minute amount of literature found. In case this scenario happens, the domains of CHIME would be considered inapplicable to understanding the hikikomori phenomenon. Methodology The reporting of this review follows the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) [28]. A search for academic peer-reviewed articles was conducted and ended on April 30, 2020, using five databases: CINAHL, PubMed, Pro- Quest, Science Direct, and Web of Science. The search terms were “hikikomori” OR “socially withdrawn youth” OR “youth social withdrawal” OR “severe social with- drawal” OR “acute social withdrawal” OR “protracted social withdrawal” OR “prolonged social withdrawal” OR “primary social withdrawal” OR “hidden youth”, accord- ing to Li and Wong [7]. In addition, search terms specific Yung et al. BMC Psychol (2021) 9:104 Page 4 of 30 to each domain of the CHIME framework were incorpo- rated in the search, e.g., “spirituality” OR “quality of life” OR “relationships” OR “hope” OR “belief in recovery” OR “empowerment” OR “strength-based” OR “focusing upon strength” OR “control over life” OR “personal responsibil- ity”, and others. Table 1 presents an example of the search strategy. Articles included in this review were in the English lan- guage, of all publication years, peer-reviewed, consisted of quantitative or qualitative research studies and case studies, included study designs that were observational or interventional in nature, and involved populations of socially withdrawn youth with a minimum duration of social withdrawal of three months. Articles that did not fit this criterion or that were commentaries, discus- sion papers, conference abstracts, letters to the editor, or reviews, were excluded from this review. And arti- cles that did not relate conceptually with the elements or topics within the CHIME framework were excluded from this review; while articles considered related were included. For example in the domain of connectedness, studies containing information about the social ties of hikikomori with their peers would reflect the concept of connectedness; therefore, would be included. The initial search yielded a total of 235 publications, with 51 in Pub- Med, 69 in ProQuest, 35 in ScienceDirect, 68 in Web of Science, and 12 from CINAHL. Titles and abstracts were screened by two reviewers and disagreements were dis- cussed until an agreement was reached. Details of the selection and exclusion processes are displayed on the PRISMA flow chart in Fig. 1. After the primary search, reference lists of the selected articles and authors with frequent publications of hikikomori research were reviewed to identify any additional articles fitting the inclusion criteria and related to both hikikomori and ele- ments within the CHIME framework. 19 articles were included from the hand search. The selection of full text articles were reviewed by all members of the research team and disagreements were discussed amongst all members. After quality appraisal of articles, data was extracted from the included articles relating to the life experiences of hikikomori or phenomena in accordance to the specific domains of CHIME for narrative synthesis, refer to Table 2. Extracted data was grouped in themes that were relevant to each other, data was summa- rized within the themes; comparisons were made either between the data or with other literature of the relevant topic depending on the availability of literature. Quality appraisal All included articles were assessed for biases and rigor in methodology using the Joanna Briggs Institute (JBI) Crit- ical Appraisal Checklist tool [29–33] and Mixed Method Appraisal Tool (MMAT) by the Department of Medicine of the McGill University [34]; and were of high quality, refer to Table 3. The JBI contains separate checklists for quasi-experimental (9 criteria), case–control (10 criteria), cohort (11 criteria), analytical and prevalence cross-sec- tional studies (8 and 9 criteria), case reports or series (8 and 10 criteria), and qualitative (10 criteria). Each com- ponent of the checklist can be rated as yes, no, unclear, or not applicable. The MMAT contains six sections in the Table 1 Search Terms for PubMed Domain of connectedness (“hikikomori” OR “socially withdrawn youth” OR “youth social withdrawal” OR “severe social withdrawal” OR “acute social withdrawal” OR “protracted social withdrawal” OR “prolonged social withdrawal” OR “primary social withdrawal” OR “hidden youth”) AND (connectedness OR relationships OR commu- nity OR peers OR friends OR friendship) Domain of hope and optimism (“hikikomori” OR “socially withdrawn youth” OR “youth social withdrawal” OR “severe social withdrawal” OR “acute social withdrawal” OR “protracted social withdrawal” OR “prolonged social withdrawal” OR “primary social withdrawal” OR “hidden youth”) AND (hope OR optimism OR “belief in recovery” OR “motivation to change” OR “positive thinking” OR “valuing success” OR dreams OR aspirations) Domain of identity (“hikikomori” OR “socially withdrawn youth” OR “youth social withdrawal” OR “severe social withdrawal” OR “acute social withdrawal” OR “protracted social withdrawal” OR “prolonged social withdrawal” OR “primary social withdrawal” OR “hidden youth”) AND (identity OR gender OR “sexual orientation” OR culture OR “dimensions of identity” OR “rebuilding positive identity” OR “redefining positive identity” OR stigma) Domain of meaning in life (“hikikomori” OR “socially withdrawn youth” OR “youth social withdrawal” OR “severe social withdrawal” OR “acute social withdrawal” OR “protracted social withdrawal” OR “prolonged social withdrawal” OR “primary social withdrawal” OR “hidden youth”) AND (“meaning in life” OR “meaning of mental illness” OR “mental illness experience” OR “meaning of experience” OR “social roles” OR “social goals” OR “rebuilding life” OR spirituality OR “quality of life”) Domain of empowerment (“hikikomori” OR “socially withdrawn youth” OR “youth social withdrawal” OR “severe social withdrawal” OR “acute social withdrawal” OR “protracted social withdrawal” OR “prolonged social withdrawal” OR “primary social withdrawal” OR “hidden youth”) AND (empowerment OR “strength-based” OR “focus- ing upon strength” OR “control over life” OR “personal responsibility”) Yung et al. BMC Psychol (2021) 9:104 Page 5 of 30 Fig. 1 PRISMA flow diagram of search details through different phases checklist; for mixed method studies only four sections are used (initial screening section, followed by sections 1; 2, 3, or 4; and 5) with ratings of yes, no, or can’t tell. studies are presented in accordance with each domain of the CHIME framework. Domain of connectedness Results A total of 44 studies were identified, 21 of which were quantitative, 9 qualitative, 3 mixed methods, and 11 case studies or series. They were published between the years 2004 and 2020. The studies came from various coun- tries, although the majority were conducted in Japan. All of studies are listed and described in Table 2. Below, the Connectedness refers to the link with peers, relation- ships, being part of the community, and receiving sup- port from peers and others in the CHIME framework [24]. In this domain 23 articles were found related to connectedness in hikikomori. These articles included two interventional studies; six cross-sectional studies; one mixed-method studies; one longitudinal study; nine Yung et al. BMC Psychol (2021) 9:104 Page 6 of 30 ± e r o f e b s t n a p c i t r a p i l l a r o f s e r o c S F A G l l a r e v O y a d - o t - y a d f o l e v e l r e h g H i e r o c s r e h g H i l a w a r d h t i w l i a c o s f o n o i t a r u d m u m n M i i - r e t c a r a h c l a w a r d h t i w l a c n i i l c n o a t a d y n O l : ) D S ± s r u o h n a e M ( e g a s u t e n r e t n I i s e r o c S g n n o i t c n u F t n e m s s e s s A l a b o G l s h t n o m 3 s a d e fi i t n e d i d e t c a r t x e e r e w s c i t s i 0 4 3 . 0 2 5 . ) 0 0 1 – 0 d e r o c s ; l e a c s s u o u n i t n o C ( ) F A G ( - r e h t o h c y s p n o i t a t i s i v e m o h e h t r e t f a d n a ± ± n o i t n e v r e t n i t s o p s e r o c s F A G n i e g n a h c o n i g n n o i t c n u f l i a c o s / k r o w / l o o h c s . 1 0 0 0 < p , . 2 3 1 . 4 3 5 t s o p o t 1 1 1 . d e w o h s h t u o y n w a r d h t i w y l l i a c o s ± . 6 4 4 e r P f o % 8 8 4 . n i i y t l u c ffi d e t a r e d o m 0 6 – 1 5 f o e r o c S i g n n o i t c n u f l i a c o s / k r o w / l o o h c s n i : ) D S n a e M ( n o i t n e v r e t n i y p a t n e m r i a p m i s u o i r e s y n a 0 5 – 1 4 f o e r o c S = = = i g n n o i t c n u f = p , 4 6 0 . = = f l l o s e u a v - P d n a s n o i t a e r r o C y b s p h s n o i t a e R l i = 1 9 9 0 . 0 0 2 0 . 7 4 9 0 . = = p p p : s h t n o m 4 : s h t n o m 8 : s h t n o m 2 1 y r o t s i h l a c n i i l i c d n a a t a d c h p a r g o m e d o c o S i l s e fi ’ s t n e i t a p e h t m o r f d e t c a r t x e e r e w g n i s s e s s a e r i a n n o i t s e u q d e p o e v e d - f l e S l : i s t n o p e m i t g n w o i l l o f e h t t a s k r o w t e n i s p h s n o i t a e r l f o k r o w t e n l i a c o s : o t y p a r e h T y a P l i r o m o k i i k H s e i t i v i t c a c i t u e p a r e h t n i n o i t a p c i t r a p i = = = = 0 0 0 0 0 . = p , 0 6 0 . i g n e b - l l e w l i a c o s o h c y s P ) 0 8 – 0 2 d e r o c s ; s m e t i i t r e k L - 8 2 ( : n o i s s e r g e R l i a c h c r a r e H 3 i l e v e L t n e m r e w o p m e f o s l e v e l 0 0 0 0 0 . p , 6 5 9 . t n e m r e w o p m E e r i a n n o i t s e u Q l a t i p a C = e r o c s r e h g H i l i l a c g o o h c y s P l a c i t s i t a t s o n y p a r e h T y a P l ) 6 2 1 – 1 2 d e r o c s ; s m e t i i t r e k L - 4 2 ( 0 0 0 0 0 . p , 9 5 0 . t n e m r e w o p m E . l l a t e s r e g o R r e p e a c S t n e m r e w o p m E d e t r o p e r e r e w s e r o c s e a c s o N l e c n a c fi n g i s i l i l a c g o o h c y s p r e h g h i = e r o c s r e h g H i i g n e b - l l e w y d u t s e h t n i d e s u s e a c s l t n e m e r u s a e m o N n o i t n e v r e t n i t s o p e t a p c i t r a p o t i - h t i w y l l i a c o s r o f d e t r o p e r s e r o c s e a c s o N l - h t i w l i a c o s e r u s a e m o t t i n u k k a s a r o N % 1 6 4 y b . l a w a r d h t i w l i a c o s n i e s a e r c e D d n a a d h c U i r e p s a t s i l k c e h c y x o r p a d e s U h t u o y n w a r d ) s r e w s n a o n / s e y m e t i - 4 ( l a w a r d : n o i t n e v r e t n i D E B - C i t s o p d e r e t s i n m d a y e v r u s d e p o e v e d - f l e S l s s e n g n i l l i w n i e s a e r c n i , i y t e x n a n i e s a e r c e D n o i t n e v r e t n i l i a c o s n i i e c n e r e ff d t n a c fi n g i s y i l l a c i t s i t a t s o N g n i s s e s s a e r i a n n o i t s e u q d e p o e v e d - f l e S l . 1 8 1 l s e a m e f d n a s e a m l f i o a t a d c h p a r g o m e D 4 2 f o e g a n a e m a h t i w , n a p a J m o r f % 5 5 2 . : e g e l l o C / y t i s r e v n U i : ) % ( l e v e l n o i t a c u d E % 4 4 3 . : h g H i i r o n u J % 5 9 3 . : l o o h c S h g H i 9 4 – 0 2 e g a , n a p a J m o r f i r o m o k i i k h , i a t a d c h p a r g o m e d o c o s i l s e a m e f d n a s e a m l f i o a t a d c h p a r g o m e D l g n i s s e s s a y e v r u s d o h e s u o h e c a f - o t - e c a F s r a e y 9 4 – 0 2 e c n e d i s e R y t i n u m m o C , ) 9 1 i c i r t a h c y s p d n a a t a d c h p a r g o m e d o c o s i i y r o t s i h c i r t a h c y s p d n a i , s m o t p m y s n o n o i t a m r o n f i d e t i c i l e s t s i r t a h c y s P i s n o i t a t l u s n o c g n i r u d y r o t s i h 5 8 , 2 5 2 s r a e y 1 2 – 2 1 e s e n h C i 8 1 1 , 4 8 3 4 5 , 6 3 1 s r a e y 9 2 – 2 1 e s e n h C i 2 0 2 , 1 7 1 s r a e y 5 3 – 6 1 e s e n a p a J 4 , 1 ± h s i n a p S . 1 9 3 e s e n a p a J 5 , 4 1 ± 4 5 . . 2 4 2 e s e n a p a J ) 0 9 1 = N ( s t l u d A n w a r d h t i W y l l i a c o S ] 9 3 [ . l a t e r o m A - n ó g a a M l ) 2 0 5 = N ( h t u o Y n e d d H i ] 5 7 [ n a h C ) 3 7 3 = N ( h t u o Y i r o m o k i i k H ) 5 = N ( = N ( i r o m o k i i k H ) 1 4 6 1 = N ( i r o m o k i i k H ) 7 3 3 = N ( ] 8 1 [ . l a t e a m a y o k o Y ] 4 7 [ . l a t e w a L i s e d u t s l a n o i t c e s - s s o r C ] 2 6 [ . l a t e a m a y o K ] 0 4 [ . l a t e o d n o K i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( ± n a e r o K 0 1 , 1 3 5 1 . ± . 3 6 1 , 5 3 . . 4 6 1 l / e d d M i , ) 1 4 ) 9 3 2 = = N ( s t n e d u t S l o o h c S h g H i N ( h t u o Y n w a r d h t i W y l l i a c o S s t n a p i c i t r a P ) N ( i s e d u t s n o i t n e v r e t n I ] 3 [ . l a t e e e L ) s ( r o h t u A ) r a e y ( i w e v e r e h t n i d e d u c n l i i s e d u t s l l A 2 e l b a T Yung et al. BMC Psychol (2021) 9:104 Page 7 of 30 ± ± ± ± ± ± 7 5 1 . ± ± 5 1 2 . s v 2 6 1 . ± 2 0 4 . . 1 4 7 s v 4 4 4 . 9 2 0 1 . : e r o c s e t i s o p m o C ) 7 – 1 d e r o c s i ; t r e k L - 1 ( e a c s b u s l 7 3 2 . . 1 2 3 s v 1 2 2 . 8 5 3 . : e r o c s t u o g n k c o L i d e r o c s ; m e t i i l t r e k L - 1 ( e a c s b u s g n i r o n g I : i e r o c s p h s n o i t a e r l f o s s o l i g n n e t a e r h T ) 7 – 1 . 5 1 2 s v 9 0 2 . 1 6 1 . 0 0 3 . : e r o c s g n i r o n g I ± . 5 0 2 s v 4 9 1 . 1 7 3 . d e r o c s ; m e t i i l t r e k L - 1 ( e a c s b u s n e t a e r h T t u o - k c o L ) 7 – 1 . 1 0 0 < p , 1 9 3 0 . : l l o o h c s e d d m o t i t n e m t s u d A j . 1 0 0 < p . 1 0 0 < p , 2 0 3 0 . : n o i t c e e r j l a t n e r a P , 0 0 4 0 . : t n e m h c a t t a l a n r e t a m l t n e a v b m A i . 5 0 0 < p , 5 8 2 0 . : s s e n y h S , 5 0 4 0 . : i p h s n o i t a e r l f o s s o l d e n e t a e r h T . 1 0 0 < p . 1 0 0 < p , 7 8 2 0 . : n o i t c e e r j r e e P . 5 0 0 < p , 9 1 2 0 . : g n i r o n g I p u o r g r e e p o t g n i t s u d a h t i j w y t l u c ffi D i y t i s n e t n i / e c n a b r u t s i d r e h g h i = e r o c s ) 7 – 1 d e r o c s ; m e t i t r e k L - 1 ( i k r o w r e h g H i . s n o i s r e v e s e n a p a J l l A : l s e u a v - P d n a n o i t a e r r o C l i r o m o k i i k H l e a c S l o o h c S o t t n e m j t s u d a a M l ± 5 8 1 . 0 2 3 . ) 8 – 1 ± 0 5 4 . : l . 1 4 2 s v 1 3 2 . 5 8 3 . : n o i t c e e r j r e e P ) s e r o c s e a c s b u s l f o m u S ( e r o c s e t i s o p m o C o o h c s o t t n e m j t s u d a a M l d e r o c s ; m e t i i l t r e k L - 1 ( e a c S n o i t c e e R r e e P j 6 7 9 . ± . 9 8 6 4 s v 7 2 2 1 . ± 1 7 0 . ± . 9 0 2 s v 0 7 0 . s v 5 7 0 . ± ± 8 0 2 . 1 2 2 . : e r o c s e c n a d o v A i ) 4 – 1 d e r o c s ; s m e t i t r e k L - 8 ( i t n e m h c a t t a l a n r e t a M l e a c s b u s t n e m h c a t t a t n a d o v a / e r u c e s n i I : e r o c s e c n e a v b m A l i l e a c s b u s l t n e m h c a t t a t n e a v b m a / e r u c e s n i I e r o c s n o i t c e e r j 1 5 0 . ± l a t n e r a P 1 5 1 . ) 4 – 1 d e r o c s ; s m e t i t r e k L - 7 ( i i l e a c S r o v a h e B g n i t c e e R j l a t n e r a P l s e a m e f l d n a s e a m n o a t a d c h p a r g o m e D i l i a c o s d o o h d l i h c , y r o t s i h l i l a c g o o h c y s p % 8 4 2 . : e g e l l o C / y t i s r e v n U i , s m o t p m y s i r o m o k i i k h , i a t a d c h p a r g o m 6 3 e g a n a e m a h t i w , n a p a J m o r f d n a , y r o t s i h l i l a c g o o h c y s p l a t n e r a p , s s a c l s u o i r a v r o f s t n e d u t s y t i s r e v n U s v i i r o m o k i i k H d e r o c s ; s m e t i i l t r e k L - 6 1 ( e a c S s s e n y h S t i a r T s r a e y 9 5 0 2 . s e c i t c a r p g n i r a e r d l i h c , s r a e y 4 8 2 2 . % 4 1 7 . : l o o h c S h g H i i - e d o c o s g n i s s e s s a n a p a J f o s a e r a l a r u r 5 , 0 1 : ) D S 3 8 2 5 . ± n a e M ( l s e a c s : e r o c s s s e n y h S l e a c S t n e m h c a t t A l a n r e t a M ) 0 8 – 6 1 e s e n a p a J 0 1 , 4 1 i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( ± l r a e c n u s a w t i , r e v e w o h ; n o i t a c u d e f o s r a e y 7 1 . . 7 1 1 h t i w d n a , 7 2 f o e g a n a e m a t a d l a c n i i l c e v i t c e p s o r t e r 5 1 , 2 1 d e d u c n l i n o i t a c u d e n o a t a d e h t r e h t e h w s r a e y l o o h c s e r p e h t : ) % ( l e v e l n o i t a c u d E y e v r u s d e s a b - n o i t a u p o p e c a f - o t - e c a F l % 8 3 . : h g H i i r o n u J d n a , n a b r u , n a t i l o p o r t e m n i d e t c u d n o c 6 8 . ± . 1 0 3 ± e s e n a p a J , . 1 1 1 . 3 6 3 e c n e d i s e R y t i n u m m o C , ) 5 1 n o i t u b i r t s i d e h t n o a t a d c h p a r g o m e d o c o S i i i a t a d c h p a r g o m e d o c o s n o n o i t a m r o n f i I 5 7 . . 4 7 2 l i a c o S h t i w s t n e i t a P , ) 7 2 a h t i w , n a p a J m o r f l s e a m e f d n a s e a m l f o m o r f d e t i c i l e e r e w y r o t s i h c i r t a h c y s p d n a i ± e s e n a p a J = r e d r o s i D y t e x n A i s t n a p i c i t r a P ) N ( ) s ( r o h t u A ) r a e y ( N ( i r o m o k i i k H ] 4 6 [ . l a t e a t a g a N ) d e u n i t n o c ( 2 e l b a T ) 4 1 1 = N ( = ) 3 9 6 = N ( N ( i r o m o k i i k H ] 3 6 [ i m a k a w a K d n a a d e m U s t n e d u t S y t i s r e v n U i , ) 4 2 = N ( i r o m o k i i k H ) 9 5 = N ( i i ] 7 3 [ e k c D d n a g e i r K Yung et al. BMC Psychol (2021) 9:104 Page 8 of 30 8 3 2 . ± 5 9 0 . : i s n o i t a m n a / s c m o c g n d a e R i i 3 1 1 . 0 9 0 . : l n o i s i v e e t g n h c t a W i ± 9 0 2 . ± 9 9 0 . 0 4 0 . : l l a w e h t g n c a F / g n i i l d I 5 6 0 . : i g n d a e r r e h t O 3 0 5 . ± 1 1 3 . ± ± 3 0 1 . : e s u e l i b o m l / t e b a T 0 9 1 . : g n i t a E ± 7 9 4 . 9 0 5 . : e s u r e t u p m o C ± : ) D S ± n a e M ( x e d n I k r o w t e N l i a c o S 0 8 1 . ± 9 7 2 . l l s e u a v - P d n a s n o i t a e r r o C L O Q h t u o Y n e d d H i , 0 5 8 0 - . = l a w a r d h t i w l i a c o s f o l e v e l 0 0 0 0 0 . , 0 5 5 0 . = l a w a r d h t i w l i a c o s f o l e v e l w o L 0 0 0 0 0 . = h g H i p = p d e t r o p e r t o n s e r o c s F E R B - L O Q O H W e f i L f o y t i l a u Q n o i t a z i n a g r O h t l a e H d l r o W e f i L f o y t i l a u Q r e h g H i = e r o c s r e h g H i ) F E R B - L O Q O H W ( f e i r b - e a c s l ) 2 1 1 – 0 g n i r o c s ; m e t i i t r e k L - 8 2 ( ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( s r a e y 0 3 – 2 1 e s e n h C i 5 1 2 , 3 7 3 i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M n o i t u b i r t s i d e h t n o a t a d c h p a r g o m e d o c o S i i i a t a d c h p a r g o m e d o c o s n o n o i t a m r o n f i I . 3 8 1 l s t l u d A d e y o p m E s v T E E N s v i r o m o k i i k H l e a c S k s i R i r o m o k i i k H T E E N i n a p S m o r f l s e a m e f d n a s e a m l f o m o r f d e t i c i l e e r e w y r o t s i h c i r t a h c y s p d n a i s n o i t a t l u s n o c n o a t a d l a c n i i l c ) D S n a e M ( ) 2 1 – 0 g n i r o c s ; s m e t i - 2 ( s n o i t i b m a r a e c n U l ± ± 7 3 1 . . 6 2 4 s v 1 4 1 . . 5 0 5 s v 0 6 1 . 8 4 5 . ± : s n o i t i b m a r a e c n U l l r a e c n u f o l e v e l r e h g H i = s n o i t i b m a e r o c s r e h g H i : ) D S ± 9 9 1 . 3 8 7 . : i g n p e e S l , s e i t i v i t c a y l i a d , i a t a d c h p a r g o m e d o c o s i % 5 3 : i d e n a p m o c c A % 3 : l o o h c s / l a t n e m i t n e s d n a y l i m a F % 8 : s d n e i r f d n a y l i m a F : ) % ( s g n i t u o n o g n o G i % 8 3 : e n o A l % 9 1 : e n o N % 3 6 : y l i m a F % 8 : s d n e i r F s d r o c e r t n e i t a p m o r f d e t c a r t x e s c i t s i l ) y n o y c n e u q e r F ( : ) % i ( e n e g y h r o o P % 7 2 : e n o N % 3 3 : s e Y % 7 6 : o N s r u o h n a e M ( s e i t i v i t c a y l i a D f o s e p y T s s e s s a o t e r i a n n o i t s e u q d e p o e v e d - f l e S l 2 6 3 . ± h c n e r F 3 1 , 3 5 ± e s e n h C i 2 0 9 1 . ± ± = ; 3 3 . ± ± = 0 5 3 . . 3 9 1 s v , 1 0 0 0 . p e f i L f o y t i l a u Q r e h g H i e r o c s r e h g H i = p u o r g l o r t n o C s v ) s e i t i d b r o m o c i l d e p o e v e d y w e n l , d e t r o p e r t o n g n i r o c s i i n a n a r k U . 7 3 1 d n a 1 0 0 0 . p ; 0 7 2 . . 7 1 1 i ) e n a r k U m o r f e a c s l 1 2 , 4 1 i c i r t a h c y s p t u o h t i w d n a h t i w ( i r o m o k i i k H , l e a c s m e t i l - 0 1 ( e a c S e f i L f o y t i l a u Q n a b a h C s r a e y 0 4 – 8 1 ) 8 2 ± h s i n a p S . 0 0 4 s r a e y 9 3 – 0 2 e s e n a p a J 3 4 , 1 2 1 – , ) 6 8 k r o w t e N l i a c o S e m y S – n a m k r e B d e fi d o M i ) 7 – 0 d e r o c s ; s m e t i o n / s e y 7 ( x e d n I s s e n d e t c e n n o c r e h g h i = e r o c s r e h g H i y r o t s i h h t l a e h d n a 2 4 , 2 6 = N ( N ( T E E N , ) 4 1 1 = ) 5 2 5 7 = s t l u d A d e y o p m E l s t n a p i c i t r a P ) N ( h t u o Y n e d d H i ) 8 8 5 = N ( i r o m o k i i k H ) 4 6 1 = N ( ) d e u n i t n o c ( 2 e l b a T ] 0 6 [ o L d n a n a h C ) s ( r o h t u A ) r a e y ( ] 9 5 [ . l a t e r o m A - n ó g a a M l N ( i r o m o k i i k H ] 8 5 [ t i k n u k k a s a r o N d n a a d h c U i = N ( p u o r g l o r t n o C , ) 5 3 = N ( i r o m o k i i k H ] 1 6 [ a v o k n a r F ) 6 6 = N ( i r o m o k i i k H ) 4 0 1 = N ( ] 1 4 [ . l a t e n e u Y : h t i w ) % ( i d e n a t n a m i i s p h s n o i t a e R l - r e t c a r a h c l a w a r d h t i w l a c n i i l c n o a t a D 5 7 4 . . 2 3 2 h t u o Y n w a r d h t i W y l l i a c o S ] 8 3 [ . l a t e c a i l u a h C Yung et al. BMC Psychol (2021) 9:104 Page 9 of 30 i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( - r e p r e t n I % ( i r o m o k i i k h - n o n s v i r o m o k i i k H / s e y 4 y b d e s s e s s a s e i t l u c ffi d i l a n o s r e p r e t n I l : ) e u a v - p ; s e i t l u c ffi d i l a n o s : s n o i t s e u q o n 1 0 0 0 < . ; . 0 6 3 s v 1 4 7 . l l a r e v O ) 1 Q ( s r e h t o g n i t e e m f o r a e F ) % ( i r o m o k i i k h f o s s a c l 1 0 0 0 < . 1 0 0 0 < . ; ; . 3 8 2 s v 7 1 5 . . 6 4 1 s v 4 3 5 . 1 0 0 0 < . 1 0 0 0 < . ; ; . 1 8 s v 2 6 3 . . 1 7 s v 3 8 4 . = = = = = 1 Q 2 Q 3 Q 4 Q l i a c o S l e p o e p r a i l i m a f g n i t e e m t u o b a y t e x n A i ) 2 Q ( f o n o i s s e r p m i ’ l s e p o e p t u o b a y t e x n A i ) 4 Q ( s p u o r g o t n i l d n e b t o n n a C ) 3 Q ( f l e s e n o s r a e y 9 3 – 5 1 e s e n a p a J 0 2 , 8 3 i r o m o k i i k H - n o N , ) 8 5 = N ( i r o m o k i i k H ) 4 2 0 3 = N ( s t n a p i c i t r a P ) N ( ) d e u n i t n o c ( 2 e l b a T ] 3 4 [ a r u m o N d n a g n o Y ) s ( r o h t u A ) r a e y ( ± ± ± ± , 4 1 2 . ± 5 5 5 . , 8 0 2 . ± 9 5 5 . , 3 5 2 . ± 9 8 5 . , 5 4 2 . ± 0 0 6 . : t r o p p u S g n g n o e B l i 4 4 2 . 5 0 6 . : t r o p p u S m e e t s E - f l e S 9 3 2 . 8 3 5 . ; s m e t i i t r e k L - 4 ( e a c s b u s l t r o p p u S e b g n a T l i i t r e k L - 4 ( e a c s b u s l ) 2 1 – 0 d e r o c s ; s m e t i t r o p p u S m e e t s E - f l e S ) 2 1 – 0 d e r o c s 3 7 1 . 0 2 6 . ) 2 1 – 0 d e r o c s ; s m e t i o t , 6 0 2 . ± 3 9 2 . , 0 8 1 . ± 9 7 2 . : ± : ) D S n a e M ( I N S t r o p p u s l i a c o s r e h g h i e r o c s r e h g H i i 3 – 1 s t n o p e m T i k r o w t e N l i a c o S e m y S – n a m k r e B d e fi d o M i = 7 8 1 . ± 9 0 3 . ) 7 – 0 d e r o c s ; s m e t i o n / s e y 7 ( x e d n I s s e n d e t c e n n o c r e h g h i = e r o c s r e h g H i , 7 8 1 . ± 0 9 6 . , 6 1 2 . ± , 3 5 1 . ± 9 2 6 . , 1 7 1 . ± 0 2 6 . , 9 9 5 . ± . 3 6 4 2 , 0 3 6 . ± 0 6 4 2 . : ± 1 8 6 . : t r o p p u S l a s i a r p p A ± 9 8 6 . 5 7 4 2 . : t r o p p u S e b g n a T l i 8 1 2 . 1 1 7 . i t r e k L - 4 ( e a c s b u s l t r o p p u S g n g n o e B l i ; s m e t i i t r e k L - 4 ( e a c s b u s l t r o p p u S l a s i a r p p A ) 2 1 – 0 d e r o c s ) 8 4 – 0 d e r o c s 2 4 , 2 6 i 3 – 1 s t n o p e m T i ; s m e t i i t r e k L - 2 1 ( n o i s r e v t r o h S - ) L E S I ( t s i L : ) D S n a e M ( L E S I l n o i t a u a v E t r o p p u S l a n o s r e p r e t n I e s e n h C i 2 6 3 . ± e s e n h C i 2 0 9 1 . % 6 7 7 . : l e d d M i % 4 3 . : r e p p U % 0 9 1 . : r e w o L : n o i t a c u d E g n i s s e s s a y e v r u s e n i l l n o d e p o e v e d - f l e S 0 6 0 . % 0 0 9 . : l e v e L r o e h c a B l - h t i w l i a c o s , i a t a d c h p a r g o m e d o c o s i % 5 8 1 . : e m o c n i l e d d M i % 6 3 . : e m o c n i h g H i % 6 2 7 . : e m o c n i w o L : ) % ( f o s a e r a n i d e v L i y r o t s i h c i r t a h c y s p d n a i i , r o v a h e b l a w a r d 0 9 , 8 7 ± e s e n a w a T i ) 8 5 2 = = N ( 2 8 8 2 . - n o N , ) 8 6 1 N ( s t l u d A n w a r d h t i W y l l i a c o S ] 2 4 [ . l a t e u W s t l u d A n w a r d h t i W y l l i a c o S i r o m o k i i k H ) 4 0 1 = N ( i s e d u t s l i a n d u t i g n o L ] 5 3 [ . l a t e n e u Y Yung et al. BMC Psychol (2021) 9:104 Page 10 of 30 d e t r o p e r e r e w s e r o c s e a c s o N l 0 0 0 0 0 . 0 0 0 0 0 . p , 3 7 0 . p , 6 6 0 . = = = = = = . 5 0 0 < p , 3 1 0 . s r e h c a e T r e p s a e a c S s g n l i l i b S h t i i w s p h s n o i t a e R l . 5 0 0 < p , 6 1 0 . s r e e P : ) ? d e r o c s ; s m e t i - 7 ( ’ s g n e Z d n a g n e Z s g n i l b S i : ) ? d e r o c s ; s m e t i - 8 ( ’ s g n e Z d n a g n e Z ; g n e D r e p s a s r e h c a e T h t i i w s p h s n o i t a e R l : ) ? d e r o c s ; s m e t i - 7 ( ’ s a M d n a u X u X ; g n e D r e p s a s r e e P h t i i w s p h s n o i t a e R l i l t r e k L - 0 1 ( e a c S m e e t s E - f l e S g r e b n e s o R : ) ? d e r o c s ; s m e t i - 9 ( ’ s a M d n a ) 0 3 – 0 d e r o c s ; s m e t i : h t i i l w p h s n o i t a e r d o o g a h t i l w s e u a v - P d n a i a t a d c h p a r g o m l s n o i t a e r r o C n i h t u o Y n e d d H i f o m e e t s e - f l e S i - e d o c o s g n i s s e s s a y e v r u s d e p o e v e d - f l e S l l e s o c g n m r o i f o t s s e n n e p o n a d n a r o f i s p h s n o i t a e r l l a n o i t o m e 3 9 2 . ± e s e n h C i 1 1 1 2 . s t n e r a P l r e p s a e a c S s t n e r a P h t i i w s p h s n o i t a e R l 9 1 1 , 4 4 2 i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( a , s s e n e v i s s e r g g a e v i s s a p f o l e v e l r e h g h i a e t a c d n i i s e r o c s r e h g h i l ; r o o C m r o F l t s i g o o h c y s p l a c n i i l c a y b e h t o t ’ s n o i t o m e s e n o t s u d a o t y c n e d n e t j n o i t a n i l c n i n a , l e p o e p d n a t n e m n o r i v n e s n o i t o m e f o n o i s s e r p x e e h t s s e r p p u s o t s n o i t a u t i s l i a c o s n i n e k a h s g n i l e e f n e h w l d e t a e r - e r u t x e t f o r e b m u n l a t o t ; T m u S d e e n a e t a c d n i i s e r o c s r e h g h i , s e s n o p s e r 3 3 0 0 . = p , 2 3 0 . ± . 1 1 0 s v 7 6 0 . ± 0 5 0 . = : T m u S ± , 4 1 1 . ± . 9 3 1 s v 8 6 1 . ± : ) D S n a e M ( i r o m o k i i k h - n o N s v i r o m o k i i k H ; - k n i - 0 1 m e t s y S e v i s n e h e r p m o C h c a h c s r o R 0 5 2 . l : r o o C m r o F . l a t e i h s a h a k a T r e p s a d e r o c s , s m e t i t o b l 3 9 8 . ± 4 9 7 3 . ± e s e n a p a J , 3 3 9 . 4 1 3 3 . i r o m o k i i k H - n o N , ) 2 2 7 3 0 0 . p l d e r o c s d n a n o i t a u p o p e s e n a p a J e h t r o f 0 1 , 2 1 s t n a p i c i t r a P ) N ( = n ( i r o m o k i i k H ) 8 1 = N ( ) d e u n i t n o c ( 2 e l b a T i s e d u t s l o r t n o c – e s a c t o l i P ] 5 6 [ . l a t e i k u s t a K ) s ( r o h t u A ) r a e y ( h t u o Y n e d d H i ) 3 6 3 = N ( i s e d u t s d o h t e m - d e x i M ] 0 6 [ o L d n a n a h C : ) D S ± 0 6 0 . : ) D S ± 4 7 3 . ± 8 1 1 . ± 9 5 2 . ± n a e M ( s e r o c s e f i L f o y t i l a u Q l a t o T g n i s s e s s a e r i a n n o i t s e u q d e p o e v e d - f l e S l m e e t s e - f l e s h g h i 5 1 , m e e t s e - f l e s r e h g h i e r o c s r e h g H i - f l e s w o l ; 5 1 < m e e t s e = = = ≥ 3 4 0 . 7 2 0 . ± 2 0 3 . : g n i l l e s n u o c e n i l n O 2 5 2 . : g n i l l e s n u o c e n ffl O i f o s e s u d n a a t a d c h p a r g o m e d o c o s i i i s e c v r e s g n i l l e s n u o c : g n i l l e s n u o c d e t a r g e t n I e f i L f o y t i l a u Q n o i t a z i n a g r O h t l a e H d l r o W 9 2 1 . ± 2 4 0 . n a e M ’ ( e r o C s n o s r e P g n u o Y ± 6 5 1 . : g n i l l e s n u o c e n i l n O 2 6 0 . : g n i l l e s n u o c e n ffl O i i n a w a T ) F E R B - L O Q O H W ( f e i r b - e a c s l r e t f a g n e b - l l i e w d e r u s a e m ( n o i s r e V i ) s e c v r e s g n i l l e s n u o c g n v e c e r i i : g n i l l e s n u o c d e t a r g e t n I ; s m e t i i t r e k L - 0 1 ( E R O C s n o s r e P g n u o Y ’ d e r e ff o g n i l l e s n u o c e n i l n o : s t l u s e r i w e v r e t n I ) 0 4 – 0 d e r o c s i - a d e m r o f y t i n u t r o p p o d e r e ff o g n i l l e s n u o c i s w e v r e t n i d e r u t c u r t s - i m e S e v i t a t i l a u Q i e n ffl o e l i i h w n o i t a c n u m m o c r o f m r o f t a p l s t l u s e r e v i t i s o p e r o c s r e w o L = d n a h t u o y n e e w t e b s t c fl n o c g n i r u d n o i t i - l u f e s u & s e g a t n a v d a d e v e c r e p g n i s s e s s a i y l i m a f r i e h t g n i l e s n u o c f o s m r o f e e r h t f o s s e n s r a e y 4 2 – 2 1 e s e n h C i 8 1 1 , 4 8 3 h t u o Y n e d d H i ) 2 0 5 = N ( ] 6 7 [ n a h C Yung et al. BMC Psychol (2021) 9:104 Page 11 of 30 t s e B ; ) % 0 0 1 ( s d n e i r f d o o G ; ) . % 0 0 3 ( s d n e i r F t s e B ; ) % 0 0 1 ( s d n e i r f d o o G ; ) . % 0 3 4 ( s d n e i r F ; ) % 2 3 ( . l w o n k y n o u o y e p o e P l : i e c o v e t a v i r P ) % 0 0 1 ( s d n e i r f ; ) % 2 3 ( . l w o n k y n o u o y e p o e P l : t x e t e t a v i r P ) % 0 0 1 ( s d n e i r f t s e B ; ) . % 6 1 9 ( i n o i t a c n u m m o c r o f s m r o f ) % 0 0 1 ( s d n e i r f t s e B ; ) % 0 0 1 ( s d n e i r f d o o G ; ) . % 0 9 4 ( s d n e i r F l i a c o s f o s l e v e l r e h g h i = s e r o c s r e h g H i y c a m i t n i s d n e i r f d o o G ; ) % 0 6 ( . s d n e i r F ; ) % 0 ( w o n k f o e g a s u e s o o h c h t u o y w o h g n i s s e s s a l y n o u o y e p o e P l : g n i t e e m a r e m a c c i l b u P i s w e v r e t n i d e r u t c u r t s - i m e s e v i t a t i l a u Q ; ) % 2 3 ( . l w o n k y n o u o y e p o e P l : i e c o v c i l b u P ) 0 7 1 – 7 1 : d e r o c s ; s m e t i 7 1 ( t r u o c f e L ; ) % 0 0 1 ( l w o n k y n o u o y e p o e P l : t x e t c i l b u P i p h s d n e i r f d n a n o i t a c n u m m o c f i o s m r o f t s e B ; ) % 0 0 1 ( s d n e i r f d o o G ; ) % 0 0 1 ( s d n e i r F s l e v e l ) % 0 0 1 ( s d n e i r f & r e l l i M l r e p s a e a c S y c a m i t n I l i a c o S r e l l i M : d e s u n o i t a c n u m m o c f i o s m r o F g n i s s e s s a s e r i a n n o i t s e u q d e t c u r t s n o c - f l e S ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( s r a e y 0 3 – 2 1 e s e n h C i – i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M s t n a p i c i t r a P ) N ( h t u o Y n e d d H i ) 7 5 3 = N ( ) d e u n i t n o c ( 2 e l b a T ) s ( r o h t u A ) r a e y ( ] 5 5 [ n a h C s d n e i r f d o o G ; ) . % 0 7 1 ( s d n e i r F ; ) % 0 ( w o n k l y n o u o y e p o e P l : g n i t e e m a r e m a c e t a v i r P ) % 0 0 1 ( s d n e i r f ) % 0 0 1 ( s d n e i r f t s e B ; ) . % 6 2 9 ( l t o n e a c S y c a m i t n I l i a c o S r e l l i M r o f s e r o c S d e t r o p e r n o i t i r t t a % 0 5 : y p a r e h t o h c y s P . t s i p a r e h t e h t f o y t i l i b a i l e r e h t g n i t s e t s h t n o m 2 1 – 6 t n e p S e t a r t s i p a r e h t f o t s u r t s i m t n e i t a P s n o i t a t l u s n o c d n a s r a e y 2 m u m n m e m i i : i t y r e v o c e R s d r o c e r l a c n i i l c m o r f d e t c a r t x e a t a d l a c n i i l C d e n fi n o c : i s r o v a h e b g n w o i l l o f e h t d e t i b h x E i e m i t f j o y t i r o a m e h t t n e p s , e m o h t a f l e s l i a c o s n i i e g a g n e t o n d d m o o r n w o n , i , l e c y c e k a w / p e e l s d e s r e v e r , i s p h s n o i t a e r l , n a m O m o r f e a m a f l i o a t a d c h p a r g o m e D y l i m a f h t i w t c a t n o c d e s u f e r d n a 4 2 e g a d e n fi n o c : i s r o v a h e b g n w o i l l o f e h t d e t i b h x E i n i e m i t f j o y t i r o a m e h t t n e p s , e m o h t a f l e s l i a c o s n i i e g a g n e t o n d d d n a m o o r n w o , & s e c i t c a r p e n e g y h r o o P i . i s p h s n o i t a e r l s e l t t o b / s r a j n i d e t a c e f e d / d e t a n i r u e h t m o r f e a m a f l i o a t a d c h p a r g o m e D 0 3 e g a , s e t a t S d e t i n U e s a c t n e i t a p a m o r f d e t c a r t x e a t a d l a c n i i l C e s a c t n e i t a p a m o r f d e t c a r t x e a t a d l a c n i i l C s r a e y 5 1 2 . e s e n a p a J 0 1 , 5 2 s r a e y 4 2 i n a m O l e a M 1 n a c i r e m A s r a e y 0 3 l e a M 1 i r o m o k i i k H ) 5 3 = N ( i r o m o k i i k H ) 1 = N ( i r o m o k i i k H ) 1 = N ( i s e i r e s / s e d u t s e s a C ] 3 7 [ i r o t t a H ] 7 4 [ . l a t e o t o m a k a S ] 9 4 [ o e T Yung et al. BMC Psychol (2021) 9:104 Page 12 of 30 y l i m a f h t i w t c a t n o c d e s u f e r d n a , i s p h s n o i t n i e m i t f j o y t i r o a m e h t t n e p s , e m o h t a f l e s l - a e r l i a c o s n i e g a g n e t o n d d m o o r n w o i , d e n fi n o c : i s r o v a h e b g n w o i l l o f e h t d e t i b h x E i s e s a c t n e i t a p m o r f 7 5 . 0 3 . 5 3 . ± ± ± 3 4 . : e r o c s s d n e i r F . 5 0 1 . 4 5 5 ± : ) D S ± n a e M l ( e a c S s s e n i l e n o L A L C U d e r o c s ; s m e t i i t r e k L - 3 ( e a c s b u s l l a d o h % 0 6 5 . : l e v e l n o i t a c u d e ; s e i r t n u o c ) 0 8 – 0 2 g n i r o c s r u o f s s o r c a m o r f , l s e a m e f d n a s e a m l f o , s m e t i i t r e k L - 0 2 ( e a c S s s e n l i l e n o L A L C U n o i t u b i r t s i d e h t n o a t a d c h p a r g o m e D i n o i t a o s i l l i a c o s 2 1 > e r o c s l l a r e v o l a t o T = e v o b a r o e e r g e d s ’ r o e h c a b l s s e n i l e n o l f o l e v e l r e h g H i e r o c s r e h g H i ) 5 1 – 0 s d n e i r F ) 5 1 – 0 = d e t c a r t x e a t a d l a c n i i l C e m i t f j o y t i r o a m e h t t n e p s , e m o h t a f l e s l i a c o s n i i e g a g n e t o n d d m o o r n w o n , i t c a t n o c e c a f - o t - e c a f d e d o v a i , i s p h s n o i t a e r l , l e c y c e k a w / p e e l s d e s r e v e r , s r e h t o h t i w n a t i s i v o t h t n o m a e c n o e m o h t f e l d n a i c n i l c t n e i t a p t u o , n a p a J m o r f e a m a f l i o a t a d c h p a r g o m e D 9 3 e g a , l y a t I m o r f l s e a m e f o w t f i o a t a d c h p a r g o m e D s r a e y 4 e m i t y r e v o c e r y p a r e h t o h c y s P 3 1 e g a d e n fi n o c : i s r o v a h e b g n w o i l l o f e h t d e t i b h x E i e s a c t n e i t a p a m o r f d e t c a r t x e a t a d l a c n i i l C d e t s u a h x e : i s r o v a h e b g n w o i l l o f e h t d e t i b h x E i n i r a e f , s r e h t o h t i w l l l e w e t a e r o t y t i l i b a n i , i l s p h s n o i t a e r n a t n a m o t i i t r o ff e m o r f o t e c n e d fi n o c o n , i y t e c o s t l u d a g n i r e t n e , f l e s m h i f o d e m a h s a t l e f , i y t e c o s h t i w e p o c i g n e b m h i f i i o s n o n p o s r e h t o d e r a e f d n a l d e y o p m e n u n i e m i t f j o y t i r o a m e h t t n e p s , e m o h t a f l e s d e n fi n o c : i s r o v a h e b g n w o i l l o f e h t d e t i b h x E i l i a c o s n i i e g a g n e t o n d d d n a m o o r n w o , , i n a p S m o r f e a m a f l i o a t a d c h p a r g o m e D 5 2 e g a i s p h s n o i t a e r l e s a c t n e i t a p a m o r f d e t c a r t x e a t a d l a c n i i l C e s a c t n e i t a p a m o r f d e t c a r t x e a t a d l a c n i i l C e s e n a p a J s r a e y 5 2 l e a M 1 s r a e y 5 2 h s i n a p S l e a M 1 i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( : ) D S ± n a e M l ( e r o c s e a c S 6 - S N S L ) 0 3 – 0 g n i r o c s , m e t i i t r e k L - 6 ( 6 - S N S L s r a e y 9 4 – 8 1 7 9 . : e r o c s l l a r e v O s s e n d e t c e n n o c l i a c o s g n i r u s a e M i n a d n I & , n a e r o K , e s e n a p a J , n a c i r e m A 4 5 . : e r o c s y l i m a F d e r o c s ; s m e t i i t r e k L - 3 ( e a c s b u s y l l i m a F 7 , 9 2 s r a e y 3 1 n a i l a t I l s e a m e F 2 e s e n a p a J s r a e y 9 3 l e a M 1 s t n a p i c i t r a P ) N ( i r o m o k i i k H ) 1 = N ( i r o m o k i i k H ) 1 = N ( i r o m o k i i k H ) 6 3 = N ( i r o m o k i i k H ) 2 = N ( i r o m o k i i k H ) 1 = N ( ) d e u n i t n o c ( 2 e l b a T ] 1 5 [ . l a t e a w u S ) s ( r o h t u A ) r a e y ( ] 5 4 [ . l a t e o r e e v O j ] 6 3 [ . l a t e o e T ] 6 4 [ i i r e n a R ] 4 4 [ . l a t e o t a K Yung et al. BMC Psychol (2021) 9:104 Page 13 of 30 - e r p n o i t n e v r e t n i i g n h c a o c d e s a b - h t g n e r t S i l t r e k L - 0 1 ( e a c S m e e t s E - f l e S g r e b n e s o R : s e r o c s t s o p ) 0 3 – 0 d e r o c s ; s m e t i 5 2 o t 6 1 : l e a c S m e e t s E - f l e S g r e b n e s o R - f l e s r e h g h i e r o c s r e h g H i 6 o t 7 1 : l e a c S s s e r t s i D l i l a c g o o h c y s P r e l s s e K m e e t s e - f l e s h g h i 5 1 , m e e t s e . 4 3 o t 8 1 . : l e a c S y t i l a t i V e v i t c e b u S j - f l e s w o l ; 5 1 < m e e t s e = = = ≥ s g n i l e e f s e r u s a e m l ( e a c S y t i l a t i V e v i t c e b u S j g n i r o c s , i d e z i g r e n e g n e b r o s s e n t r e a f l o - r e t n i n e h w r e d n e g e t i s o p p o e h t h c a o r p p a 9 1 e g a , l e a m a f i o a t a d c h p a r g o m e D . d e t s e d e n fi n o c : i s r o v a h e b g n w o i l l o f e h t d e t i b h x E i e m i t f j o y t i r o a m e h t t n e p s , e m o h t a f l e s o t y t i l i b a n i , s t n e r a p n i t s u r t s i , m y t e x n a i , y l i l i a c o s n i i e g a g n e t o n d d m o o r n w o n , i m a f h t i w t c a t n o c d e s u f e r , i s p h s n o i t a e r l m o o r o t y r t n e k c o b o t e r u t i n r u l f d e s u d n a , a i t a o r C m o r f e a m a f l i o a t a d c h p a r g o m e D t c a t n o c d o v a o t i 4 2 e g a d e n fi n o c : i s r o v a h e b g n w o i l l o f e h t d e t i b h x E i e s a c t n e i t a p a m o r f d e t c a r t x e a t a d l a c n i i l C l y n o d e t c a r t x e a t a d e v i t p i r c s e D e m i t f j o y t i r o a m e h t t n e p s , e m o h t a f l e s i t o n d d d n a , s e m a g r e t u p m o c g n y a p l i d e t c e g e N l . i s p h s n o i t a e r l l i a c o s n i e g a g n e i a t a d c h p a r g o m e D . i e n e g y h d n a e r a c - f l e s 5 3 e g a , l i z a r B m o r f e a m a f l o o t n r u t e r o t n o i t a v i t o m : s e c n e i r e p x e e f i L d n a t c a t n o c e t i s h 0 2 m o r f a t a d l a c i r i p m E i i l g n n a p x e o t s e i t l u c ffi d i ; é m u s e r n o s n o i t - a c fi i l a u q f o k c a l i g n v a h ; e c n e d fi n o c - f l e s d e l l a c n o i t n e v r e t n i f o e s u ; y t i t n e d i l i a c o s f o k c a l e s u a c e b e p o e p o t l l s e v e s m e h t d o o W e c a p S e e r F r o m e e t s e - f l e s n i i g n k c a l , e r u l i a f f o r a e f ; y t i t n e d i f o e s n e s e v i t i s o p a k c a l , e m o c r e v o t r o p p u s n i s r e d a e l p u o r g d n a ff a t s d n a , i g n h t y n a o d o t e b a n u l , i y t e x n a f o s g n i l e e f p u o r g l t o n d u o c y e h t l s e c a t s b o t u b y t e c o s i , i r o m o k i i k h h t i i w g n w e v r e t n i i e v i t a t i l a u q l i a c o s : i s r o v a h e b g n w o i l l o f e h t d e t i b h x E i e s a c t n e i t a p a m o r f d e t c a r t x e a t a d l a c n i i l C - s a e m l ( e a c S s s e r t s i D l i l a c g o o h c y s P r e l s s e K g n i r o c s , n o i s s e r p e d d n a y t e x n a s e r u i s l e v e l s s e r t s i d r e w o l s e r o c s r e w o L ) 0 5 – 0 1 y t i l a t i v r e h g h i y t i l a t i v r e h g h i e r o c s r e h g H i e r o c s r e h g H i ) 4 – 0 = = = ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( e s e n a p a J s r a e y 7 1 l e a M 1 s r a e y 9 1 l e a M – 1 s r a e y 4 2 n a i t a o r C l e a M 1 s r a e y 5 3 n a i l i z a r B l e a M 1 s r a e y 7 6 2 . e s e n a p a J – i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M p u o r g d e x M i ) 1 N ( = s t n a p i c i t r a P ) N ( i r o m o k i i k H ) 1 = N ( ) d e u n i t n o c ( 2 e l b a T ] 7 7 [ . l a t e a m u g u s t a M ) s ( r o h t u A ) r a e y ( i r o m o k i i k H ] 6 4 [ i i r e n a R i r o m o k i i k H ) 1 = N ( i r o m o k i i k H ) 1 = N ( i r o m o k i i k H ) ? = N ( ] 8 4 [ . l a t e c i l i S ] 2 7 [ . l a t e a z o R i s e d u t s e v i t a t i l a u Q ] 6 5 [ i o n g O Yung et al. BMC Psychol (2021) 9:104 Page 14 of 30 l e p o e p n i t s u r t f o k c a l : f o s e c n e i r e p x e e f i L - a t i l a u q d n a h c r a e s e r d e fi f l o a t a d l a c i r i p m E ’ s 0 3 - d M i i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( n w a r d h t i w y l l i i a c o s d e n a p m o c c a e v a h i g n v o v n l i n o i t n e v r e t n i f o e s u ; n o s r e p r e h t o s r e k r o w l i a c o s , t n e m e g a g n e r e k r o w l i a c o s e h t g n i t e e m f o n o i t n e t n i o n t u b , e n i l n o i s w e v r e t n i i s p h s n o i t a e r e t a m l i t n i : s e c n e i r e p x e e f i L e v i t a t i l a u q p u o r g s u c o f d n a l i a u d v d n i i g n w e v r e t n i i e v i t I s r a e y 4 2 – 3 1 e s e n h C i 1 2 , 7 6 e s e n a p a J l e a M 1 ) 8 8 = N ( h t u o Y n w a r d h t i W y l l i a c o S ] 4 5 [ i g n Y d n a g n o W s t n a p i c i t r a P ) N ( i r o m o k i i k H ) 1 = N ( ) d e u n i t n o c ( 2 e l b a T ) s ( r o h t u A ) r a e y ( ] 2 5 [ o k e n a K f o e s n e s g n d v o r p o t i i s g n i t u o o t h t u o y y t i r u c e s l a s r e v e r f o s k c a b t e s h t i w ” s s e c o r p o y - o y “ e b o t s d e e n y r e v o c e r ; d e t o n s s e r g o r p n i ; s s e c o r p r a e n i l - n o n a ; h t u o y f o e c a p t a a y r e v o c e r f o s s e c o r p : s n o i t a d n e m m o c e R n i e g a t s t n a t r o p m i n a s i t s u r t g n d i l i u b e r e h w g n i t r a t s d n a ; s s e c o r p y r e v o c e r r i e h t t a s i t n e i l c s n o i t c a r e t n i e c a f - o t - e c a f : s e c n e i r e p x e e f i L d n a a t a d l a c n i i l c m o r f a t a d l a c i r i p m E g n i t i s i v e m o h f o n o i t n e v r e t n i e s u ; s h t n o m l a s u f e r n e d d u s e v a h n a c , h t u o y n w a r d h t i w / s e i t i v i t c a l i a c o s n i t r a p e k a t o t h t u o y m o r f r e k r o w l i a c o s g n i r u d t c a t n o c e c a f - o t - e c a f r o f d l i h c h t i w t c a t n o c e c a f - o t - e c a f o n n a e v a h T E E N e m o s t u b , i r o m o k i i k h d e r e - d i s n o c e b n a c T E E N : a t a d y r a t n e m e p p u S l o t y t i v i t i s n e s , y c a v i r p s d e e n t n e i l c n e h w n o i t i n g o c e r , s s e n e r a w a d n a y t i v i t i s n e s s e r i u q e r g n i t i s i v e m o h : s n o i t a d n e m m o c e R e f i l l i a c o s e v i t c a t n e m e g a g n e e r n i i l s e u c s e d v o r p s g n d n u o r r u s i s ’ t n e i l c i n a c h c h w s t s e r e t n i / s e b b o h o t n o i t a e r l i e g a g n e e r g n i r u d n o i s s u c s i d r o f i s c p o t e b t n e i l c f o g n i l l a c e m a n e v i t i s n e s : e c n e i r e p x e e f i L e v i t a t i l a u q p u o r g s u c o f d n a l i a u d v d n i e h t , i r o m o k i i k h f o e r u t a n g n d u c e s - f l e s l i ; e m o h e h t n i i d e d v o r p e r a c f o y t i r o a m j e h t e r e h w s t r a t s “ f o s n o i t a d n e m m o c e r ” s i t n e i l c ; ” h t u o y n e d d h r o s y u g i l a w a r d h t i w “ f o i s w e v r e t n i d a h r e h t o m , s r e b m e m y l i m a f h t i w y n o l e r u t a r e t i l I s r a e y 4 2 – 5 1 e s e n h C i – ) 0 3 = N ( h t u o Y n w a r d h t i W y l l i a c o S ] 0 7 [ g n o W s r a e y 4 2 – 3 1 e s e n h C i 0 5 , 2 0 2 ) 2 5 2 = N ( h t u o Y n w a r d h t i W y l l i a c o S ] 6 [ g n o W Yung et al. BMC Psychol (2021) 9:104 Page 15 of 30 i - a c n u m m o c e v i t c e ff e n i , s s e n e v i t i t e p m o c d n a y t i t n e d i f o e s n e s e v i t i s o p a k c a l , n o i t - h t o m o r f s l a s i a r p p a e v i t a g e n , e c n e d fi n o c - n o n g n v a h i , e r u t u f r o f s s e n s s e e p o h l f o s t h g u o h t i i r o g n k n h t e v i t a g e n , f l e s d n a s r e l e b a n u , s n o i t c a r e t n i l i a c o s f o r a e f , f l e s f o y c a u q e d a n i f o s g n i l e e f , b o j a e r u c e s o t o o t y t e c o s i i l s n a p m o c , e c n e t e p m o c n i r o l o d o t e b a n u t l e f , r i a f n u d n a g n d n a m e d i , e m o h t a t n e m e n fi n o c s t i b h x e i , i g n h t y n a f o s s o l , l e p o e p t s u r t o t y t i l i b a n i d n a r a e f d n o c e s d n a n o s r e p t s r fi m o r f m u r o f s g n i l e e f , m e e t s e - f l e s w o l , l e p o e p n i t s u r t s e c n e i r e p x e n o s r e p ) 3 o t e u d t e n r e t n I e h t i r e v o s p h s n o i t a e r g n l i t e n r e t n i m o r f n o i t c a r t x e a t a d l a c i r i p m e ) 5 N ( m o r f t n e m e g a r u o c n e g n v e c e r i i , d e t s u r t y e h t e n o e m o s g n i t e e m m o r f i s p h s n o i t a e r l g n i r i p s n i d n a e v i t i s o p : e c n e i r e p x e e f i L i s w e v r e t n i e v i t a t i l a u q e n o - o t - e n O t a h t g n i l e e f , d e k i l d n a w e n k y e h t e n o e m o s g n v i i l d n a d e d y e h t i f i l e r a c d u o w e n o o n l i a m n a n a e k i l i d o v a o t d e r i s e d s t n e r a p : e c n e i r e p x e e f i L f o e t i s n o s w e v r e t n i i 0 5 m o r f a t a d l a c i r i p m E s s e n l l i l a t n e m f o g n i l e b a l d n a a m g i t s l a t n e m d n a i r o m o k i i k h , s p u o r g t r o p p u s - o k i i k h a d l i h c r i e h t l l a c o t e s o h c e r o f e r e h t - l i h c , s t n e r a p h t i w s e i t i n u m m o c s s e n l l i - o k i i k h e h t t a h t i l m a c t o n d d i ; s i s o n g a d i n o i t a v r e s b o d e fi d n a l i c i r t a h c y s p a g n v e c e r i i r e t f a n e v e , i r o m ff a t s m a r g o r p d e t a e r - i r o m o k l i i k h , n e r d l t n a s a e p s a w e c n e i r e p x e i r o m - h s i l b a t s e t u o b a s u o i t u a c : s e c n e i r e p x e e f i L d n a s w e v r e t n i i e v i t a t i l a u q e n o - o t - e n O s r a e y 0 5 – 9 2 e s e n a p a J 1 , 3 e s e n a p a – J – – i s g n d n fi y e K d e s u s t n e m u r t s n I r e d n e g d n a , y t i c i n h t e , e g a n a e M ) e l a m e F s v e l a M ( , ) D S ± s r a e y n a e m ( i r o m o k i i k H d e r e v o c e R ) 4 = N ( s t n a p i c i t r a P ) N ( ) d e u n i t n o c ( 2 e l b a T ) s ( r o h t u A ) r a e y ( ] 1 1 [ n a a T j i r o m o k i i k H ) ? = N ( ] 1 7 [ i n e t s n b u R i i r o m o k i i k H ] 7 5 [ o k e n a K d n a g n o Y i s t n a p c i t r a p m u r o f e n s t n e d n o p s e r y x o r P ) 0 6 1 = N ( = = i l n O N ( t e g o t e b a n u l , i s p h s n o i t a e r l r e e p o n g n v a h i e h t e s u , t n e v e e f i l c i t a m u a r t a d e c n e i r e p x e d n a y t i t n e d i - f l e s e v i t i s o p d n fi o t t e n r e t n I f o e s n e s , m o d e r o b , s t h g u o h t r o s g n i l e e f e v i t a g e n g n v a h i , g n i r i p s n i d n a e v i t i s o p , l e p o e p n i t s u r t f o k c a l , s r e h t o h t i w g n o a l , d l r o w e d i s t u o e h t h t i w h c u o t g n i s o l , t c a t , s r e h t o h t i w t c a t n o c e c a f - o t - e c a f i g n k c a l - n o c l i a c o s e l t t i l h t i w , e m o h t a t n e m e n fi i s w e v r e t n i t c a t n o c l i a c o s e l t t i l - n o c n o d e r e t n e c e y t s e f i l l : s e c n e i r e p x e e f i L e v i t a t i l a u q d e r u t c u r t s - i m e s e n o - o t - e n O s r a e y 9 2 – 4 1 e s e n h C i 8 , 2 2 ) 0 3 = N ( h t u o Y n w a r d h t i W y l l i a c o S ] 3 5 [ g n o W d n a i L f o s s o l a , e m o h t a g n y a t s d n a t s i r e g n o l o n j - y o n e e m o s , s e m a g r e t u p m o c n i t s e r e t n i m o r f d e e r f g n i l e e f d n a n o i s u c e s l r i e h t g n i s e n i l e m i t d n a s t n a r t s e r i t n e n o p m o c e s i r p r e t n e d e s a b y t i n u m m o c D E B - C , e f i l f o y t i l a u q L O Q , i g n n a r t i r o , n o i t a c u d e , t n e m y o p m e n l i t o n T E E N , i n o i t a v e d d r a d n a t S D S , s r u o h h , s t n a p i c i t r a p f o r e b m u n l a t o T N , s u s r e v s v Yung et al. BMC Psychol (2021) 9:104 Page 16 of 30 case studies, reports or case series; and four qualitative studies. Two studies that measured connectedness in hikiko- mori reported low levels of connectedness. One study using the Modified Berkman–Syme Social Network Index (SNI), which measures social ties and involve- ment in relationships, reported the low SNI score of 2.79 ± 1.80 out of 7 (Table 2) in hikikomori [35]. This was also reflected in another study using a different measure- ment tool, in which low scores for social connectedness (9.7 ± 5.7 out of 30.0 in the Lubben Social Network Scale (LSNS)-6 questionnaire) were also found (Table 2) [36]. In a study of relationships with peers and family, hikiko- mori were found to have experienced more rejection from peers and parents, had a greater tendency towards shyness, and experienced a higher level of maladjust- ment to school when compared with university students (Table 2) [37]. The scores of hikikomori compared to uni- versity students were 52.83 ± 12.27 versus 46.89 ± 9.76 for shyness, 2.21 ± 0.70 versus 2.09 ± 0.71 for maternal avoidance, 10.29 ± 4.44 versus 7.41 ± 4.02 for parental rejection, 3.85 ± 2.31 versus 2.41 ± 2.15 for peer rejec- tion, and 4.50 ± 1.62 versus 3.20 ± 1.85 for maladjust- ment to school, respectively (Table 2). Low to moderate correlations at r = 0.219–0.400 (p < 0.05–p < 0.01) with the aspects of shyness, ambivalent maternal attachment, adjustment to middle school, parental rejection, parents threatening a loss of relationship with or ignoring their child, and peer rejection were reported as having been experienced by hikikomori [37]. They further indicate that hikikomori had a low level of connectedness with others due to a lack of support from peers or parents. This, along with their shy temperament, would add to their difficulties in initiating or building relationships. The types of relationships hikikomori maintained and asocial behaviors they exhibited have also been inves- tigated. In two studies, 19.0- 34.2% of hikikomori were found to have no relationships at all; while 57.4–63.0% still maintained relationships with others, mostly family members, (Table 2) [38, 39]. They exhibited social with- drawal behaviors, with the duration ranging from three months [3] to twenty-five years [40]. In Chauliac et al.’s [38] study, 27.0% of hikikomori did not leave home, while 35.0–38.0% still went on outings, either alone or accom- panied. Typical daily activities of hikikomori, in terms of number of hours, were: 7.83 ± 1.99 sleeping, 5.09 ± 4.97 using the computer, 3.11 ± 5.03 using a tablet or mobile phone, and 1.90 ± 1.03 eating, while the remainder of their time was spent watching television, reading comics or animations, reading, idling, or facing the wall (Table 2) [41]. Hikikomori were also reported to have difficul- ties with interpersonal relationships, social interactions, or fitting into society [42, 43], and experienced peer rejection (Table 2) [37, 42]. In one study, 74.1% of hikiko- mori had difficulties with interpersonal relationships and social anxieties. For example, 36.2% feared meeting people, 48.3% were anxious about meeting with familiar people, 51.7% were worried about people’s impression of them, and 53.4% could not blend into groups (all p < 0.001 when compared with non-hikikomori between the ages of 15–39) (Table 2) [43]. Hikikomori showed disconnect- edness with peers and society, had limited relationships and most of those were with family [38, 39], and experi- enced social anxiety [43], making it difficult for them to establish relationships. The case studies were of eleven hikikomori, nine males and two females, who had been in social withdrawal for 2–20 years, were aged 13–40, and whose behaviors were reflective of those reported in the above quantitative studies, namely, confining themselves at home, spending the majority of their time in their room, and not engaging in any social relationships or avoiding face-to-face con- tact with others (Table 2) [44–49] and having social anxi- eties [50]. One hikikomori felt exhausted from effort in maintaining relationships, was unable to relate well with others, feared entering adult society, and had no confi- dence in coping with society [51]. In three of the cases, the hikikomori refused to have contact with their family members [46–48] with one hikikomori using furniture to block entry to his room to avoid contact [48]. One of the hikikomori would leave home once a month for appointments at an outpatient clinic [44]. While another reported of mistrusting their parents and inability in approaching the opposite gender when interested [50]. A reversed sleep/wake cycle of being awake in the evening and sleeping during the day was reported [44, 47]. The results from the four qualitative studies were con- sistent with previous reports of disconnected behavior, which described hikikomori’s losing touch with the out- side world, having no peer relationships, or being unable to get along with others [6, 52, 53]. In one case, a mother could not see her child face-to-face for months [6]. These four studies further explored the underlying reasons behind the social withdrawal of hikikomori. Some of the reasons given were a lack of trust in people [52, 53], hav- ing experienced a traumatic life event [6, 52, 53] such as bullying, the death of a family member, or the divorce of one’s parents. In a study by Wong and Ying [54], some hikikomori were reported to be conducting intimate rela- tionships online, but had no intention of meeting those people in person. While Chan [55] found, the higher the friendship or intimacy level, the more forms of online communication would be shared between the youth and that peer. Although the dynamics of these online rela- tionships have not been explored, which may indicate a knowledge gap. Yung et al. BMC Psychol (2021) 9:104 Page 17 of 30 Domain of hope and optimism Hope and optimism involve a belief in one’s ability to recover, find the motivation to change, have hope-inspir- ing relationships, think positively and value success, and have dreams and aspirations [24]. In this domain five studies relating to hope and optimism in hikikomori were found, four of which were qualitative studies and one a cross-sectional study. In the current literature, the belief of individuals in their ability to recover from the hikikomori lifestyle has not been explored. There have been reports on the moti- vation of hikikomori to change and return to society; however, there were obstacles that they could not over- come, such as feelings of anxiety or their lack of qualifi- cations to list on their resume [2, 56]. Some hikikomori formed positive and inspiring relationships from meet- ing someone they trusted, received encouragement from someone they knew and liked [2], or met someone on the Internet [53]. In contrast, the hikikomori in Yong and Kaneko’s [57] study were cautious about establishing rela- tionships over the Internet due to fear and an inability to trust people. Hikikomori are seen as people who engage in negative self-appraisals and thinking [56, 57]. They do not talk about success, but rather exude a sense of failure. In two of the five qualitative studies, hikikomori harbored feelings of hopelessness about their future [57]. They thought that they would be unable to secure a job, felt inadequate or incompetent, and complained that society was too demanding and unfair [57]. Hikikomori felt una- ble to do anything [56, 57] and had a fear of failure [56]. Their dreams and aspirations have not been explored in a qualitative context, however, in a cross-sectional study hikikomori were found to have high scores in the aspect of unclear ambitions about the future. A newly developed scale was used in that study, which measured scores on the unclear ambitions for the future of three participant groups: hikikomori between the ages of 20 and 39, age- matched NEET (people not in education, employment, or training), and working adults; it was found that of the three groups, hikikomori ranked the highest in having unclear ambitions for the future [58]. Domain of identity This domain involves the following: the multiple dimen- sions of identity, rebuilding or redefining a positive sense of identity, and overcoming stigma [27]. The multiple dimensions of identity, as applied towards hikikomori, would be gender, ethnicity, culture, religion, social class, personal identity and attributes, and sexual orientation. In this domain 29 articles were found, including 13 cross- sectional studies; seven case reports, studies or series; one mixed-methods study; one longitudinal study; one pilot case control study; one interventional study; and five qualitative studies. Of the 13 cross-sectional studies on hikikomori, involv- ing a total number of 1719 hikikomori, the majority were males (1043 (60.7%) to 564 (32.8%) females), although this distribution may be due to the sampling approach used in the studies; one study did not report their gender distributions. By contrast, Wu et al. [42] reported slightly more females (n = 90, 53.6%) than males (n = 78, 46.4%). More studies are needed to further explore gender dis- tribution ratios and if there are behavioural differences between male and female hikikomori. Hikikomori are not confined to any specific race or nationality, and Ameri- can, Brazilian, Chinese, French, Italian, Japanese, Korean, Oman, Spanish, Taiwanese, and Ukrainian hikikomori have been featured in cross-sectional studies or case reports (Table 2) [3, 38, 39, 41, 43, 45, 47, 49, 50, 59– 63]. No studies were found on the culture or religion of hikikomori; however, hikikomori were reported to have high levels of computer or Internet use, ranging from 5.09 ± 4.97 to 5.20 ± 3.40 h per day [3, 35]. Further inves- tigations may be considered to identify whether long durations of computer use are part of hikikomori culture and to determine what are the cultural norms of hikiko- mori. Studies of religion and culture may be conducted to uncover more about the phenomenon; however, they would not contribute towards the recovery of hikiko- mori. Seven studies reported on the dimension of social class. The majority of hikikomori are reported to have a high school level of education or above [36, 40, 42, 63]. However, in two studies it was unclear what educational level they had achieved: in their study, Yong and Nomura [43] reported that the majority of hikikomori had fin- ished school but did not report on their level of educa- tion; while Nagata et al. [64] reported the average years of education received being 11.7 ± 1.7, but it is unclear whether the preschool years were included in that fig- ure. Two studies reported that the majority of hikikomori belonged to the middle class [43, 60], while Wu et al. [42] reported that the majority lived in low-income areas. As only a few studies on the social class of hikikomori have been conducted, these might not be representative findings. With regard to personal attributes and identity, hikiko- mori were found to have higher levels of passive-aggres- siveness, a tendency to adjust their emotions to the environment and people, an inclination to suppress the expression of emotions when feeling shaken in social sit- uations, and to need and be open to emotional relation- ships [65] when compared to non-hikikomori using the Rorschach Comprehensive System (Table 2). Three quali- tative studies reported that hikikomori are lacking of self-esteem or self-confidence [2, 56, 57]. Also reported Yung et al. BMC Psychol (2021) 9:104 Page 18 of 30 as features of hikikomori are non-competitiveness, inef- fective communication, identity issues because of nega- tive appraisals from people [57], or difficulties explaining themselves to people because they lacked a social identity such as a title to an occupational or student status [56]. A longitudinal study found that hikikomori had low to moderate scores for interpersonal support, belonging- ness, and self-esteem, with Interpersonal Support Evalu- ation List scores (ISEL) of 24.60 ± 6.30 out of 48, scores of 6.00 ± 2.45 out of 12 for belongingness, and 5.59 ± 2.08 out of 12 for self-esteem [35], as shown in Table 2. No comparisons were made with age-matched youth with- out social withdrawal in the study. However, this com- pared to delinquent age matched Chinese [66] and United States freshman students [67] (26.19 ± 4.38 and 38.30 ± 6.82, respectively) would seem low. This ISEL has not been used normal youth population studies in Asia. No studies were found that explored the sexual orienta- tion of hikikomori. On rebuilding and redefining a positive self-identity, the previously mentioned studies on personal iden- tity found that hikikomori lack a positive sense of iden- tity [2, 56, 57] and need to rebuild it. In a case study, it was reported that a hikikomori exhibited behaviors of work refusal, felt ashamed of himself and feared of being labelled unemployed [51]. While some hikikomori may use the Internet to find a positive self-identity [53], relationships were also found to be important to their self-esteem [60]. Self-esteem refers to the positive and negative viewpoints individuals have of themselves [68]; and having high self-esteem may result in beliefs of being good, worthy, and positively viewed by others [69], there- fore a positive sense of identity. In Chan and Lo’s [60] study, they found that self-esteem was highly correlated with relationships with parents: r = 0.73, p = 0.0000, sib- lings: r = 0.66, p = 0.0000, teachers: r = 0.13, p < 0.05, and peers: r = 0.16, p < 0.05, and that the higher the associa- tion, the higher was the self-esteem of the hikikomori. It has been acknowledged that hikikomori are sensitive to people calling them names, such as: “hidden youth” and “withdrawn guys” [70]. However, in the dimension of stigma, only one study was found. The qualitative study, from Japan, found that parents desired to avoid stigma and the label of mental illness, and therefore chose to continue to call their child a hikikomori, even after their child received a psychiatric diagnosis [71]. There have been no other studies on stigma and hikikomori, which suggests that there is a need for more studies to be con- ducted on this subject. Meaning in life Meaning in life refers to the meaning of mental illness experiences, spirituality, quality of life, meaningful social roles and goals, and the rebuilding of one’s life [24]. As not all hikikomori have a comorbid mental illness, in this study the dimension of the meaning of mental ill- ness experiences will instead refer to the meaning of the hikikomori experience. In this domain a total of 11 arti- cles were found, comprising three cross-sectional stud- ies, five qualitative studies, and three case reports. In understanding the life experiences of hikikomori, an international cross-sectional study reported that hikiko- mori experience high levels of loneliness, at 55.4 ± 10.5 out of 80 on the University of California Los Angeles (UCLA) Loneliness Scale (Table 2), in comparison to a score of 40.0 for normal controls found in other stud- ies in these same countries [36]. Five qualitative stud- ies, each with 4–30 subjects selected conveniently from outreach programs, non-profit organizations, or online forums, reported on the lifestyles, feelings, and thoughts of hikikomori. They found a lifestyle centered on confine- ment at home, with little social contact, and having nega- tive feelings or thoughts [2, 53, 57]. Commonly reported were feelings of low self-esteem, a lack of confidence, hopelessness, and a loss of trust in people [2, 57]. Also reported were of the feeling that no one would care if they died, that they were living like an animal [2], a fear of social interactions [57], boredom, the sense that they could no longer stand staying at home, a loss of interest in computer games [53], and that they did not claim that the hikikomori experience was pleasant [71]. In contrast, some reported feeling freed from restraints and timelines and enjoying their seclusion [53]. Poor hygiene behaviors were reported in a minority of hikikomori [38, 49, 72], with one person defecating and urinating in jars or bot- tles in his room [49]. No studies were found on spirituality and meaning- ful social roles or goals in relation to hikikomori. Two cross-sectional studies were found measuring quality of life. One study reported lower quality of life scores, as measured using the Chaban Quality of Life Scale, for hikikomori with or without psychiatric comorbidities, with scores of 11.7 ± 2.70 and 13.7 ± 3.3 respectively, both with a statistical significance of p = 0.001, when compared with the control group of non-hikikomori, at 19.3 ± 3.50 (Table 2) [61]. On the contrary, Chan and Lo’s [4] study using the World Health Organization Quality of Life scale, reported an improvement in the quality of life of hidden youth as the time spent being socially with- drawn increased, with an overall correlation of r = 0.550; p = 0.0000. This may indicate a positive adjustment in the well-being of hidden youth. However, those with a higher degree of social withdrawal were seen to have a lower quality of life, with an overall correlation of r = − 0.850; p = 0.0000 (Table 2) [4]. This would indicate that social withdrawal is not a positive factor in the quality of life of Yung et al. BMC Psychol (2021) 9:104 Page 19 of 30 s i s y l a n a l a c i t s i t a t s ? d e s u a n i d e r u s a e m ? y a w e b a i l l e r e t a i r p o r p p a s a W s e m o c t u o e r e W y n a n i d e d u l c n i f o s e m o c t u o s t n a p i c i t r a p e h t e r e W p u w o l l o f s a W d n a e t e p m o c l e r e w , t o n f i s e c n e r e ff d i s n o s i r a p m o c s p u o r g n e e w t e b e h t n i d e r u s a e m ? y a w e m a s y l e t a u q e d a p u d n a d e b i r c s e d ? d e z y l a n a f o s m r e t n i w o l l o f r i e h t s t n e m e r u s a e m e m o c t u o e h t f o e h t t s o p d n a / n o i t n e v r e t n i ? e r u s o p x e e r p h t o b e r e h t e r e W l e p i t l u m s l o o t T A M M d n a I B J h t i w w e v e r n i i d e d u c n l i l s e c i t r a f o l a s i a r p p a y t i l a u Q 3 e l b a T I B J n i t n e m s s e s s a r o f a i r e t i r C l a t n e m i r e p x e - i s a u q r o f I B J e r e h t s a W l o r t n o c a s t n a p i c i t r a p e h t e r e W s t n a p i c i t r a p e h t s i t a h w y d u t s e h t e r e W e h t n i r a e l c t i s I ? p u o r g y n a n i d e d u l c n i y n a n i d e d u l c n i t a h w d n a ’ e s u a c ‘ y d u t S r a l i m i s g n i v i e c e r , e r a c / t n e m t a e r t e h t n a h t r e h t o f o n o i t n e v r e t n i r o e r u s o p x e s n o s i r a p m o c ? t s e r e t n i ? r a l i m i s s n o s i r a p m o c ? ’ t c e ff e ‘ e h t s i ✓ ✓ A / N ✓ ✓ ? ? ? ? ? ✓ ✓ X ✓ ✓ ✓ ✓ ✓ ? ? ✓ ✓ ✓ ✓ ✓ ✓ X ✓ X X ✓ ✓ ✓ ✓ ✓ e t a i r p o r p p a s a W ? d e s u s i s y l a n a l a c i t s i t a t s g n o l t s e r e t n i d i l a v , d r a d n a t s e r u s o p x e f o d o i r e p e h t s a W a n i d e s s e s s a s e m o c t u o e r e W e b o t h g u o n e ? l u f g n n a e m i s e s a c r o f y a w ? s l o r t n o c d n a l e b a i l e r d n a i g n d n u o f n o c h t i w l a e d o t ? d e t a t s s r o t c a f ? d e fi i t n e d i d n a s e s a c r o f l e b a i l e r d n a s r o t c a f y a w e m a s e h t d i l a v , d r a d n a t s ? s l o r t n o c ? y a w i s e g e t a r t s i g n d n u o f n o c n i d e r u s a e m e r e W e r e W e r u s o p x e s a W e r u s o p x e s a W a n i d e r u s a e m a i r e t i r c e m a s e h t e r e W r o f d e s u n o i t a c fi i t n e d i d n a s e s a c f o ? s l o r t n o c ✓ A / N ✓ X X ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ l e b a r a p m o c n a h t r e h t o e c n e s e r p e h t n i e s a e s i d f o e h t r o s e s a c f o e c n e s b a n i e s a e s i d ? s l o r t n o c e h t e r e W s p u o r g ? ✓ ✓ ✓ ✓ ✓ r o m A - n ó g a a M l ] 9 3 [ . l a t e ] 4 7 [ . l a t e w a L . l a t e a m a y o k o Y ] 3 1 [ . l a t e e e L ] 5 7 [ n a h C ] 8 1 [ l o r t n o c e s a c r o f I B J l e b a r a p m o c n a h t r e h t o e c n e s e r p e h t n i e s a e s i d f o e h t r o s e s a c f o e c n e s b a n i e s a e s i d ? s l o r t n o c e h t e r e W s p u o r g y d u t S ✓ . l a t e i k u s t a K ] 5 6 [ Yung et al. BMC Psychol (2021) 9:104 Page 20 of 30 ? d e s u s i s y l a n a e t a i r p o r p p a l a c i t s i t a t s i s e g e t a r t s s s e r d d a o t l e t e p m o c n i p u w o l l o f ? d e z i l i t u s a W e r e W w o l l o f s a W , l e t e p m o c p u , t o n f i d n a w o l l o f o t s s o l d e b i r c s e d p u o t s n o s a e r ? d e r o p x e l d n a o t s e m o c t u o r o f h g u o n e g n o l e b o t t n e i c ffi u s ? r u c c o e h t e r e w d n a d e t r o p e r e h t s a W e m i t p u w o l l o f n i d e r u s a e m d n a d i l a v a ? y a w e b a i l l e r s e m o c t u o e h t e r e W s t n a p i c i t r a p h t i w l a e d o t s r o t c a f n i d e r u s a e m e h t f o e e r f i g n d n u o f n o c ? d e fi i t n e d i d n a d i l a v a e h t e r e W / s p u o r g i s e g e t a r t s i g n d n u o f n o c e r e W e r e W e r u s o p x e e h t s a W t a e m o c t u o f o t r a t s e h t ? y d u t s e h t s r o t c a f ? d e t a t s ? y a w e b a i l l e r s e r u s o p x e d e r u s a e m e h t e r e W o t y l r a l i m i s l e p o e p n g i s s a d n a d e s o p x e d e s o p x e n u ? s p u o r g h t o b o t s p u o r g o w t d n a r a l i m i s e h t e r e W d e t i u r c e r e h t m o r f e m a s ? n o i t a l u p o p y d u t S t r o h o c l a c i t y l a n a r o f I B J ) d e u n i t n o c ( 3 e l b a T s i s y l a n a l a c i t s i t a t s e t a i r p o r p p a s a W s e m o c t u o e h t e r e W i s e g e t a r t s e r e W i g n d n u o f n o c e r e W , e v i t c e j b o e r e W e r u s o p x e e h t s a W h t i w l a e d o t ? d e fi i t n e d i s r o t c a f d r a d n a t s a n i d e r u s a e m e h t d n a s t c e j b u s y d u t s e h t e r e W a i r e t i r c e h t e r e W n i n o i s u l c n i r o f y d u t S ✓ ✓ ✓ ✓ ✓ ✓ X X ✓ ✓ ✓ ] 5 3 [ . l a t e n e u Y l a n o i t c e s - s s o r c l a c i t y l a n a r o f I B J ? d e s u a n i d e r u s a e m ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ? y a w ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ l e b a i l e r d n a d i l a v ? d e t a t s s r o t c a f i g n d n u o f n o c r o f d e s u a i r e t i r c f o t n e m e r u s a e m ? n o i t i d n o c e h t l e b a i l e r d n a d i l a v d e b i r c s e d g n i t t e s l y l r a e l c e p m a s e h t ? y a w ? l i a t e d n i ? d e n fi e d X X X X X ✓ X X ? X X X X X ✓ X X ? ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ] 4 6 [ . l a t e a t a g a N ] 4 [ o L d n a n a h C - s a r o N d n a a d h c U i ] 8 5 [ t i k n u k k a ] 3 6 [ . l a t e a d e m U ] 1 4 [ . l a t e n e u Y a r u m o N d n a g n o Y ] 2 4 [ . l a t e u W ] 3 4 [ ] 7 3 [ i i e k c D d n a g e i r K ] 0 4 [ . l a t e o d n o K e s n o p s e r e h t s a W e s n o p s e r w o l e h t d e g a n a m e t a r ? y l e t a i r p o r p p a , e t a u q e d a e t a r s a w , t o n f i d n a ✓ ✓ A / N ✓ e t a i r p o r p p a l a c i t s i t a t s ? s i s y l a n a l e b a i l e r , d r a d n a t s ? s t n a p i c i t r a p l l a r o f y a w n o i t a c fi i t n e d i e h t ? n o i t i d n o c e h t f o l ? e p m a s d e fi i t n e d i e h t f o e g a r e v o c t n e i c ffi u s h t i w a n i d e r u s a e m r o f d e s u s d o h t e m d e t c u d n o c s i s y l a n a e r e h t s a W n o i t i d n o c e h t s a W d i l a v e r e W a t a d e h t s a W d e b i r c s e d g n i t t e s e h t d n a s t c e j b u s y d u t s e h t e r e W ? l i a t e d n i l e p m a s e h t s a W ? e t a u q e d a e z i s ? y a w e t a i r p o r p p a ? n o i t a l u p o p t e g r a t n a n i l d e p m a s y d u t s e r e W s t n a p i c i t r a p e t a i r p o r p p a e m a r f l e p m a s e h t s a W e h t s s e r d d a o t y d u t S l a n o i t c e s - s s o r c e c n e l a v e r p r o f I B J ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ A / N ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ . l a t e r o m A - n ó g a a M l ] 2 6 [ . l a t e a m a y o K ] 9 5 [ ] 8 3 [ . l a t e c a i l u a h C ] 1 6 [ a v o k n a r F Yung et al. BMC Psychol (2021) 9:104 Page 21 of 30 l a c i t s i t a t s s a W r a e l c e r e h t s a W ? e t a i r p o r p p a s i s y l a n a g n i t n e s e r p e h t ) s ( c i n i l c / ) s ( e t i s f o g n i t r o p e r c i h p a r g o m e d ? n o i t a m r o f n i p u w o l l o f r o f o s t l u s e r y l r a e l c s e s a c n o i t a m r o f n i s c i h p a r g o m e d l a c i n i l c f o e h t f o e h t f o e h t f o ? d e t r o p e r ? s t n a p i c i t r a p n i s t n a p i c i t r a p s e m o c t u o g n i t r o p e r r a e l c g n i t r o p e r r a e l c e h t e r e W e r e h t s a W e r e h t s a W ? y d u t s e h t e s a c e h t d D i e v a h s e i r e s l e t e p m o c ? s t n a p i c i t r a p f o n o i s u l c n i e s a c e h t d D i e v i t u c e s n o c f o n o i s u l c n i e v a h s e i r e s ? s t n a p i c i t r a p n o i t a c fi i t n e d i e h t f o s t n a p i c i t r a p l l a e h t n i d e d u l c n i r o f n o i t i d n o c ? s e i r e s e s a c d i l a v e r e W s d o h t e m r o f d e s u r o f y a w e b a i l l e r s t n a p i c i t r a p l l a e h t n i d e d u l c n i ? s e i r e s e s a c n i d e r u s a e m n i n o i s u l c n i r o f , d r a d n a t s a ? s e i r e s e s a c e h t n o i t i d n o c e h t s a W a i r e t i r c r a e l c e r e h t e r e W y d u t S ) d e u n i t n o c ( 3 e l b a T s e i r e s e s a c r o f I B J ? s n o s s e l y a w a e k a t e s a c e h t s e o D e d i v o r p t r o p e r d e t a p i c i t n a n u r o d e fi i t n e d i s t n e v e ? d e b i r c s e d d n a ) s m r a h ( s t n e v e e s r e v d a e r e W n o i t i d n o c l a c i n i l c ? d e b i r c s e d y l r a e l c - t s o p e h t s a W n o i t n e v r e t n i ) s ( n o i t n e v r e t n i t n e m t a e r t r o ) s ( e r u d e c o r p e h t d n a s d o h t e m t n e m s s e s s a r o s t s e t e h t s a W c i t s o n g a i d e r e W n o i t i d n o c l a c i n i l c t n e r r u c e h t s a W n o i t a t n e s e r p n o t n e i t a p e h t f o ? d e b i r c s e d y l r a e l c y l r a e l c s t l u s e r ? d e b i r c s e d y l r a e l c s ’ t n e i t a p e h t s a W y l r a e l c y r o t s i h d n a d e b i r c s e d a s a d e t n e s e r p ? e n i l e m i t ? d e b i r c s e d y l r a e l c s ’ t n e i t a p e r e W c i h p a r g o m e d s c i t s i r e t c a r a h c y d u t S ✓ ✓ ✓ ? ✓ ✓ ✓ ✓ ✓ i s e d u t S / s t r o p e R e s a C r o f I B J ✓ ] 6 3 [ . l a t e o e T ✓ ✓ X ✓ ✓ ✓ X ✓ ✓ ✓ ✓ ✓ X ✓ X X X ✓ ✓ X X X ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ? ✓ X ✓ ✓ ? ? ? ? ✓ ✓ ? d e b i r c s e d ✓ ✓ X ✓ ✓ X ✓ ? ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ? X ✓ X ? ✓ ✓ X ] 7 4 [ . l a t e o t o m a k a S i k u z u S d n a a w u S ] 3 7 [ i r o t t a H . l a t e o r e j r e v O ) 4 1 0 2 ( ] 6 4 [ i i r e n a R ] 4 4 [ . l a t e o t a K ] 6 4 [ i i r e n a R ] 1 5 [ ] 9 4 [ o e T . l a t e a m u g u s t a M ] 7 7 [ ) 9 1 0 2 ( . l a t e ć i l i S ] 2 7 [ . l a t e a z o R Yung et al. BMC Psychol (2021) 9:104 Page 22 of 30 t r o p e r h c r a e s e r a i r e t i r c t n e r r u c , s e c i o v e h t n o e h t m o r f w o fl t n e c e r r o f , r o y l e t a u q e d a d n a , h c r a e s e r s n o i s u l c n o c e h t n i n w a r d i o t g n d r o c c a r i e h t d n a r e h c r a e s e r e h t l a c i h t e , s t n a p i c i t r a p f o e c n e u fl n i e h t o D h c r a e s e r e h t s I e r A e h t s I e h t g n i t a c o l r e h c r a e s e r r o y l l a r u t l u c t n e m e t a t s a e r e h t s I l y g o o d o h t e m h c r a e s e r e h t l y g o o d o h t e m h c r a e s e r e h t l y g o o d o h t e m h c r a e s e r e h t l y g o o d o h t e m h c r a e s e r e h t y t i u r g n o c n e e w t e b e r e h t s I y t i u r g n o c n e e w t e b e r e h t s I y t i u r g n o c n e e w t e b e r e h t s I y t i u r g n o c n e e w t e b e r e h t s I ? a t a d e h t f o n a y b l a v o r p p a e t a i r p o r p p a l a c i h t e f o ? y d o b , n o i t a t e r p r e t n i e c n e d i v e e r e h t r o , s i s y l a n a s i d n a i , s e d u t s ? d e t n e s e r p e r , a s r e v - e c i v ? d e s s e r d d a ? y l l a c i t e r o e h t e h t d n a e h t d n a n o i t a t e r p r e t n i ? s t l u s e r f o n o i t a t n e s e r p e r f o s i s y l a n a d n a ? a t a d d e s u s d o h t e m e h t d n a t c e l l o c o t ? a t a d r o n o i t s e u q ? s e v i t c e j b o e h t d n a h c r a e s e r l a c i h p o s o l i h p e v i t c e p s r e p e h t d n a h c r a e s e r l ? y g o o d o h t e m y t i u r g n o c n e e w t e b d e t a t s e h t e r e h t s I y d u t S ) d e u n i t n o c ( 3 e l b a T e v i t a t i l a u Q r o f I B J y t i l a u q e h t e v i t a t i t n a u q a i r e t i r c e v i t a t i l a u q d n a s d o h t e m e h t f o n o i t i d a r t ? d e v l o v n i h c a e f o y l e t a u q e d a ? d e s s e r d d a s t l u s e r h c r a e s e r e h t e h t s s e r d d a ? n o i t s e u q h c r a e s e r ? n o i t s e u q d e t a r g e t n i s d o h t e m d e r e t s i n m d a i r e w s n a o t o t n g i s e d ? d e d n e t n i s a / e r u s o p x e d n a ? n o i t n e v r e t n i ? n o i t a t e r p r e t n i d n a s i s y l a n a , n o i t c e l l o c ? a t a d y b s s e r d d a o t h c r a e s e r e h t ? s n o i t s e u q ? n o i t s e u q h c r a e s e r e h t ? n o i t s e u q o t e r e h d a n e e w t e b ? s n o i t a t e r p r e t n i y l e v i t c e ff e d e x i m a e r u s o p x e ? s i s y l a n a d n a e m o c t u o e h t ? n o i t a l u p o p , s e c r u o s a t a d d e t a i t n a t s b u s ? a t a d e h t e t a u q e d a r e w s n a o t h c r a e s e r e h t s t n e n o p m o c d n a r e h t e g o t t h g u o r b s t n e n o p m o c e l a n o i t a r e h t s i , d o i r e p r o f d e t n u o c c a e m o c t u o e t a i r p o r p p a e v i t a t n e s e r p e r n e e w t e b s t l u s e r f o y l e t a u q e d a n o i t c e l l o c a t a d h c a o r p p a w o l l a a t a d h c r a e s e r y d u t s e h t f o s e i c n e t s i s n o c n i l l a r e v o o t n i y d u t s e h t f o g n i s u r o f / n o i t n e v r e t n i n g i s e d e h t n i ? a t a d h t o b g n d r a g e r i t e g r a t e h t f o e v i t a t i l a u q y l t n e i c ffi u s m o r f d e v i r e d s d o h t e m e t a i r p o r p p a s s e r d d a o t ? s n o i t s e u q t n e r e ff d i s e c n e g r e v i d y l e t a u q e d a t n e r e ff d i e t a u q e d a y d u t s e h t s r e d n u o f n o c e t e p m o c l s t n e m e r u s a e m s t n a p i c i t r a p e c n e r e h o c n o i t a t e r p r e t n i e h t o D e r A s t l u s e r e h t e r A e h t e r A n a e r e h t s I g n i r u D e h t e r A e r e h t e r A e r A e h t e r A e r e h t s I e h t s I e h t e r A i s g n d n fi e v i t a t i l a u q e v i t a t i l a u q d e t c e l l o c r a e l c e h t e r A e h t s I e h t o D e r e h t e r A y d u t S ✓ ✓ ✓ ✓ ✓ ✓ X ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ d n a ] 0 6 [ o L n a h C T A M M r o f t n e m s s e s s a r o f a i r e t i r C d o h t e m d e x i m r o f T A M M ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ X ? ? ? ✓ ? ✓ ✓ ? ✓ ✓ ✓ ? ✓ ✓ ? ✓ ✓ ? ? ? ? ? ? ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ i g n Y d n a g n o W ] 6 5 [ i o n g O ] 2 5 [ o k e n a K ] 6 [ g n o W ] 0 7 [ g n o W ] 1 1 [ n a a T j ] 4 5 [ ] 1 7 [ i n e t s n b u R i ] 7 5 [ o k e n a K d n a g n o Y g n o W d n a i L ] 3 5 [ Yung et al. BMC Psychol (2021) 9:104 Page 23 of 30 y t i l a u q e h t e v i t a t i t n a u q a i r e t i r c e v i t a t i l a u q d n a s d o h t e m e h t f o n o i t i d a r t ? d e v l o v n i h c a e f o y l e t a u q e d a ? d e s s e r d d a s t l u s e r h c r a e s e r e h t e h t s s e r d d a ? n o i t s e u q h c r a e s e r ? n o i t s e u q d e t a r g e t n i s d o h t e m d e r e t s i n m d a i r e w s n a o t o t n g i s e d ? d e d n e t n i s a / e r u s o p x e d n a ? n o i t n e v r e t n i ? n o i t a t e r p r e t n i d n a s i s y l a n a , n o i t c e l l o c ? a t a d y b s s e r d d a o t h c r a e s e r e h t ? s n o i t s e u q ? n o i t s e u q h c r a e s e r e h t ? n o i t s e u q o t e r e h d a n e e w t e b ? s n o i t a t e r p r e t n i y l e v i t c e ff e d e x i m a e r u s o p x e ? s i s y l a n a d n a e m o c t u o e h t ? n o i t a l u p o p , s e c r u o s a t a d d e t a i t n a t s b u s ? a t a d e h t e t a u q e d a r e w s n a o t h c r a e s e r e h t s t n e n o p m o c d n a r e h t e g o t t h g u o r b s t n e n o p m o c e l a n o i t a r e h t s i , d o i r e p r o f d e t n u o c c a e m o c t u o e t a i r p o r p p a e v i t a t n e s e r p e r n e e w t e b s t l u s e r f o y l e t a u q e d a n o i t c e l l o c a t a d h c a o r p p a w o l l a a t a d h c r a e s e r y d u t s e h t f o s e i c n e t s i s n o c n i l l a r e v o o t n i y d u t s e h t f o g n i s u r o f / n o i t n e v r e t n i n g i s e d e h t n i ? a t a d h t o b g n d r a g e r i t e g r a t e h t f o e v i t a t i l a u q y l t n e i c ffi u s m o r f d e v i r e d s d o h t e m e t a i r p o r p p a s s e r d d a o t ? s n o i t s e u q t n e r e ff d i s e c n e g r e v i d y l e t a u q e d a t n e r e ff d i e t a u q e d a y d u t s e h t s r e d n u o f n o c e t e p m o c l s t n e m e r u s a e m s t n a p i c i t r a p e c n e r e h o c n o i t a t e r p r e t n i e h t o D e r A s t l u s e r e h t e r A e h t e r A n a e r e h t s I g n i r u D e h t e r A e r e h t e r A e r A e h t e r A e r e h t s I e h t s I e h t e r A i s g n d n fi e v i t a t i l a u q e v i t a t i l a u q d e t c e l l o c r a e l c e h t e r A e h t s I e h t o D e r e h t e r A y d u t S ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ X X ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ] 6 7 [ n a h C ] 5 5 [ n a h C T A M M r o f t n e m s s e s s a r o f a i r e t i r C ) d e u n i t n o c ( 3 e l b a T d o h t e m d e x i m r o f T A M M . l o o T l a s i a r p p A d o h t e M d e x i M T A M M , ; l o o T l a s i a r p p A e t u t i t s n I s ’ g g i r B e n n a o J , I B J ; l e b a c i l p p a t o N , A / N ; l l e t t ’ n a C r o r a e l c n U , ) ? ( k r a M n o i t s e u Q ; o N , X ; s e Y ✓ , Yung et al. BMC Psychol (2021) 9:104 Page 24 of 30 all hikikomori, and that the severity of the act of social withdrawal could be a factor in whether it is a positive experience. In the dimension of rebuilding life, a quali- tative study found that hikikomori had difficulties with stepping out of their room at first and initially felt a sense of insecurity [54]. Empowerment The domain of empowerment encompasses personal responsibility, control over life, and a focus on strengths [24]. In this domain five studies were found, two inter- ventional studies, a mixed method study, a qualitative study, and a case study; of which most were focused on interventions for hikikomori. No studies were found that explored the issue of control over life in relation to hikikomori. Studies on how hikikomori experience all dimensions of empowerment are lacking. All interven- tions exploring empowerment will be discussed in the section on interventions using elements of CHIME. Interventions for Hikikomori using elements within the CHIME framework Intervention, case studies and qualitative studies were found using specific elements within the CHIME Frame- work, such as rebuilding a positive self-identity, social roles or life, and empowerment through the interven- tions, namely, Free Space Wood [56], psychotherapy [50, 73], social worker engagement [54, 74], play therapy [75], and C-BED [18]. An ethnographic study was conducted out of a private support group operating out of an alternative school named Free Space Wood in Japan; it reported to offer hikikomori a new social environment to rebuild their social roles and identity [56]. However, the effectiveness of the program was not reported. Of studies explor- ing the use of psychotherapy to help hikikomori rebuild their lives, showed that a long duration was needed until recovery was achieved, and that it was difficult to con- vince hikikomori to stay in the therapy program. In the case studies, a duration of two to four years was reported before recovery was achieved [50, 73]. In Lee et al.’s [3] study, 41 hikikomori received a mean number of 2.8 ses- sions of psychotherapy with home visits. Close to 50% did not show any improvements in their Global Assess- ment Functioning scores post-intervention, as shown in Table 2. In Hattori’s [73] study, which had a 50% attrition rate, hikikomori were reported to have spent the first six months to one year testing the reliability of the therapist, before rejecting the therapist due to mistrust [73]. This might suggest that the effectiveness of psychotherapy for this group still needs to be evaluated. The engagement of social workers using empowerment was examined in two studies. In a quasi-experimental intervention study, social workers engaged young peo- ple online by identifying their strengths and resources for achieving goals and coping. A portion of the par- ticipants were socially withdrawn youth, and the study reported a 4.61% decrease in social withdrawal after the intervention. However, scores from scales measuring such aspects as emotional distress, perceived social sup- port, problem-solving skills, and attitudes towards seek- ing help were not reported for the socially withdrawn [74]. The exhibited behaviors of high computer usage and histories of social anxieties, may have led health prac- titioners to attempt to engage hikikomori through the online platform. And a study evaluating forms of coun- selling to engage hidden youth found using an integrated approach with online and offline counselling showing the highest positive outcomes for total means scores of quality of life (3.74 ± 0.60) when compared to a singular approach of using only one method (online 3.02 ± 0.43; offline 2.52 ± 0.27) (Table 2). Moreover, online counsel- ling was mentioned to offer a platform for communica- tion, while offline counselling offered opportunity for mediation during conflicts between youth and their fam- ily [76] although the study did not investigate or report on recovery [76]. A qualitative study reported that social workers enabled socially withdrawn youth to reengage with the outside world by accompanying them to out- ings to provide a sense of security [54]. And behaviors of feeling insecure when reentering society with sudden set- backs or refusal of social worker re-engagement; there- fore, recovery would need to be at the pace of the user and would be a non-linear process [54]. Although there were no reports of recovery, this approach might be ben- eficial for such youth. Two other intervention studies and a case study were also found focusing on the use of empowerment and a strength-based approach. An interventional study by Chan [75], which investigated whether play therapy with online games would empower and improve the psycho- logical well-being of hidden youth, reported significant correlations between play therapy and empowerment, with r = 0.59; p < 0.05, and the individual’s positive psy- chological state of development, at r = 0.600; p < 0.05. A hierarchical regression analysis has further suggested that empowerment is a strong predictor of the positive outcomes of play therapy. In a pilot intervention study called C-BED, an online platform was used to empower peer-to-peer shared learning and dialogue with online modules. The five hikikomori who participated exhibited a decrease in anxiety and an increased willingness to par- ticipate socially; however, no measurement scales were used in the study [18]. A case study of one hikikomori using strength-based coaching reported that the subject had returned to school and showed an improvement in Yung et al. BMC Psychol (2021) 9:104 Page 25 of 30 Table 4 Applicability of the CHIME framework to assessing hikikomori care Domain Connectedness Support by peers or others Relationships Part of community Hope and optimism Belief in recovery Motivation to change Hope-inspiring relationships Positive thinking and valuing success Having dreams and aspirations Identity Dimensions of identity: Race Gender Social class Culture Religion Sexual orientation Personal attributes / characteristics / identity Rebuilding/redefining a positive sense of identity Overcoming stigma Meaning in life Meaning of hikikomori experiences: Duration of social withdrawal Activities of daily life Feelings & Thoughts Spirituality Quality of life Meaningful life and social roles/goals Rebuilding life Empowerment Personal responsibility Control in life Focusing upon strengths Applicable Not studied ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ scores in the Rosenberg Self-Esteem Scale and Subjec- tive Vitality Scale, and a decrease in score in the Kessler Psychological Distress Scale following the intervention. The results of the individuals from pre-intervention to post-intervention were 16 to 25 (scoring 0–30), 1.8 to 3.4 (scoring 0–4), and 17 to 6 (scoring 10–50), respectively [77]. Although improved psychological well-being was observed from a strength-based focus, testing still needs to be conducted using a much larger sample. All of the interventions mentioned may be possible options for recovery, but further evaluation is needed since many of the studies had small sample sizes or did not have a comparable comparison group, and some did not report on the effectiveness of the intervention. Researchers have identified factors such as the internet [53] and having relationships [60], giving hikikomori a positive sense of identity or better self-esteem, these may be areas that interventions could be designed to target. However, more studies on rebuilding the positive self- identity of hikikomori are needed to better understand the phenomenon and to determine whether more factors are involved. Further explorations into interventions are needed as the mechanism to recovery is still unknown. Examples would be from a one-year longitudinal study that was not an interventional study with a resulting increase in social connectedness over time (SNI from 2.79 ± 1.80, 2.93 ± 2.06, to 3.09 ± 1.87) (Table 2), and to a level of recovery that was sufficient to allow a return to the workforce of almost 50% of the hikikomori in the study [35]. These results arose from an unintentional intervention administered by social workers during home visits to hikikomori during the period of the study, which involved the use of archived patient records [35]. How- ever, in an interventional study of a home visitation ther- apy program offered by psychiatrists to hikikomori over a period of one year, no statistically significant difference was seen in any of the outcome variables of their study, which included social networking [39]. This suggests that further exploration is needed to understand what mecha- nisms help to improve the ability of hikikomori to con- nect with other people. Discussion The CHIME framework for personal recovery was appli- cable to understanding the life experiences of the hikiko- mori population. Forty-four studies were reviewed, most of which were quantitative in nature (75%) with close to one third being cross-sectional studies (29.5%). To apply this framework to hikikomori, slight modifications were made to the framework, including the domain of iden- tity. The multiple dimensions of identity were expanded to encompass gender, religion, social class, personal identity, and attributes. In the domain of meaning, the meaning of the illness experience was replaced by the meaning of the hikikomori experience. Modifications to the framework are displayed in Fig. 2. Thematic overlap occurred between some domains, such as connectedness, identity, and meaning. The iden- tity of being a hikikomori was defined in relation to the hikikomori’s lack of connectedness with society and lack of meaningful social roles; however, this did not affect the application of the framework. By using the framework to assess the literature, hikikomori were shown to be dis- connected from peers and society, to have relationships Yung et al. BMC Psychol (2021) 9:104 Page 26 of 30 Fig. 2 Modified CHIME framework for personal recovery in Hikikomori Care limited mostly to family, to experience social anxiety, engage in negative self-appraisal, feel a lack of identity, exhibit a mistrust of people, and have high levels of pas- sive aggressiveness and a shy temperament. Those lead- ing a hikikomori lifestyle were not limited to a specific race. More studies with a larger sample size need to be conducted to determine differences in gender distribu- tion and social class. Although quality of life seemed to increase with time spent in withdrawal, higher levels of social withdrawal led to a lower quality of life. The CHIME framework was able to encompass most of the aspects relating to hikikomori. However, the limita- tions of the framework were its inability to address the building of trust and the characteristic of non-linearity in the recovery process. For hikikomori, building trust is an important stage in their recovery process [54, 73]. Long durations of six months to a year spent in building rapport have been reported [73]. Without the establish- ment of rapport, the recovery process will not begin. The building of trust could be incorporated as a dimension in Yung et al. BMC Psychol (2021) 9:104 Page 27 of 30 the CHIME framework for hikikomori. The term “yo-yo process” has been used to describe the process of recov- ery for hikikomori and the setbacks that they experience, such as a reversal in their progress [54] or their sudden refusal to take part in social activities or have face-to- face contact [6], which has been reported by social work- ers during reengagement work. This indicates that the recovery process is not linear or a matter of taking one step after another, but rather is a non-linear process. For hikikomori, non-linearity can occur at different levels and phases of recovery; however, this aspect is not appar- ent when using the CHIME framework and needs to be incorporated. An additional limitation of the framework is the inability to address aspects of spatiality in hikiko- mori care. Due to the self-secluding nature of hikikomori, the majority of care is currently provided in the home setting and starts where the client is [54, 70]. Home visits require sensitivity and awareness [6] from the healthcare team, since they are entering into each of the clients’ pri- vate sphere. Sensitivity to each client’s surroundings can provide clues about his or her hobbies or interests, which can be used as the base for initiating interaction and dis- cussion for clientele reengagement [6]. The reengagement process can also help show recognition to the clients in view of their need for privacy [6]. Spatiality is a compo- nent specific to hikikomori reengagement which without the clients can hardly be outreached at their homebound comfort zone in the first place; however, it is not a dimen- sion or domain in the CHIME framework. Consideration has to be given for accommodating the special needs of hikikomori. In addition, in the CHIME framework there is no differentiation of the level of importance of each dimension. With regard to hikikomori care, two com- ponents are of great importance: relationship dynamics and activity (type or level). These two components can be used to measure the level of connectedness and dis- tinguish the progress in the recovery achieved by individ- uals. Both categories lead to the ultimate goal, which is to reconnect with society. Consideration has to be given for incorporating the domains of relational dynamics and activity into the CHIME framework, or for adopt- ing them as major sub-domains under the domain of connectedness. The CHIME framework is applicable to focus on study- ing the psychosocial aspects of the hikikomori lifestyle and to identify areas in hikikomori research studies that have not been explored. A summary of the applicability of hikikomori research and of the areas that are yet to be studied is provided in Table 4. Studies on hikikomori in relation to the CHIME framework were found predomi- nantly in the domains of Connectedness (the major- ity, totaling 22 out of 44 studies), Identity, and Meaning in Life; however, literature is lacking on the domains of Hope and Optimism and Empowerment. One pos- sible direction of research in the future is to focus on exploring aspects of these domains to understand what could be of benefit to hikikomori. This may lead to rec- ommendations or interventions helpful for hikikomori care. Further work is needed to clarify details on the sociodemographic and ethnographic characteristics of hikikomori such as their gender distribution, gender differences, sexual orientation, culture, and religion; to understand the dynamics of the intimate relationships of hikikomori through qualitative research; to under- stand the meaning of recovery for a hikikomori and what enacted and felt stigmas would be present through tak- ing a qualitative approach; to understand what motivates hikikomori to work towards recovery or what they value through taking a qualitative or mixed-methods approach; to understand what the meaningful social roles or goals are for hikikomori and how they establish them through a qualitative enquiry that may lead to interventional designs; to understand the meaning of control over life and personal responsibilities for hikikomori through qualitative research; to explore more aspects of empow- erment to aid their recovery; to understand what meas- ures can to be taken to help hikikomori overcome their sense of failure or fear of failure rather than valuing suc- cess defined in terms of climbing up the social ladder, to help them towards recovery, to determine what measures need to be taken to improve this situation and whether they would help towards recovery; and to figure out what components of future interventions may be useful for recovery and welcomed by hikikomori. A further explo- ration of these areas could lead to greater understanding and improve hikikomori care. Limitations A known limitation to this study was by using the search formula as per Li and Wong’s [9] systematic review, there may be the risk of missing articles; which was mentioned in their study due to the formula omitting the search terms of “social isolation” and “non-engaged”. How- ever, the search terms used according to the domains of CHIME were able to locate a substantial amount of pub- lications exploring the hikikomori phenomenon prior to exclusion process. A second limitation to this study was from confining the search of relevant publications to the English language but excluding those in Japanese in par- ticular, where the phenomenon of hikikomori started to take place; thus risking the chance for missing hikikomori research of publications in languages other than English. A third limitation was due to the busy schedules and time conflicts between team members, and the logistics when screening of articles, it did not allow for a useful Kappa coefficient to be produced. Lastly, trial registration of this Yung et al. BMC Psychol (2021) 9:104 Page 28 of 30 review was not completed; as the review was in the stage of dissemination of findings when understanding that registration into the registry required completion before any synthesis of data. Conclusion The CHIME framework is applicable to the hikikomori population and can encompass most aspects of their life experiences; however, future modifications may be needed to include the three major domains of spatiality, relational dynamics, and activity; as well as the dimen- sions or sub-domains of trust building and non-linearity. Through the use of the framework, the hikikomori life- style is shown to be characterized by: disconnection from peers and society, limited relationships largely confined within the family, social anxiety, negative self-appraisal, a lack of identity, a mistrust of people, a high level of passive aggressiveness, and shyness. Hikikomori are not confined to a specific race, and higher levels of social withdrawal would tend to lead to a lower quality of life. Using the CHIME framework, many gaps in knowledge about hikikomori could be identified in the literature, such as those about gender distribution, behavioral differ- ences between males and females, sexual orientation, cul- ture, religion, and the dynamics of intimate relationships; the meaning of recovery for hikikomori, the impact of stigma on them, and how they can be motivated towards recovery or into taking up valued, meaningful social roles or goals; the meaning for hikikomori to have empower- ment and control over their life and personal responsibil- ities; and any other components of future interventions considered useful and welcomed by hikikomori. More understanding of these issues is needed to improve hikikomori care. Considerations could be made for future use of CHIME in hikikomori care, as individuals may be entrapped in the lifestyle for long periods of time [44] and recovery has been reported to take a minimum of two years [46, 73]; the framework could possibility target areas that would need attention or improvement during the time of entrapment in the lifestyle. However, the full applicability of the CHIME framework towards recovery from the hikikomori lifestyle would need further verifi- cation; further studies with recovered hikikomori could verify if the domains of CHIME were involved towards their recovery. Abbreviations CHIME: Connectedness, Hope and Optimism, Identity, Meaning in Life, Empowerment; PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analysis; JBI: Joanna Briggs Institute; MMAT: Mixed Method Appraisal Tool NEET: not being in education, training, or employment; SNI: Modified Berkman–Syme Social Network Index; ISEL: Interpersonal Support Evaluation List scores; UCLA: University of California Los Angeles; vs: Versus; WHOQOL- BREF: World Health Organization Quality of Life scale-brief; QOL: Quality of life; C-BED: Community Based Enterprise Component; LSNC: Lubben Social Net- work Scale; GAF: Global Assessment Functioning Scores; N/A: Not applicable. Acknowledgements Not applicable. Authors’ contributions JY and AM conceived the study and methodology. JY carried out the literature search, data analysis, collating of results, and wrote the manuscript. AM and VW provided supervision to JY and worked together in conceptualizing the results. All authors reviewed, edited and approved the final manuscript for submission. Funding This research did not receive any specific grants or funding to conduct this review. Availability of data and materials All data generated or analysed during this study are included in this published article. Declarations Ethical approval and consent for participation Not applicable. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1 A130, School of Nursing, The Hong Kong Polytechnic University, Hung Hom, Kowloon, HKSAR, China. 2 AAB1028, Department of Social Work, Hong Kong Baptist University, Kowloon Tong, Kowloon, HKSAR, China. 3 PQ426, School of Nursing, The Hong Kong Polytechnic University, Hung Hom, Kowloon, HKSAR, China. 4 GH507, School of Nursing, The Hong Kong Polytechnic Univer- sity, Hung Hom, Kowloon, HKSAR, China. Received: 3 February 2021 Accepted: 24 June 2021 References 1. 2. 3. Saito T. Shakaiteki hikikomori: owaranai shishunki (Social withdrawal: a neverending adolescence). Tokyo: PHP Shinsho; 1998. Tajan N. Japanese post-modern social renouncers: an exploratory study of the narratives of Hikikomori subjects. Subjectivity. 2015;8(3):283–304. Lee YS, Lee JY, Choi TY, Choi JT. Home visitation program for detecting, evaluating and treating socially withdrawn youth in Korea. Psychiatry Clin Neurosci. 2013;67(4):193–202. https:// doi. org/ 10. 1111/ pcn. 12043. 4. Chan H-y, Lo T-w. Quality of life of the hidden youth in Hong Kong. ARQOL. 2014;9(4):951–69. https:// doi. org/ 10. 1007/ s11482- 013- 9279-x. 5. Ministry of Health Labour and Welfare. 10-dai, 20-dai wo chuushin to shita ‘hikikomori’ wo meguru chiiki seishin hoken katsudou no gaidorain (Community mental health intervention guidelines aimed at socially withdrawn teenagers and young adults). Japan: Tokyo; 2003. 6. Wong V. Youth locked in time and space? Defining features of social with- drawal and practice implications. J Soc Work Pract. 2009;23(3):337–52. https:// doi. org/ 10. 1080/ 02650 53090 31026 92. 7. Kato TA, Kanba S, Teo AR. Defining pathological social withdrawal: pro- posed diagnostic criteria for hikikomori. World Psychiatry. 2020;19(1):116– 7. https:// doi. org/ 10. 1002/ wps. 20705. 8. Hamasaki Y, Pionnié-Dax N, Dorard G, Tajan N, Hikida T. Identifying social withdrawal (hikikomori) factors in adolescents: understanding the hikiko- mori spectrum. Child Psychiatry Hum Dev. 2020. https:// doi. org/ 10. 1007/ s10578- 020- 01064-8. Yung et al. BMC Psychol (2021) 9:104 Page 29 of 30 9. Li TMH, Wong PWC. Youth social withdrawal behavior (hikikomori): a systematic review of qualitative and quantitative studies. Aust N Z J Psy- chiatry. 2015;49(7):595–609. https:// doi. org/ 10. 1177/ 00048 67415 581179. 10. Kato TA, Kanba S, Teo AR. Hikikomori: multidimensional understanding, assessment, and future international perspectives. Psychiatry Clin Neuro- sci. 2019;73(8):427–40. https:// doi. org/ 10. 1111/ pcn. 12895. Synthesis. JBI, 2020. https:// synth esism anual. jbi. global. Accessed 25 Janu- ary 2021. 31. Munn Z, Moola S, Lisy K, Riitano D, Tufanaru C. Methodological guid- ance for systematic reviews of observational epidemiological studies reporting prevalence and incidence data. Int J Evid Based Healthc. 2015;13(3):147–53. 11. Tajan N. Social withdrawal and psychiatry: a comprehensive review of 32. Munn Z, Barker T, Moola S, Tufanaru C, Stern C, McArthur A, Stephenson Hikikomori. Neuropsychiatr Enfance Adolesc. 2015;63(5):324–31. https:// doi. org/ 10. 1016/j. neure nf. 2015. 03. 008. M, Aromataris E. Methodological quality of case series studies. JBI Evid Synth. 2020. https:// doi. org/ 10. 11124/ JBISR IR-D- 19- 00099. 12. Teo AR. A new form of social withdrawal in Japan: a review of hikikomori. Int J Soc Psychiatry. 2010;56(2):178–85. https:// doi. org/ 10. 1177/ 00207 64008 100629. 33. Lockwood C, Munn Z, Porritt K. Qualitative research synthesis: methodo- logical guidance for systematic reviewers utilizing meta-aggregation. Int J Evid Based Healthc. 2015;13(3):179–87. 13. Stip E, Thibault A, Beauchamp-Chatel A, Kisely S. Internet addiction, hikikomori syndrome, and the prodromal phase of psychosis. Front Psychiatry. 2016;7:6. https:// doi. org/ 10. 3389/ fpsyt. 2016. 00006. 14. Wong P, Yu RWM, Li TMH, Lai SLH, Ng HYH, Fan WTW. Efficacy of a multicomponent intervention with animal-assisted therapy for socially withdrawn youth in Hong Kong. Soc Anim. 2017. https:// doi. org/ 10. 1163/ 15685 306- 12341 462. 15. Nishida M, Kikuchi S, Fukuda K, Kato S. Jogging therapy for hikikomori social withdrawal and increased cerebral hemodynamics: a case report. Clin Pract Epidemiol Ment Health. 2016;12:38–42. https:// doi. org/ 10. 2174/ 17450 17901 61201 0038. 16. Kato TA, Teo AR, Tateno M, Watabe M, Kubo H, Kanba S. Can Pokémon GO rescue shut-ins (hikikomori) from their isolated world? Psychiatry Clin Neurosci. 2017;71(1):75–6. https:// doi. org/ 10. 1111/ pcn. 12481. 17. Antai-Otong D. Psychosocial recovery and rehabilitation. Nurs Clin N Am. 2016;51(2):287–97. https:// doi. org/ 10. 1016/j. cnur. 2016. 01. 011. 18. Yokoyama T, Kawasaki A, Mitsumiya T, Xu C. Impact of online workshop for youth empowerment: applying C-BED to Hikikomori support in Japan. Educ Q Rev. 2019;2(4):857–65. https:// doi. org/ 10. 31014/ aior. 1993. 02. 04. 112. 19. Saito T, Angles J. Hikikomori: adolescence without end. Minneapolis: University of Minnesota Press; 2013. 20. American Psychology Association. Recovery principles, https:// www. apa. 34. Hong QN, Pluye P, Fàbregues S, Bartlett G, Boardman F, Cargo M, et al. Mixed Methods Appraisal Tool (MMAT), version 2018. Registration of Copyright (#1148552), Canadian Intellectual Property Office, Industry Canada. 35. Yuen JWM, Wong VCW, Tam WWS, So KW, Chien WT. A one-year prospec- tive follow-up study on the health profile of hikikomori living in Hong Kong. Int J Environ Res Public Health. 2019. https:// doi. org/ 10. 3390/ ijerp h1604 0546. 36. Teo AR, Fetters MD, Stufflebam K, Tateno M, Balhara Y, Choi TY, et al. Iden- tification of the hikikomori syndrome of social withdrawal: psychosocial features and treatment preferences of four countries. Int J Soc Psychiatry. 2015;61(1):64–72. https:// doi. org/ 10. 1177/ 00207 64014 535758. 37. Krieg A, Dickie JR. Attachment and hikikomori: a psychosocial develop- mental model. Int J Soc Psychiatry. 2011;59(1):61–72. https:// doi. org/ 10. 1177/ 00207 64011 423182. 38. Chauliac N, Couillet A, Faivre S, Brochard N, Terra J-l. Characteristics of socially withdrawn youth in France: a retrospective study. Int J Soc Psy- chiatry. 2017;63(4):339–44. https:// doi. org/ 10. 1177/ 00207 64017 704474. 39. Malagón-Amor Á, Martín-López LM, Córcoles D, González A, Bellsolà M, Teo AR, et al. A 12-month study of the hikikomori syndrome of social withdrawal: clinical characterization and different subtypes proposal. Psy- chiatry Res. 2018;270:1039–46. https:// doi. org/ 10. 1016/j. psych res. 2018. 03. 060. org/ monit or/ 2012/ 01/ recov ery- princ iples; 2012 [accessed 1 April 2020]. 40. Kondo N, Sakai M, Kuroda Y, Kiyota Y, Kitabata Y, Kurosawa M. General 21. PSR/RPS. Psychosocial Rehabilitation/Réadaptation Psychosociale Canada: Core principles and values, https:// psrrp scana da. ca/ files/ pdf/ core% 20car ds. pdf; 2020. Accessed 1 April 2020. 22. Smith EJ. The strength-based counseling model. The Couns Psychol. 2006;34(1):13–79. https:// doi. org/ 10. 1177/ 00110 00005 277018. 23. Spence GB, Oades LG. Coaching with self-determination theory in mind: using theory to advance evidence-based coaching practice. IJEBCM. 2011;9(2):37–55. 24. Leamy M, Bird V, Le Boutillier C, Williams J, Slade M. Conceptual frame- work for personal recovery in mental health: systematic review and nar- rative synthesis. Br J Psychiatry. 2011;199(6):445. https:// doi. org/ 10. 1192/ bjp. bp. 110. 083733. 25. Brijnath B. Applying the CHIME recovery framework in two culturally diverse Australian communities: qualitative results. Int J Soc Psychiatry. 2015;61(7):660–7. https:// doi. org/ 10. 1177/ 00207 64015 573084. 26. Stickley T, Wright N, Slade M. The art of recovery: outcomes from par- ticipatory arts activities for people using mental health services. J Ment Health. 2018;27(4):367–73. https:// doi. org/ 10. 1080/ 09638 237. 2018. 14376 09. 27. Bird V, Leamy M, Tew J, Le Boutillier C, Williams J, Slade M. Fit for purpose? Validation of a conceptual framework for personal recovery with current mental health consumers. Aust N Z J Psychiatry. 2014;48(7):644–53. https:// doi. org/ 10. 1177/ 00048 67413 520046. 28. Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med. 2009;6(7):e1000097. https:// doi. org/ 10. 1371/ journ al. pmed1 000097. condition of hikikomori (prolonged social withdrawal) in Japan: psychi- atric diagnosis and outcome in mental health welfare centres. Int J Soc Psychiatry. 2011;59(1):79–86. https:// doi. org/ 10. 1177/ 00207 64011 423611. 41. Yuen JWM, Yan YKY, Wong VCW, Tam WWS, So KW, Chien WT. A Physical Health Profile of Youths Living with a “Hikikomori” Lifestyle. Int J Environ Res Public Health. 2018. https:// doi. org/ 10. 3390/ ijerp h1502 0315. 42. Wu AFW, Catmur C, Wong PW, Lau JY. The presence, characteristics and correlates of pathological social withdrawal in Taiwan: an online survey. Int J Soc Psychiatry. 2020;66(1):84–92. https:// doi. org/ 10. 1177/ 00207 64019 882724. 43. Yong R, Nomura K. Hikikomori is most associated with interpersonal relationships, followed by suicide risks: a secondary analysis of a national cross-sectional study. Front Psychiatry. 2019;10:247. https:// doi. org/ 10. 3389/ fpsyt. 2019. 00247. 44. Kato TA, Kanba S, Teo AR. A 39-year-old “adultolescent”: understanding social withdrawal in Japan. Am J Psychiatry. 2016;173(2):112–4. https:// doi. org/ 10. 1176/ appi. ajp. 2015. 15081 034. 45. Ovejero S, Caro-Cañizares I, de León-Martínez V, Baca-Garcia E. Prolonged social withdrawal disorder: a hikikomori case in Spain. Int J Soc Psychiatry. 2014;60(6):562–5. https:// doi. org/ 10. 1177/ 00207 64013 504560. 46. Ranieri F. When social withdrawal in adolescence becomes extreme: the “hikikomori” phenomenon in Italy. Psychiatr i Psychol Klin. 2015;15(3):148– 51. https:// doi. org/ 10. 15557/ PiPK. 2015. 0023. 47. Sakamoto N, Martin RG, Kumano H, Kuboki T, Al-Adawi S. Hikikomori, is it a culture-reactive or culture-bound syndrome? Nidotherapy and a clini- cal vignette from Oman. Int J Psychiatry Med. 2005;35(2):191–8. https:// doi. org/ 10. 2190/ 7WEQ- 216D- TVNH- PQJ1. 29. Tufanaru C, Munn Z, Aromataris E, Campbell J, Hopp L. Chapter 3: 48. Silić A, Vukojević J, Čulo I, Falak H. Hikikomori silent epidemic: a case Systematic reviews of effectiveness. In: Aromataris E, Munn Z (Editors). JBI Manual for Evidence Synthesis. JBI, 2020. https:// synth esism anual. jbi. global. Accessed 25 January 2021. 30. Moola S, Munn Z, Tufanaru C, Aromataris E, Sears K, Sfetcu R, Currie M, Qureshi R, Mattis P, Lisy K, Mu P-F. Chapter 7: Systematic reviews of etiol- ogy and risk. In: Aromataris E, Munn Z (Editors). JBI Manual for Evidence study. Res Psychother. 2019;22(2):317–22. https:// doi. org/ 10. 4081/ ripppo. 2019. 377. 49. Teo AR. Social isolation associated with depression: a case report of hikikomori. Int J Soc Psychiatry. 2012;59(4):339–41. https:// doi. org/ 10. 1177/ 00207 64012 437128. Yung et al. BMC Psychol (2021) 9:104 Page 30 of 30 50. Ranieri F. Psychoanalytic psychotherapy for hikikomori young adults and adolescents. Br J Psychother. 2018;34(4):623–42. https:// doi. org/ 10. 1111/ bjp. 12398. 51. Suwa M, Suzuki K. The phenomenon of “hikikomori” (social with- drawal and the socio-cultural situation in Japan today. J Psychopathol. 2013;19:191–8. 52. Kaneko S. Japan’s “socially withdrawn youths” and time constraints in Japanese Society: management and conceptualization of time in a support group for hikikomori. Time Soc. 2006;15(2–3):233–49. https:// doi. org/ 10. 1177/ 09614 63X06 067034. 53. Li TMH, Liu L, Wong PWC. Withdrawal experience and possible way- outs from withdrawal behavior in young people. Qual Soc Work. 2018;17(4):537–55. https:// doi. org/ 10. 1177/ 14733 25016 688369. 54. Wong V, Ying W. Social withdrawal of young people in Hong Kong: a social exclusion perspective. Hong Kong J Soc Work. 2006;40(01n02):61– 91. https:// doi. org/ 10. 1142/ S0219 24620 60000 64. 55. Chan GH. Intimacy, friendship, and forms of online communication among youth in Hong Kong. Comput Hum Behav. 2020;111(2020): 106407. https:// doi. org/ 10. 1016/j. chb. 2020. 106407. 56. Ogino T. Managing categorization and social withdrawal in Japan: rehabilitation process in a private support group for hikikomorians. Int J Jpn Sociol. 2004;13(1):120–33. https:// doi. org/ 10. 1111/j. 1475- 6781. 2004. 00057.x. 57. Yong RKF, Kaneko Y. Hikikomori, a phenomenon of social withdrawal and isolation in young adults marked by an anomic response to coping diffi- culties: a qualitative study exploring individual experiences from first- and second-person perspectives. Open J Prev Med. 2016;6(1):1–20. https:// doi. org/ 10. 4236/ ojpm. 2016. 61001. 58. Uchida Y, Norasakkunkit V. The NEET and Hikikomori spectrum: assessing the risks and consequences of becoming culturally marginalized. Front Psychol. 2015;6:1117–1117. https:// doi. org/ 10. 3389/ fpsyg. 2015. 01117. 59. Malagón-Amor Á, Córcoles-Martinez D, Martín-López LM, Pérez- Solá V. Hikikomori in Spain: a descriptive study. Int J Soc Psychiatry. 2015;61(5):475–83. https:// doi. org/ 10. 1177/ 00207 64014 553003. 60. Chan GH-Y, Lo TW. Family relationships and the self-esteem of hidden youth: a power dynamics perspective. J Fam Issues. 2016;37(9):1244–66. https:// doi. org/ 10. 1177/ 01925 13X14 537479. 61. Frankova I. Similar but different: psychological and psychopathological features of primary and secondary hikikomori. Front Psychiatry. 2019. https:// doi. org/ 10. 3389/ fpsyt. 2019. 00558. Int J Soc Psychiatry. 2011;59(1):73–8. https:// doi. org/ 10. 1177/ 00207 64011 423184. 65. Katsuki R, Inoue A, Indias S, Kurahara K, Kuwano N, Funatsu F, et al. Clarifying deeper psychological characteristics of hikikomori using the rorschach comprehensive system: a pilot case-control study. Front Psy- chiatry. 2019;10:412. https:// doi. org/ 10. 3389/ fpsyt. 2019. 00412. 66. Zhang S, Wang H, Chen C, Zhou J, Wang X. Efficacy of Williams LifeSkills Training in improving psychological health of Chinese male juve- nile violent offenders: a randomized controlled study. Neurosci Bull. 2015;31(01):53–60. https:// doi. org/ 10. 1007/ s12264- 014- 1492-6. 67. Dahlen ER, Martin RC. The experience, expression, and control of anger in perceived social support. Pers Individ Differ. 2005;39(2):391–401. https:// doi. org/ 10. 1016/j. paid. 2005. 01. 019. 68. Rosenberg M. Conceiving the self. New York: Basic Books; 1979. 69. Jhangiani R. & Tarry H. Chapter 3: The feeling self: self-esteem. Principles of Social Psychology-1st international ed. https:// opent extbc. ca/ socia lpsyc hology/ chapt er/ the- feeli ng- self- self- esteem/; 2014. [accessed 24 March 2021]. 70. Wong V. Social withdrawal as invisible youth disengagement. Int J Sociol Soc Policy. 2012;32(7/8):415–30. https:// doi. org/ 10. 1108/ 01443 33121 12490 57. 71. Rubinstein E. Emplotting hikikomori: Japanese parents’ narratives of social withdrawal. Culture Med Psychiatry Int J Cross Cult Health Res. 2016;40(4):641–63. https:// doi. org/ 10. 1007/ s11013- 016- 9495-6. 72. Roza TH, Spritzer DT, Lovato LM, Passos IC. Multimodal treatment for a Brazilian case of hikikomori. Braz J Psychiatry. 2020;42(4):454–6. https:// doi. org/ 10. 1590/ 1516- 4446- 2020- 0804. 73. Hattori Y. Social withdrawal in Japanese Youth: a case study of thirty-five hikikomori clients. J Trauma Pract. 2006;4(3–4):181–201. https:// doi. org/ 10. 1300/ J189v 04n03_ 01. 74. Law YW, Kwok CL, Chan PY, Chan M, Yip P. Online social work engage- ment and empowerment for young internet users: a quasi-experiment. J Affect Disord. 2019;250:99–107. https:// doi. org/ 10. 1016/j. jad. 2019. 02. 061. 75. Chan GHY. Application and effectiveness of play therapy using an online- game intervention for hidden youth. Br J Soc Work. 2019;50(7):2116–34. https:// doi. org/ 10. 1093/ bjsw/ bcz129. 76. Chan GH. A comparative analysis of online, offline, and integrated counselling among hidden youth in Hong Kong. Chi Child Youth Serv Rev. 2020;114(2020): 105042. https:// doi. org/ 10. 1016/j. child youth. 2020. 105042. 62. Koyama A, Miyake Y, Kawakami N, Tsuchiya M, Tachimori H, Takeshima T. Lifetime prevalence, psychiatric comorbidity and demographic correlates of “hikikomori” in a community population in Japan. Psychiatry Res. 2010;176(1):69–74. https:// doi. org/ 10. 1016/j. psych res. 2008. 10. 019. 63. Umeda M, Kawakami N. Association of childhood family environments 77. Matsuguma S, Kawashima M, Tsubota K. Applying strengths from the vir- tual to the real world: Strength intervention for Hikikomori youth: a case study. International Positive Psychology Association, https:// www. ippan etwork. org/ 2018/ 08/ 15/ apply ing- stren gths- from- the- virtu al- to- the- real- world/; 2018 [accessed 15 April 2020]. with the risk of social withdrawal (’hikikomori’) in the community popula- tion in Japan. Psychiatry Clin Neurosci. 2012;66(2):121–9. https:// doi. org/ 10. 1111/j. 1440- 1819. 2011. 02292.x. 64. Nagata T, Yamada H, Teo AR, Yoshimura C, Nakajima T, van Vilet I. Comor- bid social withdrawal (hikikomori) in outpatients with social anxiety disorder: Clinical characteristics and treatment response in a case series. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1371_journal.pcbi.1011115	RESEARCH ARTICLE Modeling the impact of xenointoxication in dogs to halt Trypanosoma cruzi transmission Jennifer L. Rokhsar1,2,3☯, Brinkley RaynorID Michael Z. LevyID 4,6, Ricardo Castillo-NeyraID 4,6* 4☯, Justin Sheen4,5, Neal D. Goldstein2, 1 Faculty of Health and Medical Sciences, University of Surrey, Guildford, United Kingdom, 2 Department of Epidemiology and Biostatistics, Dornsife School of Public Health, Drexel University, Philadelphia, Pennsylvania, United States of America, 3 ORISE Fellow, Emerging Leaders in Data Science and Technologies Program Fellowship, National Institute of Allergy and Infectious Diseases (NIAID), NIH, United States of America, 4 Department of Biostatistics, Epidemiology, and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, United States of America, 5 Department of Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey, United States of America, 6 One Health Unit, School of Public Health and Administration, Universidad Peruana Cayetano Heredia, Lima, Peru ☯ These authors contributed equally to this work. * cricardo@upenn.edu Abstract Background Chagas disease, a vector-borne parasitic disease caused by Trypanosoma cruzi, affects millions in the Americas. Dogs are important reservoirs of the parasite. Under laboratory conditions, canine treatment with the systemic insecticide fluralaner demonstrated efficacy in killing Triatoma infestans and T. brasiliensis, T. cruzi vectors, when they feed on dogs. This form of pest control is called xenointoxication. However, T. cruzi can also be transmit- ted orally when mammals ingest infected bugs, so there is potential for dogs to become infected upon consuming infected bugs killed by the treatment. Xenointoxication thereby has two contrasting effects on dogs: decreasing the number of insects feeding on the dogs but increasing opportunities for exposure to T. cruzi via oral transmission to dogs ingesting infected insects. Objective Examine the potential for increased infection rates of T. cruzi in dogs following xenointoxication. Design/Methods We built a deterministic mathematical model, based on the Ross-MacDonald malaria model, to investigate the net effect of fluralaner treatment on the prevalence of T. cruzi infec- tion in dogs in different epidemiologic scenarios. We drew upon published data on the change in percentage of bugs killed that fed on treated dogs over days post treatment. Parameters were adjusted to mimic three scenarios of T. cruzi transmission: high and low disease prevalence and domestic vectors, and low disease prevalence and sylvatic vectors. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Rokhsar JL, Raynor B, Sheen J, Goldstein ND, Levy MZ, Castillo-Neyra R (2023) Modeling the impact of xenointoxication in dogs to halt Trypanosoma cruzi transmission. PLoS Comput Biol 19(5): e1011115. https://doi.org/10.1371/ journal.pcbi.1011115 Editor: Eric H. Y. Lau, The University of Hong Kong, CHINA Received: June 6, 2022 Accepted: April 19, 2023 Published: May 8, 2023 Peer Review History: PLOS recognizes the benefits of transparency in the peer review process; therefore, we enable the publication of all of the content of peer review and author responses alongside final, published articles. The editorial history of this article is available here: https://doi.org/10.1371/journal.pcbi.1011115 Copyright: This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication. Data Availability Statement: All relevant data are within the manuscript and its Supporting Information files. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 1 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission Funding: BHR was supported through an NIH grant 5-T32-AI-070077-14. RCN was supported by NIH-NIAID grant 1K01AI139284, https://www.nih. gov/. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. Results In regions with high endemic disease prevalence in dogs and domestic vectors, prevalence of infected dogs initially increases but subsequently declines before eventually rising back to the initial equilibrium following one fluralaner treatment. In regions of low prevalence and domestic or sylvatic vectors, however, treatment seems to be detrimental. In these regions our models suggest a potential for a rise in dog prevalence, due to oral transmission from dead infected bugs. Conclusion Xenointoxication could be a beneficial and novel One Health intervention in regions with high prevalence of T. cruzi and domestic vectors. In regions with low prevalence and domestic or sylvatic vectors, there is potential harm. Field trials should be carefully designed to closely follow treated dogs and include early stopping rules if incidence among treated dogs exceeds that of controls. Author summary Chagas disease, caused by the parasite Trypanosoma cruzi, is transmitted via triatomine insect vectors. In Latin America, dogs are a common feeding source for triatomine vectors and subsequently an important reservoir of T. cruzi. One proposed intervention to reduce T. cruzi transmission is xenointoxication: treating dogs with oral insecticide to kill triato- mine vectors in order to decrease overall T. cruzi transmission. Fluralaner, commonly administered to prevent ectoparasites such as fleas and ticks, is effective under laboratory conditions against the triatomine vectors. One concern with fluralaner treatment is that rapid death of the insect vectors may make the insects more available to oral ingestion by dogs; a more effective transmission pathway than stercorarian, the usual route for T. cruzi transmission. Using a mathematical model, we explored 3 different epidemiologic scenar- ios: high prevalence endemic disease within a domestic T. cruzi cycle, low prevalence endemic disease within a domestic T. cruzi cycle, and low prevalence endemic disease within a semi-sylvatic T. cruzi cycle. We found a range of beneficial to detrimental effects of fluralaner xenointoxication depending on the epidemiologic scenario. Our results sug- gest that careful field trials should be designed and carried out before wide scale imple- mentation of fluralaner xenointoxication to reduce T. cruzi transmission. Introduction Chagas disease, a vector-borne neglected tropical disease, affects millions of people in Latin America [1,2]. The disease, caused by the protozoan Trypanosoma cruzi, has no specific treat- ment, and there are no vaccines available for a large-scale public health intervention; therefore, strategies to control and eliminate Chagas disease have targeted the insect vectors, triatomine bugs [3]. Importantly, transmission of T. cruzi occurs in two main cycles [4]. The sylvatic cycle involves small wild mammals acting as animal reservoirs and sylvatic bugs bringing the para- site into households, infecting humans and domiciliary mammals. The domestic cycle involves the colonization of household structures by triatomine bugs and the transmission of the para- site to and from humans and domiciliary mammals. However, there are regions where these PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 2 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission two cycles overlap, and some authors recognize the existence of a peri-urban cycle [4]. The presence of multiple wild mammalian reservoirs makes elimination virtually impossible in the sylvatic cycle [5], but within the domestic cycle dogs are much more accessible reservoirs than wild animals to target with One Health interventions for elimination of T. cruzi transmission and reduction of Chagas disease in humans [6]. Triatomine bugs acquire infection through blood meals from mammals that contain infec- tive forms of T. cruzi in their bloodstream [7]. By contrast hosts can become infected with T. cruzi through several avenues, including congenital and oral, but the most common and important is vector-borne transmission [8]. Oral transmission through predation of infected vectors is thought to be the most frequent mechanism of infection among hosts in the sylvatic cycle [9–14] and many people have been infected orally in focalized outbreaks in Latin Amer- ica [9,15]. The probability of transmission due to oral vector ingestion is estimated to be about 1000 times greater than vectorial transmission [16,17] and T. cruzi parasites in feces outside of the bug are viable (infectious) for up to 48 hours (e.g., in fruit juices) [17]. Dogs, important res- ervoirs for Chagas disease in the domestic cycle [18–21], can occasionally get infected through the oral route [22]. Dogs are key reservoirs in the urban and sylvatic cycles of T. cruzi because they are very common within households in Latin America [6,23], they have longer lifespans compared to other important animal reservoirs such as guinea pigs [24,25], they act as bridges between both cycles [25], and dogs tend to have high infection rates, are very infectious, and have high rates of contact with both vectors and humans [20]. In Latin America, reports on canine sero- prevalence in areas where natural infection occurs concentrate between 8–28%, with extremes of 1.5% and 83.3% [26–28]. Infected dogs are also 100 times more likely to infect susceptible triatomes than infected adult humans and 12 times more infectious than infected children [29]. In Brazil, Triatoma brasiliensis, one of the primary vectors of T. cruzi, overlaps geographi- cally with areas where dogs are important reservoirs of the disease [30,31]. In addition, Tria- toma infestans, one of the primary insect vectors of T. cruzi in South America, shows consistent preference for dogs over other domestic animals [29]. The strong preference triato- mine vectors have for dogs can be exploited via xenointoxication–a targeted vector control strategy where insecticides are administered to peri-domestic and domestic animals (e.g., dogs) to suppress insect infestations. For instance, by targeting dogs with topical insecticides (or insecticide impregnated collars), dogs are effectively turned into baited lethal traps [6]. Interventions on the dog population to eliminate T. cruzi transmission have been evaluated for decades. Mathematical models of Chagas disease have shown that removal of infected dogs from a household containing infected people could stop disease transmission (excluding rein- troduction) [32], but culling the dog population would be, at the very least, socially unaccept- able and hypothesized to have inconclusive results [6]. Recent experimentation treating dogs with oral or topically applied insecticides showed promising efficacy at killing triatomines [33–35]; in particular, fluralaner, a relatively new isoxazoline oral insecticide commonly used to prevent tick and flea infestations, proved especially effective in killing bugs when they fed on dogs under laboratory conditions [33] and is being considered for Chagas control programs [36]. As the unit cost for indoor residual insecticide treatment in a rural house is quite high [37–40] and can be met with low levels of community participation [41–44], treatment of canine reservoirs with insecticide could prove a useful alternative or complementary strategy to reduce T. cruzi infection in people. Additionally, due to the scarcity of insecticide for public health usage [45], treatment of canines with a safe, long-lasting, effective insecticide such as fluralaner potentially could prove a valuable tool in the face of pyrethroid shortage [6,33]. However, given that T. cruzi can be easily transmitted orally through the ingestion of triato- mines [9,13,15,46–49], there is potential for a counterproductive effect: dogs could consume PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 3 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission the infected bugs killed by the treatment [27,35,50,51], increasing infection rates in the dog population. Xenointoxication as an intervention for Chagas disease could have unexpected conse- quences. The use of fluralaner could potentially reduce T. cruzi transmission by reducing the number of infectious bugs; however, it is also possible that the use of fluralaner could increase T. cruzi transmission by making infectious bugs killed by treatment more orally available to dogs. In this study, we developed a deterministic model of T. cruzi transmission dynamics that accounts for both vector-borne transmission and transmission via ingestion of T. cruzi- infected triatomines in dog populations. We used the model to investigate the effects the inter- vention will have on the prevalence of infections among insects and dogs under a variety of epidemiologic scenarios. Results Pretreatment model In regions affected by the domestic cycle and high prevalence of disease, prior to the adminis- tration of fluralaner treatment, equilibrium prevalence for dogs was 53.68% and for bugs it was 54.48% at approximately 10,000 days (27.4 years) (S1 Fig). In regions affected by the syl- vatic cycle and low prevalence of disease, prior to the administration of treatment, equilibrium prevalence for dogs was 23.64% and for bugs it was 38.81% at approximately 20,000 days (54.8 years) (S1 Fig). As the ratio of bugs to dogs in the population goes from 5–100, population dynamics switch from one where the proportion of infected bugs exceeds the proportion of infected dogs to the reverse. Pretreatment, the parameter with the largest impact on transmis- sion dynamics is dogs’ lifespan; in populations where dogs live for � 3 years, there are higher rates of overall infection for both bug and dog populations. These parameters and the potential xenointoxication interventions (e.g., number of treatments) can be modified in our interactive visualization application found at https://jrokh.shinyapps.io/NewExternalBugs/. Treatment model: Domestic Vectors We explored several different aspects of treatment, including the frequency of treatment and the length between treatments. A single treatment of fluralaner after population equilibrium resulted in a sharp decline of the proportion of infected bugs and a simultaneous increase in the proportion of infected dogs immediately after treatment (Fig 1). The rise in the proportion of infected dogs is followed by a gradual decline and a rise back to equilibrium levels. The sharp decline in the proportion of infected bugs also rises back to equilibrium levels. The per- centage of bugs consumed by dogs will be a function of both individual dog behavior and accessibility, i.e., bugs die in a location that is accessible to the dog; therefore, we varied the percentage of dead bugs consumed by dogs. In this simulation, we assumed that dogs con- sumed 80% of bugs killed with fluralaner treatment; in simulations with this parameter set to 20% and 50%, trends remained the same (S2 Fig). As could be expected, if the dogs consumed a greater number of the bugs, the initial rise in the proportion of infected dogs is greater, fol- lowed by a shallower decline in the days post-treatment (DPT). We examined the effects of administering canine fluralaner treatment once a year for 4–6 years (Fig 2A and 2B). Similar to the effect of single treatment with fluralaner, immediately fol- lowing administration, the proportion of infected dogs rises followed by a gradual decline. Also, at each successive treatment, there is a corresponding rise in the proportion of infected dogs; however, these peaks remain less than pretreatment equilibrium prevalence. Likewise, each treatment corresponds to a sharp decline in the proportion of infected bugs; as the treat- ment effect wears off, the proportion of infected bugs rises more rapidly than the infected PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 4 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission Fig 1. Single fluralaner treatment in a high prevalence region. Proportions of dogs and bugs infected with T. cruzi after single administration of fluralaner treatment at equilibrium (27.4 years) in a region of high prevalence of endemic disease and domestic vectors was simulated. https://doi.org/10.1371/journal.pcbi.1011115.g001 dogs, but infection levels still remain less than equilibrium prevalence. Treating every year pre- vents the infection prevalence in both dogs and bugs to reach prior equilibrium levels; the effect of successive yearly treatment allows for a “stair step” effect, where each peak in dog infection prevalence at treatment administration is smaller than the peak prior. We also explored setting the triatomine birthrate to zero, allowing the triatomine population to crash after xenointoxication treatment. As expected, we found that with no vectors to transmit T. cruzi, the proportion of infected dogs declines within years. Manufacturer’s instructions call for oral fluralaner to be given to dogs once every 12 weeks (approximately 90 days) [52]. When fluralaner is given according to this frequency (Fig 2C and 2D), we observe a similar “stair step” effect; there is an initial spike in dog infections, but in subsequent treatments these peaks are smoothed out; even after treatment is stopped, the proportion of infected dogs continues to trend downwards for a period of time before the infection levels begin to climb back towards pretreatment equilibrium levels. Giving treat- ments at this frequency also suppressed the infected bug proportion from rising between treat- ments. Levels of infection in the bug population remain low for a period of time following the last treatment before returning back to pretreatment levels. In these simulations, we assumed that dogs consumed 80% of bugs killed with fluralaner treatment; in simulations with this parameter set to 20% and 50%, trends remained the same (S3 Fig). PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 5 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission Fig 2. Multiple fluralaner treatments in a high prevalence region. Treatment scenarios were simulated for equilibrium populations of bugs and dogs in a region of high prevalence of endemic disease and domestic vectors. Annual administration of fluralaner for both 4 years (A) and 6 years (B) was simulated, as well as administration every 90 days (veterinary recommendation) for one year (C) and for two years (D). https://doi.org/10.1371/journal.pcbi.1011115.g002 We examined the effects of treatment on areas with domestic vectors and a low prevalence of disease (Fig 3). In regions with a low prevalence of disease (m = 15) and dogs average life- span = 3 years, fluralaner treatment is marked by the initial increase in prevalence of infected dogs (at time of treatment), but -unlike regions with a high prevalence of disease- the infection peak does not gradually decline; rather, it forms an elevated plateau followed by a gradual decline back to equilibrium infection levels. In regions of low disease prevalence (m = 7) and dogs with longer lifespans (average lifespan = 6 years), the initial spike in dog infection preva- lence continues to rise for a period of time; with each successive treatment, the proportion of infected dogs rises higher than the peak prior (Fig 3B). When percentages of dead bugs con- sumed by dogs is lower (at 20% and 50% instead of 80%), the level of infected dogs decline with fluralaner treatment when dogs have a 3-year life span (S4A–S4C Fig). When dogs have a PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 6 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission Fig 3. Fluralaner treatment schemes in low prevalence regions. Simulations were conducted to explore the effect of fluralaner treatment of regions of low prevalence of endemic disease and domestic vectors in equilibrium; we explored a range of dog average lifespan from 3 years (A-C) to 6 years (D-F). Treatment scenarios include one time treatment (A, D), annual treatment for 4 years (B, E), and treatment every 90 days for 1 year (C, F). https://doi.org/10.1371/journal.pcbi.1011115.g003 6-year lifespan, infected dog numbers trend up after fluralaner treatment assuming a percent- age of 50% as well as 80% but trend down with 20% (S4D–S4F Fig). Treatment model: Semi-sylvatic vectors We simulated regions with lower disease prevalence and semi-sylvatic vectors for both the baseline average dog lifespan (Fig 4A–4C) as well as the 6-year life span (Fig 4D–4F). Similar to the prior models with low disease prevalence, administration of fluralaner leads to a rise in dog infection prevalence, which increases with successive treatments. The effect is particularly apparent where dogs have longer lifespans (Fig 4D–4F); although bug infection experiences a sharp decline upon treatment administration, with repeated treatments, the bug infection prevalence rebounds to levels above the pretreatment equilibrium values. The semi-sylvatic low-prevalence model is sensitive to the proportion of bugs eaten with trends being similar to those of the non-semi-sylvatic cycle low-prevalence model (S5 Fig). Discussion Our models indicate that in regions with high disease prevalence and domestic vectors treat- ment of dogs with fluralaner could provide an effective complementary community-level treat- ment of T. infestans infestations, similar to what lab experiments suggest [34]. In regions with high prevalence of household infestations, even if dogs were to consume large numbers of T. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 7 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission Fig 4. Fluralaner treatment schemes in low prevalence regions with semi-sylvatic transmission. Simulations were conducted to explore the effect of fluralaner treatment of regions of low prevalence of endemic disease and domestic vectors as well as semi-sylvatic vectors in equilibrium; we explored a range of dog average lifespan from 3 years (A-C) to 6 years (D-F). Treatment scenarios include one time treatment (A, D), annual treatment for 4 years (B, E), and treatment every 90 days for 1 year (C, F). https://doi.org/10.1371/journal.pcbi.1011115.g004 cruzi-infected bugs, our models suggest that levels of canine infection would drop below pre- treatment levels following the initial rise due to oral consumption. Even more promising, treat- ment appears to be beneficial if given at yearly intervals, which would be more cost-effective, and likely have higher community participation rates, than treating every 3 months. As a point of comparison, the price of fluralaner for a medium size dog in Peru is 22.20 USD [53] and the minimum wage in the same country is 275 USD/month [54]. Our model ignored seasonality; it could be possible to time yearly treatments leading to a slower resurgence of the vector popu- lation the following year, similar to timing of spraying campaigns [55,56]. The findings of these simulations are supported by a placebo controlled before-and-after efficacy trial of flura- laner administration to dogs in Chaco Province, Argentina (a region with high prevalence of domestic vectors/household infestation); the authors demonstrated that site infestation and domicile bug abundance plummeted over the months posttreatment [57]. In contrast, our findings suggest that in regions with low disease prevalence and domestic or sylvatic bug populations, especially in regions where dogs have longer lifespans, careful attention needs to be given to the potential of unintended consequences of xenointoxication on T. cruzi transmission. In these regions, when dogs are able to consume a large percentage of the bugs (50% or more based on our sensitivity analysis for 6-year life-spans) our models suggest that infection levels in dogs (and in some situations infection in bugs), end up higher than pretreatments levels. As the rise in dog infection prevalence occurred either at treatment PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 8 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission administration or shortly thereafter, when bug infection prevalence is very low, we can say that this expected increase is due to canine consumption of bugs killed by treatment. Whether a dog would consume a bug killed by treatment would depend on 1) dog behavior and 2) whether the bugs would be able to conceal themselves prior to succumbing to the treatment. It was reported that nearly all the bugs that fed on treated dogs between 4–60 DPT died within 24 hours of exposure [33], and we assumed that dogs consumed 80% of bugs killed by treatment and that consumption happened immediately upon death, but we report results from 20% and 50% consumption in the supplement. The consequence of relaxing this assump- tion, and having the dogs consume a smaller percentage, would reduce the risk of oral infec- tion, in some cases making treatment beneficial in regions with lower disease prevalence and domestic vectors. Detailed data on the time distribution from bugs feeding on dogs to death would improve estimates for the percentage of bugs that dogs would be able to consume. Our findings warrant further lab experiments and small field trials before launching large xenoin- toxation-based elimination programs. Before utilizing fluralaner in regions with low disease prevalence and domestic or sylvatic bug populations, especially in regions where dogs have longer lifespans, lab studies could inform how quickly bugs died within 24-hour post feeding period to refine the estimates of risk of oral transmission post xenointoxication. Better yet, ran- domized controlled field trials could be designed to closely follow treated dogs, conduct con- tinuous interim analysis, and include early stopping rules if it turns out that treated dogs are becoming infected at rates higher than controls. Current vector control for T. infestans is based on insecticide spray and threatened by the emergence of pyrethroid resistant bugs [58]. Under experimental conditions, fluralaner proved efficacious against both pyrethroid susceptible and resistant 5th-stage nymphs [33]. In fact, between 4–60 DPT, regardless of pyrethroid susceptibility status, almost all bugs were killed after feeding on treated dogs; it would not be until 90–120 DPT that cumulative mortality declined at a greater rate for susceptible bugs than resistant, and these results were found to not be statistically significant [33]. We incorporated the data from Laiño et al. on the percent- age of bugs killed after feeding on treated dogs for both the 5th-stage susceptible and resistant nymphs into the Shiny web application, but the difference in model outcome was negligible, regardless of the parameter set. As fluralaner is a relatively new isoxazoline compound, approved for use in the United States in 2014 (Food and Drug Administration [FDA], 2014), literature review resulted in no information regarding possible fluralaner resistance. Isoxazoline compounds are potent inhib- itors of γ-aminobutyric acid (GABA)-gated chloride channels (GABACls) [59]. Previous phar- macological profiles regarding cyclodiene resistance in Drosophila spp. demonstrated that resistance to cyclodiene conferred broad cross resistance to compounds blocking GABACls [60]; It has been noted that use of novel chloride channel antagonists as insecticides should be managed carefully in order to prevent the rapid development of field resistance [60]. As flura- laner has shown promise in regards to vector control in regions where T. infestans have resis- tance to pyrethroids [33], careful consideration should go into planning and implementation of community-level canine fluralaner treatment programs to avoid selecting for vectors that develop resistance toward isoxazoline compounds. There was some uncertainty inherent in several parameter estimates. Our model, describing household T. cruzi transmission dynamics, is sensitive to the parameter m, the ratio of the number of vectors feeding on any given host; households with a smaller ratio demonstrated unfavorable outcomes with fluralaner treatment when dogs consumed 4 out of 5 of the killed bugs. Yet in small households, populations of domestic animals can be unstable, creating unpredictable fluctuations in this ratio [61]. Likewise, our model only assumed one host, dogs; in a real-world context, the effectiveness of fluralaner treatment on reducing T. infestans PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 9 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission infestation would depend in part on the availability of alternative hosts, including humans, chickens and untreated dogs [33]. Experimental studies reported the majority of fed bugs were fully engorged after feeding on fluralaner treated dogs [33,34], making it unlikely that flurala- ner has a repellent effect which could divert bug feeding towards humans [6]. Field studies in Argentina suggest the fraction of domestic T. infestans with a blood meal on dogs ranging upward of 65%, and that the more bugs fed on dogs the less they fed on humans [62]; it is likely that even with alternative hosts available fluralaner could potentially reduce T. cruzi transmis- sion in regions with high disease prevalence and household T. infestans infestations. Chagas disease dynamics are complex and vary much geographically. From our results, it is clear that the impact of fluralaner on halting T. cruzi transmission depends on a combination of parasite prevalence, insect abundance, and type of triatomine vectors (domestic vs. sylvatic bugs). We developed a Shiny web application to allow users to alter the transmission and treat- ment parameters and examine the results according to local conditions. For our models we used the simplifying assumption that dogs have a constant rate of infectiousness and only leave the infected compartment through death. But similar to humans, dogs experience acute and chronic phases of infection [63]; it is during the acute phase that parasitemia is highest. Taking into account varying reports on the duration of parasitemia (Machado et al., 2001) [63,64], the potential for reactivation, and reports of “super-shedders” in other species (guinea pigs) [65], we countered the assumptions of homogeneity and temporal scales of transmission by reducing the probability of transmission between dog and bug from the reported 0.49 [29] to 0.28. Our model demonstrates the potential for canine fluralaner treatment to reduce T. cruzi transmission in regions with high disease prevalence and domestic vectors; fluralaner treat- ment could be used as a complementary, community-level intervention to reduce T. infestans populations in infested households and could be done as infrequently as once a year. On the other hand, in low endemic regions and regions with sylvatic bugs, canine treatment with flur- alaner could potentially increase infection prevalence in both dog and bug populations via canine oral consumption of vectors killed by treatment. These simulations, though a simplified version of reality, highlight the need for well-designed studies to investigate the conditions under which fluralaner xenointoxication, a promising One Health intervention, is an effective control strategy against Chagas disease. Methods Model construction We conducted a simulation study, using an adaptation of the classic Ross-MacDonald malaria model. Oral predation of triatomines is a well-characterized transmission route of T. cruzi [9– 17,22]; through simulation we explored the potential effects of transmission of T. cruzi via ingestion by dogs of triatomines killed by fluralaner treatment. Laboratory evidence suggests that metacyclic trypomastigotes are viable in dead triatomines days after triatomine death [66]; additionally, there are some reports of human infection through contamination of fruit juices with dead triatomines [9,11,15,17]. In our model, we explore scenarios in which we assume that dogs consume dead triatomines and that the dead triatomines ingestion shortly after bug death (within a few days) results in transmission scenarios similar to oral predation. This model included the following simplifying assumptions: the host (dog) population is assumed to be homogenous and constant. The vector (bug) population was also assumed to be homoge- nous but differed from the classic Ross-MacDonald malaria model [67] as a vector birth rate was incorporated to balance the impact of fluralaner treatment (to avoid having the bug popu- lation “crash” shortly after administration of insecticide). For simplicity we parameterized the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 10 / 19 PLOS COMPUTATIONAL BIOLOGYTable 1. Parameter values for modified Ross-MacDonald model simulations. Parameter Description (unit) proportion of dogs infected proportion of triatomines infected Expected number of bites on dogs per triatomine Equilibrium triatomine density per dog (triatomines per dog) Length of the incubation period (days) Daily force of triatomine mortality (1/day) Transmission efficiency from infectious triatomine to susceptible dog via bite (1/ number of bites required for transmission) Probability of an infection of an uninfected triatomine by biting an infectious dog Daily force of infectious dog mortality (1/day) Maximum proportion of vectors eaten by dogs in a day Transmission efficiency from infectious triatomine to susceptible dog via oral transmission (proportion of oral infection per infected vector consumed) The maximum birthrate at carrying capacity (day/eggs laid) Carrying capacity of vectors per dog X Y a m n g b c r p k R K z Models for xenointoxication and Trypanosoma cruzi transmission Values (range for sensitivity analysis) _ _ 1/14 [1/7-1/21] 40 [10–100] 45 [10–60] 0.005 [0.001–0.01] 0.00068 [0.0005–0.001] Source _ _ [61] Estimated from other species [61] [61] [69] [70] 0.28 [0.10–0.49] Adapted from: [29] 1/(3365) [(1/(2365))-1/ (8*365)] Varied in accordance to regional variations Varied to account for difference in individual animal behavior patterns [16] [71] Varied with assumed population size 0.8 [0.1–0.99] 0.1 0.09 [0.05–0.11] 40 [high prevalence] 15 [low prevalence, 3-year lifespan] 7 [low prevalence, 6-year lifespan] Proportion of triatomines killed by fluralaner treatment Time dependent covariate values obtained from log curve [33] https://doi.org/10.1371/journal.pcbi.1011115.t001 bug population based on data on T. infestans for the domestic cycle and data on T. dimidiata for the sylvatic cycle [68]. We made a number of simplifying assumptions: we ignored vector reproductive senescence and seasonality. We assumed that there was no host recovered class (despite the possibility of both treatment and natural recovery) and grouped hosts in the acute and chronic phases of infection into a single infected class although it is known that hosts are more infectious during the acute phase of infection [64,65]. We further assumed that the only way dogs can leave the infected compartment is through death; to account for cyclic parasite- mia, the parameter used in the model for transmission probability from dogs to vectors has been halved what has been used in prior models (see Table 1). Lastly, as T. infestans primarily exhibits night-feeding behavior to avoid diurnal predators (Schofield, 1985), we assumed that oral transmission only involves the bugs killed by treatment, i.e., there is no oral transmission prior to fluralaner administration. The implications of changing these assumptions are later discussed. Pretreatment model The model considers a single species of host (dogs) and a single vector, which represents differ- ent species in different scenarios. We do not consider more complex situations with multiple vector species. All analyses were carried out in the R software environment [72] using the dif- ferential equation solver deSolve [73] and Shiny packages [74]. Red and blue lines in Fig 5 illustrate transmission dynamics among dogs and bugs prior to fluralaner treatment, with X representing the proportion of infected dogs and 1-X the proportion of dogs that are susceptible. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 11 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission Fig 5. Mathematical models of T. cruzi transmission dynamics between dogs and T. infestans in domestic and sylvatic cycles. Dashed lines represent transmission events and solid lines represent transition between states. Prior to treatment, only vectorial transmission (blue lines) is considered to transition susceptible dogs (1-X) to infectious (X). Susceptible bugs (1-X) are replenished by a logistic birth rate. After administration of fluralaner, there are two transmission routes to infect susceptible dogs: vectorial transmission as before (blue line) and oral transmission (yellow line). In the sylvatic cycle, vectorial transmission is constant due to exposure to external infectious bugs (MM). https://doi.org/10.1371/journal.pcbi.1011115.g005 Dogs move from susceptible to infectious at a rate equal to the force of infection (FOI) due to vectorial transmission, which is equivalent to the product of the bite rate (a), probability of transmission from bugs to dogs via biting (b), the proportion of infected bugs (Y) available, and the ratio of the number of vectors depending on any given host (m, ratio of bugs to dogs) in the system. Susceptible and infected dogs can leave the population through the background death rate, r; as with prior models, no disease induced mortality is assumed for dogs [10,75]. Susceptible bugs (1-Y) become infected (Y) at a rate equal to the FOI for vectors, which is the product of the bite rate (a), the probability of transmission from dogs to bugs (c), and the pro- portion of infected dogs (X); as a susceptible bug must survive the incubation period of T. cruzi to become infectious, the FOI also depends the incubation of the parasite within the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 12 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission vector (n) and the daily probability of bug mortality (g). As with dogs, susceptible and infected bugs leave the population through the background death rate (g). Prior to treatment, transmis- sion dynamics between dogs and bug are represented by the system of differential of equations: dX, the change in proportion of infected dogs, dY, the change in proportion of infected bugs, and dm, the change in the ratio of bugs to dogs in the population (Eqs 1.1, 1.2, 1.3, respec- tively): dX ¼ mabYð1 (cid:0) XÞ (cid:0) rX dY ¼ acXðe(cid:0) gn (cid:0) YÞ (cid:0) gY � dm ¼ Rm 1 (cid:0) � m K ð1:1Þ ð1:2Þ ð1:3Þ Parameter values (Table 1) were adjusted to fit observed prevalence for regions of high and low disease prevalence domestic vectors, and regions with sylvatic vectors. High and low prev- alence are used relatively to consider different geographic regions; in our simulations, high prevalence areas have a 2.5 times greater carrying capacity of triatomine per dog than that of low prevalence. Using the model and parameter values in Table 1, the impact of fluralaner treatment on bug and dog transmission dynamics were evaluated over the timescale of decades. Treatment model Data reported from Laiño et al. regarding the percentage of bugs killed after feeding on treated dogs over time were incorporated into the treatment model [33]. We assume that all dogs in a household are treated with fluralaner at a dosage in agreement with manufacturer instructions [52]. The percentage of bugs killed after feeding on treated dogs over days post treatment (DPT) was plotted and fit to a logistic curve (S6 Fig), and the asymptote, x-midpoint and scale values were extracted at timepoints 4–360 DPT. To examine the effects of treatment on differ- ent bug populations, the percentage of killed bugs after feeding on treated dogs were taken from data regarding fifth stage pyrethroid-resistant nymphs and fifth stage pyrethroid-suscep- tible nymphs [33]; analyses in this paper used the data for 5th-stage pyrethroid susceptible nymphs. The values comprising the equation of the logistic curve were incorporated into parameter z, the percentage of bugs feeding on treated bugs that are killed at a point in time, and the time dependent covariate was incorporated into the model. Treatment was initiated into the model after both the bug and dog populations reached equilibrium. To determine these values, the equation for the basic reproductive number of T. cruzi was rearranged and solved for X and Y, the values of the proportion of infectious dogs and bugs at equilibrium, respectively [61]. Parameter values for regions with semi-sylvatic bugs were calibrated to approximately values reported for T. dimidiata reported in Yucatan, Mexico [76]. Incorporating treatment, the differential equations are altered (Eqs 2.1, 2.2, 2.3) to reflect the fact that change in proportion of infected dogs is now subjected to an additional FOI due to ingestion of dead infected bugs (Fig 5). Contact between dogs and the dead bugs depends on the availability of dead bugs at a given time point; this is the product of the bite rate, a, the percentage of bugs that will die after feed- ing on a treated dogs at that given time point, z, the proportion of infected bugs Y, and the ratio of bugs to dogs in the population, m. The rate that susceptible dogs become infected via oral transmission will depend on the product of azmY, the probability of transmission via bug ingestion, k, and percentage of dead bugs consumed, p. The rate at which the ratio of bugs to PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 13 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission dogs decreases in the population is proportional to the bite rate, and the percentage of bugs killed that feed on treated dogs at a given time point (Eq 2.3), while the population is replen- ished at the rate of the bug logistic birth rate. dX ¼ ½mabY þ pkðmazYÞ�ð1 (cid:0) XÞ (cid:0) rX dY ¼ acXðe(cid:0) gn (cid:0) YÞ (cid:0) gY (cid:0) mazY � dm ¼ Rm 1 (cid:0) � m K (cid:0) maz ð2:1Þ ð2:2Þ ð2:3Þ As all bugs in the previous model are assumed to be subjected to fluralaner treatment, the model would not properly represent regions where triatomine vectors include sylvatic bugs. To account for external bugs not affected by treatment, a constant was introduced (MM) to the FOI for dogs through vectorial transmission (Eq 3.1). The constant MM represents sylvatic infected bugs that can contribute to the vectorial FOI in dogs but would not contribute to the oral FOI if killed by treatment and whose populations would not be reduced if some individu- als are killed by fluralaner (Fig 5). Values for the constant were derived by running the model without treatment and determining their impact on infection prevalence in dogs. dX ¼ ½mabY þ MM þ pkðmazYÞ�ð1 (cid:0) XÞ (cid:0) rX ð3:1Þ We performed sensitivity analyses upon input parameters based on a range of plausible val- ues found in the literature (S2–S4 Figs). We also created a Shiny web application [74] to allow users to simulate the model in a way that can capture regional variation in multiple parameters available at https://jrokh.shinyapps.io/NewExternalBugs/. All analyses were carried out assum- ing the dogs consume 80% of the bugs killed by treatment (p = 0.8). We also explored different consumption levels from 20% to 80% (S2–S5 Figs). Unless explicitly stated, all models were run using the baseline parameter values (Table 1). R code used within the shiny application and to run the different simulations overviewed here can be found in S1 Code. Supporting information S1 Fig. Proportion of infected dogs and infected T. infestans prior administration of flura- laner treatment. A corresponds to the baseline pre-treatment model in regions with high dis- ease prevalence and domestic vectors. B corresponds to the baseline pre-treatment model in regions with low prevalence and sylvatic vectors. (TIF) S2 Fig. Sensitivity analysis on proportion of consumed bugs by dogs for single fluralaner treatment in a high prevalence region (corresponds to Fig 1 in main text). (A) corresponds to 20% of bugs being consumed; (B) corresponds to 50% of bugs being consumed; (C) corre- sponds to 80% of bugs being consumed. (TIF) S3 Fig. Sensitivity analysis on proportion of consumed bugs by dogs for multiple fluralaner treatments in a high prevalence region (corresponds to Fig 2 in main text). Annual admin- istration of fluralaner for both 4 years (A) and 6 years (B) was simulated, as well as administra- tion every 90 days (veterinary recommendation) for one year (C) and for two years (D). (TIF) S4 Fig. Sensitivity analysis on proportion of consumed bugs by dogs for fluralaner treat- ments schemes in a low prevalence region (corresponds to Fig 3 in main text). We explored PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 14 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission a range of dog average lifespan from 3 years (A-C) to 6 years (D-F). Treatment scenarios include one time treatment (A, D), annual treatment for 4 years (B, E), and treatment every 90 days for 1 year (C, F). (TIF) S5 Fig. Sensitivity analysis on proportion of consumed bugs by dogs for fluralaner treat- ments schemes in a low prevalence region with semi sylvatic cycles (corresponds to Fig 4 in main text). We explored a range of dog average lifespan from 3 years (A-C) to 6 years (D-F). Treatment scenarios include one time treatment (A, D), annual treatment for 4 years (B, E), and treatment every 90 days for 1 year (C, F). (TIF) S6 Fig. The percentage of bugs killed after feeding on treated dogs over days post treat- ment. Data from Laino et al (2019) on the declining percentage of bugs killed after feeding on fluralaner treated dogs was fit to a logistic curve and incorporated into the model of T. cruzi transmission dynamics in bugs and dogs with fluralaner treatment. Initial analyses used the data from 5th stage pyrethroid susceptible nymphs. (TIF) S1 Code. Output rendered from the R markdown used to run the simulations addressed. (HTML) Acknowledgments JLR would like to acknowledge the expertise and guidance of her advisor, Dr Thersa Sweet at Dornsife School of Public Health at Drexel University. Author Contributions Conceptualization: Jennifer L. Rokhsar, Michael Z. Levy, Ricardo Castillo-Neyra. Formal analysis: Jennifer L. Rokhsar, Brinkley Raynor, Justin Sheen, Neal D. Goldstein, Michael Z. Levy, Ricardo Castillo-Neyra. Funding acquisition: Ricardo Castillo-Neyra. Investigation: Jennifer L. Rokhsar, Michael Z. Levy, Ricardo Castillo-Neyra. Methodology: Jennifer L. Rokhsar, Brinkley Raynor, Justin Sheen, Michael Z. Levy, Ricardo Castillo-Neyra. Project administration: Michael Z. Levy, Ricardo Castillo-Neyra. Resources: Michael Z. Levy. Supervision: Michael Z. Levy, Ricardo Castillo-Neyra. Visualization: Jennifer L. Rokhsar, Brinkley Raynor. Writing – original draft: Jennifer L. Rokhsar, Brinkley Raynor, Justin Sheen, Michael Z. Levy, Ricardo Castillo-Neyra. Writing – review & editing: Jennifer L. Rokhsar, Brinkley Raynor, Justin Sheen, Neal D. Goldstein, Michael Z. Levy, Ricardo Castillo-Neyra. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 15 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission References 1. Bern C. Chagas’ Disease. N Engl J Med. 2015; 373: 456–466. https://doi.org/10.1056/NEJMra1410150 PMID: 26222561 2. World Health Organization. Chagas disease in Latin America: an epidemiological update based on 2010 estimates = Maladie de Chagas en Ame´ rique latine: le point e´pide´miologique base´ sur les estima- tions de 2010. Wkly Epidemiol Rec Releve´ E´ pide´ miologique Hebd. 2015; 90: 33–44. 3. World Health Organization. Chagas disease fact sheet. 1 Apr 2021 [cited 15 Mar 2022]. Available: https://www.who.int/news-room/fact-sheets/detail/chagas-disease-(american-trypanosomiasis) 4. Carlos Pinto Dias J. Epidemiology of Chagas Disease. Chagas Disease—American Trypanosomiasis: its impact on transfusion and clinical medicine. Sao Paulo, Brazil; 1992. Available: http://www.dbbm. fiocruz.br/tropical/chagas/chapter4.html 5. 6. Finkelman J. Innovative community-based ecosystem management for dengue and Chagas disease prevention in low and middle income countries in Latin America and the Caribbean. Trans R Soc Trop Med Hyg. 2015; 109: 89–90. https://doi.org/10.1093/trstmh/tru201 PMID: 25604758 Travi BL. Considering Dogs as Complementary Targets of Chagas Disease Control. Vector Borne Zoo- notic Dis Larchmt N. 2019; 19: 90–94. https://doi.org/10.1089/vbz.2018.2325 PMID: 30102585 7. Kirchhoff LV. Chapter 1—Epidemiology of American Trypanosomiasis (Chagas Disease). In: Weiss LM, Tanowitz HB, Kirchhoff LV, editors. Advances in Parasitology. Academic Press; 2011. pp. 1–18. https://doi.org/10.1016/B978-0-12-385863-4.00001–0 8. Coura JR. The main sceneries of Chagas disease transmission. The vectors, blood and oral transmis- sions—A comprehensive review. Mem Inst Oswaldo Cruz. 2015; 110: 277–282. https://doi.org/10. 1590/0074-0276140362 PMID: 25466622 9. de Noya BA, Dı´az-Bello Z, Colmenares C, Ruiz-Guevara R, Mauriello L, Muñoz-Caldero´n A, et al. Update on oral Chagas disease outbreaks in Venezuela: epidemiological, clinical and diagnostic approaches. Mem Inst Oswaldo Cruz. 2015; 110: 377–386. https://doi.org/10.1590/0074-02760140285 PMID: 25946155 10. Kribs C. Vector consumption and contact process saturation in sylvatic transmission of T. cruzi. 2006 [cited 17 Mar 2022]. https://doi.org/10.1080/08898480600788576 11. Pacheco LV, Santana LS, Barreto BC, Santos E de S, Meira CS. Transmissão oral da doenc¸a de Cha- gas: Uma revisão de literatura. Res Soc Dev. 2021; 10: e31910212636–e31910212636. https://doi.org/ 10.33448/rsd-v10i2.12636 12. Pereira KS, Schmidt FL, Barbosa RL, Guaraldo AMA, Franco RMB, Dias VL, et al. Transmission of Chagas Disease (American Trypanosomiasis) by Food. Advances in Food and Nutrition Research. Elsevier; 2010. pp. 63–85. https://doi.org/10.1016/S1043-4526(10)59003-X PMID: 20610174 13. Shikanai-Yasuda MA, Carvalho NB. Oral Transmission of Chagas Disease. Clin Infect Dis. 2012; 54: 845–852. https://doi.org/10.1093/cid/cir956 PMID: 22238161 14. Kobylinski K, Rutledge-Connelly R. Blood Feeding Insect Series: American Trypanosomiasis—Chagas Disease: ENY-726/IN650, 7/2006. EDIS. 2006;2006. https://doi.org/10.32473/edis-in650-2006 15. Barroso Ferreira RT, Branquinho MR, Cardarelli-Leite P. Transmissão oral da doenc¸a de Chagas pelo consumo de ac¸aı´: um desafio para a Vigilaˆ ncia Sanita´ ria. Vigilaˆ ncia Sanita´ria Em Debate. 2014; 2: 358/ 160. https://doi.org/10.3395/vd.v2i4.358 16. Kribs-Zaleta C. Estimating Contact Process Saturation in Sylvatic Transmission of Trypanosoma cruzi in the United States. PLoS Negl Trop Dis. 2010; 4: e656. https://doi.org/10.1371/journal.pntd.0000656 PMID: 20436914 17. Passos LAC, Guaraldo AMA, Barbosa RL, Dias VL, Pereira KS, Schmidt FL, et al. Sobrevivência e infectividade do Trypanosoma cruzi na polpa de ac¸aı´: estudo in vitro e in vivo. Epidemiol E Servic¸os Sau´ de. 2012; 21: 223–232. https://doi.org/10.5123/S1679-49742012000200005 18. Castillo-Neyra R, Chou Chu L, Quispe-Machaca V, Ancca-Juarez J, Malaga Chavez FS, Bastos Mazuelos M, et al. The potential of canine sentinels for reemerging Trypanosoma cruzi transmission. Prev Vet Med. 2015; 120: 349–356. https://doi.org/10.1016/j.prevetmed.2015.04.014 PMID: 25962956 19. Fujita O, Sanabria L, Inchaustti A, De Arias AR, Tomizawa Y, Oku Y. Animal reservoirs for Trypano- soma cruzi infection in an endemic area in Paraguay. J Vet Med Sci. 1994; 56: 305–308. https://doi.org/ 10.1292/jvms.56.305 PMID: 8075219 20. Gu¨rtler RE, Cardinal MV. Reservoir host competence and the role of domestic and commensal hosts in the transmission of Trypanosoma cruzi. Acta Trop. 2015; 151: 32–50. https://doi.org/10.1016/j. actatropica.2015.05.029 PMID: 26051910 21. Jime´nez-Coello M, Guzma´ n-Marı´n E, Ortega-Pacheco A, Acosta-Viana KY. Serological survey of American trypanosomiasis in dogs and their owners from an urban area of Me´ rida Yucatàn, Me´ xico. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 16 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission Transbound Emerg Dis. 2010; 57: 33–36. https://doi.org/10.1111/j.1865-1682.2010.01130.x PMID: 20537099 22. Coffield DJ, Spagnuolo AM, Shillor M, Mema E, Pell B, Pruzinsky A, et al. A Model for Chagas Disease with Oral and Congenital Transmission. PLOS ONE. 2013; 8: e67267. https://doi.org/10.1371/journal. pone.0067267 PMID: 23840647 23. Mota-Rojas D, Caldero´n-Maldonado N, Lezama-Garcı´a K, Sepiurka L, Maria Garcia R de C. Abandon- ment of dogs in Latin America: Strategies and ideas. Vet World. 2021; 14: 2371–2379. https://doi.org/ 10.14202/vetworld.2021.2371-2379 PMID: 34840456 24. Levy MZ. Kindling, Logs, and Coals: The Dynamics of Trypanosoma cruzi, the Etiological Agent of Cha- gas Disease in Arequipa, Peru. Population Biology of Vector-Borne Diseases. Oxford: Oxford Univer- sity Press; 2020. https://doi.org/10.1093/oso/9780198853244.003.0012 25. Gabrielli S, Spinicci M, Macchioni F, Rojo D, Totino V, Rojas P, et al. Canine Trypanosoma cruzi infec- tion in the Bolivian Chaco. Parasit Vectors. 2018; 11: 632. https://doi.org/10.1186/s13071-018-3247-0 PMID: 30541629 26. Cardinal MV, Enriquez GF, Macchiaverna NP, Argibay HD, Ferna´ ndez M del P, Alvedro A, et al. Long- term impact of a ten-year intervention program on human and canine Trypanosoma cruzi infection in the Argentine Chaco. PLoS Negl Trop Dis. 2021; 15: e0009389. https://doi.org/10.1371/journal.pntd. 0009389 PMID: 33979344 27. Gu¨rtler RE, Lauricella M, Solarz ND, Bujas MA, Wisnivesky-Colli C. Dynamics of transmission of Trypa- nosoma cruzi in a rural area of Argentina. I—The dog reservoir: an epidemiological profile. Rev Inst Med Trop Sao Paulo. 1986; 28: 28–35. https://doi.org/10.1590/s0036-46651986000100006 PMID: 3094132 28. Epidemiology Zeledo´n R., Modes of Transmission and Reservoir Hosts of Chagas’ Disease. Ciba Foun- dation Symposium 20—Trypanosomiasis and Leishmaniasis (with Special Reference to Chagas’ Dis- ease). John Wiley & Sons, Ltd; 1974. pp. 51–85. https://doi.org/10.1002/9780470720035.ch4 29. Gurtler RE, Cecere MC, Castanera MB, Canale D, Lauricella MA, Chuit R, et al. Probability of Infection with Trypanosoma cruzi of the Vector Triatoma infestans Fed on Infected Humans and Dogs in North- west Argentina. Am J Trop Med Hyg. 1996; 55: 24–31. https://doi.org/10.4269/ajtmh.1996.55.1. TM0550010024 30. Arau´jo-Neto VT de, Honorato NRM, de Oliveira Santana R, Barbosa-Silva AN, da Matta Guedes PM, Chiari E, et al. Trypanosoma cruzi circulating among dogs and triatomines in the endemic countryside of the State of Rio Grande do Norte, Brazil. Acta Trop. 2019; 200: 105067. https://doi.org/10.1016/j. actatropica.2019.105067 PMID: 31255585 31. Costa J, Almeida CE, Dotson EM, Lins A, Vinhaes M, Silveira AC, et al. The epidemiologic importance of Triatoma brasiliensis as a chagas disease vector in Brazil: a revision of domiciliary captures during 1993–1999. Mem Inst Oswaldo Cruz. 2003; 98: 443–449. https://doi.org/10.1590/s0074- 02762003000400002 PMID: 12937751 32. Cohen JE, Gu¨ rtler RE. Modeling household transmission of American trypanosomiasis. Science. 2001; 293: 694–698. https://doi.org/10.1126/science.1060638 PMID: 11474111 33. 34. Laiño MA, Cardinal MV, Enriquez GF, Alvedro A, Gaspe MS, Gu¨rtler RE. An oral dose of Fluralaner administered to dogs kills pyrethroid-resistant and susceptible Chagas disease vectors for at least four months. Vet Parasitol. 2019; 268: 98–104. https://doi.org/10.1016/j.vetpar.2019.03.005 PMID: 30981313 Loza A, Talaga A, Herbas G, Canaviri RJ, Cahuasiri T, Luck L, et al. Systemic insecticide treatment of the canine reservoir of Trypanosoma cruzi induces high levels of lethality in Triatoma infestans, a princi- pal vector of Chagas disease. Parasit Vectors. 2017; 10: 344. https://doi.org/10.1186/s13071-017- 2278-2 PMID: 28724448 35. Reithinger R, Ceballos L, Stariolo R, Davies CR, Gu¨ rtler RE. Chagas disease control: deltamethrin- treated collars reduce Triatoma infestans feeding success on dogs. Trans R Soc Trop Med Hyg. 2005; 99: 502–508. https://doi.org/10.1016/j.trstmh.2004.11.013 PMID: 15869774 36. Queiroga TBD, Gomez LCP, de Sena ER, dos Santos WV, Ferreira HRP, de Arau´ jo-Neto VT, et al. Insecticidal efficacy of fluralaner (Bravecto) against Triatoma brasiliensis, a major vector of Trypano- soma cruzi in Brazil. Parasit Vectors. 2021; 14: 456. https://doi.org/10.1186/s13071-021-04978-x PMID: 34488865 37. Castillo-Riquelme M, Guhl F, Turriago B, Pinto N, Rosas F, Martı´nez MF, et al. The Costs of Preventing and Treating Chagas Disease in Colombia. PLoS Negl Trop Dis. 2008; 2: e336. https://doi.org/10.1371/ journal.pntd.0000336 PMID: 19015725 38. Nakagawa J, Cordo´n-Rosales C, Jua´ rez J, Itzep C, Nonami T. Impact of residual spraying on Rhodnius prolixus and Triatoma dimidiata in the department of Zacapa in Guatemala. Mem Inst Oswaldo Cruz. 2003; 98: 277–281. https://doi.org/10.1590/s0074-02762003000200019 PMID: 12764447 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 17 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission 39. Vazquez-Prokopec GM, Spillmann C, Zaidenberg M, Kitron U, Gu¨ rtler RE. Cost-Effectiveness of Cha- gas Disease Vector Control Strategies in Northwestern Argentina. PLoS Negl Trop Dis. 2009; 3: e363. https://doi.org/10.1371/journal.pntd.0000363 PMID: 19156190 40. Yoshioka K, Nakamura J, Pe´rez B, Tercero D, Pe´rez L, Tabaru Y. Effectiveness of Large-Scale Chagas Disease Vector Control Program in Nicaragua by Residual Insecticide Spraying against Triatoma dimi- diata. Am J Trop Med Hyg. 2015; 93: 1231–1239. https://doi.org/10.4269/ajtmh.15-0403 PMID: 26416118 41. Ferna´ ndez M del P, Gaspe MS, Gu¨rtler RE. Inequalities in the social determinants of health and Chagas disease transmission risk in indigenous and creole households in the Argentine Chaco. Parasit Vectors. 2019; 12: 184. https://doi.org/10.1186/s13071-019-3444-5 PMID: 31029147 42. Gonc¸alves R, Logan RAE, Ismail HM, Paine MJI, Bern C, Courtenay O. Indoor residual spraying prac- tices against Triatoma infestans in the Bolivian Chaco: contributing factors to suboptimal insecticide delivery to treated households. Parasit Vectors. 2021; 14: 327. https://doi.org/10.1186/s13071-021- 04831-1 PMID: 34134775 43. Gu¨rtler RE, Cecere MC, Lauricella MA, Petersen RM, Chuit R, Segura EL, et al. Incidence of Trypano- soma cruzi infection among children following domestic reinfestation after insecticide spraying in rural northwestern Argentina. Am J Trop Med Hyg. 2005; 73: 95–103. 44. Paz-Solda´ n VA, Bauer KM, Hunter GC, Castillo-Neyra R, Arriola VD, Rivera-Lanas D, et al. To spray or not to spray? Understanding participation in an indoor residual spray campaign in Arequipa, Peru. Glob Public Health. 2018; 13: 65–82. https://doi.org/10.1080/17441692.2016.1178317 PMID: 27189446 45. Hemingway J, Beaty BJ, Rowland M, Scott TW, Sharp BL. The Innovative Vector Control Consortium: improved control of mosquito-borne diseases. Trends Parasitol. 2006; 22: 308–312. https://doi.org/10. 1016/j.pt.2006.05.003 PMID: 16713358 46. Davis DS, Russell LH, Adams LG, Yaeger RG, Robinson RM. An experimental infection of Trypano- soma cruzi in striped skunks (Mephitis mephitis). J Wildl Dis. 1980; 16: 403–406. https://doi.org/10. 7589/0090-3558-16.3.403 PMID: 6774112 47. Roellig DM, Ellis AE, Yabsley MJ. Oral transmission of Trypanosoma cruzi with opposing evidence for the theory of carnivory. J Parasitol. 2009; 95: 360–364. https://doi.org/10.1645/GE-1740.1 PMID: 18763853 48. Silva-dos-Santos D, Barreto-de-Albuquerque J, Guerra B, Moreira OC, Berbert LR, Ramos MT, et al. Unraveling Chagas disease transmission through the oral route: Gateways to Trypanosoma cruzi infec- tion and target tissues. PLoS Negl Trop Dis. 2017; 11: e0005507. https://doi.org/10.1371/journal.pntd. 0005507 PMID: 28379959 49. Yaeger RG. Transmission of Trypanosoma cruzi infection to opossums via the oral route. J Parasitol. 1971; 57: 1375–1376. 50. Bradley K, Bergman D, Woods J, Crutcher J, Kirchhoff L. Prevalence of American trypanosomiasis (Chagas disease) among dogs in Oklahoma. J Am Vet Med Assoc. 2000;217. https://doi.org/10.2460/ javma.2000.217.1853 PMID: 11132891 51. Fiatsonu E, Busselman RE, Hamer GL, Hamer SA, Ndeffo-Mbah ML. Effectiveness of fluralaner treat- ment regimens for the control of canine Chagas disease: A mathematical modeling study. PLoS Negl Trop Dis. 2023; 17: e0011084. https://doi.org/10.1371/journal.pntd.0011084 PMID: 36693084 52. Merck Animal Health. Bravecto (Fluralaner). Merck Animal Heatlh; 2014. Available: https://us.bravecto. com/pdfs/bravecto_pi_mah.pdf 53. Administracio´ n Gestio´ n Comercial. Listado de precios Peru´. Administracio´ n Gestio´ n Comercial; 2022. 54. Sueldo mı´nimo: Gobierno aumenta de 930 a 1025 soles la remuneracio´ n mı´nima vital en el Peru´. In: infobae [Internet]. 3 Apr 2022 [cited 18 Apr 2022]. Available: https://www.infobae.com/america/peru/ 2022/04/03/sueldo-minimo-gobierno-de-pedro-castillo-aumenta-de-930-a-1025-soles-la- remuneracion-minima-vital-en-el-peru/ 55. Barbu C, Dumonteil E, Gourbière S. Optimization of Control Strategies for Non-Domiciliated Triatoma dimidiata, Chagas Disease Vector in the Yucata´ n Peninsula, Mexico. PLoS Negl Trop Dis. 2009; 3: e416. https://doi.org/10.1371/journal.pntd.0000416 PMID: 19365542 56. Zu Dohna H, Cecere MC, Gu¨rtler RE, Kitron U, Cohen JE. Re-establishment of local populations of vec- tors of Chagas disease after insecticide spraying. J Appl Ecol. 2007; 44: 220–227. https://doi.org/10. 1111/j.1365-2664.2006.01243.x PMID: 17710182 57. Gu¨rtler RE, Laiño MA, Alvedro A, Enriquez GF, Macchiaverna NP, Gaspe MS, et al. Treatment of dogs with fluralaner reduced pyrethroid-resistant Triatoma infestans abundance, Trypanosoma cruzi infec- tion and human-triatomine contact in the Argentine Chaco. Parasit Vectors. 2022; 15: 257. https://doi. org/10.1186/s13071-022-05343-2 PMID: 35831874 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 18 / 19 PLOS COMPUTATIONAL BIOLOGYModels for xenointoxication and Trypanosoma cruzi transmission 58. Forlani L, Pedrini N, Girotti JR, Mijailovsky SJ, Cardozo RM, Gentile AG, et al. Biological Control of the Chagas Disease Vector Triatoma infestans with the Entomopathogenic Fungus Beauveria bassiana Combined with an Aggregation Cue: Field, Laboratory and Mathematical Modeling Assessment. PLoS Negl Trop Dis. 2015; 9: e0003778. https://doi.org/10.1371/journal.pntd.0003778 PMID: 25969989 59. Gassel M, Wolf C, Noack S, Williams H, Ilg T. The novel isoxazoline ectoparasiticide fluralaner: selec- tive inhibition of arthropod γ-aminobutyric acid- and L-glutamate-gated chloride channels and insecti- cidal/acaricidal activity. Insect Biochem Mol Biol. 2014; 45: 111–124. https://doi.org/10.1016/j.ibmb. 2013.11.009 PMID: 24365472 60. Bloomquist JR. Cyclodiene resistance at the insect GABA receptor/chloride channel complex confers broad cross resistance to convulsants and experimental phenylpyrazole insecticides. Arch Insect Bio- chem Physiol. 1994; 26: 69–79. https://doi.org/10.1002/arch.940260106 PMID: 8054658 61. Levy MZ, Tustin A, Castillo-Neyra R, Mabud TS, Levy K, Barbu CM, et al. Bottlenecks in domestic ani- mal populations can facilitate the emergence of Trypanosoma cruzi, the aetiological agent of Chagas disease. Proc Biol Sci. 2015;282. https://doi.org/10.1098/rspb.2014.2807 PMID: 26085582 62. Gurtler RE, Cohen JE, Cecere MC, Chuit R. Shifting Host Choices of the Vector of Chagas Disease, Triatoma Infestans, in Relation to the Availability of Host in Houses in North-West Argentina. J Appl Ecol. 1997; 34: 699–715. https://doi.org/10.2307/2404917 63. Eloy LJ, Lucheis SB. Canine trypanosomiasis: etiology of infection and implications for public health. J Venom Anim Toxins Trop Dis. 2009; 15: 589–611. 64. Machado EM, Fernandes AJ, Murta SM, Vitor RW, Camilo DJ, Pinheiro SW, et al. A study of experi- mental reinfection by Trypanosoma cruzi in dogs. Am J Trop Med Hyg. 2001; 65: 958–965. https://doi. org/10.4269/ajtmh.2001.65.958 PMID: 11792006 65. Castillo-Neyra R, Borrini Mayorı´ K, Salazar Sa´nchez R, Ancca Suarez J, Xie S, Na´ quira Velarde C, et al. Heterogeneous infectiousness in guinea pigs experimentally infected with Trypanosoma cruzi. Parasitol Int. 2016; 65: 50–54. https://doi.org/10.1016/j.parint.2015.09.009 PMID: 26432777 66. Del S. Guzma´ n Marı´n E, Antonio Barrera Pe´ rez M, Eugenia Rodrı´guez Fe´lix M. Viabilidad del Trypano- soma cruzi en Triatoma dimidiata muertos. Enfermedades Infecc Microbiol Clı´nica. 1999; 19: 113–115. 67. Aron JL, May RM. The population dynamics of malaria. In: Anderson RM, editor. The Population Dynamics of Infectious Diseases: Theory and Applications. Boston, MA: Springer US; 1982. pp. 139– 179. https://doi.org/10.1007/978-1-4899-2901-3_5 68. World Health Organization. Control of Chagas disease: second report of the WHO expert committee. World Health Organization; 2002. Available: https://apps.who.int/iris/handle/10665/42443 69. Cruz-Pacheco G, Esteva L, Vargas C. Control measures for Chagas disease. Math Biosci. 2012; 237: 49–60. https://doi.org/10.1016/j.mbs.2012.03.005 PMID: 22450034 70. Catala SS, Gorla DE, Basombrio MA. Vectorial Transmission of Trypanosoma Cruzi: An Experimental Field Study with Susceptible and Immunized Hosts. Am J Trop Med Hyg. 1992; 47: 20–26. https://doi. org/10.4269/ajtmh.1992.47.20 PMID: 1636879 71. Are´valo A, Carranza JC, Guhl F, Clavijo JA, Vallejo GA. Comparison of the life cycles of Rhodnius colombiensis Moreno, Jurberg & Galvão, 1999 and R. prolixus Stal, 1872 (Hemiptera, Reduviidae, Tria- tominae) under laboratory conditions. Biomed Rev Inst Nac Salud. 2007;27 Suppl 1: 119–129. 72. R core team. R: A language and environment for statistical computing. R Foundation for Statistical Computing; 2018. Available: https://www.R-project.org/ 73. Soetaert K, Petzoldt T, Setzer RW, ddaspk.f PNB (files, dvode.f, zvode.f), et al. deSolve: Solvers for Ini- tial Value Problems of Differential Equations (“ODE”, “DAE”, ‘DDE’). 2021. Available: https://CRAN.R- project.org/package=deSolve 74. Chang W, Cheng J, Allaire JJ, Sievert C, Schloerke B, Xie Y, et al. shiny: Web Application Framework for R. 2021. Available: https://CRAN.R-project.org/package=shiny 75. Gu¨rtler RE, Kravetz FO, Petersen RM, Lauricella MA, Wisnivesky-Colli C. The prevalence of Trypano- soma cruzi and the demography of dog populations after insecticidal spraying of houses: a predictive model. Ann Trop Med Parasitol. 1990; 84: 313–323. https://doi.org/10.1080/00034983.1990.11812475 PMID: 2124438 76. Lee BY, Bartsch SM, Skrip L, Hertenstein DL, Avelis CM, Ndeffo-Mbah M, et al. Are the London Decla- ration’s 2020 goals sufficient to control Chagas disease?: Modeling scenarios for the Yucatan Penin- sula. PLoS Negl Trop Dis. 2018; 12: e0006337. https://doi.org/10.1371/journal.pntd.0006337 PMID: 29554086 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1011115 May 8, 2023 19 / 19 PLOS COMPUTATIONAL BIOLOGY
10.1371_journal.pcbi.1009322	RESEARCH ARTICLE Neurally-constrained modeling of human gaze strategies in a change blindness task Akshay JagatapID Devarajan SridharanID 1,2* 1¤a, Simran PurokayasthaID 1¤b, Hritik JainID 1¤c, 1 Centre for Neuroscience, Indian Institute of Science, Bangalore, India, 2 Computer Science and Automation, Indian Institute of Science, Bangalore, India ¤a Current address: IN Machine Learning, Amazon, World Trade Centre, Bangalore, India ¤b Current address: Department of Psychology, New York University, New York, United States of America ¤c Current address: The Data Science Institute, Columbia University, New York, United States of America * sridhar@iisc.ac.in Abstract Despite possessing the capacity for selective attention, we often fail to notice the obvious. We investigated participants’ (n = 39) failures to detect salient changes in a change blind- ness experiment. Surprisingly, change detection success varied by over two-fold across participants. These variations could not be readily explained by differences in scan paths or fixated visual features. Yet, two simple gaze metrics–mean duration of fixations and the vari- ance of saccade amplitudes–systematically predicted change detection success. We explored the mechanistic underpinnings of these results with a neurally-constrained model based on the Bayesian framework of sequential probability ratio testing, with a posterior odds-ratio rule for shifting gaze. The model’s gaze strategies and success rates closely mimicked human data. Moreover, the model outperformed a state-of-the-art deep neural network (DeepGaze II) with predicting human gaze patterns in this change blindness task. Our mechanistic model reveals putative rational observer search strategies for change detection during change blindness, with critical real-world implications. Author summary Our brain has the remarkable capacity to pay attention, selectively, to important objects in the world around us. Yet, sometimes, we fail spectacularly to notice even the most salient events. We tested this phenomenon in the laboratory with a change-blindness experiment, by having participants freely scan and detect changes across discontinuous image pairs. Participants varied widely in their ability to detect these changes. Surprisingly, two low- level gaze metrics—fixation durations and saccade amplitudes—strongly predicted suc- cess in this task. We present a novel, computational model of eye movements, incorporat- ing neural constraints on stimulus encoding, that links these gaze metrics with change detection success. Our model is relevant for a mechanistic understanding of human gaze strategies in dynamic visual environments. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Jagatap A, Purokayastha S, Jain H, Sridharan D (2021) Neurally-constrained modeling of human gaze strategies in a change blindness task. PLoS Comput Biol 17(8): e1009322. https:// doi.org/10.1371/journal.pcbi.1009322 Editor: Alireza Soltani, Dartmouth College, UNITED STATES Received: March 3, 2021 Accepted: August 4, 2021 Published: August 24, 2021 Peer Review History: PLOS recognizes the benefits of transparency in the peer review process; therefore, we enable the publication of all of the content of peer review and author responses alongside final, published articles. The editorial history of this article is available here: https://doi.org/10.1371/journal.pcbi.1009322 Copyright: © 2021 Jagatap et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: Data availability. Data associated with all figures and tables presented in the manuscript is available online at: https://doi. org/10.6084/m9.figshare.8247860. Code availability. Code for reproducing the all figures and PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 1 / 34 PLOS COMPUTATIONAL BIOLOGYtables presented in the manuscript is available online at: https://doi.org/10.6084/m9.figshare. 8247860. Funding: All the awards are received by Dr. Devarajan Sridharan (DS). The sponsors/funders and the corresponding grant numbers are listed below: Wellcome Trust-Department of Biotechnology India Alliance Intermediate fellowship – IA/I/15/2/502089 Science and Engineering Research Board Early Career award – ECR/2016/000403 Pratiksha Trust award – FG/ SMCH-19-2047 India-Trento Programme for Advanced Research (ITPAR) grant – INT/ITAL Y/ ITPAR-IV/COG/2018/G Department of Biotechnology-Indian Institute of Science Partnership Program grant Sonata Software foundation grant Tata Trusts grant The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: Devarajan Sridharan is a research consultant at Google. All other authors have declared that no competing interests exist. Modeling gaze strategies in a change blindness task Introduction We live in a rapidly changing world. For adaptive survival, our brains must possess the ability to identify relevant, changing aspects of our environment and distinguish them from irrelevant, static ones. For example, when driving down a busy road it is critical to identify changing aspects of the visual scene, such as vehicles shifting lanes or pedestrians crossing the street. Our ability to identify such critical changes is facilitated by visual attention–an essential cognitive capacity that selects the most relevant information in the environment, at each moment in time, to guide behavior [1]. Yet, our capacity for attention possesses key limitations. One such limitation is revealed by the phenomenon of “change blindness”, in which observers fail to detect obvious changes in a sequence of visual images with intervening discontinuities [2,3]. Previous literature suggests that observers’ lapses with detecting changes occur if the changes fail to draw attention; for example if the change is presented concurrently with distracting events, such as an intervening blank or transient noise patches. Change blindness, therefore, provides a useful framework for studying visual attention mechanisms and its lapses [4]. Such lapses have important real-world implications: observers’ success in change blindness tasks has been linked to their driving experience levels [5,6] and safe driving skills [7]. In the laboratory, change blindness is tested, typically, by presenting an alternating sequence of (a pair of) images that differ in one important detail (Fig 1A, “flicker” paradigm) [2,3]. Participants are instructed to scan the images, with overt eye movements, to locate and identify the changing object or feature. While many previous studies have investigated the phe- nomenon of change blindness itself [8–10], very few studies have directly identified gaze- related factors that determine observers’ success in change blindness tasks [4]. In this study, we tested 39 participants in a change blindness experiment with 20 image pairs (Fig 1A). Sur- prisingly, participants differed widely (by over 2-fold) in their success with detecting changes. To understand the reason for these striking differences in performance, first, we analyzed participants’ eye movement data, acquired at high spatial- and temporal- resolution, as they scanned each pair of images. We discovered that two key gaze metrics–mean fixation duration and the variance in the amplitude of saccades–were consistently predictive of participants’ suc- cess. Next, we developed a model of overt visual search based on the Bayesian framework of sequential probability ratio testing [11–14] (SPRT), in which subjects decided the next, most probable location for making a saccade based on a posterior odds ratio test. In our SPRT model, we also incorporated biological constraints on stimulus encoding and transformation, based on well-known properties of the visual processing pathway [15,16] (e.g. bounded firing rates, Poisson variance, foveal magnification, and saliency computation). Our neurally-constrained model mimicked key aspects of human gaze strategies in the change blindness task: model success rates were strongly correlated with human success rates, across the cohort of images tested. In addition, the model exhibited systematic variation in change detection success with fixation duration and saccade amplitude, in a manner closely resembling human data. Finally, the model outperformed a state-of-the-art deep neural net- work (DeepGaze II [17]) in predicting probabilistic patterns in human saccades in this change blindness task. We propose our model as a benchmark for mechanistic simulations of visual search, and for modeling human observer strategies during change detection tasks. Results Fixation and saccade metrics predict change detection success 39 participants performed a change blindness task (Fig 1A). Each experimental session con- sisted of a sequence of trials with a different pair of images tested on each trial. Images PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 2 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task Fig 1. Gaze metrics predict success in a change blindness experiment. A. Schematic of a change blindness experiment trial, comprising a sequence of alternating images (A, A’), displayed for 250 ms each, with intervening blank frames (B) also displayed for 250 ms (“flicker” paradigm), repeated for 60 s. Red circle: Location of change (not actually shown in the experiment). All 20 change image pairs tested are available in Data Availability link. B. Distribution of success rates of n = 39 participants in the change blindness experiment. Red and blue bars: good performers (top 30th percentile; n = 9) and poor performers (bottom 30th percentile; n = 12), respectively. Inverted triangles: Cut-off values of success rates for classifying good (red) versus poor (blue) performers. C. Classification accuracy, quantified with area- under-the-curve (AUC), for classifying trials as hits versus misses (left horizontal line) and performers as good versus poor (right horizontal line), obtained with a support vector machine classifier. Violin plots: Null distributions of classification accuracies based on a permutation test (�� p<0.001). Error bars: Clopper-Pearson binomial confidence intervals. D. Feature selection measures for identifying the most informative features that distinguish good from poor performers. From top to bottom: Fisher score, Information gain, Change in area-under-the-curve (AUC) and bag of decision trees (for details, see Feature Selection Metrics in the Materials and Methods). Brighter colors indicate more informative features. Solid red outline: most informative feature in the fixation feature subgroup (left); dashed red outline: most informative feature in the saccade feature subgroup (right). FD—fixation duration, SA—saccade amplitude, SD—saccade duration, SPS—saccade peak speed. μ and σ2 denote mean and variance of the respective parameter. E. Distribution of mean fixation duration (μFD, in milliseconds) across 19 change images for good performers (x- axis) versus poor performers (y-axis); one change image pair, successfully detected by all performers, was not included in these analyses (see text). Each data point denotes average value of μFD, across each category of performers, for each image tested. Dashed diagonal line: line of equality. p-value corresponds to significant difference in mean fixation duration between good and poor performers. F. Same as in E, but comparing variance of saccade amplitudes (in squared degrees of visual angle) for good versus poor performers. Other conventions are the same as in panel E. https://doi.org/10.1371/journal.pcbi.1009322.g001 presented included cluttered, indoor or outdoor scenes (see Data Availability link). To ensure uniformity of gaze origin across participants, each trial began when subjects fixated continu- ously on a central cross for 3 seconds. This was followed by the presentation of the change blindness image pair: alternating frames of two images, separated by intervening blank frames (250 ms each, Fig 1A). Of the image pairs tested, 20 were “change” image pairs, in that these differed from each other in one of three key respects (S1 Table): (i) size of an object changing; (ii) color of an object changing or (iii) change involving the appearance (or disappearance) of PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 3 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task an object. The remaining (either 6 or 7 pairs; Materials and Methods) were “catch” image pairs, which comprised an identical pair of images; data from these “catch” trials were not ana- lyzed for this study (Materials and Methods; complete change image set in Data Availability link). Change- and catch- image pairs were interleaved and tested in the same pseudorandom order across subjects. Subjects were permitted to freely scan the images to detect the change, for up to a maximum of 60 seconds per image pair. They indicated having detected the change by foveating at the location of change for at least 3 seconds. A response was marked as a “hit” if the subject was able to successfully detect the change within 60 seconds, and was marked as a “miss” otherwise. We observed that participants varied widely in their success with detecting changes: success rates varied over two-fold–from 45% to 90%–across participants (Fig 1B). These differences may arise from innate differences in individual capacities for change detection as well as other experimental factors (see Discussion). Nonetheless, we tested if individual-specific gaze strate- gies when scanning the images could explain these variations in change detection success. First, we ranked subjects in order of their change detection success rates. Subjects in the top 30th (n = 9) and bottom 30th (n = 12) percentiles were labelled as "good" and "poor" perform- ers, respectively (Fig 1B). This choice of labeling ensured robust differences in performance between the two classes: change detection success for good performers varied between 75% and 90%, whereas that for and poor performers varied between 45% and 61%. Nevertheless, the results reported subsequently were robust to these cut-offs for selecting good and poor per- formers (see S1 Fig for results based on performance median split). Next, we selected four gaze metrics from the eye-tracker: saccade amplitude, fixation duration, saccade duration and sac- cade peak speed (justification in the Materials and Methods) and computed the mean and the variance of these four metrics for each subject and trial. These eight quantities were employed as features in a classifier based on support vector machines (SVM) to distinguish good from poor performers (Materials and Methods). One image pair (#20), for which all participants correctly detected the change, was excluded for these analyses (Figs 1–3). Classification accuracy (area-under-the-curve/AUC) for distinguishing good from poor performers was 79.9% and significantly above chance (Fig 1C, p<0.001, permutation test, Materials and Methods). We repeated these same analyses, but this time classifying each trial as a hit or miss. Classification accuracy was 77.7% and, again, significantly above chance (Fig 1C, p<0.001). Taken together, these results indicate that fixation- and saccade- related gaze metrics contained sufficient information to accurately classify change detection success. Next, we identified gaze metrics that were the most informative for classifying good versus poor performers. This analysis was done separately for the fixation and saccade metric subsets: these were strongly correlated within each subset and uncorrelated across subsets (S2A Fig). For each metric, we performed feature selection with four approaches–Fisher score [18], AUC change [19] and Information Gain [20] and bag of decision trees (OOB error) [21]. A higher value of each selection measure reflects a greater importance of the corresponding gaze metric for classifying between good and poor performers. Among fixation metrics, mean fixation duration was assigned higher importance based on three out of the four feature selection mea- sures (Fig 1D, solid red outline). Among the saccade metrics, variance of saccade amplitudes was assigned highest importance, based on all four feature selection measures (Fig 1D, dashed red outline). We confirmed these results post hoc: mean fixation duration was significantly higher for good performers, across images (Fig 1E; p = 0.0015, Wilcoxon signed rank test), whereas variance of saccade amplitude was significantly higher for poor performers (Fig 1F; p<0.001, Wilcoxon signed rank test). We considered the possibility that the differences in fixation duration and saccade ampli- tude variance between good and poor performers could arise from differences in multiple, PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 4 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task distinct modes of these, respective distributions. Nonetheless, statistical tests provided no sig- nificant evidence for multimodality in either fixation duration or saccade amplitude distribu- tions for either class of performers (S2B Fig) (Hartigan’s dip test for unimodality; fixation duration: p>0.05, in 8/9 good performers with median p = 0.74, and in 8/12 poor performers with median p = 0.31; saccade amplitudes: p>0.05, in 9/9 good performers with median p = 0.99 and in 10/12 poor performers with median p = 0.99). In sum, these results indicate that two key gaze metrics–mean fixation duration and vari- ance of saccade amplitude–were strong and sufficient predictors of change detection success in a change blindness experiment. Next, we tested if more complex features of eye movements–such as scan paths, fixation maps or fixated object features–differed systematically between good and poor performers. Scan path data is challenging to compare across individuals because scan paths can vary in terms of both the number and sequence of image locations samples. We compared scan paths across participants by encoding them into a “string” sequences (Materials and Methods). Briefly, fixation points for each image were clustered, with data pooled across subjects, and individual subjects’ scan paths were encoded as strings based on the sequence of clusters vis- ited across successive fixations (Fig 2A and 2B). We then quantified the deviation between scan paths for each pair of subjects using the edit distance [22]. Median scan path edit dis- tances were not significantly different between good and poor performer pairs (Fig 2C, p = 0.14, Wilcoxon signed rank test). We also tested if the median inter-category edit distance between the good and poor performer categories would be higher than the median intra-cate- gory edit distance among the individual (good or poor) performer categories (Fig 2D). These edit distances were also not significantly different (p>0.1, one-tailed signed rank test). Second, we asked if fixation “maps”–two-dimensional density maps of the distribution of fixations [23]–were different across good and poor performers. For each image, we correlated fixation maps across every pair of participants (Materials and Methods). Again, we observed no significant differences between fixation map correlations between good- and poor- per- former pairs (Fig 2E, p = 0.29, Wilcoxon signed rank test), nor significant differences between intra-category (good vs. good and poor vs. poor) fixation map correlations and inter-category (good vs. poor) correlations (Fig 2F, p>0.1, one-tailed signed rank test). Third, we asked if overall statistics of saccades were different across good and poor per- formers. For this, we computed the probabilities of saccades between specific fixation clusters (“domains”), ordered by the most to least fixated locations on each image (Materials and Methods). The saccade probability matrix, estimated by pooling scan paths across each cate- gory of participants, is shown in Fig 3A (average across n = 19 image pairs). Visual inspection of the saccade probability matrices revealed no apparent differences between the good and poor performers (difference in S3A Fig). In addition, we tested if we could classify between good and poor performers based on individual subjects’ saccade probability matrices. Classifi- cation accuracy with an SVM based on saccade probability matrix features (~56.67%, Fig 3B) was not significantly different from chance (p>0.1, permutation test). Fourth, we tested whether good and poor performers differed in terms of fixated image fea- tures, as estimated with principal components analysis (Fig 3C, Materials and Methods). These fixated features typically comprised horizontal or vertical edges at various spatial frequencies, and were virtually identical between good and poor performers (Fig 3D, first six principal fea- tures for each class). We observed significant correlations across components of identical rank between good and poor performers (median r = 0.22, p<0.001, across top n = 150 components that explained ~80% of the variance). Similar correlations were obtained with fixated features obtained with the saliency map [24] (median r = 0.20, p<0.001, S3B Fig). PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 5 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task Fig 2. Scan paths and fixation maps do not distinguish good from poor performers. A. (Left) Representative image used in the change blindness experiment (Image #6 in Data Availability link). (Right) Clustering of the fixation points based on the peak of the fitted BIC (n = 13) profile. Fixation points in different clusters are plotted in different colors. Black fixations occurred in fixation sparse regions that were not included in the clustering. Black arrows show a representative scan path–a sequence of fixation points. The character “string” representation of this scan path is denoted on the right side of the image. B. Variation in the Bayesian Information Criterion (BIC; y-axis) with clustering fixation points into different numbers of clusters (x-axis; Materials and Methods). Circles: Data points. Gray curve: Bi-exponential fit. C. Distribution of edit distances among good performers (x-axis) versus edit distances among poor performers (y-axis). Each data point denotes median edit distance for each image tested (n = 19). Other conventions are the same as in Fig 1E. D. Distribution of intra-category edit distance (y-axis), among the good or among the poor performers, versus the inter-category edit distance (x-axis), across good and poor performers. Red and blue data: intra-category edit distance for good and poor performers respectively. Each data point denotes the median for each image tested (n = 19). Other conventions are the same as in panel C. E. Same as panel C, but comparing Pearson correlations of fixation maps among good (x-axis) and poor performers (y-axis). Other conventions are the same as in panel C. F. Same as in panel D, but comparing intra- versus inter-category Pearson correlations of fixation maps. Other conventions are the same as in panel D. G. Distribution of time to first fixation within the region of change (in seconds) for good performers (x-axis) versus poor performers (y-axis). Other conventions are the same as in panel C. H. Same as in E, but comparing time to detect change (in seconds) for good versus poor performers. Other conventions are the same as in panel G. https://doi.org/10.1371/journal.pcbi.1009322.g002 Fifth, we tested whether good and poor performers differed systematically in the spatial dis- tributions of fixations relative to the change location, before change was detected. For this, we computed the frequency of fixations and the total fixation duration, based on the distance of fixation relative to the center of the change location (binned in concentric circular windows of increasing radii, in steps of 50 pixels, Materials and Methods). We observed no systematic dif- ferences in the distributions of either total fixation duration, or frequency of fixations, relative to the change location between good and poor performers (S4 Fig; p = 0.99 for fixation dura- tion, p = 0.97 for fixation frequency, Kolmogorov-Smirnov test). In other words, the spatial distribution of fixations, relative to the change location, was similar between good and poor performers. Finally, we tested whether good and poor performers differed in the time to first fixation on the region of change, or the time to detect changes (on successful trials). Again, we observed no significant differences in the distributions of either time to first fixation, or time to detect changes, between good and poor performers (Fig 2G and 2H; p = 0.08 for time to first fixation PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 6 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task Fig 3. Saccade probabilities and fixated features are similar across good and poor performers. A. Average saccade probability matrices for the good performers (top; red outline) and poor performers (bottom; blue outline). These correspond to probabilities of making a saccade between different “domains” (1–4), each corresponding to a (non-contiguous) collection of image regions, ordered by frequency of fixations: most fixated regions (domain 1) to least fixated regions (domain 4). Cell (I, j) (row, column) of each matrix indicates the probability of saccades from domain j to domain i. B. Classification accuracy for classifying good versus poor performers based on the saccade probability matrix features, using a support vector machine classifier. Other conventions are as in Fig 1D. Error bars: s.e.m. C. Identifying low-level fixated features across good and poor performers. 112x112 image patches were extracted, centered around each fixation, for each participant; each point in the 112x112 dimensional space represents one such image patch. Principal component analysis (PCA) was performed to identify low-level spatial features explaining maximum variance among the fixated image patches, separately for good and poor performers. D. Top 6 principal components, ranked by proportion of variance explained, corresponding to spatial features explaining greatest variance explained across fixations, for good performers (left panels) and poor performers (right panels). These spatial features were highly correlated across good and poor performers (median r = 0.20, p<0.001, across n = 150 components). https://doi.org/10.1371/journal.pcbi.1009322.g003 in change region, p = 0.28 for time to detect change, signrank test). Taken together with the previous analysis, these results indicate that poor performers fixated as often and as close to areas near the change, but simply failed to detect these changes successfully. Overall, these analyses indicate that relatively simple gaze metrics like fixation durations and saccade amplitudes predicted successful change detection. More complex metrics like scan paths, fixated image features or the spatial distribution of fixations, were not useful indi- cators of change detection success. In other words, “low-level” gaze metrics, rather than “high- level” scanning strategies, determined participants’ success with change detection. A neurally-constrained model of eye movements for change detection We developed a neurally-constrained model of change detection to explain these empirical trends in the data. Briefly, our model employs the Bayesian framework of Sequential Probabil- ity Ratio Testing (SPRT) framework [14,15] to simulate rational observer strategies when per- forming the change blindness task. We incorporated key neural constraints, based on known properties of stimulus encoding in the visual processing pathway, into the model. For ease of understanding we summarize key steps in our model’s saccade generation pipeline (Fig 4A and 4B), first; a detailed description is provided thereafter. In the model, distinct neural populations, with (noisy) Poisson firing statistics, encode the saliency of the foveally-magnified image at each region. During fixation, following each alter- nation (Fig 1A, either A followed by A’, or vice versa) the model computes a posterior odds PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 7 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task Fig 4. A Bayesian model of gaze strategies for change detection. A. Schematic showing a typical fixation across the pair of images (A, A’) and an intervening blank. B. Detailed steps for modeling change detection (see text for details). (Clockwise from top left) At each fixation, a Cartesian variable resolution (CVR) transform is applied to mimic foveal magnification, followed by a saliency map computation to determine firing rates at each location. Instantaneous evidence for change versus no change (log-likelihood ratio, log L(t)) is computed across all regions of the image. An inverse CVR transform is applied to project the evidence back into the original image space, where noisy evidence is accumulated, (sequential probability ratio test, drift-diffusion model). The next fixation point is chosen using a softmax function applied over the accumulated evidence (Et). To model human saccadic biases, a distribution of saccade amplitudes and turn angles is imposed on the evidence values prior to selecting the next fixation location (polar plot inset). C. A representative gaze scan path following model simulation (cyan arrows). Colored squares: specific points of fixation (see panel D). Grid: Fine divisions over which the image was sub-divided to facilitate evidence computation. Green (1), blue (2) and red (3) squares denote first (beginning of simulation), intermediate (during simulation) and last (change detection) fixation points, respectively. D. Evidence accumulated as a function of time for the same three representative regions as in panel D; each color and number denotes evidence at the corresponding square in panel C. When the model fixated on the green or blue squares (in panel C), the accumulated evidence did not cross the threshold for change detection. As a result, the model continued to scan the image. When the model fixated on the red square (in the change region), the accumulated evidence crossed threshold (horizontal, dashed gray line) and the change was detected. https://doi.org/10.1371/journal.pcbi.1009322.g004 ratio for change versus no change at each region and at each instant of time (Eq 1), and accu- mulates this ratio as “evidence” (Eq 2, Results). If the accumulated evidence exceeds a prede- termined (positive) threshold for change detection at the location of fixation, the model is deemed to have detected the change. If, the accumulated evidence dips below a predetermined (negative) threshold for “no change” at the fixated location, the observer terminates the cur- rent fixation. The next fixation location is chosen based on a stochastic (softmax) decision rule (Eq 3), with the probability of saccade to a region being proportional to the accumulated evi- dence at that region. Note that both images—odd and even—must be included in these com- putations to generate each saccade. The model continues scanning over the images in the sequence until either the change is detected or until the trial duration has elapsed (as in our experiment), whichever occurs earlier. Neural representation of the image pair. At the onset of each fixation, the image was magnified foveally based on the center of fixation [25], with the Cartesian Variable Resolution (CVR) transform [26] (Materials and Methods; S5 Fig). Next, a saliency map was computed with the frequency-tuned salient region detection method [24] for each image of each pair. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 8 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task Table 1. Model parameters and their default values. Parameters Time bin Image duration Trial duration Temperature Decay factor Decay scale Noise scale Prior odds ratio Change threshold “No change” threshold Threshold decay Symbol Δt τ – T γ β W P Fc Fn z Foveal magnification factor FMF Value 25 ms 10Δt 60 s 0.01 0.004 Description Unit of time for the model Duration for which each image or blank is shown Total duration of trial Modulates stochasticity of next saccade Decay of the evidence with time (inversely related) 4.0 grid units Spatial range of evidence decay U(-5, 5) Models noise in evidence accumulation 0.1 100 -20 0 0.05 Prior odds of change to no change Threshold to determine change Threshold to determine “no change”. Decay rate of no-change threshold Magnification of the fixated region on the fovea according to the CVR transform Firing rate bounds Firing rate prior λmin, λmax μf 5, 120 spikes/bin Minimum and maximum firing rates 3 spikes/bin Expected difference in firing rates https://doi.org/10.1371/journal.pcbi.1009322.t001 Saliency computation was performed on the foveally-magnified image, rather than on the orig- inal image, to mimic the sequence of these two computations in the visual pathway; we denote these foveally-transformed saliency values as S and S’, for each image (A) and its altered ver- sion (A’), respectively. Each image was partitioned into a uniform 72x54 grid of equally-sized regions. We index each region in each image pair as A1, A2 . . ., AN and A’1, A’2. . ., A’N, respectively (N = 3888). Distinct, non-overlapping, neural populations encoded the saliency value (Si, Si’) in each region of each image. While in the brain, neural receptive fields typically overlap, we did not model this overlap here, for reasons of computational efficiency (Materials and Methods). The firing rates for each neural population were generated from independent Poisson processes. The average firing rate for each region λi was modeled as a linear function of the average saliency of image patch falling within that region as: liðSiÞ ¼ lmin þ ðlmax (cid:0) lminÞhSk i ik, where i is the saliency value of the kth pixel in region Ai, and the angle brackets denote an average Sk across all pixels in that region. In other words, when the change between images A and A’ occurred in region i, the difference in firing rates between λi and λ’i was proportional to the difference in saliency values across the change. We modeled each change detection trial (total duration T, Table 1), as comprising of a large number of time bins of equal duration (Δt, Table 1). At every time bin, the number of spikes from each neural population was drawn a Poisson distribution whose mean was determined by the average saliency of all pixels within the region. At the end of each fixation, the model either indicated its detection of change, thereby terminating the simulation, or shifted gaze to a new location. The precise criteria for signaling change versus shifting gaze are described next. For ease of description, we depict a typical fixation in Fig 4A. The first image of the pair (say, A) persists m time-bins from the onset of the current fixation. Next, a blank epoch occurs from m+1 to p time-bins. Following this, the second image of the pair (A’) appears for an inter- val from p+1 to n time bins, until the end of fixation. We denote the number of spikes pro- t. Xi and Yi represent the total number of spikes duced by neural population i at time t by wi produced by neural population i when fixating at the first and second images respectively, dur- tÞ; Y i ¼ Sn ing the current fixation. Thus, Xi ¼ Sm tÞ. We denote the number of spikes in the blank period as Bi ¼ Sp t ¼ 0: For simplicity, we assume that no spikes t¼1ðwi t¼mþ1wi t¼pþ1ðwi PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 9 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task occurred during the blank period (Bi = 0), although this is not a strict requirement, as the key model computations rely on relative rather than absolute firing rates. In sum, the observer must perform change detection with a noisy neural representation derived from saliency map of the foveally-magnified image. Modeling change detection with an SPRT rule. The observer faces two key challenges with change detection in this change blindness task. First, were the images not interrupted by a blank, a simple pixel-wise difference of firing rates over successive time epochs would suffice to localize the change. For example, computing \|hXii−hYii\| (where \|x\| denotes the absolute value of x, and angle brackets denote average over many time bins), and testing if this differ- ence is greater than zero at any region i, suffices to identify the location of change. On the other hand, such an operation does not suffice when images are interleaved with a blank, as in change blindness tasks. For example, a pixel-wise subtraction of each image from the blank (\|hXii−hBii\| or \|hYii−hBii\|) yields large values at all locations of the image. Therefore, when images are interrupted by a blank, information about the first image must be maintained across the blank interval and compared with second image following the blank, for detecting the change. Second, even if no blank occurred between the images, a pixel-wise differencing operation would not suffice, due to the stochasticity of the neural representation: a non-zero difference in the number of spikes from a particular region, i (\|Xi−Yi\|) is not direct evidence of change at that location. In other words, the observer’s strategy for this change blindness task must take into account both the occurrence of the blank between the two images, as well as the, stochasticity in the Poisson neural representation of the image, for successfully detecting changes. To address both of these challenges, we adopt an SPRT-based search rule. First, we compute the difference in the number of spikes between the first and second image at each region Ai in the image. We denote the random variable indicating this difference by Zi = Xi−Yi, and its value at end of time bin t as z. We then compute a likelihood ratio for change (C) versus no change (N), as: ð Li t; z Þ ¼ pðZiðtÞ ¼ zjCÞ pðZiðtÞ ¼ zjNÞ ð1Þ Specifically, the observer tests if the observed value of Zi was more likely to arise from two generating processes (Change, C), or could from a single, underlying generating process (No Change, N). This computation is performed at each time step following the onset of the second image (t>p) of each pair. Details of computing this likelihood ratio for Poisson processes are provided in the Materials and Methods; for our model this computation involves an infinite sum, which we calculate using Bessel functions and efficient analytic approximations [27]. The functional form of the log-likelihood ratio resembles a piecewise linear function of firing rate differences (S6A and S6B Fig, see next section), which can be readily achieved by the output of simple neural circuits [28–30]. Second, the observer integrates the “evidence” for change at location Ai, by accumulating the logarithm of the likelihood ratio log(Li(t)), along with the log of the prior odds ratio (Pi), as in the SPRT framework. EiðtÞ ¼ ð1 (cid:0) giðtÞÞEiðt (cid:0) 1Þ þ logðLiðtÞÞ þ logðPiÞ þ WiðtÞ ð2Þ where γi 2[0,1] is a decay parameter for evidence accumulation at location Ai, which simulates “leaky” evidence accumulation [15,31] with larger values of γi, indicating greater “leak” in evi- dence accumulation, Pi is the prior odds ratio of change to no change (P(C)/P(N)) at each loca- tion, Wi(t) represents white noise, sampled from a uniform distribution (Table 1), to mimic PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 10 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task noisy evidence accumulation [32]. Here we assume that the prior ratio is constant across time and space, but nonetheless study the effect of varying prior ratios on model performance (next section). Both of these features–leak and noise in evidence accumulation–are routinely incor- porated in models of human decision-making [31], and are grounded in experimental obser- vations in brain regions implicated in decision-making [15]. Evidence accumulation occurs in the original, physical space of the image, and not in the CVR transformed space (Fig 4B). Note that this formulation of an SPRT decision involves evaluating and integrating fully the loga- rithm of the Bayesian posterior odds ratio (product of the prior odds and likelihood ratio, Pi × Li(t)). Evidence accumulation is performed for each region in the image; Ei(t) for each region is calculated independently of the other regions. If the accumulated evidence Ei(t) crosses a posi- tive threshold, Fc (Table 1), the observer stops scanning the image and region Ai, at which the threshold Fc was crossed, is declared the “change region”. If, on the other hand, the accumu- lated evidence crosses a negative (no-change) threshold Fn (Table 1), the observer terminates the current fixation and determines the next region to fixate, Ak, based on a softmax probabil- ity function: pk ¼ e Ek T =SN i¼1e Ei T ð3Þ where Ei is the evidence value for region i, N is the number of regions in the image, and T is a temperature parameter which controls the stochasticity of the saccade (decision) policy (Mate- rials and Methods; see also next section). For selecting the next point of gaze fixation, we also matched directional saccadic biases typically observed in human data [33] (Fig 4B, described in Materials and Methods section on “Comparison of model performance with human data”). In some simulations we also decayed the no-change threshold (Fn) with different decay rates (z; Table 1) and studied its effect on model performance. Because we observed virtually no false alarms (signaling a no-change location as change) in our experimental data (0.06% of all trials; Materials and Methods) we did not model decay in the change threshold (Fc), which would have yielded significantly more false alarms. Note that although we have not explicitly modeled inhibition-of-return (IOR), this feature emerges naturally from the evidence accumulation rule in the model. Following each fixation, the accumulated evidence for no-change decays gradually (Eq 2), thereby reducing the proba- bility that subsequent fixations occur, immediately, at the erstwhile fixated location. This fea- ture encourages the model to explore the image more thoroughly. We illustrate gaze shifts by the model in an exemplar change blindness trial (Fig 4C and 4D). The model’s scan path is indicated by cyan arrows showing a sequence of fixations, ultimately terminating at the change region. When the model fixated, initially, on regions with no change (Fig 4C, squares with green/1 or blue/2 outline), transient evidence accumulation occurred either favoring a change (positive fluctuations) or favoring no change (negative fluctuations) (Fig 4D, green and blue traces, respectively). In each case, evidence decayed to baseline values rapidly during the blank epochs, when no new evidence was available, and the accumulated evidence did not cross threshold. Finally, when the model fixated on the change region (Fig 4C, square with red out- line/3), evidence for a change continued to accumulate, until a threshold-crossing occurred (Fig 4D, red trace, threshold: dashed gray line). At this point, the change was deemed to have been detected, and the simulation was terminated. Model trends resemble qualitative trends in human experimental data We tested the effect of key model parameters on change detection performance, to test for qualitative matches with our experimental findings. We simulated the model and measured PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 11 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task Fig 5. Effect of model parameters on change detection success. A. Change in model performance (success rates, % correct) with varying the relative interval of the images and blanks, measured in units of time bins (Δt = 25 ms/time bin; Table 1), while keeping the total image+blank interval constant (at 50 time bins). Positive values on the x-axis denote larger image intervals, as compared to blanks, and vice versa, for negative values. Blue points: Data; gray curve: sigmoid fit. B. Same as in panel (A), but with varying the maximum decay factor (γ; Eq 2). Curves: Sigmoid fits. C. Same as in panel (A) but with varying the firing rate prior (μf) for image pairs with the lowest (blue; bottom 33rd percentile) and highest (red; top 33rd percentile) magnitudes of firing rate changes. Curves: Smoothing spline fits. Colored squares: μf corresponding to the center of area of the two curves. D. Same as in panel (A), but with varying the mean fixation duration (μFD; measured in time bins, Δt = 25 ms/ time bin). (Inset; lower) Variation of μFD with prior ratio of change to no change (P(C:NC)). (Inset; upper) Same as lower inset but with varying threshold decay rate z (Table 1). E. Same as in panel (A), but with varying saccade amplitude variance (σ2 SA with the softmax function temperature parameter (T) (see text for details). F. Same as in panel (A), but with varying saccade amplitude variance (σ2 SA with the foveal magnification factor (FMF). Other conventions in B-F are the same as in panel A. Error bars (all panels): s.e.m. SA). (Inset) Variation of σ2 SA). (Inset) Variation of σ2 https://doi.org/10.1371/journal.pcbi.1009322.g005 change detection performance by varying each model parameter in turn (Table 1, default val- ues), while keeping all other parameters fixed at their default values. For these simulations we employed the frequency-tuned salient region detection method [24] to generate the saliency map. The first three simulations (Fig 5A–5C) tested whether the model performed as expected based on its inherent constraints. The last three simulations (Fig 5D–5F) evaluated whether emergent trends in the model matched empirical observations regarding gaze metrics in our study (Fig 1E and 1F). The results reported represent averages over 5–10 repetitions of each simulation. First, we tested the effect of varying the relative durations of the image and the blank, while keeping their overall presentation duration (image+blank) constant. Note that no new evi- dence accrues during the blanks, whereas decay of accumulated evidence continues. Therefore, extending the duration of the blanks, relative to the image, should cause a substantial deterio- ration in the performance of the model. The simulations confirmed this hypothesis: PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 12 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task performance deteriorated (or improved) systematically with decreasing (or increasing) dura- tions of the image relative to the blank (Fig 5A). Second, we tested the effect of varying the magnitude of the decay factor (γ, Table 1). Decreasing γ prolongs the (iconic) memory for evidence relevant to change detection; γ = 1 represents no memory (immediate decay; no integration) of past evidence, whereas γ = 0 indi- cates reliable memory (zero decay; perfect integration) of past evidence (refer Eq 2). Model success rates were at around 80% for γ = 0 and performance degraded systematically with increasing γ (Fig 5B); in fact, the model was completely unable to detect change for γ values greater than around 0.2, suggesting the importance of the transient memory of the image across the blank for successful change detection. Third, we tested the effect of varying μf, the prior on the magnitude of the difference between the firing rates (across the image pair) in the change region (Fig 5C). For this, we divided images into two extreme subsets (highest and lowest 1/3rd), based on a tercile (three- way) split of firing rate magnitude differences. The performance curve for the highest tercile (largest firing rate differences in change region) of images was displaced rightward relative to the performance curve for the lowest tercile (smallest firing rate differences). Specifically, μf corresponding to the center of area of the performance curves was systematically higher for the images with higher firing rate differences (Fig 5C, colored squares). Fourth, we tested the effect of varying mean fixation duration (μFD)–a key parameter identi- fied in this study as being predictive of success with change detection. The mean fixation dura- tion is not a parameter of the model. We, therefore, varied the mean fixation duration, indirectly, by varying the prior odds ratio (P) and the decay rate (z) of the no-change threshold (Fn). A lower prior odds ratio of change to no-change biases evidence accumulation toward the no-change threshold, leading to shorter fixations (and vice versa; Fig 5D, lower inset). On the other hand, a higher decay rate of the no-change threshold leads to a greater probability of bound crossing of the evidence in the negative direction, again leading to shorter fixations (and vice versa; Fig 5D; upper inset). In either case, we found that decreasing (increasing) the mean fixation duration produced systematic deterioration (improvement) in the performance of the model (Fig 5D). These results recapitulate trends in the human data, indicating that increased fixation duration may be a key gaze metric indicating change detection success. SD)–the other key parameter we had identified as being predictive of change detection success. Again, because the variance of the saccade amplitude is not a parameter of the model, we varied this, indi- rectly, by varying the temperature (T) parameter in the softmax function: a higher temperature value leads to random sampling from many regions of the image, thereby increasing s2 SD whereas a low temperature value leads to more deterministic sampling, thereby reducing s2 SD (Fig 5E, inset). With increasing saccade variance, performance dropped steeply (Fig 5E). Fifth, we tested the effect of varying the saccade amplitude variance (s2 Finally, we also explored the effect of varying the foveal magnification factor (FMF) across a two-fold range. Saccade amplitude variance decreased robustly as the FMF increases (Fig 5F, inset) (see Discussion). As with the previous simulation, we observed a systematic decrease in performance with increasing saccade amplitude variance (Fig 5F), again, recapitulating trends in the human data. Taken together, these results show that gaze metrics that were indicative of change detec- tion success in the change blindness experiment also systematically influenced change detec- tion performance in the model. Specifically, the two key metrics indicative of change detection success in humans, namely, fixation duration and variance of saccade amplitude, were also predictive of change detection success in the model. These effects could be explained by chang- ing specific, latent parameters in the model (e.g. decay rate of the no-change threshold, prior PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 13 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task ratios, foveal magnification factors). Our model, therefore, provides putative mechanistic links between specific gaze metrics and change detection success in the change blindness task. Model performance mimics human performance quantitatively In addition to these qualitative trends, we sought to quantify similarities between model and human performance in this change blindness task. For this analysis, we modeled biases inher- ent in human saccade data (S7 Fig) by matching key saccade metrics in the model–amplitude and turn angle of saccades–with human data (Figs 6A and 7A, r = 0.822, p<0.01; see Materials and Methods section on “Comparison of model performance with human data”). For these simulations, and subsequent comparisons with a state-of-the-art deep neural network model (DeepGaze II) [17] we used the saliency map generated by the DeepGaze network rather than Fig 6. Comparison between human and model performance. A. (Left) Joint distribution of saccade amplitude and saccade turn angle for human participants (averaged over n = 39 participants). Colorbar: Hotter colors denote higher proportions. (Right) Same as in the left panel, but for model, averaged over n = 40 simulations. B. Correlation between change detection success rates for human participants (x-axis) and the model (y-axis). Each point denotes average success rates for each of the 20 images tested, across n = 39 participants (human) or n = 40 iterations (model). Error bars denote standard error of the mean across participants (x-axis) or simulations (y-axis). Dashed gray line: line of equality. C. Average absolute deviation from human performance of the sequential probability ratio test (SPRT) model (Model, leftmost bar), for a control model in which evidence decayed rapidly (Control 1, γ = 1; second bar from left), for a control model in which the stopping rule was based on the derivative of the posterior odds ratio (Control 2; third bar from left), or for a control model which employed a random search strategy (Control 3, T = 104; rightmost bar). p-values denote significance levels following a paired signed rank test, across n = 20 images (�p < 0.05). https://doi.org/10.1371/journal.pcbi.1009322.g006 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 14 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task Fig 7. Comparison between human, model and Deep Gaze II performance. A. Distribution of saccade amplitudes for human participants (yellow), sequential probability ratio test (SPRT) model (red) and the Deep Gaze II neural network (blue). B. Top 10 clusters of human fixations, ranked by cumulative fixation duration (rows/columns 1–10). Increasing indices correspond to progressively lower cumulative fixation duration. C. Saccade probability matrix (left) averaged across all images and all participants, (middle) for simulations of the sequential probability ratio test (SPRT) model, and (right) for the Deep Gaze II neural network. D. Distribution, across images, of the correlations (r-values) of saccade probability matrices between human participants and sequential probability ratio test (SPRT) model (left) and human participants and Deep Gaze II neural network (right). p-value indicates pairwise differences in these correlations across n = 20 images. https://doi.org/10.1371/journal.pcbi.1009322.g007 the frequency-tuned salient region detection algorithm, so as to enable a direct comparison between our model and DeepGaze. As a first quantitative comparison, we tested whether image pairs in which human observ- ers found difficult to detect changes (S2C Fig), were also challenging for the model. For this, we compared the model’s success rates across images with observers’ success rates in the change blindness experiment. Remarkably, the model’s success rates, averaged across 40 inde- pendent runs, correlated significantly with human observers’ average success rates (Fig 6B, r = 0.476, p = 0.034, robust correlations across n = 20 images). We compared the Bayesian SPRT search rule, as specified in our model, against three alter- native control models, each with a different search strategy or stopping rule: (i) a model in which evidence decayed rapidly, so that the decision to signal change was based on the instan- taneous posterior odds ratio alone; (ii) a model in which the stopping rule was based on cross- ing a threshold “rate of change” of the posterior odds ratio, and (iii) a model that employed a random search strategy (Materials and Methods). For each of these models, the average abso- lute difference in performance with the human data was significantly higher, compared with that of the original model (Fig 6C; p<0.05 for 2/3 control models; Wilcoxon signed-rank test). Moreover, none of the control model’s success rates correlated significantly with human observers’ success rates (r = 0.09–0.42, p>0.05, for all 3/3 control models; robust correlations). Finally, we tested whether model gaze patterns would match human gaze patterns beyond that achieved by state-of-the-art fixation prediction with a deep neural network: DeepGaze II PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 15 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task [17]. First, we quantified human gaze patterns by computing the probability of saccades pair- wise among the top 10 clusters with the largest number of fixations (e.g. Fig 7B) for each image. We then compared these human saccade probability matrices (Fig 7C, left) with those derived from simulating the model (Fig 7C, middle) as well as with those generated by the DeepGaze network (Fig 7C, right). For the latter, saccades were simulated using the same soft- max rule as employed in our model (Eq 3) along with inhibition-of-return [34] (Materials and Methods); in addition, for each image, we identically matched the distribution of fixation durations between DeepGaze and our model (Materials and Methods). The model’s saccade probability matrix (Fig 7C, middle) closely resembled the human sac- cade probability matrix (Fig 7C, left), indicating that model was able to mimic human saccades patterns closely. On the other hand, the DeepGaze saccade matrix (Fig 7C, right) deviated sig- nificantly from the human saccade probability matrix. Confirming these trends, we observed significantly higher correlations between the human saccade probabilities and our model’s sac- cade probabilities (Fig 7D, left) as compared to those with DeepGaze’s saccade probabilities (Fig 7D, right) (human-SPRT model: median r = 0.51, human-DeepGaze II: median r = 0.14; p<0.01 for significant difference in correlation values across n = 20 images, signed rank test). These results were robust to the underlying saliency map in our SPRT model: replacing Deep- Gaze’s saliency map with the frequency-tuned salient region detection method yielded nearly identical results (S8A and S8B Fig). The chief reason for these differences was readily apparent upon examining the saccade amplitude distributions across the human data, our SPRT model and DeepGaze: whereas the human and model distributions contained many short saccades, the DeepGaze distribution contained primarily long saccades (Fig 7A). Consequently, we repeated the comparison of sac- cade probabilities limiting ourselves to the range of saccade amplitudes in the DeepGaze model. Again, we found that our model’s saccade probabilities were better correlated with human saccade probabilities (S8C and S8D Fig) (human-SPRT model: median r = 0.29, human-DeepGaze II: median r = 0.10; p<0.001). We propose that these differences occurred because DeepGaze saccades are generated based on relative saliencies of different regions across the image, whereas saliency computation, per se, may be insufficient to model human saccade strategies in change blindness tasks or, in general, in change detection tasks. In summary, change detection success rates were robustly correlated between human par- ticipants and the model. Moreover, our model outperformed a state-of-the-art deep neural network in predicting gaze shifts among the most probable locations of human gaze fixations in this change blindness task. Discussion The phenomenon of change blindness reveals a remarkable property of the brain: despite the apparent richness of visual perception, the visual system encodes our world sparsely. Stimuli at locations to which attention is not explicitly directed are not effectively processed [4]. Even salient changes in the visual world sometimes fail to capture our attention and remain unde- tected. Visual attention, therefore, plays a critical role in deciding the nature and content of information that is encoded by the visual system. In a laboratory change blindness experiment, we observed that participants varied widely in their ability to detect changes. These differences cannot be directly attributed to participants’ inherent change detection abilities. Nevertheless, a recent study evaluated test-retest reliability in change blindness tasks, and found that observers’ change detection performance was rela- tive stable over periods of 1–4 weeks [35]. In our study, participants whose fixations lasted marginally longer, on average, and whose saccades were less spatially variable, were best able PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 16 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task to detect changes. Given the intricate link between mechanisms for directing eye-movements and those governing visual attention [36–38], our results suggest the hypothesis that spatial attention shifts more slowly in time (higher fixation durations), and less erratically in space (lower saccade variance), in order to enable participants to detect changes effectively. To explain our experimental observations mechanistically, we developed, from first princi- ples, a neurally-constrained model based on the Bayesian framework of sequential probability ratio testing [15,31]. Such SPRT evidence accumulation models have been widely employed in modelling human decisions [31], and also appear to have a neurobiological basis [15]. In our model, we incorporated various neural constraints including foveal magnification, saliency maps, Poisson statistics in neural firing and human saccade biases. Even with these constraints, the model was able to faithfully reproduce key trends in the human change detection data, both qualitatively and quantitatively (Figs 5 and 6). The model’s success rates correlated with human success rates, and the model reproduced key saccade patterns in human data, outper- forming competing control models (Figs 6B and 6C, 7C and S8). On the one hand, our study follows a rich literature on human gaze models, that fall, loosely, into two classes. The first class of “static” models use information in visual saliency maps [23,37,39] to predict gaze fixations. These saliency models, however, do not capture dynamic parameters of human eye fixations, which are important for understanding strategies underlying visual exploration in search tasks, like change blindness tasks. The second class of “dynamic” models seek to predict the temporal sequence of gaze shifts [40–43]. Nevertheless, these approaches were developed for free-viewing paradigms, and comparatively few studies have focused on gaze sequence prediction during search tasks [44,45]. On the other hand, sev- eral previous studies have developed algorithms to address the broader problem of “change point” detection [46–48]. Yet, none of these algorithms are neurally-constrained (e.g. foveal magnification, Poisson statistics), and none models gaze information or saccades. To the best of our knowledge, ours is the first neurally-constrained model for gaze strategies in change blindness tasks, and developing and validating such a model is a central goal of this study. Specifically, our model outperformed a state-of-the-art deep neural network (DeepGaze II), in terms of predicting saccade patterns in this change blindness task. Yet, a key difference must be noted when comparing our model with DeepGaze. Our model relies on a decision rule based on posterior odds for generating saccades: For this, it must compare evidence for change versus no change across the two images. In our simulations, in contrast, the DeepGaze model generates saccades independently on the two images, without comparing them. Based on these simulations, we found that our model’s gaze patterns provided a closer match to human data compared to gaze patterns from DeepGaze (Figs 7C and 7D and S8). Because DeepGaze is a model tailored for predicting free-viewing saccades, this comparison serves only to show that even a state-of-the-art free-viewing saliency prediction algorithm is not suffi- cient to accurately predict gaze patterns in the context of a change detection (or change blind- ness) task. In other words, saccades made with the goal of detecting changes are likely to be different from saccades made in free-viewing conditions. Our model exhibited several emergent behaviors that matched previous reports of human failures in change blindness tasks. First, the model’s success rates improved systematically as the blank interval was reduced (Fig 5A); this trend mimics previously-reported patterns in human change blindness tasks, in which shortening the interval of the intervening blank improves change detection performance [4]. Second, the model’s success rates improved sys- tematically with reducing the evidence decay rate across the blank (Fig 5B). In other words, retaining information across the blank was crucial to change detection success. This result may have intriguing links with neuroscience literature, which has shown that facilitating neu- ral activity in oculomotor brain regions (e.g. the superior colliculus) during the blank epoch PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 17 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task counters change blindness [49]. Third, the model’s ability to detect changes improved when its internal prior (expected firing rate difference) aligns with the actual firing rate difference at the change region (Fig 5C). These results may explain a results from a previous study [50], which found that familiarity with the context of the visual stimulus was predictive of change detection success. Finally, the model provided mechanistic insights about key trends observed in our own experiments, specifically, a critical dependence of success rates on mean fixation durations and the variance of saccade amplitudes (Fig 5D, 5E and 5F). Fixation durations in the model varied systematically either with altering the prior odds ratio or the decay rate of the no-change bound. Note that the prior odds ratio corresponds to an individual’s prior belief in the prior probability of change to no change. The lower this ratio, the higher the degree of belief in no change, and the sooner the individual seeks to break each fixation. In our model, this was achieved by having the prior ratio bias evidence accumulation toward the no-change (nega- tive) bound. Similarly, faster decay of the no-change bound, possibly reflecting a stronger “urgency” to break fixations, resulted in faster bound crossing and, therefore, shorter fixations. Regardless of the mechanism, shorter fixation durations resulted in impaired change detection performance (Fig 5E), providing a putative mechanistic link between fixation durations and change detection success in the experimental data. In addition, saccade amplitude variance modulated systematically with changes in the foveal magnification factor (FMF). With higher foveal magnification the model is, perhaps, able to better distinguish features in regions proxi- mal to the fixation location, and saccade to them, thereby resulting in overall shorter saccades, and lower variance. Moreover, the higher foveal magnification, enables analyzing the region of change with higher resolution, thereby leading to better change detection performance. As a consequence performance degraded systematically with increased saccade amplitude variance (Fig 5F), the common underlying cause for each being the change in the foveal magnification factor. This provides a plausible mechanism for higher variance of saccade amplitudes in “poor” performers. We implemented three control models in this study. The first control model—in which evi- dence decayed rapidly (γ = 1)—mimics the scenario of rapidly decaying short-term memory; this model signals the change based on threshold crossing of the instantaneous, rather than the accumulated, posterior odds. In the second control model, we employed an alternative stop- ping rule: a rapid, large change in the posterior odds ratio sufficed to signal the change. Such a “temporally local” stopping rule obviates the need for evidence accumulation (short-term memory) and may be implemented by neural circuits that act as temporal change detectors (differentiators). The third control model mimicked a random saccade strategy, with a high temperature parameter (T = 104) of the softmax function. This model establishes baseline (chance) levels of success, if an observer were to ignore model evidence and saccade randomly to different locations on each image, and arrive at the change region “by chance”. Each of these control models fell short of our SPRT model in terms of their match to human performance. Nonetheless, our SPRT model can be improved in a few ways. First, saliency maps in our model were typically computed with low-level features (e.g. Fig 5; the frequency tuned salient region method). Incorporating more advanced saliency computations (e.g. semantic saliency) [51] into the saliency map could render the model more biologically realistic. Second, although our neurally-constrained model provides several biologically plausible mechanisms for explaining our experimental observations, it does not identify which of these mechanisms is actually at play in human subjects. To achieve this objective, model parameters may be fit with maximum likelihood estimation [52] or Bayesian methods for sparse data (e.g. hierarchical Bayesian modelling)[53]. Yet, in its current form, such fitting is rendered challenging because PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 18 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task the model is not identifiable: multiple parameters in the model (e.g. prior ratios or decay of the no-change threshold) produce similar effects on specific gaze parameters (e.g. fixation dura- tions, Fig 5D). Future extensions to the model, for example, by measuring and modeling more gaze metrics for constraining the model, may help overcome this challenge. Such model-fitting will find key applications for identifying latent factors contributing to inter-individual differ- ences in change detection performance. Our simulations have interesting parallels with recent literature. With a battery of cognitive tasks Andermane et al. [35] identified two factors that were critical for predicting change detection success: “visual stability”–the ability to form stable and robust visual representation– and “visual ability”–indexing the ability to robustly maintain information in visual short term memory. Other studies have identified associated psychophysical factors, including attentional breadth [54] and visual memory [55] as being predictive of change detection success. We pro- pose that (higher) fixation durations and (lower) variability of saccade amplitudes may both index a (higher) “visual stability” factor, indexing the ability to form more stable visual repre- sentations. In contrast, the temporal decay factor (Table 1, γ) and spatial decay scale (Table 1, β) may correspond to visual memory and attentional breadth, respectively; each could com- prise key components of the “visual ability” factor, indexing robust maintenance of informa- tion in short-term memory. Our model provides a mechanistic test-bed to systematically explore the contribution of each of these factors and their constituent components to change detection success in change blindness experiments. A mechanistic understanding of the behavioral and neural processes underlying change blindness will have important real-world implications: from safe driving [56] to reliably verify- ing eyewitness testimony [57]. Moreover, emerging evidence suggests that change blindness (or a lack thereof) may be a diagnostic marker of neurodevelopmental disorders, like autism [58–60]. Our model characterizes gaze-linked mechanisms of change blindness in healthy individuals and may enable identifying the mechanistic bases of change detection deficits in individuals with neurocognitive disorders. Materials and methods Ethics statement Informed written consent was obtained from all participants. The study was approved by the Institutional Human Ethics Committee (IHEC) at the Indian Institute of Science (IISc), Bangalore. Experimental protocol We collected data from n = 44 participants (20 females; age range 18–55 yrs) with normal or corrected-to-normal vision and no known impairments of color vision. Of these, data from 4 participants, who were unable to complete the task due to fatigue or physical discomfort, were excluded. Data from one additional participant was irretrievably lost due to logistical errors. Thus, we analyzed data from 39 participants (18 females). Images were displayed on a 19-inch Dell monitor at 1024x768 resolution. Subjects were seated, with their chin and forehead resting on a chin rest, with eyes positioned roughly 60 cm from the screen. Each trial began when subjects continually fixated on a central cross for 3 sec- onds. This was followed by presentation of the change image pair sequence for 60s: each frame (image and blank) was 250 ms in duration. The trial persisted until the subjects indicated the change by fixating at the change region for at least 3 seconds continuously (“hit”), or if the maximum trial duration (60 s) elapsed and the subjects failed to detect the change (“miss”). An online algorithm tracked, in real-time, the location of the subjects’ gaze and signaled the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 19 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task completion of a trial based on whether they were able to fixate stably at the location of change. Each subject was tested on either 26 or 27 image pairs, of which 20 pairs differed in a key detail (available in Data Availability link); we call these the “change” image pairs. The remaining image pairs (7 pairs for 30 subjects and 6 pairs for 9 subjects) contained no changes (“catch” image pairs); data from these image pairs were not analyzed for this study (except for comput- ing false alarm rates, see next). To avoid biases in performance, the ratio of “change” to “catch” trials was not indicated to subjects beforehand, but subjects were made aware of the possibility of catch trials in the experiment. We employed a custom set of images, rather than a standard- ized set (e.g.[4]), due to the possibility that some subjects might have been familiar with change images used in earlier studies. Overall, the proportion of false-alarms–proportion of fixations with durations longer than 3s in catch trials–was negligible (~0.06%, 17/32248 fixations across 264 catch trials) in this experiment. To further confirm if the subjects indeed detected the change on hit trials, a post- session interview was conducted in which each subject was presented with one of each pair of change images in sequence and asked to indicate the location of perceived change. The post- session interview indicated that about 5.7% (31/542) of hit trials were not recorded as such; in these cases, the total trial duration was 60 s indicating that even though the subject fixated on the change region, the online algorithm failed to register the trial as a hit. In addition, 2.9% (7/238) of miss trials, in which the subjects were unable to detect the location of change in the post-session interview, ended before the full trial duration (60 s) had elapsed; in these cases, we expect that subjects triggered the termination of the trial by accidentally fixating for a pro- longed duration near the change. We repeated the analyses excluding these 4.8% (38/780) trials and observed results closely similar to those reported in the text. Finally, eye-tracking data from 0.64% (5/780) trials were corrupted and, therefore, excluded from all analyses. Subjects’ gaze was tracked throughout each trial with an iViewX Hi-speed eye-tracker (Sen- soMotoric Instruments Inc.) with a sampling rate of 500 Hz. The eye-tracker was calibrated for each subject before the start of the experimental session. Various gaze parameters, includ- ing saccade amplitude, saccade locations, fixation locations, fixation durations, pupil size, sac- cade peak speed and saccade average speed, were recorded binocularly on each trial, and stored for offline analysis. Because human gaze is known to be highly coordinated across both eyes, only monocular gaze data was used for these analyses. Each session lasted for approxi- mately 45 minutes, including time for instruction, eyetracker calibration and behavioral testing. SVM classification and feature selection based on gaze metrics We asked if subjects’ gaze strategies would be predictive of their success with detecting changes. To answer this question, as a first step, we tested if we could classify good versus poor performers (Fig 1C) based on their gaze metrics alone. As features for the classification analy- sis, we computed the mean and variance of the following four gaze metrics: saccade amplitude, fixation duration, saccade duration and saccade peak speed recorded by the eyetracker. We did not analyze two other gaze metrics acquired from the eyetracker: saccade average speed and pupil diameter for these analyses. Saccade average speed was highly correlated with sac- cade peak speed across fixations (r = 0.93, p<0.001), and was a redundant feature. In addition, while pupil size is a useful measure of arousal [61], it is often difficult to measure reliably, because slight, physical movements of the eye or head may cause apparent (spurious) changes in pupil size that can be confounded with real size changes. Before analysis, feature outliers were removed based on Matlab’s boxplot function, which considers values as outliers if they are greater than q3 + w × (q3 –q1) or less than q1 –w × (q3 –q1), where q1 and q3 are the 25th PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 20 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task and 75th percentiles of the data, respectively, and setting w = 1.5 provides 99.3 percentile cov- erage for normally distributed data. To avoid biases in estimating gaze metrics for good versus poor performers this last fixation at the change location (a minimum of 3 seconds of data) were removed from the eyetracking data for all “hit” trials before further analyses. Following outlier removal, these eight measures were employed as features in a classifier based on support vector machines (SVM) to classify good from poor performers (fitcsvm func- tion in Matlab). The SVM employed a polynomial kernel, and other hyperparameters were set using hyperparameter optimization (OptimizeHyperParameters option in Matlab). Features from each image were included as independent data points in feature space. Classifier perfor- mance was assessed with 5-fold cross validation, and quantified with the area-under-the-curve (AUC [62]). For these analyses, we included gaze data from all but one image (Image #20, see Data Availability link), in which every subject detected the change correctly. Significance levels (p-values) of classification accuracies were assessed with permutation testing by randomly shuffling the labels of good and poor performers across subjects 100 times and estimating a null distribution of classification accuracies; significance values correspond to the proportion of classification accuracies in the null distribution that were greater than the actual classifica- tion accuracy values. A similar procedure was used for SVM classification of trials into hits and misses except that, in this case, class labels were based on whether the trial was a hit or a miss, and permutation testing was performed by shuffling hit or miss labels across trials. Because we employed summary statistics (e.g. mean, variance) of the gaze metrics in these fea- ture selection analysis, we tested for unimodality of the logarithm of the respective gaze metric distributions with Hartigan’s dip test for unimodality [63]. Next, we sought to identify gaze metrics that best distinguished good from poor performers. For this we employed four standard metrics—Fisher score [18], AUC change [19] and Infor- mation Gain [20] and bag of decision trees [21]–which quantify the relative importance of each feature for distinguishing the two groups of subjects (Fig 1D). A detailed description of these metrics is provided next. Feature selection metrics (i) Fisher score computes the “quality” of features based on their extent of overlap across clas- ses. In a two-class scenario, Fisher Score for the jth feature is defined as, F jð Þ ¼ � � 1 nðþÞ(cid:0) 1 ð�xðþÞ j (cid:0) �xjÞ2 þ ð�xð(cid:0) Þ Þ2 þ � (cid:0) �xðþÞ j j 1 nð(cid:0) Þ(cid:0) 1 (cid:0) �xjÞ2 � i¼1 ðxð(cid:0) Þ Snð(cid:0) Þ i;j i¼0 ðxðþÞ SnðþÞ i;j (cid:0) �xð(cid:0) Þ j Þ2 ð4Þ where, �xj is the average value of the jth feature. Similarly �xðþÞ are the average of jth fea- ture for the positive and negative category respectively. Here �xðþÞ i;j denote the jth feature of ith sample-index for each category, with n(+) and n(-) being the number of positive and neg- ative instances respectively. A more discriminative feature has a higher Fisher score. i;j and �x ð(cid:0) Þ and �xð(cid:0) Þ j j (ii) AUC change describes the change in area-under-the-curve (AUC) with removing each feature in turn. The AUC (A) is the area under the ROC curve, plotted by varying the discrimi- nation threshold and plotting the True Positive Rate (TPR) as a function of the False Positive Rate (FPR). A ¼ R 1 x¼0TPRðFPR(cid:0) 1ðxÞÞdx ð5Þ A more discriminative feature’s absence produces a higher deterioration in classification accuracy. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 21 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task (iii) Information gain is a classifier-independent measure of the change in entropy upon partitioning the data based on each feature. A more discriminative feature has a higher infor- mation gain. Given binary class labels Y for a feature X, the entropy of Y (E(Y)) is defined as, EðYÞ ¼ (cid:0) pþlogðpþÞ (cid:0) p(cid:0) logðp(cid:0) Þ where, p+ is the fraction of positive class labels and p− is the fraction of negative class labels. The Information Gain given Y for a feature X is given by, IG X; Yð Þ ¼ E Yð Þ (cid:0) min i nX>divðiÞEðYX>divðiÞÞ þ nX<divðiÞEðYX<divðiÞÞ nX>divðiÞ þ nX<divðiÞ div ið Þ ¼ XsortedðiÞ þ Xsortedði þ 1Þ 2 ð6Þ ð7Þ where, nX>div(i) and nX<div(i) is the number of entries of X greater than and less than div(i), YX>div(i) and YX<div(i) are the entries of Y for which the corresponding entries of X are greater than and less than div(i) respectively and Xsorted(i) indicates a feature vector with its values sorted in ascending order. A more discriminative feature has a higher Information Gain. (iv) Out-of-bag error based on a bag of decision trees is an approach for feature selection using bootstrap aggregation on an ensemble of decision trees. Rather than using a single deci- sion tree this approach avoids overfitting by growing an ensemble of trees on independent bootstrap distributions drawn from the data. The most important features are selected by out- of-bag estimates of feature importance in the bagged decision trees (OOB error). We used the Treebagger function, as implemented in Matlab, with saccade and fixation features as inputs to the model, which classified if the data belonged to a good or poor performer. The number of trees was set to 6, with all other hyperparameters set to their default values. Analysis of scan paths and fixated spatial features We compared scan paths and low-level fixated (spatial) features across good and poor per- formers. To simplify comparing scan paths across participants, we adopted the following approach: we encoded each scan path into a finite length string. As a first step, fixation maps were generated to observe where the subjects fixated the most. Very few fixations occurred in object-sparse regions (e.g. sky), or had uniform color or texture, like the walls of a building (Fig 2A). In contrast, many more fixations around crowded regions with more intricate details. For each image, fixation points of all subjects were clustered, and each cluster was assigned a character label. The entire scan path, comprising a sequence of fixations, was then encoded as a string of cluster labels. Before clustering fixation points, we sought to minimize the contributions of regions with very low fixation density. To quantify this we adopted the following approach: Let xi be a fixa- tion point and let Dr denote the average Euclidean distance of xi from the set of other fixation xi Þ(cid:0) 1 denote the inverse of Dr . Now, we distributed xi points which are at a radius r from it. Let ðDr xi all the fixation points uniformly on the image; let U denote this set. We find the point yi in U that was closest (in Euclidean distance) to xi, and compute ðDr yi Þ(cid:0) 1=ðDr density at the fixation point xi was defined as rðxiÞ ¼ ðDr yi xi density less than 1 indicate regions which were sampled with less density than that corre- sponding to a uniform sampling strategy. These fixation points with very low fixation density Þ(cid:0) 1; as before. Then, the fixation Þ(cid:0) 1. Thus, all points with PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 22 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task were grouped into a single cluster since these occurred in regions that were explored relatively rarely. For these analyses r was set to 40 pixels, although the results were robust to variations of this parameter. The remaining fixation points were clustered using k-means clustering algorithm. The main challenges in working with the k-means algorithm are with: (i) deciding the num- ber of clusters (k) and ii) deciding initial cluster centers. To overcome these, we utilized the Bayesian Information Criterion (BIC) employed in the context of x-means clustering [64]: this allowed us to determine the optimum k. For each k ranging from 1 to 50, a BIC score was com- puted. Following smoothing, k corresponding to the highest BIC score was selected as the opti- mum cluster count. Once the number of clusters was fixed, the initial cluster centers were fixed using an iterative approach: For each iteration, initial cluster centers were selected using the k-means++ algorithm [65] and the values which gave the highest BIC score were selected as the initial cluster centers. Using the k and initial centers identified with these approaches, the fixation points were clustered for each image (Fig 2A, right). Once these clusters were iden- tified for each image, we employed four approaches for the analysis of scan-paths and fixated spatial features. First, we computed the edit distance between scan paths [22]. Briefly, the edit distance pro- vides an intuitive measure of the dissimilarity between two strings. It corresponds to the mini- mal number of “edit” operations—insertions, deletions or substitutions—that are necessary to transform one string into the other. For each image, the edit distance between the scan paths of each pair of subjects was calculated and normalized (divided) by the longer scan path length of the pair; this was done to normalize for differences in scan path length across subjects. A distribution of normalized edit distances was calculated among the good performers, and among the poor performers, across images. Median edit distance of each category of perform- ers was compared against the other, with a Wilcoxon signed rank test. However, note that the lack of a significant difference would only indicate that good performers and poor performers, each, followed similarly-consistent strategies. Therefore, to test whether these strategies were indeed significantly different between good and poor performers, we compared the median edit distance among the good (or poor) performers (intra-category edit distance) with the median edit distance across good and poor performers (inter-category edit distance), for all images, with a one-tailed signed rank test. Second, we computed the probabilities of making a saccade among specific types of clusters, which we call “domains”. Clusters obtained for each image were sorted in descending order of cumulative fixation duration. These were then grouped into four “domains”, based on quar- tiles of fixation duration, and ordered such that the first domain had the highest cumulative fixation duration (most fixated domain) and the last domain had the least cumulative fixation duration (least fixated domain). We then computed the probability of making a saccade from each domain to the other. We denote these saccade probabilities as: P(ik, jk+1), which repre- sents the probability of making a saccade from domain i at fixation k to domain j at fixation k +1. We tested if the saccade probabilities among domains were different between good and poor performers by using saccade probability matrices as vectorized features in a linear SVM analysis (other details as described in section on “SVM classification and feature selection based on gaze metrics”). Third, we computed the correlation between fixation distributions over images. Each image was divided into 13x18 tiles, and a two-dimensional histogram of fixations was computed for each image and participant. Binning at this resolution yielded non-empty bins for at least 15% of the bins; results reported were robust to finer spatial binning. The vectorized histograms of fixations were correlated between every pair of performers for each image, and median corre- lations compared across the two categories of performers, with a Wilcoxon signed rank test. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 23 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task As before, we also compared the median fixation correlations among the good (or poor) per- formers with the median fixation correlations across good and poor performers (intra- versus inter-category), for all images, with a one-tailed signed rank test. Fourth, we tested whether good and poor performers fixated on distinct sets of low-level spatial features in the images. For this, we identified spatial features that explained the greatest amount of variance in fixated image patches across good and poor performers. Specifically, image patches of size 112x112 pixels around each fixation point, corresponding to approxi- mately 4˚ of visual angle were extracted from each image for each participant and converted to grayscale values using the rgb2gray function in Matlab, which converts RGB images to gray- scale by eliminating the hue and saturation information while retaining the luminance. Two sets of fixated image patches was constructed separately for the good and poor performers. Each of these image patch sets was then subjected to Principal Component Analysis (PCA), using the pca function in Matlab, to identify low-level features in the image which occurred at the most common points of fixation across each group of subjects (Fig 3D). We, next, sorted the PCA feature maps based on the proportion of explained variance, and correlated each pair of sorted maps across good and poor performers; in the Results, we report average correlation values across the top 150 principal component maps. We did not attempt an SVM classifica- tion analysis based on PCA features, because of the high dimensionality of the extracted PC maps (~104), and the low number of data points in our experiment (~800). We also performed the same analysis after transforming each image into a grayscale saliency map using the fre- quency tuned salient region detection algorithm [24]. The same analyses were repeated for spatial features extracted from good and poor performers’ fixated image patches. Fifth, we tested if good and poor performers differed in terms of the spatial patterns of their fixations relative to the change region. For this, we computed the fixation frequency (counts) and the total fixation duration for each participant, based on the distance relative to the center of the change location, binned in concentric circular windows of increasing radii, in steps of 50 pixels. Each of these metrics were normalized by the respective parameter for each image and pooled together, separately for the good and poor performers, and compared between the two classes of performers with the Kolmogorov-Smirnov test (S4 Fig). Finally, to test if good and poor performers differed in terms of their latencies to fixate on the change region, we also compared the time to first fixation on the region of change, or the time from trial initiation to detect changes (on successful trials) for good and poor performers (Fig 2G and 2H). Model simulations and choice of parameters The model was simulated with a sequence of operations, as shown in Fig 4B. The model has been fully described in the Results. In these simulations, the CVR transformation that mimics foveal magnification was performed before the saliency map was computed (see next section). This sequence mimics the order of operations observed in the brain: foveal magnification occurs at the level of the retina, whereas saliency computation occurs at the level of higher brain structures like the superior colliculus [66] or the parietal cortex [67]. Saliency maps were computed using the frequency tuned salient region detection algorithm [24]. Because of this sequence of operations, we needed to re-compute the saliency map for each image for every possible location of fixation (at the pixel level): an operation that is computationally unfeasible on a standard desktop system. To expedite the computation, we represented each image in a reduced 864x648 pixel space and divided each image into a grid of non-overlapping patches or regions (72x54; Fig 4B), such that each patch covered 12x12 pixels. For two images of portrait orientation (Images #10 and #19; in Data Availability link), the same operations were done except that x- and y- grid resolutions were interchanged. We then pre-computed CVR PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 24 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task transforms and saliency maps for each of these pre-computed grid centers and performed sim- ulations based on these region-based representations of the images. Model parameters used for the simulations are specified in Table 1. Model parameters were not fit to human behavioral data, for example, using maximum-likelihood estimation. Rather, we selected model parameters so that they either matched the parameters used in the experi- ment (e.g. image and blank durations), or matched human metrics. We describe next the spe- cific justification for choice of each model parameter listed in Table 1. The time bin (Δt) was specified as 25 ms; larger and smaller values resulted in less or more frequent evaluations of the evidence (Eq 1) producing correspondingly faster or slower accu- mulation of the evidence. The image and blank durations (τ) were fixed at 10 time bins (250 ms), matching their durations in the actual experiment. The trial duration was fixed to 2400 bins (60 s), again matching the actual experiment. The temperature (T) parameter was set to ensure a similar range of saccade amplitude variance in the model, as in the human data. The decay factor (γ), which determines how quickly accumulated evidence “decays” over time, and decay scale (β), which governs the spatial extent of evidence accumulation, were set to default values that enabled the model to match average human performance across all images. Then, their values were varied over a wide range to test the effect of these parameters on model suc- cess with change detection. The spatial distribution of the decay parameter at each region was specified based on a two-dimensional Gaussian function, with its peak at the region of fixation; therefore, γi at each location is a function of time and depends on the current region being fix- ated. Noise scale (W), which controls the noise added during the evidence accumulation pro- cess, and threshold (Fc), which controls the threshold value of evidence needed for reporting a change (Fig 4D), were set so that their respective values ensured negligibly low false-positive rates (< 2%), overall. The prior odds ratio (P) and “no change” threshold (Fn) were set to val- ues that provided an approximate match to the median human fixation durations. Firing rate bounds (λmin, λmax) for encoding saliency were between 5 and 120 spikes per time bin. This corresponds to an overall population firing rate range of 0.2–4.8 kHz, which, assuming around 50 units in the neural population encoding each region, works out to a firing rate in the range of 4–96 Hz per neuron; these numbers mimic the biologically-observed range of firing rates for SC neurons (~5–100 Hz, White et al. 2017; their Fig 3). The firing rate prior (μf) was set to 3 spikes per bin, and the effect of varying this parameter on performance was also tested (Fig 5C). Finally, we used a third-order Taylor series approximation to the softmax function to achieve a softer saturation of this function. Note that these model parameter values were cho- sen based on human gaze metrics, or average task performance, but never based on task per- formance in individual images, to avoid circularity when correlating model performance with human performance across images (see Materials and Methods section “Comparison of model performance with human data”). Human saccade sequences tend to be biased in terms of the amplitude of individual sac- cades, and the angles between successive saccades (S7 Fig); these biases likely reflect properties of the oculomotor system that generates these saccades [33]. Because these saccade properties are not emergent features of our model, we matched the human saccade turn angle and ampli- tude distributions in the model. This was done by multiplying the map of evidence accumu- lated with the human saccade amplitude and turn angle distribution, before imposing the softmax function for computing saccade probabilities (Fig 4B). The effect of this bias was that the model generated scan-paths which qualitatively resembled human scan-paths (e.g. Fig 4C). Again, we sought to match only human saccade statistics in the model, and not task perfor- mance, when imposing this saccade bias to avoid circularity when computing the correlation between model and human performance in the change blindness task (see Materials and Methods section “Comparison of model performance with human data”). We repeated the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 25 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task simulations without imposing human saccade biases on the model, and obtained nearly identi- cal results. Cartesian variable resolution (CVR) transform. We modeled a key biological feature of visual representations of images, in terms of differences between foveal and parafoveal represen- tations. When a particular region is fixated, the representation of the fixated region, which is mapped onto the fovea, is magnified whereas the representation of the peripheral regions are correspondingly attenuated (S5 Fig). We modeled this using the Cartesian Variable Resolution (CVR) transform, which mimics known properties of visual magnification in humans [26]. The enhanced sensory representation of the foveated (fixated) region was modeled accord- ing to the following mathematical transformation of the image. We considered the foveated pixel to be the origin, denoted by (x0, y0) in the original image. An arbitrary point in the image, denoted by (x, y) is at a distance from the origin given by, dx = x−x0 and dy = y−y0. The following logarithmic transformation was then performed: dvx ¼ lnðbdx þ 1ÞSfx; dvy ¼ lnðbdy þ 1ÞSfy ð8Þ where, β is a constant (= 0.05) that determines central magnification, and Sfx and Sfy (= 200) are scaling factors along x and y directions, respectively; results reported were robust to mod- est variations of these parameter values. The final coordinates of the CVR transformed image are given by: x1 = x0+dvx and y1 = y0+dvy. Computation of the likelihood ratio (Li(t; z)) We provide here a detailed derivation of Eq 1 in the Results, involving computation of the likelihood ratio Li(t; z) for change versus no change at each region Ai. At each fixation, the model is faced with evaluating evidence for two hypotheses: change (C) versus no change (N). Note that the true difference between the firing rates of the generating processes at the change region is not known to the model, apriori; this corresponds to the fact that, in our experiment, the observer cannot know the precise magnitude or nature of the change occur- ring in each change image pair, apriori. We posit that the model expects to observe a firing rate difference of ±μf between the means of the two Poisson processes associated with the change region; this represents the apriori expectation of the magnitude of change for human observers. Here, we model this prior as a singleton value, although it is relatively straightfor- ward to extend the model to incorporate priors drawn from a specified density function (e.g. Gaussian). Let Xi and Yi denote the number of spikes observed in the m and n−p time-bins that the model fixates on the two images (A or A’), respectively (Fig 4A). Let λi denote the mean firing rate observed during this fixation, up until the current time bin; for this derivation, we posit that λi is measured in units of spikes per time bin; measuring λi in units of spikes per second simply requires multiplication by a scalar factor (Table 1), which does not impact the following deriva- tion. The model estimates the mean firing rate over the fixation interval as λi = (Xi+Yi)/(m+n−p). Note that this estimate of the mean firing rate is updated during each fixation across time bins. For hypothesis C to be true, Xi would be a sample from a Poisson process with mean, Γ1 = m(λi+μf) or Γ1 = m(λi−μf) and Yi would be a sample from a Poisson process with mean, Γ2 = (n −p)(λi−μf) or Γ2 = (n−p)(λi+μf), respectively. Similarly, for hypothesis N to be true, Xi would be a sample from a Poisson process with mean, Γ1 = m(λi) and Yi would be a sample from an identical Poisson process with mean, Γ2 = (n−p)(λi). For detecting changes, we assume that the model computes only the difference in the number of spikes, Zi = Yi−Xi, between the two images, rather than keeping track of the precise number of spikes generated by each image. The observed difference Zi could, therefore, be positive or negative. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 26 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task For hypothesis C (occurrence of change), the likelihood of observing a specific value of the difference in the number of spikes across the two images Zi = z is given as: P Zi ¼ zjC ð Þ ¼ 1 2 ðPðZi ¼ zjXi � rðmðli þ mf ÞÞ; Y i � rððn (cid:0) pÞðli (cid:0) mf ÞÞ þ PðZi ¼ zjXi � rðmðli (cid:0) mf ÞÞ; Y i � rððn (cid:0) pÞðli þ mf ÞÞÞ ð9Þ where ρ denotes the Poisson distribution. Here, we have assumed that the prior probabilities of encountering image A or A’ when the fixation lands in a given region are equal (the 1/2 fac- tor). Similarly, for hypothesis N (no change), the likelihood of observing a specific difference in the number of spikes, z, is given as: PðZi ¼ zjNÞ ¼ PðZi ¼ zjX � rðmliÞ; Y � rððn (cid:0) pÞliÞ The likelihood ratio of hypotheses, change versus no change, is computed as: ð Li z; t Þ ¼ PðZiðtÞ ¼ zjCÞ PðZiðtÞ ¼ zjNÞ ð10Þ ð11Þ We next expand these expressions with the analytical form of the Poisson distribution, P X ¼ k; X � rðlÞ , and marginalize over all values of Xi = x and Yi = x+z. These ð Þ ¼ e(cid:0) llk k! calculations involve computing an infinite sum which can be efficiently solved using Bessel functions. Specifically, the infinite sum in our calculation can be computed using the identity: ffiffi Þ where I is a modified Bessel function of the first kind. c S1 y¼0 � � ¼ p cy y!ðyþzÞ! p (cid:0) zI z:2 ffiffi ð c With some algebra, we can show that: (i) when Zi t � 0: � ðmðn(cid:0) pÞðl2 i � (cid:0) m2 f ÞxÞ x!ðxþzÞ! B1S1 x¼0 Li Tð Þ ¼ B2S1 z¼0 ðmðn(cid:0) pÞl2Þx x!ðxþzÞ! ¼ B1c(cid:0) z B2c(cid:0) z 1 Izð2c1Þ 2 Izð2c2Þ q where c1 ¼ 1 ¼ 0:5m(cid:0) z B0 q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi mðn (cid:0) pÞðl2 i (cid:0) m2 f Þ " ðli (cid:0) mf Þ(cid:0) ze(cid:0) ðmðl1þmf Þþðn(cid:0) pÞðli(cid:0) mf ÞÞ þðli þ mf Þ(cid:0) ze(cid:0) ðmðli(cid:0) mf Þþðn(cid:0) pÞðliþmf ÞÞ and c2 ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi mðn (cid:0) pÞl2 # i , and B2 ¼ ðn (cid:0) pÞzlz i e(cid:0) liðmþn(cid:0) pÞ: (ii) when Zi t < 0: where � 1S1 B0 x¼0 ðmðn(cid:0) pÞðl2 i (cid:0) m2 f ÞxÞ x!ðxþzÞ! � Li Tð Þ ¼ B0 2 S1 z¼0 ðmðn(cid:0) pÞl2Þx x!ðxþzÞ! ¼ B0 1cz 2cz B0 1Izð2c1Þ 2Izð2c2Þ 1 ¼ 0:5m(cid:0) z B0 " ðli þ mf Þ(cid:0) ze(cid:0) ðmðl1þmf Þþðn(cid:0) pÞðli(cid:0) mf ÞÞ þðli (cid:0) mf Þ(cid:0) ze(cid:0) ðmðli(cid:0) mf Þþðn(cid:0) pÞðliþmf ÞÞ # ; B2 ¼ m(cid:0) zl(cid:0) z i e(cid:0) liðmþn(cid:0) pÞ and c1 and c2 are the same as before. ð12Þ ð13Þ ð14Þ PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 27 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task We computed the value of the Bessel function using the Matlab function besseli. When val- ues of x and z were large or disproportionate, Matlab’s floating point arithmetic could not compute these expressions correctly; in this case, we employed variable precision arithmetic (vpa in Matlab). In addition, d for extreme values of x and z we adopted the following approximations: i. For sufficiently large values of x: I z; x ð Þ � exffiffiffiffiffi p 2px � 1 (cid:0) � 4z2(cid:0) 1 8x ii. For sufficiently large values of z: I z; x ð Þ � x 2ð Þz Gðzþ1Þ Note that the model makes the following assumptions: (i) the model makes a change versus no-change decision based on the difference of spike counts (Zt i ¼ z), rather than by keeping track of the absolute spike counts produced by each image (see next); (ii) the model estimates the average firing rate based on the number of spikes produced until that time-bin λi = (Xi+Yi)/(n−p+m); (iii) the model has a discrete, single valued, prior on the change in firing rates μf; this prior is different from the actual difference in firing rates across the two images, which is computed based on the difference in their, respective, salience values during the simu- lation (see also Fig 5C). Comparison of model performance with human data Using the saccade generation model shown in Fig 4B, we simulated the model 100 times using the same images as employed in the human change blindness experiment (Fig 1A). All stochas- tic parameters (evidence noise, fixation durations) were resampled with fresh random ‘seeds’ for each iteration of the model. We then computed the accuracy of the model as the proportion of times the model detected the change—fixation on change region until threshold crossing (Fig 4D)—versus the proportion of times the model failed to detect the change region. These proportions of correct detections were then compared for human performance (average across n = 39 participants) versus model performance (n = 100 iterations), across images, using robust correlations [68]. For these analyses, we employed the state-of-the-art DeepGaze II net- work [17] for generating the saliency map. Next, we performed control analyses to compare the SPRT model with three other change detection models, each with particular differences in search strategy or stopping rule. First, we tested a model that failed to integrate evidence effectively by setting γ = 1 in the evidence inte- gration step (Eq 2). Such a model completely ignores past evidence and makes decisions based solely on instantaneous posterior odds ratio (Li(t)Pi). Second, we tested a model with an alter- native stopping rule in which the change was detected based on the derivative of the posterior odds ratio (difference of log (Li Pi) between two successive timesteps) crossing a threshold. For these two models, threshold values for terminating the simulation were determined based on two pilot runs across all 20 images; thresholds were chosen such that the models provided a negligible proportion of false-alarm (<0.01%) comparable with our experimental data. Third, we tested a model in which evidence computation and accumulation were intact, but the model selected the next location of saccade with a random strategy. This was achieved by set- ting a high value of the temperature parameter (T) in the final softmax function (T = 104), which resulted in a nearly uniform probability, across the image, of selecting the next fixation (“random searcher”). For all three models, we identically matched the timing and distribution of fixation interval durations with our standard SPRT model. The distribution of absolute dif- ferences in performance between the human data and our model across images, and the corre- sponding distributions for control models were compared with paired signed rank tests (Fig 6C). PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 28 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task Finally, we tested the model’s ability to predict human gaze shift strategies. For this, we employed the following approach. First, we identified the top 10 fixated clusters in each image. Next, we constructed a saccade probability matrix between every pair of clusters among these ten clusters (Fig 7C, rows/columns 1–10) in the human data, by combining fixation data across all n = 39 participants. The model was then simulated 40 times, and the average probability of saccades between the same clusters for each image was computed for the model. These 10x10 saccade probability matrices were then linearized and compared between the model and human data using Pearson’s correlations (Fig 7D, left). Comparison of model performance with DeepGaze We also compared the model’s ability to predict human gaze patterns with that of DeepGaze II [17]. DeepGaze is among the top-ranked algorithms for human gaze prediction, and is based on a deep learning model for fixation prediction which employs features extracted from the VGG-19 network, another deep learning neural network trained to identify objects in an image. For this comparison, the model was simulated with all of the same steps as in Fig 4B, except that no likelihood ratio was computed, and no evidence accumulated. Rather, saccades occurred stochastically based on the same softmax rule as employed in our algorithm (Eq 3), but based on DeepGaze II saliency values alone. Again, saccade probability matrices were com- pared between the DeepGaze II prediction and human data using Pearson’s correlations (Fig 7D, right). To enable a fair comparison with DeepGaze we incorporated the following additional fea- tures in the DeepGaze model simulations. First, inhibition-of-return (IOR) is an emergent fea- ture of our model (see Results). We, therefore, incorporated IOR in the DeepGaze model as well [34]. IOR was implemented as a Gaussian patch (G) centered on the current fixation (x, y) with a standard deviation (σ) of 20 pixels. The amplitude of G was scaled up by a time depen- dent factor (tanh(0.05 t)), so that the impact of IOR increased progressively over the course of the trial. IOR values were accumulated in a spatial map with a discount factor of 0.25 across successive timesteps (IOR(x, y, t) = (0.25 � IOR(x, y, t—1)) + G(x, y; σ)). IOR values were clipped between 0 and 1, and the complement of IOR map was multiplied with the foveally- magnified saliency map before computing the next location of fixation. Second, because the DeepGaze model was not accumulating evidence for change, there was no clear termination criterion. Therefore, we identically matched the timing and distribution of fixation interval durations (timesteps for each fixation) with our SPRT model. This was accomplished by initi- ating and terminating each fixation in the DeepGaze model at the exact same times when these were initiated or terminated in the SPRT model, respectively. Third, to ensure that both the model and DeepGaze produced saccades with the same level of stochasticity we identically matched the temperature parameter in the softmax function (Eq 3) for deciding the next sac- cade location. Lastly, we also performed comparisons with the human data by limiting the sac- cade amplitude range for comparison. The SPRT model (and humans) make many short saccades, whereas DeepGaze primarily makes long saccades (Fig 7B). Therefore, we performed a control analysis, comparing the human data, SPRT model and DeepGaze considering only saccades with amplitudes greater than the 10th percentile of those generated by the DeepGaze model (S8C and S8D Fig). Supporting information S1 Fig. Re-analysis of gaze metrics by re-classifying good and poor performers based on a median split of performance. From top to bottom row: Re-analysis of the data shown in Figs 1 and 2 (main text), except that “good” and “poor” performers were defined based on a median PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 29 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task split of the data. Other conventions are the same as in the corresponding figure panels in the main text. (TIF) S2 Fig. Gaze metrics predictive of success, distributions of gaze metrics and variance in success rates across images. A. Pair-wise correlations among the eight gaze metrics used as features in classification analysis of good versus poor performers (Fig 1C, main text). Gray squares: non-significant correlations. Colored square: significant correlations at p<0.01 with Bonferroni correction for multiple comparisons. Abbreviations are as in Fig 1C (main text). B. Saccade amplitude (left) and fixation duration (right) distributions for representative partici- pants (ID-s in each subplot title). Red fits: Mixture of Gaussians model. p-value in title of each subplot indicates significance level for deviation from unimodality per Hartigan’s dip test (smaller p-values represent greater evidence of bi/multi-modailty). C. Success rates of human observers on the change blindness trial images (n = 20), sorted by the proportion of hits. Error bars denote standard error of the mean performance across participants. (TIF) S3 Fig. Fixated features for good and poor performers. A. Difference between the average saccade probability matrices for the good and poor performers (good minus poor). Other con- ventions are the same as in Fig 3A (main text). Note that these differences are 3 orders of mag- nitude smaller than the values in Fig 3A (main text). B. Same as in Fig 3D (main text) except that fixated features were identified following PCA on 112x112 patches extracted from a saliency map, rather than the grayscale image. The saliency map was generated with the fre- quency tuned saliency algorithm [24]. Other conventions are the same as in Fig 3D main text. (TIF) S4 Fig. Distribution of fixations, relative to change location, for good and poor performers. A. Distribution of frequency of fixations, binned based on the distance of fixation relative to the center of the change location, separately for good (red) and poor (blue) performers. B. Same as in panel A but for the total fixation duration. (TIF) S5 Fig. Mimicking foveation in the model. Illustration of foveal magnification with the Carte- sian Variable Resolution (CVR) transform for a hypothetical fixation (highlighted by the cir- cle) on one of the images used in the change blindness task (Image #6, S1 Table). (TIF) S6 Fig. Dependence of the likelihood ratio (L(t; z)) on mean firing rate and firing rate prior. A. Likelihood ratio (L(t; z)) as a function of spike count difference between the first and second image (z, Eq 1; main text) for different values of the mean firing rate, λ = 4 . . . 10 spikes/bin. The number of time bins for which the first and second images were fixated (m and n−p, respectively) have each been fixed to 5 bins, and the firing rate difference prior, μf fixed at 3 spikes/bin. Curves of progressively lighter shades: increasing values of the mean fir- ing rate. B. Same as in A, but for different values of the firing rate difference prior, μf = 1, 3, 5 . . . 13 spikes/bin and mean firing rate λ fixed at 40 spikes/bin. Curves of progressively lighter shades: increasing values of μf. (TIF) S7 Fig. Mimicking Saccade Turn Angle distribution. Polar heat map indicating the distribu- tion of human saccade amplitudes and turn angles. The arrow indicates the location of the last saccade. The histogram was computed using data from all (n = 39) participants and all (n = 20) images. The bias against right angled turns is apparent. The distribution was PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 30 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task smoothed both along the radial and angular directions, for display purposes only. (TIF) S8 Fig. Model saccade probability matrices, and correlations with human data (control analyses). A-B. Same as in Fig 7C and 7D (main text), except with replacing DeepGaze’s saliency algorithm with the frequency-tuned salient region detection algorithm. C-D. Same as in Fig 7C and 7D (main text) except including only saccades whose amplitude was at least as large (or greater) than the 10th percentile of saccade amplitudes generated by the DeepGaze model (Fig 7B, main text, dashed vertical line). For C, the saccade probability matrix was nor- malized by its range for visualization purposes only. Other conventions are the same as in Fig 7C and 7D (main text). (TIF) S1 Table. List of images employed in the change blindness task. (DOCX) Acknowledgments We thank Ranit Sengupta for help with compiling the results and Guruprasath Gurusamy for help with preparing figures. We also thank Prof. Veni Madhavan for generously sharing the eye tracker employed in these experiments. Author Contributions Conceptualization: Devarajan Sridharan. Data curation: Akshay Jagatap, Simran Purokayastha. Formal analysis: Akshay Jagatap, Hritik Jain. Funding acquisition: Devarajan Sridharan. Investigation: Simran Purokayastha. Methodology: Akshay Jagatap, Simran Purokayastha, Hritik Jain. Supervision: Devarajan Sridharan. Writing – original draft: Devarajan Sridharan. Writing – review & editing: Devarajan Sridharan. References 1. Carrasco M. Visual attention: The past 25 years. Vision Res. 2011; 51(13):1484–525. https://doi.org/10. 1016/j.visres.2011.04.012 PMID: 21549742 2. Simons DJ, Rensink RA. Change blindness: Past, present, and future. Trends Cogn Sci. 2005; 9(1):16– 20. https://doi.org/10.1016/j.tics.2004.11.006 PMID: 15639436 3. Simons DJ, Ambinder MS. Change blindness: Theory and consequences. Curr Dir Psychol Sci. 2005; 14(1):44–8. 4. Rensink RA, O’Regan JK, Clark JJ. To see or not to see: The need for attention to perceive changes in scenes. Psychol Sci. 1997; 8(5):368–73. 5. Zhao N, Chen W, Xuan Y, Mehler B, Reimer B, Fu X. Drivers’ and non-drivers’ performance in a change detection task with static driving scenes: is there a benefit of experience? Ergonomics. 2014; 57 (7):998–1007. https://doi.org/10.1080/00140139.2014.909952 PMID: 24787715 6. Crundall D. The Deceleration Detection Flicker Test: A measure of experience? Ergonomics. 2009; 52 (6):674–84. https://doi.org/10.1080/00140130802528337 PMID: 19424922 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 31 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task 7. Beanland V, Filtness AJ, Jeans R. Change detection in urban and rural driving scenes: Effects of target type and safety relevance on change blindness. Accid Anal Prev. 2017; 100:111–22. https://doi.org/10. 1016/j.aap.2017.01.011 PMID: 28130981 8. Tse PU. Mapping visual attention with change blindness: New directions for a new method. Cogn Sci. 2004; 28(2):241–58. https://doi.org/10.1016/j.cogsci.2003.12.002 PMID: 18797511 9. O’Regan JK, Rensink RA, Clark JJ. Change-blindness as a result of ‘mudsplashes.’ Nature. 1999; 398 (6722):34. https://doi.org/10.1038/17953 PMID: 10078528 10. Droll JA, Gigone K, Hayhoe MM. Learning where to direct gaze during change detection. J Vis. 2007; 7 (14):6. https://doi.org/10.1167/7.14.6 PMID: 18217801 11. Bogacz R, Brown E, Moehlis J, Holmes P, Cohen JD. The physics of optimal decision making: a formal analysis of models of performance in two-alternative forced-choice tasks. Psychol Rev. 2006; 113 (4):700. https://doi.org/10.1037/0033-295X.113.4.700 PMID: 17014301 12. Gold JI, Shadlen MN. Representation of a perceptual decision in developing oculomotor commands. Nature. 2000; 404(6776):390. https://doi.org/10.1038/35006062 PMID: 10746726 13. Smith PL, Ratcliff R. Psychology and neurobiology of simple decisions. Trends Neurosci. 2004; 27 (3):161–8. https://doi.org/10.1016/j.tins.2004.01.006 PMID: 15036882 14. Wald A. Sequential analysis. Courier Corporation; 1973. 15. Gold JI, Shadlen MN. The neural basis of decision making. Annu Rev Neurosci. 2007;30. https://doi. org/10.1146/annurev.neuro.29.051605.113038 PMID: 17600525 16. Kira S, Yang T, Shadlen MN. A neural implementation of Wald’s sequential probability ratio test. Neu- ron. 2015; 85(4):861–73. https://doi.org/10.1016/j.neuron.2015.01.007 PMID: 25661183 17. Ku¨ mmerer M, Wallis TSA, Bethge M. DeepGaze II: Reading fixations from deep features trained on object recognition. arXiv. 2016. 18. Boser BE, Guyon IM, Vapnik VN. A training algorithm for optimal margin classifiers. Fifth Annual Work- shop on Computational learning theory. 1992;144–52. 19. Nasrabadi NM. Pattern recognition and machine learning. J Electron Imaging. 2007; 16(4):49901. 20. Yang Y. A comparative study on feature selection in text categorization. IEEE Conf International Confer- ence on Machine Learning. 1997;412–20. 21. Hothorn T, Lausen B. Double-bagging: Combining classifiers by bootstrap aggregation. Pattern Recog- nit. 2003; 36:1303–9. 22. Cristino F, Mathoˆ t S, Theeuwes J, Gilchrist ID. ScanMatch: A novel method for comparing fixation sequences. Behav Res Methods. 2010; 42(3):692–700. https://doi.org/10.3758/BRM.42.3.692 PMID: 20805591 23. Itti L, Koch C, Niebur E. A model of saliency-based visual attention for rapid scene analysis. IEEE Trans Pattern Anal Mach Intell. 1998; 20(11):1254–9. 24. Achanta R, Hemami S, Estrada F, Susstrunk S. Frequency-tuned salient region detection. In: Computer Vision and Pattern Recognition, 2009;1597–604. 25. White BJ, Berg DJ, Kan JY, Marino RA, Itti L, Munoz DP. Superior colliculus neurons encode a visual saliency map during free viewing of natural dynamic video. Nat Commun. 2017; 8:14263. https://doi. org/10.1038/ncomms14263 PMID: 28117340 26. Wiebe KJ, Basu A. Modelling ecologically specialized biological visual systems. Pattern Recognit. 1997; 30(10):1687–703. 27. Abramowitz M, Stegun IA. Handbook of Mathematical Functions with Formulas, Graphs, and Mathe- matical Tables (Applied Mathematics Series 55). Natl Bur Stand Washington, DC. 1964; 28. Izhikevich E M. Dynamical Systems in Neuroscience: The Geometry of Excitability and Bursting. Dyn. Syst. 2007. 29. Heiberg T, Kriener B, Tetzlaff T, Einevoll GT, Plesser HE. Firing-rate models for neurons with a broad repertoire of spiking behaviors. J Comput Neurosci. 2018; 45(2):103–32. https://doi.org/10.1007/ s10827-018-0693-9 PMID: 30146661 30. Dayan P, Abbott LF. Theoretical Neuroscience: Computational And Mathematical Modeling of Neural Systems. Massachusetts Institute of Technology Press. 2005. 31. Ratcliff R, Smith PL, Brown SD, McKoon G. Diffusion decision model: current issues and history. Trends Cogn Sci. 2016; 20(4):260–81. https://doi.org/10.1016/j.tics.2016.01.007 PMID: 26952739 32. Busemeyer JR, Townsend JT. Decision field theory: A dynamic-cognitive approach to decision making in an uncertain environment. Psychol Rev. 1993; 100(3):432. https://doi.org/10.1037/0033-295x.100.3. 432 PMID: 8356185 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 32 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task 33. Roberts J, Wallis G, Breakspear M. Fixational eye movements during viewing of dynamic natural scenes. Front Psychol. 2013; 4:797. https://doi.org/10.3389/fpsyg.2013.00797 PMID: 24194727 34. Itti L, Koch C. A saliency-based search mechanism for overt and covert shifts of visual attention. Vision Res. 2000; 40(10):1489–506. https://doi.org/10.1016/s0042-6989(99)00163-7 PMID: 10788654 35. Andermane N, Bosten JM, Seth AK, Ward J. Individual differences in change blindness are predicted by the strength and stability of visual representations. Neurosci Conscious. 2019; 2019(1). https://doi. org/10.1093/nc/niy010 PMID: 30697440 36. Deubel H, Schneider WX. Saccade target selection and object recognition: Evidence for a common attentional mechanism. Vision Res. 1996; 36(12):1827–37. https://doi.org/10.1016/0042-6989(95) 00294-4 PMID: 8759451 37. Hoffman JE, Subramaniam B. The role of visual attention in saccadic eye movements. Attention, Per- ception, Psychophys. 1995; 57(6):787–95. https://doi.org/10.3758/bf03206794 PMID: 7651803 38. Moore T, Fallah M. Control of eye movements and spatial attention. Natl Acad Sci. 2001; 98(3):1273 LP–1276. https://doi.org/10.1073/pnas.021549498 PMID: 11158629 39. Judd T, Ehinger K, Durand F, Torralba A. Learning to predict where humans look. IEEE International Conference on Computer Vision. 2009;2106–13. 40. Hacisalihzade SS, Stark LW, Allen JS. Visual perception and sequences of eye movement fixations: A stochastic modeling approach. IEEE Trans Syst Man Cybern. 1992; 22(3):474–81. 41. Peters RJ, Itti L. Beyond bottom-up: Incorporating task-dependent influences into a computational model of spatial attention. IEEE Conference on Computer Vision and Pattern Recognition. 2007;1–8. 42. Wang W, Chen C, Wang Y, Jiang T, Fang F, Yao Y. Simulating Human Saccadic Scanpaths on Natural Images. IEEE Conference on Computer Vision and Pattern Recognition. 2011;441–8. 43. Adeli H, Vitu F, Zelinsky GJ. A Model of the Superior Colliculus Predicts Fixation Locations during Scene Viewing and Visual Search. J Neurosci. 2017; 37(6):1453–67. https://doi.org/10.1523/ JNEUROSCI.0825-16.2016 PMID: 28039373 44. Boccignone G. Advanced statistical Methods for eye movement analysis and modeling: A gentle intro- duction. arXiv. 2015; 45. Boccignone G, Ferraro M. Modelling gaze shift as a constrained random walk. Phys A Stat Mech its Appl. 2004; 331(1):207–18. 46. Adams RP, Mackay D. Bayesian Online Changepoint Detection. arXiv. 2007; 47. Matteson DS, James NA. A Nonparametric Approach for Multiple Change Point Analysis of Multivariate Data. J Am Stat Assoc. 2014; 109(505):334–45. 48. Gerrit J.J. van den Burg CKIW. An Evaluation of Change Point Detection Algorithms. arXiv. 2020; 49. Cavanaugh J, Wurtz RH. Subcortical modulation of attention counters change blindness. J Neurosci. 2004; 24(50):11236–43. https://doi.org/10.1523/JNEUROSCI.3724-04.2004 PMID: 15601929 50. Werner S, Thies B. Is “Change Blindness” Attenuated by Domain-specific Expertise? An Expert-Nov- ices Comparison of Change Detection in Football Images. Vis cogn. 2000; 7(1–3):163–73. 51. Stirk JA, Underwood G. Low-level visual saliency does not predict change detection in natural scenes. J Vis. 2007; 7(10):3. https://doi.org/10.1167/7.10.3 PMID: 17997672 52. Brox T. Maximum Likelihood Estimation BT—Computer Vision: A Reference Guide. Springer. 2014;481–2. 53. Britten GL, Mohajerani Y, Primeau L, Aydin M, Garcia C, Wang W-L, et al. Evaluating the Benefits of Bayesian Hierarchical Methods for Analyzing Heterogeneous Environmental Datasets: A Case Study of Marine Organic Carbon Fluxes. Front Environ Sci. 2021; 9:28. 54. Pringle HL, Irwin DE, Kramer AF, Atchley P. The role of attentional breadth in perceptual change detec- tion. Psychon Bull Rev. 2001; 8(1):89–95. https://doi.org/10.3758/bf03196143 PMID: 11340871 55. Angelone B, Severino S. Effects of individual differences on the ability to detect changes in natural scenes. J Vis. 2008; 8. 56. Caird JK, Edwards CJ, Creaser JI, Horrey WJ. Older driver failures of attention at intersections: Using change blindness Methods to assess turn decision accuracy. Human Factors & Ergonomics Society. 2005;235–49. https://doi.org/10.1518/0018720054679542 PMID: 16170936 57. Davies G, Hine S. Change Blindness and Eyewitness Testimony. J Psychol. 2007; 141:423–34. https:// doi.org/10.3200/JRLP.141.4.423-434 PMID: 17725074 58. Smith H, Milne E. Reduced change blindness suggests enhanced attention to detail in individuals with autism. J Child Psychol Psychiatry. 2009; 50(3):300–6. https://doi.org/10.1111/j.1469-7610.2008. 01957.x PMID: 19309329 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 33 / 34 PLOS COMPUTATIONAL BIOLOGYModeling gaze strategies in a change blindness task 59. Fletcher-Watson S, Leekam S, Connolly B, Collis J, Findlay J, McConachie H, et al. Attenuation of change blindness in children with autism spectrum disorders. British Journal of Dev Psychol. 2012; 30 (3):446–58. https://doi.org/10.1111/j.2044-835X.2011.02054.x PMID: 22882373 60. Remington A, Campbell R, Swettenham J. Attentional status of faces for people with autism spectrum disorder. Autism. 2011 Jun 24; 16(1):59–73. https://doi.org/10.1177/1362361311409257 PMID: 21705475 61. Wang C-A, Baird T, Huang J, Coutinho JD, Brien DC, Munoz DP. Arousal Effects on Pupil Size, Heart Rate, and Skin Conductance in an Emotional Face Task. Front Neurol. 2018; 9:1029. https://doi.org/10. 3389/fneur.2018.01029 PMID: 30559707 62. Bradley AP. The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recognit. 1997; 30(7):1145–59. 63. Hartigan JA, Hartigan PM. The Dip Test of Unimodality. Ann Stat. 1985 Mar 1; 13(1):70–84. 64. Pelleg D, Moore AW, others. X-means: Extending K-means with Efficient Estimation of the Number of Clusters. IEEE International Conference on Machine Learning. 2000;727–34. 65. Arthur D, Vassilvitskii S. k-means++: The advantages of careful seeding. ACM-SIAM Symposium on Discrete Algorithms. 2007;1027–35. 66. Knudsen EI. Neural Circuits That Mediate Selective Attention: A Comparative Perspective. Trends Neu- rosci. 2018; 41(11):789–805 https://doi.org/10.1016/j.tins.2018.06.006 PMID: 30075867 67. Bisley JW, Goldberg ME. Attention, Intention, and Priority in the Parietal Lobe. Annu Rev Neurosci. 2010; 33(1):1–21. 68. Pernet C, Wilcox R, Rousselet G. Robust Correlation Analyses: False Positive and Power Validation Using a New Open Source Matlab Toolbox. Front Psychol. 2013; 3:606. https://doi.org/10.3389/fpsyg. 2012.00606 PMID: 23335907 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1009322 August 24, 2021 34 / 34 PLOS COMPUTATIONAL BIOLOGY
10.1186_s12964-020-00677-9	Yao et al. Cell Commun Signal (2020) 18:187 https://doi.org/10.1186/s12964-020-00677-9 RESEARCH Open Access Pathogenic effects of inhibition of mTORC1/ STAT3 axis facilitates Staphylococcus aureus-induced pyroptosis in human macrophages Ruiyuan Yao1†, Yuhao Chen1,2†, Huifang Hao1†, Zhixin Guo1, Xiaoou Cheng1, Yuze Ma1, Qiang Ji1, Xiaoru Yang1, Yanfeng Wang1, Xihe Li1,3* and Zhigang Wang1* Abstract Background: Pyroptosis is a recently identified pathway of caspase-mediated cell death in response to microbes, lipopolysaccharide, or chemotherapy in certain types of cells. However, the mechanism of how pyroptosis is regulated is not well-established. Methods: Herein, the intracellular bacteria were detected by staining and laser confocal microscopy and TEM. Live/ dead cell imaging assay was used to examine macrophage death. Western blot and immunohistochemical staining were used to examine the protein changes. IFA was used to identify typical budding vesicles of pyroptosis and the STAT3 nuclear localization. SEM was used to observe the morphological characteristics of pyroptosis. ELISA was used to detect the level of inflammatory cytokines. Pyroptosis was filmed in macrophages by LSCM. Results: S. aureus was internalized by human macrophages. Intracellular S. aureus induced macrophage death. S. aureus invasion increased the expression of NLRP3, Caspase1 (Casp-1 p20) and the accumulation of GSDMD-NT, GSDMD-NT pore structures, and the release of IL-1β and IL-18 in macrophages. Macrophages pyroptosis induced by S. aureus can be abrogated by blockage of S. aureus phagocytosis. The pyroptosic effect by S. aureus infection was pro- moted by either rapamycin or Stattic, a specific inhibitor for mTORC1 or STAT3. Inhibition of mTORC1 or STAT3 induced pyroptosis. mTORC1 regulated the pyroptosic gene expression through governing the nuclear localization of STAT3. mTORC1/STAT3 axis may play a regulatory role in pyroptosis within macrophages. Conclusions: S. aureus infection induces human macrophage pyroptosis, inhibition of mTORC1/STAT3 axis facili- tates S. aureus-induced pyroptosis. mTORC1 and STAT3 are associated with pyroptosis. Our findings demonstrate a regulatory function of the mTORC1/STAT3 axis in macrophage pyroptosis, constituting a novel mechanism by which pyroptosis is regulated in macrophages. Keywords: Pyroptosis, Staphylococcus aureus, mTORC1, STAT3 Correspondence: lixihe@saikexing.com; lswzg@imu.edu.cn †Ruiyuan Yao, Yuhao Chen, Huifang Hao have contributed equally to this work 1 State Key Laboratory of Reproductive Regulation and Breeding of Grassland Livestock, School of Life Sciences, Inner Mongolia University, Hohhot 010070, China Full list of author information is available at the end of the article Background Pyroptosis is a lytic form of caspase-dependent cell death, which encompasses caspase-1, -4, -5, and -11. Caspase-1 is activated by various canonical inflammasomes, and caspase-4/5/11 recognizes cytosolic bacterial lipopoly- saccharide directly, both of which trigger pyroptosis [1]. © The Author(s) 2020. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Yao et al. Cell Commun Signal (2020) 18:187 Page 2 of 18 In the canonical pathway, intracellular bacteria upregu- late caspase-1, which then cleaves gasdermin D to pro- duce N-terminal GSDMD fragment (GSDMD-NT) [2]. GSDMD-NT forms pores in the membrane that drive swelling and membrane rupture, and the cytokines IL-1β and IL-18 are released through GSDMD-NT pore structures [3]. Pyroptosis can defend against microbial challenges and endogenous threats by eliminating such compromised cells [4]. Many factors cause pyroptosis, including lipopolysaccharide, chemotherapy drugs, TNF- a, 17β-estradiol (E2), and bacteria [5–7]. As extracellular pathogens, certain types of bacteria can invade a variety of mammalian nonprofessional phago- cytes and can be engulfed by professional phagocytes, such as neutrophils and monocytes, and survive in them [8, 9]. The uptake of bacteria by nonprofessional phago- cytes is mediated by adhesins [10]. Fibronectin-binding proteins, a major type of adhesin, mediate the internali- zation of adherent bacteria [11]. Professional phagocytic cells, such as neutrophils and macrophages, are designed to engulf microorganisms and clear debris—a process that can be divided conceptually into formation of the phagosome and subsequent evolution into a degradative compartment [12]. The monocyte-derived macrophages (MDMs) are separated into two specific phenotypes, classically activated macrophages (M1) and alternatively activated ones (M2) [13], Macrophages differentiate into M1 induced by IFN-γ and LPS, and the M1macrophages participate in the removal of pathogens during infec- tion, induce M1 phenotype polarization which have main roles in host defense against various microbial pathogens and increased tumoricidal activity [14, 15]. Therefore, M1-polarized human MDMs were used as a cell model for macrophages function [16, 17]. Several pathogens can invade cells and survive intra- cellularly for various periods [18]. The success of intra- cellular pathogens is attributed to their ability to inhibit phagocytosis by preventing opsonophagocytosis and blocking specific signaling pathways [19], such as avoid- ing delivery to the lysosome and release into the cyto- plasm and arresting phagosome maturation, creating an optimal niche for replication [20]. Staphylococcus aureus, a gram-positive bacterium and human pathogen that causes a wide range of illnesses, from skin infections to severe pneumonia and sepsis, can invade cells and trigger pyroptosis in human mac- rophages [7]. However, the mechanism by which it is reg- ulated in professional phagocytes is unknown. Mammalian target of rapamycin complex 1 (mTORC1) is a central coordinator in eukaryotic cells with regard to cell growth and metabolism with environmental inputs [21], autophagy, apoptosis, necroptosis, and other forms of cell death [22]. mTORC1 regulates autophagy by inhibiting ULK and the nuclear translocation of tran- scription factor EB (TFEB) [23, 24] and suppresses apop- tosis in pterygium by controlling Beclin1-dependent autophagy by targeting Bcl-2 [25]. Among the intrinsic forms of cell death, pyroptosis has received increased attention recently, by the function of mTORC1 in pyrop- tosis is unkonwn. Signal transducer and activator of transcription 3 (STAT3) is a latent transcription factor that mediates extracellular signals, such as those from cytokines and growth factors [26]. STAT3 plays an important role in programmed cell death, and inhibition of STAT3 leads directly to apoptosis [27]. In addition, autophagy is gov- erned by STAT3 activation by upregulating or down- regulating essential autophagy genes [28]. However, the relationship between STAT3 and pyroptosis has not been reported. Moreover, mTOR has been implicated in the regulation of STAT3 activation [29], and mTORC1 stimu- lates STAT3 to restrain proinflammatory responses [30]. Nevertheless, the regulatory function of mTORC1 in the expression of pyroptosic genes via STAT3 and pyroptosis is unknown. Pyroptosic macrophages are important in the defense against microbial infections, removing pathogens and rendering them susceptible to phagocytosis and kill- ing by a secondary phagocyte [31, 32]. As mentioned previously, mTORC1 and STAT3 are associated with autophagy and apoptosis. Thus, in this study, we hypoth- esized that mTORC1/STAT3 axis mediates pyroptosis via regulating expression of pyroptosic genes in human macrophages. We examined whether mTORC1 signaling regulates S. aureus-induced pyroptosis and whether the inhibition of the mTORC1/STAT3 axis causes pyroptosis in human macrophages. Collectively, our study demon- strates that mTORC1 and STAT3 have critical functions in the regulation of S. aureus-induced pyroptosis and that inactivation of the mTORC1/STAT3 axis causes pyropto- sis in human macrophages. Our results implicate a novel function of the mTORC1/STAT3 axis in regulating cell death and provide insights into the mechanism by which pyroptosis is governed in professional phagocytic cells. Methods Reagents and antibodies The anti-GSDMD (Cat# ab215191), anti-caspase-1 (Cat# ab1872), anti-S6 (Cat# ab184551), anti-4EBP1 (Cat# ab2606), anti-p-mTOR (Ser2448) (Cat# ab32028), anti- mTOR (Cat# ab10926), goat anti-rabbit (Cat# ab136817), and anti-mouse (Cat# ab205719) secondary antibod- ies were purchased from Abcam (Abcam plc 330 Cam- bridge Science Park, Cambridge, UK.) The anti-p-S6 (Ser240/244) (Cat# 5346s), anti-STAT3 (Cat# 4904s) anti- p-4EBP1 (Thr37/46) (Cat# 2855s), and anti-p-STAT3 Yao et al. Cell Commun Signal (2020) 18:187 Page 3 of 18 (Cat# 4113s) were purchased from Cell Signaling Tech- nology (Cell Signaling Technology, Inc., Beverley, MA, USA). The anti-GSDMD (Cat# abs128820) was purchased from Absin (Absin Co., Ltd. Shanghai, China). The goat FITC-conjugated anti-rabbit IgG (Cat# 115-095-003) and FITC-conjugated anti-mouse IgG (Cat# 115-095-146) were purchased from Jackson (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA). The anti-β- actin (Cat# A5441) was purchased from Sigma (Sigma- Aldrich, Inc. St. Louis, MO, USA). (PEPROTECH from PEPROTECH LPS (Cat# L2630), Stattic (Cat# S7947), CFSE (Cat# 21888), and PMA (Cat# P1585) were purchased from Sigma-Aldrich, Inc. (St. Louis, MO, USA). Rapamycin (Cat# 53123-88-9) was purchased from (Gene Opera- tion, Ann Arbor, MI, USA). IFN-γ (Cat# 121527) was Inc., purchased Rocky Hill, USA). β-mercaptoethanol (Cat# M8211) was obtained from Solarbio (Solarbio Science and Technol- ogy, Co., Ltd. Beijing, China). Dil (Cat# KGMP002) was purchased from KeyGEN (KeyGEN BioTECH, Co., Ltd. Jiangsu, China). DAPI (Cat# C1005) was acquired from Beyotime (Beyotime Biotechnology, Co., Ltd. Shanghai, China). Alexa Fluor® 594 Phalloidin (Cat# A12381) was purchased from Invitrogen (Invitrogen, Carlsbad, New Mexico, USA). Cytochalasin B (Cat# HY-16928) was pur- chased from MCE (MedChemExpress, New Jersey, Mon- mouth, USA). Cell culture leukemia mononuclear Human myeloid (THP-1) cells and THP-1-derived macrophages were cultured in RPMI-1640 medium (Hyclone Laboratories, Inc. Logan, UT, USA) that was supplemented with 10% fetal bovine serum (BI Biological Industries, Beit Haemek Israel), 100 U/mL penicillin G, 100 mg/mL streptomy- cin (Sigma-Aldrich, Inc. St. Louis, MO, USA), and 0.1% β-mercaptoethanol (Solarbio Science and Technology, Co., Ltd. Beijing, China). To induce the classical (M1) polarization program, the medium was replaced with fresh medium without 10% fetal bovine serum and with PMA (100 ng/mL) for 24 h and then supplemented with 2.5 ng/mL IFN-γ and 100 ng/mL LPS for 24 h. Cells were cultured at 37 °C in humidified air with 5% CO2. Macrophages were infected by S. aureus (ATCC25923) at a multiplicity of infection (MOI) of 25:1 (bacteria to macrophages). Spread plate method Macrophages were infected with S. aureus for 3 h at an MOI of 25, and the extracellular bacteria were killed and lysed by antibiotics and lysozyme for 2 h. Monolayer macrophages were lysed, and the number of intracellular bacteria was determined by spread plate method. Bacteria and macrophage staining The macrophages were seeded onto a slide and incu- bated overnight. Bacteria (S. aureus) were washed with PBS and then incubated with CFSE (5(6)-car- boxyfluorescein diacetate N-succinimidyl ester) at 4 °C for 15 min. The stained bacteria were centrifuged for 10 min at 3000×g at 4 °C 3 times. Macrophages were infected by the stained bacteria at an MOI of 25, washed with PBS 3 times, and then fixed with 4% para- formaldehyde for 20 min. After being treated with 1% Triton X-100 for 5 min, the macrophages were stained with Alexa Fluor® 594 Phalloidin for 1 h in the dark, washed with PBS 3 times, and counterstained with 100 μL DAPI for 3 min to assess the nuclear morphol- ogy. Finally, the slide was mounted with glycerinum for examination under a laser scanning confocal micro- scope (LSCM) (NIKON AIR, Nikon Corp., Tokyo, Japan). Immunofluorescence assay (IFA) The cells were seeded onto a slide, incubated overnight, washed with PBS, and fixed with 4% paraformaldehyde for 15 min. After treatment with 1% Triton X-100 for 10 min, the cells were blocked with 1% BSA for 1 h, stained with GSDMD/Caspase-1 primary antibody overnight at 4 °C, and incubated with FITC-labeled goat anti-rabbit/mouse IgG for 1 h at room temperature. Dil was added at 37 °C for 12 min to stain the membrane, and nuclear was stained with 100 μL DAPI for 3 min. Finally, the slide was mounted with glycerinum and examined under an LSCM (NIKON AIR, Nikon Corp., Tokyo, Japan). TEM and SEM Macrophages were infected by bacteria (S. aureus) for 3 h at an MOI of 25, and the extracellular bacteria were killed and lysed by antibiotics and lysozyme for 2 h. The infected macrophages were washed with PBS 3 times centrifuged for 10 min at 3000×g at 4 °C, and fixed with 2.5% glutaraldehyde overnight. Finally, the cells were embedded in 4% AGAR and fixed with 2.5% glutaraldehyde overnight. The samples were examined by TEM (Hitachi HT7700, Hitachi, Ltd., Tokyo, Japan) to detect intracellular bacteria. For SEM, macrophages were seeded onto a slide and incubated overnight. The macrophages were infected by bacteria or treated with rapamycin, washed with PBS, and fixed with 2.5% glu- taraldehyde for 4 h. After being air-dried, the slides were examined by SEM (Hitachi S-4800, Hitachi, Ltd., Tokyo, Japan) to detect GSDMD-NT pores on the cytomembrane. Yao et al. Cell Commun Signal (2020) 18:187 Page 4 of 18 Western blot Macrophages were washed with cold PBS and lysed in cell lysis buffer. The lysis buffer comprised 50 mM Tris (pH 7.4), 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 1% PMSF, and phosphatase inhibitors. Equal amounts (40 μg) of protein were elec- trophoresed on 10% (w/v) sodium dodecyl sulfate–pol- yacrylamide gels, transferred to polyvinylidene fluoride membranes, and incubated with the primary anti- body. Peroxidase-conjugated secondary antibody and enhanced chemiluminescence (ECL) reagent were used to detect the signals with the Western Blotting System (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA). Immunohistochemical staining Cells were seeded onto slides, incubated overnight, washed with PBS, and fixed with 4% paraformaldehyde for 15 min. For immunohistochemical detection of pro- teins (1000× magnification), antigens were detected using UltraSensitive™ S-P Immunohistochemistry hypersensitive kits (mouse/rabbit) (MXB Biotechnol- ogy Co. Ltd., Fujian, China) with hematoxylin (MXB Biotechnology Co. Ltd., Fujian, China) for counterstain- ing. A biotin-streptavidin peroxidase-based method was used to detect the primary antibodies. ELISA Levels of IL-1β and IL-18 in the supernatant was quan- tified by enzyme linked immunosorbent assay (ELISA) using NeoBioscience ELISA kits (Neobioscience Tech- nology Co. Ltd., Shenzhen, China). The amount of culture media and intracellular α-hemolysin were measured via standard Mbbiolog ELISA kit (Mbbiol- ogy Biological Technology Co. Ltd., Jiangsu, China) according to the manufacturer’s guidelines. Absorbance at 450 nm and 630 nm was read on a Varioskan Flash Multimode Reader (Thermo Fisher Scientific, Pitts- burgh, PA, U.S.A.). All measurements were performed in triplicate, and the mean value of 3 independent measurements was used for statistical analysis. Confocal microscopy and light microscopy Mounted IFA slides were observed under an LSCM (NIKON AIR, Nikon Corp., Tokyo, Japan). The mac- rophages phenotype, IHC slides, and large bubbles on the macrophages were observed under a light micro- scope (NIKON Eclipse80i, Nikon Corp., Tokyo, Japan). qRT‑PCR analysis Quantitative real-time PCR (qPCR) was performed to determine the levels of CASP1, and GSDMD in mac- rophages in the treatment and control groups. For the bacterial infection, macrophages were infected with S. aureus for 3 h at an MOI of 25, and the extracellu- lar bacteria were killed and lysed by antibiotics and lysozyme for 2 h. For the inhibitor treatment, mac- rophages were treated with 100 nM rapamycin for 6 h or 100 μM Stattic for 1 h, and the total RNA was extracted from untreated and treated cells. Total RNA was prepared with the RNAiso Plus reagent accord- ing the manufacturer’s instructions (Takara Co. Ltd., Dalian China). Briefly, the cells were washed with PBS and lysed in RNAiso Plus, and chloroform was added to the cell lysates for homogenization. The top aqueous layer was transferred to a new tube after centrifuga- tion, and isopropanol was added to the supernatant and mixed well. Total RNA was precipitated by centrifuga- tion, and the pellet was dissolved in RNase-free water. RNA quantities over 600 ng/μL and a purity of 1.90– 2.0, based on the 260/280 ratio, were used to synthesize cDNA. mRNA was reverse-transcribed with oligo (dT)12–18 primer using the AMV 1st Strand cDNA Synthesis Kit (Takara Co. Ltd., Dalian China). ACTB was selected as the internal control gene. The primer sequences were as follows (5′–3′): CASP1: forward, CGA CAA GGT CCT GAA GGA GAA GAG reverse, CGT GTG CGG CTT GAC TTG TCC ATT GSDMD: TGC TGC reverse, CTG AGG GGA TGG TGA CCG TCT TCT ACTB: forward, TCA CCA ACT GGG ACG ACA T reverse, GCA CAG CCT GGA TAG CAA C forward, TGT ACG TGG TGA CTG AGG The KAPA SYBP® FAST qPCR Kit Optimized for LightCycler® 480 (KAPA BIOSYSTEMS, Inc., Boston, MA, U.S.A.) was used for the PCR according to the manufacturer’s instructions. The program comprised an initial denaturation step at 95 °C for 5 min; 40 cycles of 95 °C for 5 s, 54 °C for 30 s, and 72 °C for 20 s; and a final extension of 72 °C for 10 min. Three technical rep- licates were run in each experiment. 2−ΔΔCT values were calculated to determine expression levels, and the qPCR results were analyzed by student’s t-test to compare the expression between untreated and treated groups. Three independent experiments were performed. Caspase‑1 activity assay The caspase-1 activity was measured using a Caspase1 Activity Assay kit (Beijing solarbio science and technol- ogy co., ltd. Beijing, China). According to the manufac- turer’s instructions, about 50 μg protein from cells was mixed with synthetic tetrapeptide Ac-YVAD-pNA and Yao et al. Cell Commun Signal (2020) 18:187 Page 5 of 18 incubated at 37 °C for 2 h. A standard curve was prepared using the pNA standard. The absorbance was determined at 405 nm with a 96-well plate reader and the caspase1 activity was normalized for total proteins of cell lysates. Live/dead cell imaging assay It is performed using the Live & Dead® Viability/Cyto- toxicity Assay Kit (US Everbright, Inc., Suzhou, China). Macrophages were infected with S. aureus and incubated with 2 μM Calcein AM and 4 μM PI solution. Incubation at room temperature for 15–20 min and observe labeled cells under Live cell Imaging System (NIKON TI-E, Nikon Corp., Tokyo, Japan.) Phagocytosis assays Macrophages were treated with 100 nM rapamycin for 6 h or 10 μg/mL Cytochalasin B for 30 min, respectively, cells and S. aureus were co-cultured at an MOI of 25 for 3 h to enable phagocytosis. Phagocytosis was stopped by placing on ice and washing the macrophages cul- tures twice to remove non-phagocytosed bacteria. The extracellular bacteria were killed and lysed by antibiotics and lysozyme. Monolayer macrophages were lysed, and the number of intracellular bacteria was determined by spread plate method. Statistical analyses Statistical analyses were conducted using SPSS PASW Statistics for Windows, v18.0 (SPSS Inc.: Chicago, IL, USA). Data were analyzed using standard paramet- ric statistics and one-way ANOVA, followed by Tukey’s method. Data are expressed as mean ± SD. The results are presented as the average of at least 3 independent experiments. Statistical significance was accepted when p ≤ 0.05. Results Phagocytosis of S. aureus by macrophages M1-polarized human MDMs increase their caspase-1 activity, release of IL-1β, and loss of cell membrane integ- rity compared with the M2 program [7, 33]. To charac- terize S. aureus-induced pyroptosis in macrophages, THP-1 cells were differentiated into macrophages in cul- ture with 100 ng/mL PMA (phorbol myristate acetate) for 24 h. Then, the macrophages were treated in differ- entiation medium with 2.5 ng/mL IFN-γ and 100 ng/mL LPS for 24 h to transform them into M1-polarized mac- rophages (Additional file 1: Figure S1a). The M1 mac- rophages released more TFN-α and IL-1β than the Mφ phenotype (Additional file 1: Figure S1b), indicating that M1-polarized macrophages were successfully induced by IFN-γ and LPS. To measure the internalization of S. aureus by mac- rophages, we infected M1 phenotype macrophages with S. aureus for 3 h and then killed and lysed the extracel- lular bacteria with antibiotics and lysozyme. The bacte- ria were evaluated intracellularly and extracellularly by bacterial colony count—7.5 ± 0.12 × 103 CFU/mL were counted in whole-cell lysates, whereas extracellular bacteria were not found in the culture medium (Addi- tional file 2: Table S1). Further, intracellularly stained bacteria were observed under a laser scanning confocal microscope (LSCM) (Fig. 1a), and under a transmission electron microscope (TEM), S. aureus adhered to the cell membrane and invaded the macrophages (Fig. 1b). These data indicate that S. aureus was internalized by the macrophages. Intracellular S. aureus induces pyroptosis of macrophages induces NLRP3 Considering S. aureus can induce NLRP3-mediated sign- aling triggering caspase-1 activation and programmed necrosis through the production of a pore-forming toxin in macrophages [7, 34], to eliminate the possibility of pore-forming toxin inflammasome- dependent pyroptosis during S. aureus infection period, α-hemolysin as the most prominent [35], we first evalu- ated the levels of α-hemolysin (Hla) in cell medium and in macrophages. The results showed that α-hemolysin was not detectable both in cell medium and macrophages (Additional file 3: Figure S2a), indicating α-hemolysin was not accumulated during the 3 h infection period. To determine whether inhibition of phagocytosis was specifically restricted to the engulfment of S. aureus and has effect on pyroptosis, macrophages were treated with Cytochalasin B, an inbibitor of phagocytosis [36], and intracellular bacteria were detected. The results showed that the number of intracellular bacteria were statisti- cally decreased in Cytochalasin B treated cells (Addi- tional file 3: Figure S2b). Furthermore, the activation of NLRP3 and Caspase-1, the expression of GSDMD-NT, IL-1β, and IL-18 triggered by S. aureus were abrogated by Cytochalasin B (Fig. 2a, b). These data indicate that blockage of S. aureus phagocytosis could abrogate the pyroptosis induction. Intracellular bacteria may lead to cell pyroptosis, which is accompanied by caspase-1 activation, follow- ing the generation of GSDMD-NT fragment to form membrane pores and the release of IL-18 and IL-1β [37]. To determine whether pyroptosis is induced by S. aureus infection in human macrophages, we first measured the expression of several markers by western blot, immunofluorescence, and immunohistochemis- try staining. NLRP3, Caspase-1 (Casp-1 p20) (Fig. 2c, d) and GSDMD-NT (Fig. 2e, f and Figure S2c), were upregulated in cells that harboured intracellular S. Yao et al. Cell Commun Signal (2020) 18:187 Page 6 of 18 Fig. 1 Staphylococcus aureus invades macrophages. S. aureus infected macrophages for 3 h, and then, the macrophages were cultured in medium supplemented with antibiotics and lysozyme to kill and lyse the extracellular bacteria. The intracellular bacteria were detected by staining and laser confocal microscopy and TEM. a The intracellular S. aureus (green) stained with CFSE was observed by laser confocal microscopy; macrophages nuclei were co-stained with DAPI (blue), and actin was stained with phalloidin (red). Scale bars represent 50 μm. b Bacteria were attached to the cell membrane and engulfed, based on micrographs obtained by TEM; several important observations are magnified. Red arrows indicate S. aureus. Scale bars represent 2 μm. a′ Control, non-infected human macrophages. b′ The bacteria were located predominantly in the plasma membrane; macrophages were allowed to engulf bacteria and initiate phagocytosis. c′ S. aureus was engulfed by macrophages via the formation of typical phagocytic cups Yao et al. Cell Commun Signal (2020) 18:187 Page 7 of 18 aureus. By ELISA, compared with control, IL-18 and IL-1β levels were significantly increased (p < 0.05) in the medium of cells that contained bacteria (Fig. 2g). Next, to examine the morphological characteristics of macrophages that were infected by S. aureus, a scan- ning electron microscope (SEM) was used to observe the GSDMD-NT pore structures on the plasma mem- brane. Pore structures with ~ 30-nm diameters were observed (Fig. 2h). Moreover, when the cell membrane was damaged, the exocytic patching mechanism was initiated to repair the membranes by removing the pores in cells with few gasdermin pores [38]. We exam- ined the typical budding vesicles that act during the shedding of damaged plasma membranes using LSCM, interesting, we found that budding vesicles were pre- sent in macrophages with intracellular S. aureus but not control cells (Fig. 2i). Further, once GSDMD pores form at levels that exceed the cell’s compensatory abili- ties, cells begin to swell, with characteristic large bub- bles that protrude from the plasma membrane [39]. Thus, a light microscope was used to examine the bubbles that formed on the macrophages—large bub- bles were observed on cells with intracellular S. aureus versus control cells (Additional file 3: Figure S2d). By video, cell swelling with characteristic large bubbles and lytic cells were seen (Video Abstract). Notably, by live/dead cell imaging assay, S. aureus was found to induce macrophage death (Additional file 4: Video 1). Taken together, these results suggest that intracellular S. aureus induces pyroptosis in human macrophages. Rapamycin promotes S. aureus‑induced pyroptosis in macrophages As discussed, mTORC1 regulates autophagy and apop- tosis. Thus, we hypothesized that S. aureus-induced pyroptosis is associated with mTORC1 signaling. We treated S. aureus-invaded cells with rapamycin and then examined the expression of caspase-1 (Casp-1 p20), NLRP3 and GSDMD-NT by western blot, immunofluo- rescence, and immunohistochemistry staining and the levels of IL-18 IL-1β in the cell medium by ELISA. Rapa- mycin upregulated the S. aureus-induced expression of caspase-1 (Casp-1 p20) and NLRP3 (Fig. 3a, b), GSDMD- NT (Fig. 3c, d and Additional file 5: Figure S3a) and the release of IL-18 and IL-1β Fig. 3e) (p < 0.01). Further, rapamycin promoted the formation of S. aureus-induced GSDMD-NT membrane pores (Fig. 3f ) and bubbles (Fig. 3g). Moreover, to assess the effect of rapamycin on macrophage uptake of S. aureus, macrophages that had engulfed bacteria were analysed by bacterial colony count. Notably, rapamycin did not influence the ability of macrophages to phagocytose bacteria (Additional file 5: Figure S3b). These data suggest that rapamycin acceler- ates S. aureus-induced pyroptosis in macrophages and that mTORC1 is involved in pyroptosis. mTORC1 inhibition causes macrophage pyroptosis Based on the data above, we speculated that rapamy- cin induces pyroptosis in macrophages. To this end, we treated cells with 100 nM rapamycin for 6 h. We first examined mTORC1 signaling and found the phospho- rylation of mTOR, S6, and 4EBP1 decreased (Additional file 6: Figure S4a), indicating that the mTORC1 pathway is inhibited by rapamycin. By western blot, immunofluo- rescence, and immunohistochemistry, NLRP3, caspase-1 (Casp-1 p20), (Fig. 4a, b) and GSDMD-NT (Fig. 4c, d and Additional file 6: Fig. S4b) were upregulated by rapa- mycin, and the release of IL-18 and IL-1β was enhanced by rapamycin (Fig. 4e) (p < 0.01). Meanwhile, using a colorimetric assay to assess caspase-1 activity during (See figure on next page.) Fig. 2 Staphylococcus aureus triggers pyroptosis in human macrophages. Cells were infected with S. aureus for 3 h (MOI 25:1), and the pyroptotic characteristics were examined, including pyroptosic protein markers, inflammatory cytokines release, and morphology. a Expression of pyroptosis-related proteins in response to S. aureus invasion and Cytochalasin B was assessed by western blot. NLRP3, caspase1-p20 and GSDMD-NT proteins were less prominent than S. aureus invasion after Cytochalasin B treated. b The levels of IL-18 and IL-1β were quantified by an ELISA. The levels of IL-18 and IL-1β were less prominent than S. aureus invasion after Cytochalasin B treated. c Western blot analysis showed S. aureus enhanced NLRP3 and caspase1-p20 expression in infected macrophages. d Immunofluorescence assay showed greater caspase 1 expression in infected macrophages than in the control. Representative confocal microscopy images of caspase 1 expression (Green) in cells that were co-stained with DAPI (blue). Scale bars represent 10 μm. e GSDMD-NT expression was examined in S. aureus-infected macrophages and controls by western blot. S. aureus invasion induced greater GSDMD-NT expression in infected macrophages. f Immunofluorescence assay showed greater GSDMD-NT expression in infected macrophages than in the control. Representative confocal microscopy images of GSDMD-NT expression (green) in cells that were co-stained with DAPI (blue). Scale bars represent 10 μm. g IL-18 and IL-1β levels were determined by ELISA. S. aureus invasion induced more release of IL-18 and IL-1β in infected macrophages. h Scanning electron microscopy of GSDMD-NT pores on plasma membrane in S. aureus-infected macrophages. Red arrow indicates GSDMD-NT pore. Scale bar, 500 nm. i GSDMD-NT (green) in cells co-stained with Dil (red) as a membrane marker. Representative confocal microscopy images of S. aureus-infected macrophages and control cells by immunofluorescence assay. The purple arrows indicate the necks of budding vesicles, indicating shedding of wounded plasma membrane in S. aureus-infected macrophages. Scale bars represent 1 μm. The resolved bands were quantified using Gel-Pro Analyzer 4.0 (Media Cybernetics, Inc., Rockville, MD, USA). Fluorescence intensity of the immunofluorescent was measured by imaging analysis software (NIS-Elements Viewer, Nikon Instruments Inc. Shanghai, China). p < 0.05; *p < 0.01. n 3 independent experiments. Error bar indicates SD = Yao et al. Cell Commun Signal (2020) 18:187 Page 8 of 18 rapamycin treatment, we found that caspase-1 activ- ity was significantly increased (Fig. 4f ). As in S. aureus- invaded cells, rapamycin induced the formation of GSDMD-NT membrane pores (Fig. 4g) by SEM. Budding vesicles from the plasma membrane were observed under an LSCM (Fig. 4h), acting as a membrane damage repair mechanism and appearing early during pyroptosis. Bub- bles were also observed under a light microscope (Fig. 4i). Yao et al. Cell Commun Signal (2020) 18:187 Page 9 of 18 Further, by video, cell swelling with characteristic large bubbles and lytic cells were seen (Additional file 7: Video 2). These data indicate that inhibition of mTORC1 causes pyroptosis in macrophages and that mTORC1 signaling has inhibitory effects on pyroptosis in macrophages. mTORC1 regulates the expression of CASP1 and GSDMD through STAT3 in macrophages The experiments above demonstrated that rapamy- cin increased the protein expression levels of caspase-1 (Casp-1 p20) and GSDMD-NT; previous study has shown that STAT3 is associated with apoptosis and autophagy and thus [28], we hypothesized that mTORC1 may regu- late these proteins through STAT3 in macrophages. To determine whether mTORC1 regulates the expres- sion of CASP1 and GSDMD via STAT3, we first treated macrophages with 100 nM rapamycin for 6 h and found that repamycin reduced the phosphorylation of STAT3, indicating STAT3 activation was inhibited (Fig. 5a). Then, we measured the mRNA levels of these pyroptosic genes by qRT-PCR, demonstrating that rapamycin upregulates CASP1 and GSDMD (Fig. 5b) (p < 0.01). These data sug- gest that mTORC1 regulates the expression of pyroptosic genes, likely through STAT3 in macrophages. STAT3 inhibition causes macrophage pyroptosis through upregulations of CASP1 and GSDMD To determine whether STAT3 affects the expression of CASP1, and GSDMD, we treated cells with Stattic, a selec- tive inhibitor of STAT3 activation and dimerization. Stattic inhibited STAT3 phosphorylation in macrophages (Fig. 6a). Stattic increased the mRNA levels of CASP1 and GSDMD (Fig. 6b), indicating that their expression was governed by STAT3 in macrophages. Further, cells were pretreated with Stattic and exposed to S. aureus produced more NLRP3, Casp-1 p20 and GSDMD-NT comparing to only Stattic treated and untreated cells (Fig. 6c). By ELISA, the levels IL-18 and IL-1β in the cell medium increased in treated cells (Fig. 6d) (p < 0.01). In addition, Stattic treatment significantly induced caspase-1 activation (Fig. 6e). These data suggest that inhibition of STAT3 induces pyropto- sis in macrophages. By video, pseudopod contracting and cell rounding (Additional file 8: Video 3), cell swelling with characteristic large bubbles and lytic cells were observed (Fig. 6f and Additional file 9: Video 4). Thus, STAT3 is criti- cal in pyroptosis, controlling the expression of pyroptosic genes in human macrophages. Inactive mTORC1 prevents the nuclear localization of STAT3 in macrophages Based on the results above, we reasoned that mTORC1 governs the nuclear localization of STAT3 to regulate the expression of pyroptosic genes. To this end, immunofluo- rescence was performed to confirm whether mTORC1 promotes the nuclear localization of STAT3. Cells were treated with rapamycin or Stattic as a positive control. The nuclear localization of STAT3 was prevented in cells that were treated with either compound, compared with con- trol (Fig. 7a, b), indicating that the nuclear localization of STAT3 is controlled by mTORC1. These results demon- strate that mTORC1 signaling negatively regulates human macrophage pyroptosis by regulating the nuclear localiza- tion of STAT3 and pyroptosic gene expression (Fig. 7c). Discussion During cell death, cell-intrinsic effector functions are coordinated to restrict infection and resolve innate and adaptive immune responses. The most widely used clas- sification of cell death had consisted of 2 types: apop- tosis and necrosis [40], but in recent years, necroptosis and pyroptosis have been confirmed [41]. Apoptotic cells retain plasma membrane integrity, versus pyrop- tosis or necroptosis is the rupture of cell lytic, which weakens the integrity of the plasma membrane and allows the influx of extracellular ions and fluid, leading to cell swelling [42]. In lytic pyroptosis, variety of cytokines are released, rendering it an inflammatory event [43]. Pyroptosis is a (See figure on next page.) Fig. 3 Rapamycin promotes Staphylococcus aureus-induced pyroptosis in macrophages. Macrophages were pretreated with rapamycin (100 nM) for 6 h and then infected with S. aureus for 3 h. The pyroptotic characteristics were determined, including pyroptosic protein markers, inflammatory cytokine release, and morphology. (a Western blot analysis of S. aureus-induced NLRP3 and caspase1-p20 expression. b Immunofluorescence assay of S. aureus-induced caspase 1 expression. Representative confocal microscopy images of caspase 1 (green) in cells that were co-stained with DAPI (blue). Scale bars represent 10 μm. c GSDMD-NT expression by western blot. d GSDMD-NT expression by immunofluorescence assay. Representative confocal microscopy images of GSDMD-NT (green) in cells that were co-stained with DAPI (blue). Scale bars represent 10 μm. e Macrophages were pretreated with rapamycin for 6 h and then infected with S. aureus for 3 h. The levels of IL-18 and IL-1β in cell culture medium were determined by ELISA. f GSDMD-NT pores on plasma membrane by scanning electron microscopy. Red arrow indicates GSDMD-NT pore. Scale bar, 500 nm. g Large bubbles by light microscopy, as indicated by red arrows in pyroptotic cells. Scale bars represent 1 μm. The resolved bands were quantified using Gel-Pro Analyzer 4.0 (Media Cybernetics, Inc., Rockville, MD, USA). Fluorescence intensity of the immunofluorescent was measured by imaging analysis software (NIS-Elements Viewer, Nikon Instruments Inc. Shanghai, China). p < 0.05; *p < 0.01. n Error bar indicates SD 3 independent experiments. = Yao et al. Cell Commun Signal (2020) 18:187 Page 10 of 18 form of caspase-mediated cell death, and its regulatory mechanism is poorly understood. Studies have demon- strated that rapamycin induces apoptosis in peritoneal carcinomatosis [44] and necrosis in cardiac cells [45], and mTORC1 negatively regulates autophagy by inhibiting ULK [24]. mTORC1 is a central regulator of apoptosis, necrosis, and autophagy. However, the regulatory func- tion of mTORC1 in pyroptosis has not been reported. In this study, we found that S. aureus induces pyroptosis in human macrophages and that the inhibition of mTORC1 by rapamycin promotes S. aureus-induced pyroptosis. Yao et al. Cell Commun Signal (2020) 18:187 Page 11 of 18 These data suggest that mTORC1 regulates pyroptosis in macrophages. Pyroptosis was initially observed in macrophages that were infected with Salmonella typhimurium or Shigella flexneri [46, 47]. Cytosol-invasive bacteria, such as Lis- teria monocytogenes, induce pyroptosis [48]. Of note, S. aureus pore-forming toxins (PFTs) induce NLRP3- mediated signaling, triggering caspase-1 activation and pyroptosis in human and mouse monocytic cells [49]. In our study, the effect of PFTs on pyroptosis have been eliminated and pyroptosis can be abrogated by block- age of S. aureus phagocytosis. These data indicate that S. aureus internalized macrophages, and intracellular S. aureus induced pyroptosis in professional phagocytic cells. Furthermore, inactive mTORC1 did not affect the ability of macrophages to phagocytose bacteria, our data may reveal the biological role of pyroptosis in S. aureus invasion. In general, plasma membrane damage can be repaired efficiently in macrophages, for which endocy- tosis, membrane patching, and extracellular budding can be used [50]. Once gasdermin pores are present in numbers that exceed the cell’s compensatory abili- ties, the cells swell and the plasma membrane separates from the cytoskeleton in large fluid-filled bubbles and cell lytic dying [39]. In the present study, GSDMD-NT pore, swelling, and membrane rupture during pyropto- sis—were observed. Further, our video showed pyrop- tosic cells swelling with characteristic large bubbles and eventually lysing with rapamycin or Stattic treatment. This morphological evidence demonstrates that S. aureus infection, mTORC1 inhibition, and STAT3 inhi- bition induce macrophage pyroptosis. In our previous work, mTORC1 signaling was initi- ated by peptidoglycan (PGN) from S. aureus in mouse macrophages [51]. In the present study, we treated S. aureus-infected macrophages with rapamycin to inhibit mTORC1 and examine how S. aureus-induced pyroptosis is altered. We were surprised to find that S. aureus-induced pyroptosis was promoted by rapa- mycin in macrophages, for which there are 2 possibili- ties: rapamycin enhances S. aureus to induce pyroptosis and mTORC1 inhibition causes pyroptosis, in which case mTORC1 inactivation is an independent factor for macrophage pyroptosis. In rapamycin-treated cells, inactivation upregulated pyroptosic pro- mTORC1 teins, including CASP1 and GSDMD, release of IL-1β, and induced macrophage pyroptosis. Several studies in cells of mice and human reported that the negative regulation of inflammasome by rapamycin. Muhamuda et al. (2017) showed that rapamycin treatment (10 μM, 24 h) of lethal Ixodes ovatus ehrlichia (IOE) -infected WT-BMM attenuated production of IL-1β, which is the gold standard readout of inflammasome activity [52]. In addition, another group have observed that mTORC1 signaling inhibition by rapamycin (100 nM, 24 h) sup- pressed IL-1β secretion in these cells monocyte-derived macrophages (MDMs) [53]. In contrast, Rojas Márquez et al. [54] showed that mTORC1 inhibition by rapam- ycin (100 nM, 90 min) upregulation of IL-1β produc- tion in macrophages; Chimin et al. [55] reported that adipocyte mTORC1 deficiency promotes adipose tissue inflammation and NLRP3 inflammasome activation in mice with raptor deletion, which are in line with our findings of mTORC1 inhibition by rapamycin (100 nM, 6 h) upregulation of IL-1β and caspase-1 productions in the present study. As a latent transcription factor, STAT3 modulates a range of target genes, including induced and repressed target genes that mediate cellular and organismal func- tions [56]. Moreover, STAT3 can be activated by LPS in human monocytes [30] and by PGN in mouse mac- rophages [51]. STAT3 regulates programmed cell death. Constitutive STAT3 activation leads directly to the induc- tion of BCL-X, which inhibits of apoptosis [25]. STAT3 functions by upregulating executes anti-autophagic (See figure on next page.) Fig. 4 Rapamycin causes pyroptosis in macrophages. Macrophages were treated with rapamycin (100 nM) for 6 h, and the pyroptotic characteristics were examined. a Western blot of NLRP3 and caspase1-p20 expression. b Immunofluorescence assay of caspase 1 expression. Representative confocal microscopy images of caspase 1 expression (green) in cells that were co-stained with DAPI (blue). Scale bars represent 10 μm. c Rapamycin upregulates GSDMD-NT expression. d Rapamycin upregulates GSDMD-NT expression in cells. Representative confocal microscopy images of caspase 1 expression (green) in cells that were co-stained with DAPI (blue) by immunofluorescence assay. Scale bars represent 10 μm. e Rapamycin increases the levels of IL-18 and IL-1β in cell culture medium. f Whole cell lysates were extracted from macrophages, Caspase-1 activity was determined by colorimetric assay and induced by rapamycin treatments. g Scanning electron microscopy of GSDMD-NT pores on plasma membrane in rapamycin-treated macrophages. Red arrow indicates GSDMD-NT pore. Scale bar, 500 nm. h GSDMD-NT (green) in cells co-stained with Dil (red) as membrane marker. Representative confocal microscopy images of rapamycin-treated macrophages and control cells by immunofluorescence assay. The purple arrows indicate the necks of budding vesicles in rapamycin-treated macrophages. Scale bars represent 1 μm. i Large bubbles by light microscopy, indicated by red arrows in pyroptotic cells. Scale bars represent 1 μm. The resolved bands were quantified using Gel-Pro Analyzer 4.0 (Media Cybernetics, Inc., Rockville, MD, USA). Fluorescence intensity of the immunofluorescent was measured by imaging analysis software (NIS-Elements Viewer, Nikon Instruments Inc. Shanghai, China). p < 0.05; *p < 0.01. n Error bar indicates SD 3 independent experiments. = Yao et al. Cell Commun Signal (2020) 18:187 Page 12 of 18 negative regulators of autophagy, such as MCL1, PIK3R1/ p55a, and PIK3R1/p50a [26], and by downregulat- ing essential autophagy genes, such as BECN1 and PIK3C3 [57]. Moreover, STAT3 negatively regulates gene expression in MEFs and cancer lines to control type I IFN-mediated antiviral response [58–60]. In fact, a whole-transcriptome profiling study showed that STAT3 acts as a transcriptional activator and suppressor, with a Yao et al. Cell Commun Signal (2020) 18:187 Page 13 of 18 Fig. 5 mTORC1 regulates the expression of CASP1 and GSDMD through STAT3 in macrophages. Macrophages were treated with rapamycin (100 nM) for 6 h, and the activity of mTORC1 and STAT3 and the expression of pyroptosic genes were examined. a Western blot analysis of mTORC1 activation and STAT3 activation. Phosphorylation of STAT3 was inhibited. b Expression of CASP1 and GSDMD, mRNA upregulated by rapamycin. The resolved bands were quantified using Gel-Pro Analyzer 4.0 (Media Cybernetics, Inc., Rockville, MD, USA). p < 0.05; *p < 0.01. n experiments. Error bar indicates SD 3 independent = comparable number of upregulated and downregulated genes in diffuse large B cell lymphoma (DLBCL) cells [61]. In our cases, STAT3 was inhibited by the selective inhibitor Stattic, resulting in the upregulation of pyrop- tosic genes, including CASP1, GSDMD and triggering pyroptosis in human macrophages. The nuclear localization of STAT3 comprises phos- phorylation, dimerization, and nuclear localization, can be blocked by Stattic [62, 63]. In addition, phosphoryla- tion of the Ser727 residue in the carboxyl transactivation domain might positively regulate STAT3 transcriptional activation; Dodd et al. demonstrated that STAT3 is phos- phorylated directly by mTORC1 on Ser727 [64–66], confirming that mTORC1 functions in STAT3 activation. In a previous study, we found that rapamycin inhibits STAT3 phosphorylation [67], indicating that mTORC1 can regulate STAT3 activation. In the present study, mTORC1 was inhibited by rapamycin, accompanied by STAT3 inactivation and upregulation of pyroptosic gene. In the regulation of pyroptosis, STAT3 consistently had inhibitory functions, in accordance with the function of mTORC1 in pyroptosis in human macrophages. Further, we found that rapamycin prevents the nuclear localiza- tion of STAT3. Thus, mTORC1 combines with STAT3 to form the mTORC1/STAT3 axis to control pyroptosis in human macrophages. Yao et al. Cell Commun Signal (2020) 18:187 Page 14 of 18 Fig. 6 Inhibition of STAT3 by Stattic upregulates expression of pyroptosic genes and causes pyroptosis in macrophages. Macrophages were treated with Stattic (100 μM) for 1 h to inhibite STAT3 activation. The expression of pyroptosic genes and pyroptosic protein markers and the release of inflammatory cytokines were examined. a Stattic inhibits STAT3 phosphorylation in cells. b Stattic enhances the expression of CASP1 and GSDMD mRNA in macrophages. c Stattic enhances NLRP3, caspase1-p20 and GSDMD expression, pyroptosis-related proteins were severally elevated in S. aureus infected cells which pretreated with Stattic. d Stattic increases the production of IL-18 and IL-1β in cell culture medium. e Caspase-1 activity was enhanced by Stattic treatment. f Stattic triggers pyroptosis. Red arrows indicate large bubbles in pyroptic cells. Scale bars represent 100 μm. The resolved bands were quantified using Gel-Pro Analyzer 4.0 (Media Cybernetics, Inc., Rockville, MD, USA). p < 0.05; p < 0.01. n experiments. Error bar indicates SD 3 independent = Yao et al. Cell Commun Signal (2020) 18:187 Page 15 of 18 Fig. 7 Inactive mTORC1 prevents the nuclear localization of STAT3 in macrophages. a Nuclear localization of STAT3 was prevented in cells treated with rapamycin. Scale bars represent 10 μm. The nuclear fluorescence intensity of STAT3 was shown. b Nuclear localization of STAT3 was prevented in cells treated with Stattic. Scale bars represent 10 μm. The fluorescence intensity of nuclear STAT3 was graphed. c The model of mTORC1 regulating pyroptosis through nuclear localization of STAT3 and pyroptosic gene expression in human macrophage. STAT3 is likely a negative regulator of pyroptosic gene expression. mTORC1 inhibition by rapamycin or STAT3 inhibition by Stattic prevents the nuclear localization of STAT3, upregulates pyrotosic gene expression, triggering human macrophage pyroptosis Yao et al. Cell Commun Signal (2020) 18:187 Page 16 of 18 Conclusion In summary, mTORC1 and STAT3 regulate human mac- rophage pyroptosis. The mTORC1/STAT3 axis is critical in pyroptosis in human macrophages. S. aureus invaded macrophages to induce pyroptosis; this process was pro- moted by rapamycin. mTORC1 regulates the expres- sion of pyroptosic genes, including CASP1 and GSDMD, through STAT3. Inhibition of mTORC1 prevents the nuclear localization of STAT3. The pathogenic effects of the inhibition of the mTORC1/STAT3 axis facilitates S. aureus-induced pyroptosis. This study demonstrates a regulatory function of the mTORC1/STAT3 axis in mac- rophage pyroptosis, constituting a novel mechanism by which pyroptosis is regulated in macrophages. Supplementary information Supplementary information accompanies this paper at https ://doi. org/10.1186/s1296 4-020-00677 -9. Additional file 1: Figure S1. PMA-induced differentiation of THP-1 to macrophages and further induction by LPS and IFN-γ into M1-polarized macrophages. THP-1 cells were treated with PMA (100 ng/mL) for 24 h to differentiate into Mφ -polarized macrophages and then stimulated with IFN-γ (2.5 ng/mL) and LPS (100 ng/mL) to polarize them into classical M1 macrophages. (a) THP-1 and THP-1-derived macrophages and Mφ and M1 macrophages. Scale bars represent 100 μm. (b) TFN-α and IL-1β in culture medium were determined by ELISA. M1 macrophages had more TFN-α and IL-1β than Mφ macrophages. p < 0.01. n 3 independent experi- ments. Error bar indicates SD. = Additional file 2: Table S1. Number of S. aureus in macrophages. Additional file 3: Figure S2. (a) The amount of α-hemolysin in culture media and intracellular were evaluated by ELISA. (b) Effect of Cytochalasin B on macrophage phagocytosis. (c) Immunohistochemical analysis of pyroptosic protein GSDMD-NT in S. aureus-infected human macrophages. Scale bars represent 10 μm. (d) Large bubbles observed by light micros- copy in S. aureus-infected macrophages. Large bubbles indicated by red arrows in pyroptotic cells. Scale bars represent 1 μm. p < 0.05; p < 0.01; NS 3 independent experiments. Error bar indicates SD. p > 0.05. n = = Additional file 4: Video 1. The Live/Dead assay was performed by simul- taneously monitoring the fluorescence after macrophages were infected with S. aureus for 3 h (MOI 25:1), Green, as live cell indicator and red, as dead cell indicator. Additional file 5: Figure S3. (a) Immunohistochemical analysis of GSDMD-NT in human macrophages. Macrophages were preincubated with rapamycin for 6 h and then infected with S. aureus. Scale bars rep- resent 10 μm. (b) Quantification of the number of CFU/ml of S. aureus in 3 cells treated with rapamycin and without rapamycin. NS independent experiments. Error bar indicates SD. p > 0.05. n = = Additional file 6: Figure S4. Rapamycin inhibits mTORC1 signaling and upregulates GSDMD-NT expression. (a) Macrophages were treated with 100 nM rapamycin for 6 h. (b) Rapamycin upregulates GSDMD-NT in macrophages. Scale bars represent 10 μm. The resolved bands were quantified using Gel-Pro Analyzer 4.0 (Media Cybernetics, Inc., Rockville, MD, USA). p < 0.05; **p < 0.01. n indicates SD. 3 independent experiments. Error bar = Additional file 7: Video 2. Macrophages were treated with 100 nM rapamycin for 6 h, and pyroptosis was filmed in macrophages by laser confocal microscopy. A representative field was recorded. Arrow indicates lysing dead cell. Additional file 8: Video 3. Macrophages were treated with 100 μM Stattic for 1 h, and pseudopod contracting, cell rounding, and lysing dead cells were filmed in macrophages by laser confocal microscopy. A representa- tive field was recorded. Arrow indicate pseudopod contracting, cell round- ing, and lysing dead cells. Additional file 9: Video 4. Macrophages were treated with 100 μM Stat- tic for 1 h, and pyroptosis was filmed in macrophages by laser confocal microscopy. A representative field was recorded. Arrow indicate lysing dead cells. Abbreviations MDMs: Monocyte-derived macrophages; THP-1: Human myeloid leukemia mononuclear; M1: Classically activated macrophages; mTORC1: Mamma- lian target of rapamycin complex 1; STAT3: Signal transducer and activator of transcription 3; PMA: Phorbol myristate acetate; IFN-γ: Interferon-γ; LPS: Lipopolysaccharide; S. aureus: Staphylococcus aureus; GSDMD: Gasdermin D; caspase-1: Cysteine-requiring aspartate protease 1; S6: Ribosomal protein S6; 4EBP1: The eukaryotic initiation factor 4E binding protein 1; β-actin: Actin beta; IL-1β: Interleukin-1 beta; IL-18: Interleukin-18; NLRP3: NLR family pyrin domain containing 3. Acknowledgements We thank Wuhan Servicebio Technology Co., Ltd for their support in the TEM studies. The authors thank Ms. Xiaoyang Jia and Dr. Xuan Wang for their gener- ous help in the live cell station assay. Authors’ contributions ZW and HH conceived this study, generated hypotheses, and designed the experiments. RY, YC, and HH performed the experiments. ZG, XC, YM, QJ, XY, and YW analyzed the data. RY, YC, and HH wrote, reviewed, and edited the manuscript. XL supervised the project. All authors read and approved the final manuscript. Funding This work was supported by the Natural Sciences Foundation of China (Nos. 31860309, 31960669, 31760675), the Natural Science Foundation of Inner Mongolia Autonomous Region of China (2019MS03022), the Science and Technology Major Project of Inner Mongolia Autonomous Region of China (No. 2020ZD15) and the Science and Technology Major Project of Inner Mon- golia Autonomous Region of China to the Stããate Key Laboratory of Repro- ductive Regulation and Breeding of Grassland Livestock (No. zdzx2018065). Availability of data and materials The datasets supporting the conclusions of this article are included within the article and its additional files. Ethics approval and consent to participate No ethics approval was required for this study that did not involve patients or patient data. Consent for publication All authors consent to publication. Competing interests The authors declare that they have no competing interests. Author details 1 State Key Laboratory of Reproductive Regulation and Breeding of Grassland Livestock, School of Life Sciences, Inner Mongolia University, Hohhot 010070, China. 2 School of Life Sciences, Jining Normal University, Jining 012000, China. 3 Research Center for Animal Genetic Resources of Mongolia Plateau, Inner Mongolia University, Hohhot 010070, China. Received: 9 March 2020 Accepted: 26 October 2020 Yao et al. Cell Commun Signal (2020) 18:187 Page 17 of 18 References 1. Shi J, Zhao Y, Wang K, et al. Cleavage of GSDMD by inflammatory cas- pases determines pyroptotic cell death. Nature. 2015;526(7575):660–5. 2. Bergsbaken T, Fink SL, Cookson BT. Pyroptosis: host cell death and inflam- 3. 4. 5. mation. Nat Rev Microbiol. 2009;7(2):99–109. Jorgensen I, Miao EA. Pyroptotic cell death defends against intracellular pathogens. Immunol Rev. 2015;265(1):130–42. Jorgensen I, Rayamajhi M, Miao EA. Programmed cell death as a defence against infection. Nat Rev Immunol. 2017;17(3):151–64. Ezquerro S, Mocha F, Frühbeck G, et al. Ghrelin reduces TNF-α-induced human hepatocyte apoptosis, autophagy and pyroptosis: role in obesity- associated NAFLD. J Clin Endocrinol Metab. 2019;104(1):21–37. 6. Wei Q, Zhu R, Zhu J. E2-Induced Activation of the NLRP3 inflammasome triggers pyroptosis and inhibits autophagy in HCC Cells. Oncol Res. 2019;27(7):827–34. 7. Accarias S, Lugo-Villarino G, Foucras G, et al. Pyroptosis of resident macrophages differentially orchestrates inflammatory responses to Staphylococcus aureus in resistant and susceptible mice. Eur J Immunol. 2015;45(3):794–806. 8. Molinari G, Rohde M, Guzman CA, et al. Two distinct pathways for the invasion of streptococcus pyogenes in non-phagocytic cells. Cell Micro- biol. 2000;2(2):145–54. 9. Gresham HD, Lowrance JH. Caver TE (2000) Survival of Staphylococ- cus aureus inside neutrophils contributes to infection. J Immunol. 2000;164(7):3713–22. 10. Foster TJ. The remarkably multifunctional fibronectin binding proteins of Staphylococcus aureus. Eur J Clin Microbiol Infect Dis. 2016;35(12):1923–31. 11. Garzoni C, Kelley WL. Staphylococcus aureus: new evidence for intracel- lular persistence. Trends Microbiol. 2009;17(2):59–65. 12. Lacoma A, Cano V, Moranta D, et al. Investigating intracellular persistence of Staphylococcus aureus within a murine alveolar macrophage cell line. Virulence. 2017;8(8):1761–75. 13. Murray PJ, Allen JE, Biswas SK, et al. Macrophage activation and polarization: nomenclature and experimental guidelines. Immunity. 2014;41(1):14–20. 14. Shirey KA, Cole LE, Keegan AD, et al. Francisella tularensis live vaccine strain induces macrophage alternative activation as a survival mecha- nism. J Immunol. 2008;181(6):4159–67. 15. Moghaddam AS, Mohammadian S, Vazini H, et al. Macrophage plastic- ity, polarization and function in health and disease. J Cell Physiol. 2018;233(9):6425–40. 16. Zhang H, Lin C, Zeng C, et al. Synovial macrophage M1 polarisation 27. Epling-Burnette PK, Liu JH, Catlett-Falcone R, et al. Inhibition of STAT3 signaling leads to apoptosis of leukemic large granular lymphocytes and decreased Mcl-1 expression. J Clin Invest. 2001;107(3):351–62. 28. Pensa S, Lloyd-Lewis B, Sargeant TJ, et al. Signal transducer and activa- tor of transcription 3 and the phosphatidylinositol 3-kinase regula- tory subunits p55α and p50α regulate autophagy in vivo. FEBS J. 2014;281(20):4557–67. 29. Kayleigh MD, Jian Y, Ming HS, et al. mTORC1 drives HIF-1α and VEGF-A signalling via multiple mechanisms involving 4E-BP1, S6K1 and STAT3. Oncogene. 2015;34(17):2239–50. 30. Weichhart T, Costantino G, Poglitsch M, et al. The TSC-mTOR signal- ing pathway regulates the innate inflammatory response. Immunity. 2008;29(4):565–77. 31. Kovacs SB, Miao EA. Gasdermins: effectors of pyroptosis. Trends Cell Biol. 2017;27(9):673–84. 32. Orning P, Weng D, Starheim K, et al. Pathogen blockade of TAK1 triggers caspase-8 dependent cleavageof gasdermin D and cell death. Science. 2018;362(6418):1064–9. 33. Alberto M, Antonio S, Silvano S, et al. The chemokine system in diverse forms of macrophage activation and polarization. Trends Immunol. 2004;25(12):677–86. 34. Xiaogang W, William JE, Jean CL, et al. Orchestration of human mac- rophage NLRP3 inflammasome activation by Staphylococcus aureus extracellular vesicle. Proc Natl Acad Sci USA. 2020;117(6):3174–84. Ichiro I, Naoko I, Georgia AW, et al. A Staphylococcus aureus pore-forming toxin subverts the activity of ADAM10 to cause lethal infection in mice. Nat Med. 2011;17(10):1310–4. 35. 36. Mattila JT, Munderloh UG, et al. Phagocytosis of the Lyme disease spirochete, Borrelia burgdorferi, by cells from the ticks, Ixodes scapularis and Dermacentor andersoni, infected with an endosymbiont, Rickettsia peacockii. J Insect Sci. 2007;7:58. 37. Jorgensen I, Lopez JP, Laufer SA, et al. IL-1β, IL-18, and eicosanoids promote neutrophil recruitment to pore-induced intracellular traps fol- lowing pyroptosis. Eur J Immunol. 2016;46(12):2761–6. 38. Rühl S, Shkarina K, Demarco B, et al. ESCRT-dependent membrane repair negatively regulates pyroptosis downstream of GSDMD activation. Sci- ence. 2018;362(6417):956–60. 39. Fink SL, Cookson BT. Caspase-1-dependent pore formation during pyrop- tosis leads to osmotic lysis of infected host macrophages. Cell Microbiol. 2006;8(11):1812–25. 40. Galluzzi L, Maiuri MC, Vitale I, et al. Cell death modalities: classification and pathophysiological implications. Cell Death Differ. 2007;14(7):1237–43. 41. Galluzzi L, Vitale I, Aaronson SA, et al. Molecular mechanisms of cell death: exacerbates experimental osteoarthritis partially through R-spondin-2. Ann Rheum Dis. 2018;77(10):1524–34. recommendations of the Nomenclature Committee on Cell Death 2018. Cell Death Differ. 2018;25(3):486–541. 17. Ortiz-Masiá D, Cosín-Roger J, Calatayud S, et al. M1 macrophages activate Notch Signalling in epithelial cells: relevance in Crohn’s disease. J Crohns Colitis. 2016;10(5):582–92. 42. Gong YN, Guy C, Olauson H, et al. ESCRT-III acts downstream of MLKL to regulate necroptotic cell death and its consequences. Cell. 2017;169(2):286–300. 18. Kubica M, Guzik K, Koziel J, et al. A potential new pathway for Staphylo- 43. Bliss-Moreau M, Chen AA, D’Cruz AA, et al. A motive for killing: coccus aureus dissemination: the silent survival of S. aureus phagocytosed by human monocyte-derived macrophages. PLoS ONE. 2008;3(1):e1409. 19. Sarantis H, Grinstein S. Subversion of phagocytosis for pathogen survival. effector functions of regulated lytic cell death. Immunol Cell Biol. 2017;95(2):146–51. 44. Song X, Dilly AK, Kim SY, et al. Rapamycin-enhanced mitomycin Cell Host Microbe. 2012;12(4):419–31. 20. Vergne I, Chua J, Lee HH, et al. Mechanism of phagolysosome biogen- esis block by viable Mycobacterium tuberculosis. Proc Natl Acad Sci USA. 2005;102(11):4033–8. 21. Saxton RA, Sabatini DM. mTOR signaling in growth, metabolism, and disease. Cell. 2017;169(2):361–71. 22. Abe K, Yano T, Tanno M, et al. mTORC1 inhibition attenuates necroptosis through RIP1 inhibition-mediated TFEB activation. Biochim Biophys Acta Mol Basis Dis. 2019;1865(12):165552. 23. Martina JA, Chen Y, Gucek M, et al. MTORC1 functions as a transcrip- tional regulator of autophagy by preventing nuclear transport of TFEB. Autophagy. 2012;8(6):903–14. 24. Ganley IG, du Lam H, Wang J, et al. ULK1.ATG13.FIP200 complex mediates mTOR signaling and is essential for autophagy. J Biol Chem. 2009;284(18):12297–305. 25. Liu Y, Xu H, An M. mTORC1 regulates apoptosis and cell proliferation in pterygium via targeting autophagy and FGFR3. Sci Rep. 2017;7(1):7339. 26. LevyDarnell DEDJ Jr. Stats: transcriptional control and biological impact. Nat Rev Mol Cell Biol. 2002;3(9):651–62. C-induced apoptotic death is mediated through the S6K1-Bad-Bak path- way in peritoneal carcinomatosis. Cell Death Dis. 2014;5:e1281. 45. Lassaletta AD, Elmadhun NY, Zanetti AV, et al. Rapamycin treatment of healthy pigs subjected to acute myocardial ischemia-reperfusion injury attenuates cardiac functions and increases myocardial necrosis. Ann Thorac Surg. 2014;97(3):901–7. 46. Monack DM, Raupach B, Hromockyj AE, et al. Salmonella typhimurium invasion induces apoptosis in infected macrophages. Proc Natl Acad Sci USA. 1996;93(18):9833–8. 47. Hilbi H, Moss JE, Hersh D, et al. Shigella-induced apoptosis is dependent on caspase-1 which binds to IpaB. J Biol Chem. 1998;273(49):32895–900. 48. Jorgensen I, Zhang Y, Krantz BA, et al. Pyroptosis triggers pore-induced intracellular traps (PITs) that capture bacteria and lead to their clearance by efferocytosis. J Exp Med. 2016;213(10):2113–28. 49. Craven RR, Gao X, Allen IC, et al. Staphylococcus aureus α-hemolysin acti- vates the NLRP3-inflammasome in human and mouse monocytic cells. PLoS ONE. 2009;4(10):e7446. 50. Jimenez AJ, Maiuri P, Lafaurie-Janvore J, et al. ESCRT machinery is required for plasma membrane repair. Science. 2014;343(6174):1247136. Yao et al. Cell Commun Signal (2020) 18:187 Page 18 of 18 51. Vangan N, Cao Y, Jia X, et al. mTORC1 mediates peptidoglycan induced 60. Garrido-Tapia M, Hernández CJ, Ascui G, et al. Stat3 inhibition by sta21 inflammatory cytokines expression and NF-kB activation in macrophages. Microb Pathog. 2016;99:111–8. increases cell surface expression of MICB and the release of soluble micb by gastric adenocarcinoma cells. Immunobiology. 2017;222(11):1043–51. 52. Muhamuda K, Mounia AEA, Jennie V, et al. MyD88-dependent inflam- masome activation and autophagy inhibition contributes to Ehrlichia- induced liver injury and toxic shock. PLoS Pathog. 2017;13(10):e1006644. Ip WKE, Hoshi N, Shouval DS, et al. Anti-inflammatory effect of IL-10 mediated by metabolic reprogramming of macrophages. Science. 2017;356(6337):513–9. 53. 54. Rojas Márquez JD, Ana Y, Baigorrí RE, et al. Mammalian Target of rapa- mycin Inhibition in Trypanosoma cruzi-infected macrophages leads to an intracellular profile that is detrimental for infection. Front Immunol. 2018;9:313. 55. Chimin P, Andrade ML, Belchior T, et al. Adipocyte mTORC1 deficiency promotes adipose tissue inflammation and NLRP3 inflammasome activa- tion via oxidative stress and de novo ceramide synthesis. J Lipid Res. 2017;58(9):1797–807. 56. You L, Wang Z, Li H, et al. The role of STAT3 in autophagy. Autophagy. 2015;11(5):729–39. 61. Lu L, Zhu F, Zhang M, et al. Gene regulation and suppression of type I interferon signaling by stat3 in diffuse large B cell lymphoma. Proc Natl Acad Sci USA. 2018;115(3):E498–505. 62. Schust J, Sperl B, Hollis A, et al. Stattic: a small-molecule inhibitor of stat3 activation and dimerization. Chem Biol. 2006;13(11):1235–42. 63. Souissi I, Najjar I, Ah-Koon L, et al. A stat3-decoy oligonucleotide induces cell death in a human colorectal carcinoma cell line by blocking nuclear transfer of stat3 and stat3-bound NF-κB. BMC Cell Biol. 2011;12:14. 64. Levy DE, Lee CK. What does stat3 do? J Clin Invest. 2002;109(9):1143–8. 65. Yokogami K, Wakisaka S, Avruch J, et al. Serine phosphorylation and maximal activation of stat3 during CNTF signaling is mediated by the rapamycin target mTOR. Curr Biol. 2000;10(1):47–50. 66. Dodd KM, Yang J, Shen MH, et al. mTORC1 drives HIF-1α and VEGF-A signalling via multiple mechanisms involving 4E-BP1, S6K1 and STAT3. Oncogene. 2015;34(17):2239–50. 67. Bao W, Wang Y, Fu Y, et al. mTORC1 regulates Flagellin-induced inflamma- 57. Miao LJ, Huang FX, Sun ZT, et al. Stat3 inhibits Beclin 1 expression tory response in macrophages. PLoS ONE. 2015;10(5):e0125910. through recruitment of HDAC3 in nonsmall cell lung cancer cells. Tumor Biol. 2014;35(7):7097–103. 58. Wang WB, Levy DE, Lee CK. Stat3 negatively regulates type I IFN-medi- ated antiviral response. J Immunol. 2011;187(5):2578–85. 59. Couto JP, Daly L, Almeida A, Knauf JA, Fagin JA, Sobrinho-Simões M, et al. Stat3 negatively regulates thyroid tumorigenesis. Proc Natl Acad Sci U S A. 2012;109(35):E2361-2370. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1186_s13567-021-00981-3	Weiße et al. Vet Res (2021) 52:112 https://doi.org/10.1186/s13567-021-00981-3 RESEARCH ARTICLE Open Access Immunogenicity and protective efficacy of a Streptococcus suis vaccine composed of six conserved immunogens Christine Weiße1, Denise Dittmar2, Beata Jakóbczak3, Volker Florian3, Nicole Schütze4, Gottfried Alber4, Kristin Klose5, Stephan Michalik2, Peter Valentin‑Weigand6, Uwe Völker2 and Christoph Georg Baums1* Abstract A vaccine protecting against different Streptococcus suis serotypes is highly needed in porcine practice to improve animal welfare and reduce the use of antibiotics. We hypothesized that immunogens prominently recognized by con‑ valescence sera but significantly less so by sera of susceptible piglets are putative protective antigens. Accordingly, we investigated immunogenicity and protective efficacy of a multicomponent vaccine including six main conserved immunogens, namely SSU0934, SSU1869, SSU0757, SSU1950, SSU1664 and SSU0187. Flow cytometry confirmed surface expression of all six immunogens in S. suis serotypes 2, 9 and 14. Although prime‑booster vaccination after weaning resulted in significantly higher specific IgG levels against all six immunogens compared to the placebo‑ treated group, no significant differences between bacterial survival in blood from either vaccinated or control animals were recorded for serotype 2, 9 and 14 strains. Furthermore, vaccinated piglets were not protected against morbid‑ ity elicited through intranasal challenge with S. suis serotype 14. As ~50% of animals in both groups did not develop disease, we investigated putative other correlates of protection. Induction of reactive oxygen species (ROS) in blood granulocytes was not associated with vaccination but correlated with protection as all piglets with >5% ROS survived the challenge. Based on these findings we discuss that the main immunogens of S. suis might actually not be a priori good candidates for protective antigens. On the contrary, expression of immunogens that evoke antibodies that do not mediate killing of this pathogen might constitute an evolutionary advantage conserved in many different S. suis strains. Keywords: TroA, OppA, Basic membrane lipoprotein, LysM, Di‑peptidyl peptidase IV, Subtilisin‑like serine protease, Nucleoside ABC transporter, Bactericidal assay, Immunogens Introduction Streptococcus suis is a very successful colonizer of mucosal surfaces in pigs. However, it is also a major porcine pathogen causing severe pathologies such as meningitis, polyarthritis, septicemia and endocarditis. Currently, 29 serotypes (cps) have been confirmed [1]. Correspondence: christoph.baums@vetmed.uni‑leipzig.de 1 Institute of Bacteriology and Mycology, Centre for Infectious Diseases, Faculty of Veterinary Medicine, Leipzig University, Leipzig, Germany Full list of author information is available at the end of the article Strains of cps1, cps2, cps1/2, cps3, cps4, cps7, cps9 and cps14 are associated with diseases and main herd prob- lems in different countries, whereas other serotypes contribute only marginally to morbidity [2]. Worldwide, S. suis cps2 is most frequently isolated from clinical cases in pigs and humans [2]. However, cps9 has become most prevalent among invasive isolates in some European countries with a large pig industry such as The Neth- erlands and Spain [2, 3]. In South America cps14 ranks third among invasive S. suis strains [4, 5]. Furthermore, cps14 is also frequently found in the United Kingdom [3, © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Weiße et al. Vet Res (2021) 52:112 Page 2 of 18 6] and contributes substantially to severe zoonotic cases in Asia [2]. As no licensed vaccine has convincingly reduced the burden of S. suis diseases in the field, different vaccina- tion approaches are currently under investigation. This includes conjugated capsule polysaccharides [7], live attenuated Salmonella enterica serovar Choleraesuis vec- tors delivering conserved surface proteins [8, 9] and dif- ferent recombinant antigens [10, 11]. A multicomponent vaccine including five recombinant antigens, which were identified in a screen for fitness genes important for colo- nization of the pig nasal epithelium, elicits partial protec- tion against a homologous cps2 challenge [12]. Various groups have been using immunoproteomics to identify conserved immunogens, which are discussed as prom- ising protective antigens. Main immunogens of virulent cps2 strains include lipoproteins, muramidase-released protein, surface antigen one and suilysin [13]. However, data demonstrating protective efficacy of these immuno- gens is very limited or even contradictory. In this study we investigated the protective efficacy of a multicomponent vaccine against cps14 in the natural host. The multicomponent vaccine included immuno- gens expressed by at least three different important S. suis cps (2, 14, 9). Sera drawn from convalescent piglets contained significantly higher levels of immunogen-spe- cific IgG antibodies than sera drawn from susceptible piglets. Although significant differences in specific anti- body levels were elicited through vaccination, protection was not observed. The heterogeneity in the clinical out- come between the different animals was used to investi- gate a new putative correlate of protection: the induction of reactive oxygen species (ROS, also described as oxida- tive burst) in blood granulocytes. Materials and methods Bacterial strains and growth conditions S. suis strain 10 is an mrp + epf + sly + cps2+ strain of sequence type 1 that has been used by different groups successfully to induce disease experimentally [14]. (cps7 + mrp +) and 16085/3b Strains 13-00283-02 (cps9 + mrp + sly +) are virulent and highly virulent strains of sequence types 27 and 94, respectively [15, 10]. Strain V3117/2 of cps14 and sequence type 1 was originally isolated from the brain of a pig with menin- gitis. Sequencing of the cpsK gene [16, 17] and serotyp- ing by slide agglutination using specific rabbit antisera (ID-Lelystad, Lelystad, The Netherlands) confirmed that strain V3117/2 is a cps14 strain. The latter was kindly conducted by Astrid de Greef (Wageningen Biovet- erinary Research, Lelystad, The Netherlands). Strain 7119-1 (mrp + cps2+) of sequence type 28 was isolated from the joint fluid of a piglet, which developed arthritis before the start of the experimental trial described below. Bacteria were cultured on Columbia agar plates sup- plemented with 6% sheep blood or in BactoTM Todd Hewitt broth (THB) at 37 °C for 24 h or overnight, if not stated otherwise. For the use in bactericidal assays, strains were grown in THB until late exponential phase and stored with 20% (v/v) glycerol at −80 °C. Escheri- chia (E.) coli was cultured in Luria–Bertani (LB) medium. If appropriate, 100 μg/mL ampicillin was added. Genotyping Genotyping of S. suis strains isolated from pigs was con- ducted by MP-PCR detecting mrp, epf, sly, arcA, gdh, cps1/14, cps2, cps7 and cps9 as well as with an epf mon- oplex variant PCR [18]. Sanger sequencing of the cpsK locus was conducted to identify cps14 among cps1 posi- tive strains [16, 17]. Multi locus sequence typing (MLST) was performed as described previously [15]. Expression and purification of recombinant (r) proteins To express recombinant SSU0934, SSU1869, SSU0757, SSU1950, SSU1664, SSU0309 and SSU0187 in E. coli, plasmids were constructed as follows. Genomic DNA of S. suis (strain 10, cps2) was used to amplify genes by PCR. The DNA fragments of ssu0934, ssu1869, ssu0757, ssu1950, ssu1664 and ssu0187 were generated using 0934(-21aa)-F/0934-R, 1869(-20aa)-F/1869-R, 757Pro- F/757Pro-R, 1950(-30aa)-F/1950-R, 1664(-31aa)-F/1664- R and 0187-F/0187(-stop)-R primer pairs, respectively (Primer sequences are listed in Additional file 1). The products of ssu0934 and ssu1950 were digested with KpnI and HindIII. The fragments of ssu1869 and ssu1664 were digested with SacI and SalI, whereas the products of ssu0757 and ssu0187 were digested with SalI and BamHI or HindIII and SacI, respectively. The gene products of ssu0934, ssu1869, ssu0757, ssu1950 and ssu1664 were cloned into pQE80L vector (Qiagen, Hilden, Germany) or pQE80L-Strep vector for ssu0187 digested with the appropriate enzymes. pQE80L-Strep vector was con- structed by introducing Strep-tag sequence C-terminal into the MCS of the pQE80L vector by PCR using prim- ers pQE80L-strep-F/pQE80L-HindIII-R. The DNA fragment of ssu0309 was synthesized and cloned into a standard vector at the restriction sites of SacI and BamHI by the company Eurofins Genom- ics GmbH. The gene was subcloned into pQE80L vector using SacI and BamHI enzymes. The cloned sequence encodes the protein sequence of SSU0309 (GenBank: CAR44738.1) from amino acid 20 to 1051. Plasmid constructs were transformed to E. coli TOP10F’ cells (Invitrogen, Thermo Fisher Scientific, Dreieich, Germany). The purified plasmid DNA was Weiße et al. Vet Res (2021) 52:112 Page 3 of 18 sequenced for verification (conducted by Eurofins MWG Operon, Ebersberg, Germany). E. coli cultures were grown shaking in 400 mL LB medium at 30 °C to an OD600nm of 0.5–0.7. Four-hour overexpression of proteins was induced by addition of IPTG (dioxane-free, Thermo Fisher Scientific, Dreieich, Germany) at a final concentration of 0.1 mM (SSU1869, SSU0757, SSU1950, SSU0309 and SSU0187) or 1 mM (SSU0934 and SSU1664). The cells were harvested by centrifugation at 9150 × g for 10 min at room tempera- ture and resuspended in LEW buffer (50 mM NaH2PO4, 300 mM NaCl, pH 8.0) to the final OD600nm of 60. To purify SSU0934, SSU1869 and SSU0757 cells were dis- rupted by sonication and centrifuged at 26 900 × g for 30 min at 4 °C to dispose cell debris. The resulting super- natant was used for the protein purification. Proteins were purified under native conditions using Ni-IDA 2000 packed columns (Protino, Macherey–Nagel, Düren, Germany) as recommended by the manufacturer. The column was washed twice with LEW buffer contain- ing 20 mM imidazole. Proteins were eluted with elution buffer (50 mM NaH2PO4, 300 mM NaCl, 250 mM imida- zole, pH 8.0). The purified proteins were dialyzed against dialysis buffer (300 mM NaCl, 9.3 mM Na2HPO4, 1.6 mM NaH2PO4, pH 7.4). To purify SSU0309 cell suspension was incubated for 20 min at room temperature after addition of lysozyme (from chicken egg white, Sigma-Aldrich, Taufkirchen, Germany) and protease inhibitor cocktail (VWR, Darm- stadt, Germany) to the final concentration of 1 mg/mL and 50 µL/1 g pellet, respectively. Cells were disrupted by sonication and centrifuged at 26 900 × g for 15 min at 4 °C to dispose cell debris. The protein was purified as described for SSU0934, SSU1869 and SSU0757 with some alterations, namely, the column was washed twice with LEW buffer and the protein was dialyzed against PBS buffer (146 mM NaCl, 9.3 mM Na2HPO4, 1.6 mM NaH2PO4). To purify SSU1950 and SSU1664, cells were disrupted by addition of BugBuster (10 × concentrated, protein extraction reagent, EMD Millipore, Merck KGaA, Darm- stadt, Germany), lysozyme and Benzonase (EMD Milli- pore) to the final concentration of one time concentrated, 1 mg/mL and 12.5 U/mL, respectively. The mixtures were incubated for 20 min at room temperature and fur- ther centrifuged at 26 900 × g for 30 min at 20 °C. Pro- teins were purified as described for SSU0934, SSU1869 and SSU0757. To purify SSU0187, cells were disrupted by sonication and processed as described for SSU0934, SSU1869 and SSU0757. SSU0187 was purified using Strep-Tactin resin (Superflow Plus from Qiagen). The supernatant was incubated with appropriate amounts of the beads with gentle shaking at 4 °C for 1 h. The mixture was loaded on a laboratory column (with 10 µm filter pore size, MoBiTec, Göttingen, Germany). The resin was washed six times with LEW buffer. The proteins were eluted with LEW buffer enriched with 2.5 mM d-desthio- biotin (Novagen, Merck KGaA, Darmstadt, Germany). The purified protein was dialyzed as described for SSU0934, SSU1869 and SSU0757. Purity of all proteins was assessed by SDS-PAGE and validated by mass spec- trometric analysis. MS sample preparation and measurement For analysis of the recombinant proteins by mass spec- trometry, 2 µg of total protein was prepared with the sin- gle pot solid-phase enhanced sample preparation (SP3) protocol [19]. Liquid chromatography electro spray ioni- zation tandem mass spectrometry (MS/MS) on a LTQ Orbitrab Velos instrument selecting TOP 20 precursor ions for CID fragmentation per cycle was performed as described earlier [20]. Data analysis was carried out using MaxQuant 1.5.3.8 [21]. Peptides were identified by search against the Uniprot E. coli database (release 08/2017) spiked with the sequences of the recombinant proteins. The following settings were used: Trypsin/P as proteolytic enzyme, two missed cleavage and methio- nine oxidation. Only peptides identified with a PSM false discovery rate (FDR) 0.01 were used for further analy- sis. Proteins were only identified, if two or more unique peptides were found per protein. The mass spectrometry analysis allowed assignment of the signal intensities pre- dominantly to the recombinant proteins: SSU0187 ≥ 97%, SSU0934 = 91%, SSU1869 = 90%, SSU1664 = 91%, SSU0757 = 88%, and SSU1950 = 84%. The residual inten- sity was distributed among various E. coli proteins. Classification of sera used for characterization of immunogens Sera used for characterization of immunogens origi- nated from various experimental infections with dif- ferent S. suis strains or from bacterin vaccination as specified in Additional files 2 and 3, respectively. The protocols for the animal experiments were either approved by the Committee on Animal Experiments of the Lower Saxonian State Office for Consumer Pro- tection and Food Safety (Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit, LAVES under the permit number 509.6-42502-04/829; 33.14-42502-04-12/0965; 33.12-42502-04-16/2305A) or the Saxony Regional Office (Landesdirektion Sachsen under the permit no. TVV14/15; TVV26/15; N01/16; TVV11/16; TVV28/16; TVV37/17). All studies were performed in strict accordance with the principles and recommendations outlined in the European Conven- tion for the Protection of Vertebrate Animals Used for Weiße et al. Vet Res (2021) 52:112 Page 4 of 18 Experimental and Other Scientific Purposes (European Treaty Series, no. 123) and the German Animal Protec- tion Law (Tierschutzgesetz). The porcine sera used to characterize the immunogens were classified as follows: “Sera pre-infection” (n = 20) were defined as sera drawn prior to experimental S. suis infection of pig- lets surviving the following challenge to the end of the observation period of at least two weeks. Sera collected at the end of this observation period are in the group “Sera post-infection”. “Sera post-infection” (n = 20) were defined as sera drawn at the end of the observation period after experimental S. suis infection (also referred to as sera of reconvalescent pigs). The length of the observation period ranged from 14 to 22 days. At the point of serum collection neither clinical nor bacteriological nor histo- logical examinations revealed any signs of an acute S. suis infection. S. suis challenge strains in these experi- ments belonged either to cps2 (n = 10), cps9 (n = 3), cps7 (n = 5) or cps14 (n = 2) (see Additional file 2). “Sera of susceptible piglets” (n = 20) were defined as sera drawn prior to an experimental S. suis infection that resulted in a fatal S. suis disease in this piglet (also including piglets euthanized due to reaching termina- tion criteria). Pathologies in these piglets included meningitis, serositis and polyarthritis as confirmed by pathohistological investigations in association with detection of the S. suis challenge strain of either cps2 (n = 8), cps9 (n = 5), cps7 (n = 4) and cps14 (n = 3) (see Additional file 2). “Hyperimmune sera” (n = 16) were defined as sera drawn from prime-booster vaccinated pigs not experi- mentally infected with S. suis. Vaccination was conducted either with a cps2 (n = 8), a cps7 (n = 2) or a cps9 (n = 6) bacterin (see Additional file 3). Determination of IgG antibody titers using bead‑based xMAP® technology The IgG antibody titers of 76 porcine sera (Additional files 2, 3) against recombinant purified S. suis proteins were determined using bead-based xMAP® technol- ogy (Luminex®) [22]. The analysis of the purified Strep- or His-tagged proteins was performed according to a modified method from Meyer et al. [23]. For detection of antigen-specific IgG a series of seven serum dilutions was investigated (50-fold, 250-fold, 1000-fold, 5000- fold, 50 000-fold, 100 000-fold and 200 000-fold). The amount of swine IgG antibodies bound to each antigen was detected via a goat anti-porcine IgG antibody (H + L) conjugated with PE (R-phycoerythrin) (6050-09, South- ernBiotech, Birmingham, USA). The data analysis was performed using the xMAPr app [23]. Generation of rabbit anti‑sera Immunizations of rabbits with recombinant variants of SSU1869, SSU0757, SSU1950 and SSU0187 and respec- tive blood collections were performed by DAVIDS Bio- technologie GmbH (Regensburg, Germany). Generation of hyperimmune sera in rabbits against SSU0934 and SSU1664 was conducted by ourselves as described [24] and was registered at the LAVES under 87A044. SDS‑PAGE and Western blot analysis SDS-Page and Western blot analysis was performed essentially as described previously [24]. Rabbit sera against proteins of interest were used in a dilution of 1:5000. The secondary antibody, goat-anti-rabbit IgG HRP (catalogue 111-035008, Dianova, Hamburg, Ger- many), was diluted 1:25 000. Detection of antigens on the surface of S. suis through flow cytometry For flow cytometry analysis, bacteria were grown in THB or porcine plasma until early stationary phase and stored in 20% (v/v) glycerol at −80 °C. Plasma was col- lected from piglets at the age of 7–9 weeks from a herd that is known to be infected with several S. suis sero- types. The collection of blood samples was approved by the Landesdirektion Sachsen (permit no. TVV40/18). Before use, plasma samples were pooled. Bacteria-glyc- erol solutions containing 6.5 × 106 CFU were centrifuged at 4 °C for 10 min at 2100 × g, supernatant was discarded and pellets were washed twice with phosphate buffered saline (PBS) to remove remaining glycerol. Pellets were blocked with 300 µL 1:100 diluted normal donkey serum (Dianova, catalogue #017-000-121) for 30 min at 4 °C on the rotator, followed by staining with primary and sec- ondary antibodies for 30 min at 4 °C rotating as well. Sera from rabbits immunized with one of the recombinant antigens of the multicomponent vaccine (1:50 dilution in PBS) served as primary antibodies. Serum of a non- immunized rabbit was used as the control. As secondary antibody we used a 1:200 diluted fluorescein isothiocy- anate (FITC)-labeled donkey anti-rabbit IgG (Dianova, catalogue #DAB-87179). Between every incubation step, pellets were washed twice with PBS. Samples were fixated with 2% (w/v) paraformaldehyde. Flow cytometry meas- urement was conducted with BD, FACS Fortessa and data were analyzed using the FlowJoTM_V10 software. Animal experiments Using a litter-matched design, 18 German landrace pig- lets from a herd known to be free of cps1, cps7, cps9, cps14 but not cps2 were divided into two groups (pla- cebo: n = 9; vaccinated: n = 9). The classification of the Weiße et al. Vet Res (2021) 52:112 Page 5 of 18 herd was based on the MP-PCR typing results [18] of S. suis isolates from the tonsils of more than 500 ani- mals over the last 16 years [in the following these piglets are referred to as specific pathogen free (spf )]. Piglets were prime-booster immunized intramuscularly at the age of four and six weeks either with the multicompo- nent vaccine or a placebo with PBS, both supplemented with 20% (v/v) Emulsigen as adjuvant. Every vaccine dose contained 150 µg of each recombinant immuno- gen (SSU0934, SSU1869, SSU0757, SSU1950, SSU1664, SSU0187). After weaning, one piglet developed acute dis- ease. This animal and the littermate were excluded from the experiment. Two weeks after booster immunization the remaining 16 animals were infected intranasally with 5 × 109 CFU of S. suis cps14 V3117/2 grown in Bacto™ Tryptic Soy Broth without dextrose (BD, Heidelberg, Germany). Monitoring of clinical parameters after infec- tion was performed every 8 h including measurement of the inner body temperature, assessment of movements and feed intake using a score sheet as described previ- ously [10] with the following modification: moderate and ceased feed intake received scores of 2 and 5, respec- tively. Piglets were only fed at these time points. Piglets exhibiting high fever (≥ 40.5 °C) combined with apathy and anorexia (over 24 h), as well as animals showing any clinical signs of acute polyarthritis or severe meningitis, were euthanized for reasons of animal welfare. Surviving piglets were sacrificed fourteen days post-infection. All animals went through the same necropsy, histopathologi- cal and bacteriological screenings as described previously [10]. Animals were infected experimentally and cared for in accordance with the principles outlined in the EU Direc- tive 2010/63/EU. All animal experiments or samplings were conducted by veterinarians and in accordance with the principles outlined in the European Convention for the Protection of Vertebrate Animals Used for Experi- mental and other Scientific Purposes and the German Animal Protection Law (Tierschutzgesetz). The animal experiment of this study was approved by the Saxony Regional Office (permit no. TVV57/18). Bactericidal assay show proliferation, whereas survival factors < 1.0 indicate killing of S. suis. Opsonophagocytosis assay Blood was collected from weaning piglets from a herd which is known to be infected with several S. suis cps (permit no. N19/14 and A09/19). Porcine neutrophils were separated using density centrifugation as described before [25]. Purified porcine neutrophils (5 × 106 in 400 µL RPMI) were mixed with 100 µL of porcine serum and 3 × 105 CFU of S. suis strains 10 or 13-00283-02 or 16085/3b and incubated for 1 h at 37 °C on a rotator. We used serum of colostrum-deprived piglets as the negative control. Hyperimmune serum from an animal vaccinated with a cps2 bacterin served as the positive control. CFU were determined before and after incubation and survival factors were calculated as mentioned before for the bac- tericidal assay. Reconstituted blood assay with pre‑adsorbed sera The cps2 S. suis isolate 7119-1 of a piglet, which devel- oped severe disease before experimental infection, was used to pre-adsorb sera from animals collected eleven days post booster immunization. For each sample, 5 mL of overnight culture in THB was centrifuged for 15 min at 6620 × g, supernatant was discarded and the pellets were washed twice with PBS. Bacteria were resuspended in 500 µL piglet sera and incubated for 1 h rotating at 4 °C. Afterwards, samples were centrifuged for 20 min at 10 000 × g and the supernatant transferred to a 0.22 µm UltrafreeMC centrifugal filter (UFC30GV0S, Merck Millipore, Merck KGaA, Darmstadt, Germany) and centrifuged for 4 min at 12 000 × g. Sera were used to reconstitute blood for bactericidal assays as follows. Hep- arinized blood of a healthy pig (11th week, permit no. A09/19) was centrifuged at 500 × g for 10 min. Plasma was removed and blood cells were washed twice with PBS and mixed 1:1 (v/v) with 100 µL of pre-adsorbed serum or untreated control serum and 3 × 105 CFU/mL of cps14 strain V3117/2. CFU were determined before and after an incubation of 2 h at 37 °C on a rotator. Survival factors were calculated as mentioned before. Prior immunization and eleven days after booster, hep- arinized blood was collected from all animals and a bactericidal assay was conducted. Five hundred micro- liters of heparinized blood (16 I. U. heparin/mL) were mixed with either 1.2 × 106 CFU of S. suis strain 10 or 6 × 106 CFU of 16085/3b or 3 × 106 CFU of V3117/2 and incubated for 2 h at 37 °C on a rotator. Before and after incubation, CFU were determined by plating of serial dilutions. Survival factors represent the ratio of CFU at 120 min to CFU at timepoint zero. Survival factors > 1.0 Detection of S. suis induced oxidative burst in granulocytes Oxidative burst was detected as previously described [26]. Briefly, S. suis cps14 strain V3117/2 was added at a concentration of 6 × 106 CFU/mL to heparinized blood samples. To determine S. suis induced signals, for each blood sample a PBS control (instead of S. suis) was used. As a positive control 0.1 µg/mL PMA (Sigma-Aldrich) stimulation was used instead of S. suis. Detection of phagocytosis in combination with oxidative burst was done with cps2-preadsorbed sera and untreated control Weiße et al. Vet Res (2021) 52:112 Page 6 of 18 sera in a reconstituted assay using heparinized blood of a healthy piglet (11th week permit no. A09/19). To remove the plasma, the blood was washed twice with PBS and was finally adjusted to three-quarter original volume with PBS. S. suis, V3117/2, prestained with Cell Trace FarRed (Thermo Fisher Scientific), was added and intensively mixed before addition to the test-sera to reach finally 6 × 106 CFU/mL. All samples were incubated at 37 °C for 15 min in a water bath. Subsequently, dihydrorho- damine 123 (Sigma-Aldrich, DHR123; 5 µg/mL) was added and incubation was continued for another 10 min at 37 °C. Finally, erythrocytes were lysed two times using erythrocyte lysis buffer (0.155 M ammonium chlo- ride, 10 mM potassium bicarbonate, 0.1 mM disodium EDTA, pH 7.2). The remaining leucocytes were washed two times with PBS and finally samples were fixed with 2% (w/v) paraformaldehyde and measured immediately by flow cytometry (BD FACSCalibur). Data analysis was conducted by gating on granulocytes (FlowJo). S. suis- induced oxidative burst was defined as the difference: % Rhodamine123 (Rho123) + granulocytes in S. suis sam- ple − % Rho123 + granulocytes in PBS control. ELISA IgM and IgG antibody levels of pre-adsorbed and con- trol sera were determined following a standard protocol as described before, but with minor modifications [15]. Briefly, Nunc Immuno MaxiSorp plates (SigmaAldrich) were coated overnight at 4 °C with 0.2% (v/v) formal- dehyde-inactivated bacteria of strain V3117/2. Conva- lescent sera of five piglets, experimentally infected with V3117/2, were mixed in equal proportions and used as reference serum to define 100 ELISA units. Plates were blocked and dilutions were made in PBS with 0.5% (w/v) bovine serum albumin (BSA) and 1% (w/v) gelatin. For IgM detection a polyclonal secondary goat-anti-porcine- IgM horseradish-peroxidase (HRP) conjugated antibody (NBP2-42699H, 1:10 000, Novus Biologicals, Wiesbaden- Nordenstadt, Germany) and for IgG detection a poly- clonal goat anti-porcine IgG HRP conjugated antibody (A100-105P, 1:10 000, Bethyl, Hamburg, Germany) was used (1 h incubation time each). Statistical analysis IgG titers of the different groups were compared using a Wilcoxon rank sum test. Multiple test adjusted (Ben- jamini–Hochberg procedure) [27] p-values (q-values) below 0.05 were considered statistically significant. All calculations and plot generation were performed in R (v 3.6.1) [28] using the packages: tidyverse (v 1.3.0) [29], helfeRlein (v 0.2.2) [30], plotly (v 4.9.2.1) [31], ggsignif (v 0.6.0) [32], ggplot2 (v 3.3.2) [33], grid (v 3.6.1) [28], gri- dExtra (v 2.3) [34], FactoMineR (v 2.3) [35], factoextra (v 1.0.7) [36], ggrepel (v 0.8.2) [37], and patchwork (v 1.0.1) [38]. Data of the bactericidal assay was analyzed with the Mann–Whitney-U test for comparison of the vaccinated versus placebo-treated group. The correlation factor between blood survival factors and S. suis-induced oxida- tive burst rates was calculated with Spearman rank corre- lation. Significant differences between pre-adsorbed and control sera in reconstituted blood assay and IgG-ELISA were determined using the Wilcoxon matched pairs test. A paired t-test was used to compare these groups in phagocytosis assay and IgM-ELISA. Probabilities lower than 0.05 were considered significant (p-value: ≤ 0.001 = ; ≤ 0.01 = ; ≤ 0.05 = ; > 0.05 = NS). Results Selection of putative protective immunogens We hypothesized that protective immunogens expressed by different serotypes should be prominently recog- nized by sera which were drawn from pigs post-infection and which elicited killing of S. suis in bactericidal or opsonophagocytosis assays. Sera drawn from suscepti- ble piglets succumbing after challenge to S. suis disease should contain much less IgG antibodies binding to these immunogens. Over the last years we have estab- lished a large biobank including sera drawn from piglets in experimental S. suis studies with a detailed documen- tation of clinics, pathologies and bacteriologies. All sera included in the study contained S. suis specific antibod- ies as they were collected from piglets colonized with S. suis. Blood and serum of these piglets were investigated in bactericidal and opsonophagocytosis assays with dif- ferent serotypes (Additional file 2). Sera were assigned to different groups as outlined in the Material and methods and Figure 1A. A collection of proteins encoded by genes conserved in the recently published genomes of strains 10 (cps2), 13-00283-02 (cps7) and 16085/3b (cps9) were expressed recombinantly (Additional file 4) [39]. Meas- urement of IgG titers revealed that post-infection sera contained significantly higher IgG levels against SSU0934, SSU1869, SSU0757, SSU1950, SSU1664 and SSU0187 than sera drawn from susceptible piglets (see Figure 1B), which prompted us to include these six proteins in the vaccine of this study. However, such significant differ- ences in binding of specific IgG in post-infection sera and sera of susceptible piglets were not recorded for all inves- tigated S. suis proteins. As an example, SSU0309, a homo- logue of the pneumococcal histidine triad protein Pht309 [40], displayed similar IgG binding in the different groups of sera (Figure 1B). Noteworthy, in vitro killing of the four investigated S. suis strains (of cps2, 7, 14 and 9) was more prominent in osponophagocytosis or bactericidal assays using sera or blood of the post-infection group Weiße et al. Vet Res (2021) 52:112 Page 7 of 18 Figure 1 Classification of porcine sera (A) used to characterize immunogens (B) included in the vaccine. Sera pre and post‑infection were drawn as indicated within the course of experimental S. suis infections of non‐vaccinated piglets that survived the experiment to the end of the observation period. Animals that developed severe disease and died or had to be euthanized shortly after infection were classified as susceptible. Sera of these animals were collected before infection. Hyperimmune sera were taken from animals prime‑booster vaccinated with a S. suis bacterin. The pie charts indicate the proportion of the indicated S. suis cps among the experimental infections used for collection of serum samples (A). Recognition of S. suis proteins SSU0934, SSU1869, SSU0757, SSU1950, SSU1664, SSU0187 and SSU0309 through specific IgG present in the indicated sera (B). Detection of serum IgG bound to the indicated antigens coupled to beads was performed using the xMAPr approach [23]. The antigens included in the vaccine recruited significantly more specific IgG in hyperimmune sera and sera post‑infection compared to sera of susceptible pigs. SSU0309 is shown as an example of an antigen that did not show these differences. Unpaired Wilcoxon rank sum test (black), paired Wilcoxon rank NS. sum test (brown) p‑value: ; > 0.05 0.001 0.01 0.05 *; ; ≤ = ≤ = ≤ = = than using sera of susceptible piglets (Additional file 2). Differences in specific IgG binding between sera drawn post- and pre-infection, which were significant for all of the six antigens but SSU1950, indicated that the selected S. suis antigens are expressed in vivo. Of note, hyperim- mune sera drawn from piglets vaccinated with bacterins also contained significantly higher IgG levels against each of the six mentioned immunogens in comparison Weiße et al. Vet Res (2021) 52:112 Page 8 of 18 to sera drawn from susceptible piglets. As outlined in Additional file 3, at least 6 of the 16 hyperimmune sera mediated killing of more than one cps in opsonophagocy- tosis assays. Based on these findings we considered these immunogens putative protective antigens. Analysis of expression of selected antigens on the surface of cps2, cps14 and cps9 In silico analysis revealed that SSU0934, SSU1869, SSU0757, SSU1950, SSU1664 and SSU0187 are likely secreted or located on the bacterial surface (Table 1). SSU1869 is identical with TroA, a virulence-associated lipoprotein that is necessary for manganese uptake [41]. SSU0757, a virulence-associated subtilisin–like pro- tease on the cell surface of S. suis, is described to be able to degrade the Aα chain of fibrinogen and to trig- ger pro-inflammatory response in macrophages [42]. In strain 16,085/3b an early stop codon within an insertion element results in a truncated ORF (SSU16085_00675) encoding a protein of only 40 kDa without an LPXTG motif [39]. The insertion element carries an IS630 trans- posase family gene (SSU16085_00676). The 140 kDa large C-terminus of SSU0757 (as annotated in P1/7) might also be expressed in 16085/3b, if UUG is used as an alterna- tive start codon (SSU16085_00677). However, this ORF lacks a signal sequence but carries a C-terminal LPXTG motif (Additional file 4). SSU1950, also known as LysM, is a peptidoglycan-binding protein described to con- tribute to protection against phagocytosis of a microglia cell line [43]. Putative functions of the other proteins are specified in Table 1. We investigated expression of these proteins on the surface of different serotypes by flow cytometry. This was conducted with each of the rabbit hyperimmune sera elicited through immunization with individual immunogens. In advance, we verified that the hyperimmune sera recognize the recombinant proteins using Western blot analysis (Additional file 5). As shown in Figure 2, flow cytometry suggests that all six immuno- gens were detectable on the surface of cps2 strain 10, cps9 strain 16085/3b and cps14 V3117/2 grown until early stationary growth in THB or porcine plasma, though antigen-specific labelling varied between strains and also between cultivation in THB and porcine plasma. Immunogenicities of the multicomponent vaccine Weaning piglets were prime-booster vaccinated with the multicomponent vaccine to investigate immunogenici- ties and protective efficacies. One piglet developed acute disease related to infection with an S. suis cps2 strain, designated 7119-1, belonging to sequence type 28 and known to be present in the original herd. It therefore had to be excluded from the experiment together with its lit- termate. At the start of the experiment, levels of specific IgG antibodies binding to the selected immunogens were rather low in vaccinated and placebo-treated piglets con- sistent with colonization-induced background antibody levels. After booster immunization the multicomponent group showed significantly higher levels of specific IgG against all six immunogens than before immunization and compared to the control group (Figure 3). We won- dered if the induction of systemic IgG levels is associated with reduced proliferation or enhanced killing of different S. suis serotypes in blood of these piglets. Accordingly, bactericidal assays were conducted with S. suis strains V3117/2, 10 and 16085/3b of cps14, 2 and 9, respectively. Significant differences in the bacterial survival factors were not recorded between placebo-treated and vacci- nated piglets eleven days post booster vaccination (Fig- ure 4), although there was a tendency to lower survival factors in the blood of vaccinated piglets for cps14, 2 and 9 (mean survival factors of 0.92 ± 1.25 vs. 0.23 ± 0.56 for cps14; 0.44 ± 0.97 vs. 0.16 ± 0.40 for cps2 and 7.56 ± 6.91 vs. 3.87 ± 3.07 for cps9, respectively). Survival factors of cps14 strain V3117/2 were substantially lower than 1 in Table 1 Immunogens selected for the multicomponent vaccine SSU in P1/7 Putative function/ortholog Localisation/tagg PSORTb (score) Theoretical molecular weight (kDa)a SSU0934 SSU1869 SSU0757 SSU1950 SSU1664 Basic membrane lipoprotein Zinc ABC transporter TroA Subtilisin‑like serine protease LysM protein Oligopeptide substrate‑binding protein OppA Signal sequenceb Lipobox (LAAC) signal sequenceb signal sequenceb LPXTG signal sequenceb Signal sequenceb Lipobox (LAAC) Unknown Membrane (9.68) Cell wall (10) Unknown Unknown SSU0187 Di‑peptidyl peptidase IV No signal sequenceb Extracellular (9.6) 36 34 168c 19 64 88 a Excluding posttranslational modifications. b Detection by SignalP3.0 (using neural networks (NN) and hidden Markov models (HMM) trained on Gram-positive bacteria). c In strain 16085/3b an early stop codon within an insertion results in an ORF of only 40 kDa and without an LPXTG motif. Reference [52] [41, 58] [42, 59, 60] [43, 61] [44, 45] [62, 63] Weiße et al. Vet Res (2021) 52:112 Page 9 of 18 Figure 2 Detection of vaccine antigens on the bacterial surface after cultivation in THB (A) or porcine plasma (B). S. suis strains 10 (cps2), 16085/3b (cps9) and V3117/2 (cps14) were grown until stationary phase and incubated with sera of rabbits immunized with the indicated vaccine antigens. In the next step bacteria were labeled with a fluorescein isothiocyanate (FITC)‑labeled anti‑rabbit antibody. After fixation with 2% (w/v) paraformaldehyde bacteria were investigated by flow cytometry using a LSR Fortessa (BD Biosciences). Data were analyzed with FlowJo software 10.1r5 (TreeStar, Ashland, OR, USA). As controls, bacteria were measured without incubation in a rabbit antiserum (w/o 1. antibody) or with a rabbit serum collected prior immunization (pre immun). The experiment was repeated three times. the vaccinated piglets but much less than in the placebo group, suggesting multicomponent-induced bactericidal activity against cps14. These results suggest some limited protection against cps14, which prompted us to conduct a cps14 challenge experiment. Protective efficacy of the multicomponent vaccine against S. suis cps14 Vaccinated (n = 8) and placebo-treated (n = 8) piglets were infected intranasally with cps14 strain V3117/2 of sequence type 1 fourteen days after booster immuniza- tion. Five vaccinated animals and four placebo animals demonstrated clinical signs of severe disease like body temperatures above 40.2 °C or anorexia within the fol- lowing 10 days (Table 2). In both groups two piglets had to be euthanized for animal welfare reasons because of signs of polyarthritis. In these four early euthanized ani- mals, the challenge strain was detected in multiple inner organs. Furthermore, one vaccinated animal showed additional signs of central nervous system dysfunction (opisthotonus, generalized tremor, ataxia). As shown in Figure 5, significant differences in mortality and morbid- ity were not recorded between immunized and placebo- treated piglets. Furthermore, detection of the challenge strain in inner organs was very similar in vaccinated and control piglets (Table 3). Pathohistological screenings revealed lesions in the brain, serosa and spleen in single animals of the placebo group. Scoring of fibrinosuppurative lesions of challenged piglets did not reveal substantial differences between the two groups (Additional file 6). In summary, the challenge experiment did not reveal any protection against disease induced through S. suis cps14 challenge. S. suis‑induced oxidative burst as a parameter to estimate protection against invasive infection with cps14 strain In a previous study we demonstrated that the S. suis induced production of ROS in granulocytes is antibody- mediated and leads to a reduced bacterial survival in blood [26]. To illuminate the role of S. suis-induced Weiße et al. Vet Res (2021) 52:112 Page 10 of 18 Figure 3 Immunization elicited significant higher titers of IgG against all six immunogens included in the vaccine. Serum was collected before prime immunization and eleven days after booster immunization with a multicomponent vaccine (n 8). Each serum was incubated with beads coupled with selected antigens. Detection of bound serum IgG was performed using the xMAPr approach [23]. Paired Wilcoxon rank sum test (black), unpaired Wilcoxon rank sum test (brown), red star indicates imputed value. p ‑v alu e : 0. 05 8) or a placebo (n * ; > 0.0 5 0.0 01 0.01 * ; NS . ; = = = = ≤ ≤ ≤ = = Weiße et al. Vet Res (2021) 52:112 Page 11 of 18 (n = 7) survived the challenge (7/7). These results confirm that induction of oxidative burst in blood granulocytes is important for protection against S. suis and can be con- sidered as a correlate of protection. However, it appears unlikely that the antibodies crucial for ROS induction were related to vaccination as we did not observe a dif- ference between vaccinated and placebo-treated piglets. We hypothesized that antibodies elicited through nat- ural infection with a cps2 strain present in the original herd might have affected the outcome of the experimen- tal infection. To investigate this, we pre-adsorbed the investigated sera from animals with an S. suis-induced ROS > 5% with the cps2 strain isolated from the dis- eased piglet prior to the trial (strain 7119-1). As shown in Figure 7, this led to significantly decreased ROS in granulocytes after phagocytosis of S. suis strain V3117/2. Furthermore, we observed an increased bacterial survival factor of V3117/2 in reconstituted blood assay using pre- adsorbed sera compared to untreated sera. In fact, using ELISA we were able to show that the pre-adsorbtion led to a significant decrease of antibodies binding to the sur- face of V3117/2, whereby the difference between treated and untreated control sera was greater in IgM than in IgG antibody levels. These results suggest that antibod- ies elicited through natural infection with a cps2 strain might have been important for the outcome of the chal- lenge experiment due to cross-reaction with cps14. Discussion Sera drawn from piglets in convalescence after S. suis infection recognize numerous immunogenic surface- associated proteins. Some of these immunogens are expressed by different serotypes that made these immu- nogens attractive candidates for cross-protective vac- cines in our opinion. In accordance with this reasoning we investigated the protective efficacy of a multicom- ponent vaccine including six different immunogens expressed by different serotypes (Figures 1 and 2; Addi- tional file 4). Of note, the included lipoprotein SSU1664, also known as oligopeptide-binding protein OppA, was already known to be an immunogenic protein of cps2 strains [44, 45]. The immunogens selected for the Figure 4 Survival of S. suis in blood drawn from vaccinated or placebo‑treated piglets. Eleven days after booster immunization with a multicomponent vaccine or application of a placebo blood was drawn from 8‑week‑old piglets (n with S. suis strains V3117/2 (cps14), 10 (cps2) and 16085/3b (cps9). Significant differences were not recorded between vaccinated and placebo‑treated piglets in these bactericidal assays using the Mann– Whitney‑U test. The survival factor represents the ratio of CFU at 120 min to CFU at timepoint zero. Bars and error bars represent mean values and standard deviations, respectively. 8 per group) and incubated = oxidative burst in granulocytes in this immunization- challenge experiment, we determined the induction of oxidative burst by the challenge S. suis strain V3117/2 in blood samples drawn 11d after booster immunization. We detected a wide range of S. suis-induced burst rates between 0 and 17% by the cps14 strain V3117/2. Between placebo-treated and vaccinated animals we found no sta- tistical difference (Figure 6A). To analyze, whether the induction of oxidative burst in blood granulocytes before challenge is related to protection, the cps14 challenged animals were divided in a convalescent (filled symbols in Figure 6B) and a susceptible subgroup for cases with euthanasia based on a high clinical score (open symbols in Figure 6B). The oxidative burst rate shows a strong negative correlation with bacterial survival factors (Spearman rS = −0.87 with p < 0.0001; Figure 6B). Nota- bly, animals with S. suis-induced oxidative burst < 5% (n = 9) show a clearly reduced survival (4/9), whereas all animals with an S. suis-induced oxidative burst > 5% Table 2 Evaluation of protection induced by multicomponent vaccine after intranasal challenge with S. suis cps14 strain V3117/2 Immunization Morbidity Mortality Mean Clinical score (SD) Clinical signs Max. body temperature (°C) CNSa Lameness No feed intake < 40 40 ≥ and ≤ 40.2 > 40.2 Placebo Vaccine 4/8 5/8 2/8 3/8 7 (11.1) 11 (12.0) 0/8 1/8 2/8 3/8 2/8 3/8 4/8 3/8 0/8 1/8 4/8 4/8 a Signs of central nervous system (CNS) dysfunction such as convulsions and opisthotonus. Weiße et al. Vet Res (2021) 52:112 Page 12 of 18 Figure 5 Mortality (A) and morbidity (B) of vaccinated and placebo‑treated piglets induced through S. suis cps14 infection. Piglets were 109 CFU of strain V3117/2 (cps14) 14 days after booster immunization. Morbidity was defined as piglets showing challenged intranasally with 5 an inner body temperature of 40.2 °C or higher. In case of high fever ( 40.5 °C), apathy and anorexia persisting over 24 h as well as in all cases of central nervous system dysfunction or clinical signs of acute polyarthritis, animals were euthanized for animal welfare reasons. All surviving piglets were sacrificed 14 days post‑infection. Significant differences between piglets vaccinated with the multicomponent vaccine (n 8) or the placebo (n 8) were not recorded. Statistical analysis was conducted with the log‑rank test. × = ≥ = multicomponent vaccine were prominently recognized by convalescence sera and hyperimmune sera, many of which mediated killing of different serotypes in bacteri- cidal assays or opsonophagocytosis assays. This made us even more believe that they are putative protective anti- gens. Our assumption seemed to be supported by the finding that IgG antibody levels against these immuno- gens in sera of susceptible piglets were significantly lower. Furthermore, we recorded significantly elevated IgG lev- els against 5 of these 6 immunogens in sera taken before experimental infection of piglets surviving the following S. suis challenge compared to sera of piglets succumb- ing to the infection (compare pre-infection sera [blue color] and sera of susceptible piglets [green color] in Fig- ure 1). Though vaccinated piglets demonstrate high IgG antibody titers against all six immunogens in contrast to placebo-treated piglets (Figure 3), experimental challenge with cps14 failed to demonstrate protection. Further- more, the results of a bactericidal assay also indicate sus- ceptibility to S. suis cps9 bacteremia in vaccinated piglets as the investigated cps9 strain proliferated in the blood of these piglets in vitro (survival factor > 1). Accordingly, we discarded the hypothesis that the six selected immuno- gens are strong protective antigens, at least not against morbidity induced by cps14 infection (Figure 5) and most likely also not against cps9 bacteremia (Figure 4). It is also questionable whether other main immunogens of S. suis are protective antigens. Muramidase-released protein (MRP) and surface antigen one (SAO) are main immuno- gens of S. suis cps2 [13, 46], but piglets with high levels of antibodies against these immunogens developed disease after experimental cps2 infection in different studies [13, 46–49]. One possible explanation for the ineffectiveness is that these immunogens might not be accessible to anti- bodies on the surface of the pathogen, at least not dur- ing bacteremia. However, our flow cytometric analysis indicates that rabbit hyperimmune sera recognized the selected immunogens on the surface of S. suis cps9 strain 16085/3b after cultivation in porcine plasma (Fig- ure 2B). Thus, it appears likely that these immunogens are expressed during bacteremia. However, immuniza- tion did not result in significant bactericidal activity of porcine blood. Most strikingly, the S. suis cps9 strain pro- liferated in blood of vaccinated piglets in vitro, though this strain shows very prominent binding of immunogen- specific IgG raised in rabbits, especially after cultivation in porcine plasma (Figure 2B). During bacteremia many S. suis bacteria are attached to phagocytes without being phagocytosed [50, 51]. This association might be crucial for breaching host barriers and therefore pathogenesis. However, it is unknown if this association with phago- cytes has an impact on antibody recognition. A limita- tion of the phenotypic characterization of S. suis by flow cytometry after cultivation in plasma is the absence of phagocytes. Many immunogens such as MRP are not only pre- sent on the bacterial surface but are also released into the environment of the bacterium [15]. Though we con- firmed that SSU0934 is surface-located, Gómez-Gascón et al. [52] identified SSU0934 in an immunosecretomic study indicating that this immunogen is also released Weiße et al. Vet Res (2021) 52:112 Page 13 of 18 8 / 1 8 / 2 8 / 2 8 / 3 8 / 2 8 / 2 8 / 1 8 / 1 8 / 1 8 / 2 8 / 1 8 / 0 8 / 2 8 / 1 8 / 4 8 / 6 8 / 2 8 / 3 d r a c o d n E f i d u fl t n o J i e F S C , n i a r B r e v i L n e e p S l d a s o r e S c g n u L l i s n o T m o r f d e t a l o s i s a w a n i a r t s e g n e l l a h c s i u s . S e h t h c i h w n i l i s t e g p f o r e b m u N b s n a g r o r e n n i 3 ≥ n i n i a r t s e g n e l l a h c e h t f o n i r o b n a g r o r e n n n o i t a l o s i e h t r o f e v i t i s o p s t e g p f o r e b m u N l i f o n o i t a l o s i e h t r o f e v i t i s o p s t e g p f o r e b m u N l i n o i t a z i n u m m I 2 / 7 1 1 3 V n i a r t s 4 1 s p c s i u s . S h t i w e g n e l l a h c l a s a n a r t n i i l r e t f a s t e g p m o r f n i a r t s e g n e l l a h c e h t f o n o i t a l o s i e R 3 e l b a T i n a n i n i a r t s e g n e l l a h c e h t i d u fl t n o i j n i r o a s o r e s 8 / 3 8 / 3 o b e c a P l i e n c c a V a h c e h T a . s n o i t c e f n i i d e t a n m e s s i d g n w o h s i i s l a m n a m o r f s e t a o s i l e v i t a t n e s e r p e r e v fi f o s u c o l K s p c f o g n i c n e u q e s r e g n a S d n a s e t a o s i l l l a f o R C P y b d e fi i t n e d i i s a w n a r t s e g n e l l . d e n e e r c s e r e w b m i l e v i t c e p s e r e h t f o s e r u t c n u p t n o i j l a n o i t i d d a s s e n e m a l f o e s a c n I . l i a m n a h c a e n i d e t a g i t s e v n i e r e w s t n o i j l a p r a c d n a l a s r a t h t o b f o s e r u t c n u P f . i d u fl l i a n p s o r b e r e C e . s l i s n o t e h t t o n t u b d r a c o d n e r o F S C , i n a r b , r e v i l , n e e p s l , g n u l o t s r e f e r n a g r o r e n n I . d e t a g i t s e v n i s a w e b o l l i a n a r c e n O b c . y t i v a c l i a d r a c i r e p r o l a e n o t i r e p , l a r u e P d l Weiße et al. Vet Res (2021) 52:112 Page 14 of 18 Figure 6 Oxidative burst of granulocytes is not associated with immunization (A), but with reduced bacterial survival (B). S. suis strain V3117/2 (cps14)‐induced oxidative burst rates of granulocytes from blood samples drawn 11 d post booster did not reveal an immunization effect of the multicomponent vaccine, but showed correlation with the respective bacterial survival factors (SF). To determine oxidative burst 106 CFU/mL V3117/2 or PBS (negative‑control) were added to the whole blood samples (11 dp booster) and induced by S. suis in granulocytes, 6 incubated for 15 min at 37 °C in water bath and for further 10 min after addition of dihydrorhodamin 123 (5 μg/mL). Following erythrocyte lysis the samples were directly measured by flow cytometry. S. suis‑induced oxidative burst rates were calculated as described in “Materials and methods”. A Comparison of S. suis‑induced oxidative burst rates of placebo‑treated (black circle) and vaccinated (grey square) animals. Statistical analysis was conducted with the unpaired t‑test (two‑tailed). B Spearman correlation between S. suis induced‑oxidative burst rates with the respective bacterial survival factors of S. suis V3117/2 (cps14) in blood of all animals (n 16) as shown in Figure 4. Animals that had to be euthanized after challenge with S. suis V3117/2 (cps14) due to a high clinical score are marked with white symbols (†n 5). × = = into the environment. Release of strong immunogens might limit binding of antibodies to the bacterial surface and induction of opsonophagocytosis in vivo. This might constitute an evolutionary advantage for the pathogen and deteriorate the protective efficacy of an immunogen- based vaccine. The protective efficacy of an S. suis vaccine does not only depend on the antigen but also on the adjuvant [47, 49] and the vaccination protocol. Though we cannot rule out that there are better adjuvants for this multicompo- nent vaccine, it should be noted that we have been using the water-in-oil adjuvant Emulsigen for a S. suis bacterin and also for a recombinant S. suis vaccine in previous studies using comparable vaccination protocols leading to significant protection against cps2 [53, 54]. Therefore, it appears unlikely that the adjuvant is the main reason for the disappointing protective efficacy. One limitation of our and other immunoproteomics studies is that the included convalescence and hyper- immune sera might mediate killing of S. suis mainly by capsule-specific antibodies and not by antibodies recog- nizing protein antigens. Accordingly, some convalescence sera included in our study mediated prominent killing in the opsonophagocytic assay only against the infection strain (e.g. sera of piglets 33 and 150; Additional file 2). Sera of piglets mediating killing of multiple serotypes generally recognize a large set of antigens that makes it very difficult to identify the crucial antigens (results not shown). In a previous study we demonstrated that ROS induc- tion in S. suis infected porcine blood is antibody and complement dependent [26]. In the present study, ROS induction in neutrophils after in vitro cps14 infection did not exhibit differences between blood drawn from placebo-treated and vaccinated piglets (Figure 6A). However, in porcine blood with high bactericidal activ- ity against S. suis cps14, we observed strong induction of ROS (> 5%). As ROS induction was not associated with vaccination, we suppose that antibodies crucial for the high ROS induction were induced through infection with strains colonizing piglets of this herd. Based on thorough screening of S. suis isolates of piglets from the original herd over many years we are convinced that this herd is free of S. suis cps1 and cps14. However, piglets of the Weiße et al. Vet Res (2021) 52:112 Page 15 of 18 Figure 7 Pre‑adsorption of sera with cps2 led to increased survival of cps14 (A) and decreased cps14‑induced oxidative burst (B). Sera collected from animals 11 days post booster immunization, which showed an V3117/2‐induced oxidative burst > 5% (n 7), were pre‑adsorbed with the cps2 strain isolated from a diseased piglet not included in the challenge study. Pre‑adsorbed and untreated control sera (ctr) were used to 106 CFU/mL of Cell Trace® reconstitute porcine blood, which was either incubated with 3 FarRed‑labbeld V3117/2 before adding dihydrorhodamine 123. Frequency of granulocytes showing both, oxidative burst and phagocytosis activity induced by cps14 strain V3117/2 was measured by flow cytometry (BD FACSCalibur and FlowJo) (B). Levels of IgG (C) and IgM (D) antibodies binding to the surface of cps14 were measured in non‑adsorbed (ctr) and pre‑adsorbed sera (cps2‑pre ads.). Mean values are indicated by horizontal lines, standard deviations by error bars. Significant differences between treated and control sera were either determined using Wilcoxon matched pair test (A 105 CFU/mL of V3117/2 for bactericidal assay (A) or 6 D) (p < 0.05, p < 0.01, p < 0.001, *p < 0.0001). C) or paired t‑test (B × × = + + original herd including the piglets used in this study were colonized with a cps2 S. suis strain of sequence type 28 which is not related to clonal complex 1 harboring the challenge strain. This strain caused disease in one piglet originally planned to be included in the challenge experi- ment. Thus, it appears likely that the other piglets devel- oped a serological response against this strain. Of note, the capsule of S. suis cps14 and cps2 share an identical sialic acid-containing side chain [55, 56]. Recognition of cps2 capsule by rabbit anti-cps2 serum is significantly reduced when sialic acid is removed [57], suggesting that the antibody response against cps2 in rabbits is mainly directed against the side chain. Therefore, one might speculate that infection with S. suis cps2 might induce capsule-specific IgM antibodies cross-reacting with the capsule of cps14. However, such cross-reaction has not been shown for porcine sera and serum raised in rab- bits against cps2 as use for agglutination generally does not cross-react with cps14 suggesting important con- formational differences [57]. In the present study, pre- adsorption of sera with the isolated cps2 strain led to a significant decrease of IgM antibody levels binding to cps14 (Figure 7) which is in accordance with cross- reacting IgM antibodies. These cross-reacting IgM anti- bodies might have made a big difference for survival in porcine blood and induction of ROS as IgM is a strong inducer of the classical complement pathway. In fact, we were able to observe increased bacterial survival and reduced cps14-induced phagocytosis and oxidative burst after pre-adsorption of sera with the isolated cps2 strain Weiße et al. Vet Res (2021) 52:112 Page 16 of 18 (Figure 7). It is, however, also possible that antibod- ies against protein antigens made a difference in ROS induction. In summary, the results of this study question the idea that strong immunogens expressed by different S. suis serotypes are a priori good candidates for protective antigens. S. suis colonizes different mucosal surfaces and tonsils during the entire life of a pig. After weaning, most conventional piglets go through an adaptive immune response against S. suis [15]. This pig-adapted pathogen has most likely gone through a substantial evolution- ary selection against antibody-mediated elimination. Recruitment of host proteins to the bacterial surface, attachment to host cells, formation of biofilms, hiding immunogens under a thick capsule and release of immu- nogens in the environment are all putative mechanisms that allow this pathogen to survive even in the presence of specific antibodies. Thus, the development of a cross- protective S. suis vaccine remains a major challenge. Abbreviations BSA: Bovine serum albumin; CFU: Colony forming unit; E.: Escherichia; FITC: Fluorescein isothiocyanate; HRP: Horseradish‑peroxidase; LB: Luria–Bertani; MLST: Multi locus sequence typing; MRP: Muramidase‑released protein; ORF: Open reading frame; PBS: Phosphate buffered saline; ROS: Reactive oxygen species; SAO: Surface antigen one; S.: Streptococcus; SDS‑Page: Sodium dodecyl sulfate polyacrylamide gel electrophoresis; SF: Survival factor; TBST: Tris‑buffered saline with 0.05% Tween 20; THB: Todd Hewitt broth. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s13567‑ 021‑ 00981‑3. Additional file 1: Oligonucleotides used for recombinant expression of S. suis antigens in E. coli. Name and sequences (5’‑3’) of oligonucleo‑ tide primers used for construction of expression vectors of recombinant proteins included in the multicomponent vaccine (SSU0934, SSU1869, SSU0757, SSU1950, SSU1664, SSU0187). Additional file 2: Origin and bactericidal activity of sera of suscep‑ tible piglets, sera pre‑infection and sera post‑infection. Information on the original experimental infection and results of bactericidal as well as opsonophagocytosis assays of sera defined as sera of susceptible pigs, sera pre‑infection and sera post‑infection. Additional file 3: Origin and activity of hyperimmune sera in opsonophagocytosis assays. Additional file 4: Multiple sequence alignment (tblastn) of antigens in different Streptococcus suis strains. Multiple sequence alignments of each antigen were conducted using sequences retrieved from the genome sequences of S. suis strains 10, 13‑00283‑02 and 16085/3b. As query served the respective sequence of strain P1/7. Dots represent identities, whereas differences are highlighted in red. Additional file 5: SDS‑PAGE and Western blot analysis of antigens included in the multicomponent vaccine. The purified proteins were run on SDS–polyacrylamide gel (A), transferred to membranes and probed with antisera raised in rabbits against each antigen (B‑G). Numbers on the left are molecular masses in kDa. Additional file 6: Scoring of fibrinosuppurative lesions of piglets challenged with S. suis cps14. Fourteen days after prime‑booster immunization with multicomponent vaccine or placebo, sixteen growing = × 109 CFU of piglets (n 8 per group) were infected intranasally with 5 S. suis cps14 V3117/2. Five vaccinated animals and four placebo animals demonstrated clinical signs of severe disease. In both groups two piglets had to be euthanized for animal welfare reasons because of signs of polyarthritis. One vaccinated animal showed additional signs of central nervous system dysfunction (opisthotonus, generalized tremor, ataxia) and had to be euthanized as well. Surviving piglets were sacrificed fourteen days post‑infection. Necropsies and histopathological screenings of the indicated tissues were conducted with all 16 piglets as described previously [10]. Acknowledgements We thank Hilde Smith (DLO, Lelystad, Netherlands) for providing S. suis strain 10. Flow cytometry was performed at the Core Unit Flow Cytometry (CUDZ) of the College of Veterinary Medicine, University of Leipzig. Vishnu Dhople is acknowledged for support in mass spectrometry. We acknowledge support from the Leipzig University within the program of Open Access Publishing. Authors’ contributions CW designed and conducted experiments. Furthermore, CW analysed the data and drafted the manuscript. DD designed and conducted xMAP® tech‑ nology‑based experiments for determination of IgG antibody titers. DD and SM analyzed related data. BJ and VF expressed and purified antigens included in the vaccine. NS and GA designed and conducted oxidative burst experi‑ ments and analyzed the data. Experimental infection and necropsies were performed by CW and CGB. KK did the histopathological screenings. CGB, VF, GA and UV conceived the study and designed experiments. PVW and CGB provided sera and designed the infection experiments used for generation of the sera included in Figure 1 as well as in Additional files 2, 3. All authors have read and approved the final manuscript. Funding Open Access funding enabled and organized by Projekt DEAL. This Project was funded by the German Federal Ministry of Education and Research within the Infect Control 2020 consortium (FKZ 03ZZ0816B to UV, FKZ 03ZZ0816C to CGB and GA and FKZ 03ZZ0816E to VF). Availability of data and materials The datasets analyzed during the current study are available from the cor‑ responding author on reasonable request. Declarations Ethics approval and consent to participate All animal experiments or samplings were conducted by veterinarians and in accordance with the principles outlined in the European Convention for the Protection of Vertebrate Animals Used for Experimental and Other Scientific Purposes and the German Animal Protection Law (Tierschutzgesetz). The pro‑ tocols for the animal experiments, from which sera used for characterization of immunogens originate, were either approved by the Committee on Animal Experiments of the Lower Saxonian State Office for Consumer Protection and Food Safety (Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit, LAVES under the permit number 87A044, 509.6‑42502‑ 04/829; 33.14‑42502‑04‑12/0965; 33.12‑42502‑04‑16/2305A) or the Landesdi‑ rektion Sachsen (permit no. TVV14/15; TVV26/15; N01/16; TVV11/16; TVV28/16; TVV37/17), which included approval through the registered committee for animal experiments. The collection of blood samples from pigs was approved by the Landesdirektion Sachsen (permit no. TVV40/18 and A09/19, respec‑ tively). The animal experiment of this study was approved by the Saxony Regional Office (permit no. TVV57/18), which includes approval through the registered committee for animal experiments. Competing interests The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Ceva Innovation Center GmbH is a company developing veterinary vaccines. VF and BJ are employees of Ceva Innovation Center GmbH. Furthermore, research of CGB and GA is funded by Ceva Innovation Center GmbH. Weiße et al. Vet Res (2021) 52:112 Page 17 of 18 Author details 1 Institute of Bacteriology and Mycology, Centre for Infectious Diseases, Faculty of Veterinary Medicine, Leipzig University, Leipzig, Germany. 2 Department of Functional Genomics, Interfaculty Institute for Genetics and Functional Genomics, University Medicine Greifswald, Greifswald, Germany. 3 Ceva Innovation Center GmbH, Dessau‑Rosslau, Germany. 4 Institute of Immunol‑ ogy, Centre for Infectious Diseases, Faculty of Veterinary Medicine, Leipzig University, Leipzig, Germany. 5 Institute of Pathology, Faculty of Veterinary Medicine, Leipzig University, Leipzig, Germany. 6 Department of Infectious Dis‑ eases, Institute for Microbiology, University of Veterinary Medicine Hannover, Hannover, Germany. Received: 23 March 2021 Accepted: 20 May 2021 References 1. Kerdsin A, Akeda Y, Hatrongjit R, Detchawna U, Sekizaki T, Hamada S, Gottschalk M, Oishi K (2014) Streptococcus suis serotyping by a new multi‑ plex PCR. J Med Microbiol 63:824–830 2. Goyette‑Desjardins G, Auger J‑PP, Xu J, Segura M, Gottschalk M (2014) Streptococcus suis, an important pig pathogen and emerging zoonotic agent‑an update on the worldwide distribution based on serotyping and sequence typing. Emerg Microbes Infect 3:e45 3. Wisselink HJ, Smith HE, Stockhofe‑Zurwieden N, Peperkamp K, Vecht U (2000) Distribution of capsular types and production of muramidase‑ released protein (MRP) and extracellular factor (EF) of Streptococcus suis strains isolated from diseased pigs in seven European countries. Vet Microbiol 74:237–248 4. Martinez G, Pestana de Castro AF, Ribeiro Pagnani KJ, Nakazato G, Dias da Silveira W, Gottschalk M (2003) Clonal distribution of an atypical MRP + phenotype of virulent Streptococcus suis serotype 2 EF and suilysin strains in Brazil. Can J Vet Res 67:52–55 + , 5. Costa ATR, Lobato FCF, Abreu VLV, Assis RA, Reis R, Uzal FA (2005) Serotyp‑ ing and evaluation of the virulence in mice of Streptococcus suis strains isolated from diseased pigs. Rev Inst Med Trop Sao Paulo 47:113–115 6. Wileman TM, Weinert LA, Howell KJ, Wang J, Peters SE, Williamson SM, Wells JM, Langford PR, Rycroft AN, Wren BW, Maskell DJ, Tucker AW (2019) Pathotyping the Zoonotic Pathogen Streptococcus suis: novel genetic markers to differentiate invasive disease‑associated isolates from non‑disease‑associated isolates from England and Wales. J Clin Microbiol 57:e01712‑e1718 8. 7. Goyette‑Desjardins G, Calzas C, Shiao TC, Neubauer A, Kempker J, Roy R, Gottschalk M, Segura M (2016) Protection against Streptococcus suis sero‑ type 2 infection using a capsular polysaccharide glycoconjugate vaccine. Infect Immun 84:2059–2075 Li Q, Lv Y, Li Y, Du Y, Guo W, Chu D, Wang X, Wang S, Shi H (2020) Live attenuated Salmonella enterica serovar Choleraesuis vector delivering a conserved surface protein enolase induces high and broad protec‑ tion against Streptococcus suis serotypes 2, 7, and 9 in mice. Vaccine 38:6904–6913 Li YA, Ji Z, Wang X, Wang S, Shi H (2017) Salmonella enterica serovar Choleraesuis vector delivering SaoA antigen confers protection against Streptococcus suis serotypes 2 and 7 in mice and pigs. Vet Res 48:89 10. Rieckmann K, Seydel A, Klose K, Alber G, Baums CG, Schütze N (2019) Vac‑ cination with the immunoglobulin M‑degrading enzyme of Streptococcus suis, IdeSsuis, leads to protection against a highly virulent serotype 9 strain. Vaccine X 3:100046 9. 11. Segura M (2015) Streptococcus suis vaccines: candidate antigens and progress. Expert Rev Vaccines 14:1587–1608 12. Brockmeier SL, Loving CL, Nicholson TL, Wang J, Peters SE, Weinert L, Chaudhuri R, Seilly DJ, Langford PR, Rycroft A, Wren BW, Maskell DJ, Tucker AW (2018) Use of proteins identified through a functional genomic screen to develop a protein subunit vaccine that provides significant protection against virulent Streptococcus suis in pigs. Infect Immun 86:e00559‑e617 13. Kock C, Beineke A, Seitz M, Ganter M, Waldmann K‑H, Valentin‑Weigand P, Baums CG (2009) Intranasal immunization with a live Streptococcus suis isogenic ofs mutant elicited suilysin‑neutralization titers but failed to induce opsonizing antibodies and protection. Vet Immunol Immuno‑ pathol 132:135–145 14. Vecht U, Wisselink HJ, Van Dijk JE, Smith HE (1991) Virulence of Strep‑ tococcus suis type 2 strains in newborn germfree pigs depends on phenotype. Infect Immun 60:550–556 15. Rieckmann K, Seydel A, Szewczyk K, Klimke K, Rungelrath V, Baums CG (2018) Streptococcus suis cps7: an emerging virulent sequence type (ST29) shows a distinct, IgM‑determined pattern of bacterial survival in blood of piglets during the early adaptive immune response after weaning. Vet Res 49:48 16. Roy D, Athey TBT, Auger JP, Goyette‑Desjardins G, Van Calsteren MR, Takamatsu D, Okura M, Teatero S, Alcorlo M, Hermoso JA, Segura M, Gottschalk M, Fittipaldi N (2017) A single amino acid polymorphism in the glycosyltransferase CpsK defines four Streptococcus suis serotypes. Sci Rep 7:4066 17. Athey TBT, Teatero S, Lacouture S, Takamatsu D, Gottschalk M, Fittipaldi N (2016) Determining Streptococcus suis serotype from short‑read whole‑ genome sequencing data. BMC Microbiol 16:162 18. Silva LMG, Baums CG, Rehm T, Wisselink HJ, Goethe R, Valentin‑Weigand P (2006) Virulence‑associated gene profiling of Streptococcus suis isolates by PCR. Vet Microbiol 115:117–127 19. Sielaff M, Kuharev J, Bohn T, Hahlbrock J, Bopp T, Tenzer S, Distler U (2017) Evaluation of FASP, SP3, and iST protocols for proteomic sample prepara‑ tion in the low microgram range. J Prot Res 16:4060–4072 20. Bhardwaj G, Dörr M, Sappa PK, Ameling S, Dhople V, Steil L, Klingel K, Empen K, Beug D, Völker U, Felix SB, Hammer E (2017) Endomyocardial proteomic signature corresponding to the response of patients with dilated cardiomyopathy to immunoadsorption therapy. J Proteomics 150:121–129 21. Cox J, Mann M (2008) MaxQuant enables high peptide identification rates, individualized p.p.b.‑range mass accuracies and proteome‑wide protein quantification. Nat Biotechnol 26:1367–1372 22. Dunbar SA (2006) Applications of Luminex® xMAPTM technology for rapid, high‑throughput multiplexed nucleic acid detection. Clin Chim Acta 363:71–82 23. Meyer TC, Schmidt F, Steinke J, Bröker BM, Völker U, Michalik S (2020) Technical report: xMAPr‑high‑dynamic‑range (HDR) quantification of antigen‑specific antibody binding. J Proteomics 212:103577 24. Seele J, Singpiel A, Spoerry C, von Pawel‑Rammingen U, Valentin‑ Weigand P, Baums CG (2013) Identification of a novel host‑specific IgM protease in Streptococcus suis. J Bacteriol 195:930–940 25. Seele J, Beineke A, Hillermann L‑M, Jaschok‑Kentner B, Von Pawel‑Ram‑ mingen U, Valentin‑Weigand P, Baums CG (2015) The immunoglobulin M‑degrading enzyme of Streptococcus suis, IdeSsuis, is involved in comple‑ ment evasion. Vet Res 46:45 26. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B 57:289–300 27. Rungelrath V, Öhlmann S, Alber G, Schrödl W, von Köckritz‑Blickwede M, de Buhr N, Martens A, Baums CG, Schütze N (2019) Survival of Streptococ- cus suis in porcine blood is limited by the antibody‑ and complement‑ dependent oxidative burst response of granulocytes. Infect Immun 88:e00598‑e619 28. R Core Development Team (2019) R: a language and environment for statistical computing. Austria, Vienna 29. Wickham H, Averick M, Bryan J, Chang W, McGowan L, François R, Grol‑ emund G, Hayes A, Henry L, Hester J, Kuhn M, Pedersen T, Miller E, Bache S, Müller K, Ooms J, Robinson D, Seidel D, Spinu V, Takahashi K, Vaughan D, Wilke C, Woo K, Yutani H (2019) Welcome to the Tidyverse. J Open Source Softw 4:1686 30. Gepp J, Luettgau D, Bleier A, Krabel T, Albers M (2018) helfRlein: R‑helper functions. R package version 0.2.2. https:// github. com/ STATW ORX/ helfR lein 31. Sievert C (2020) Interactive web‑based data visualization with R, plotly, and shiny. Chapman and Hall 32. Ahlmann‑Eltze C (2019) ggsignif: significance brackets for ‘ggplot2’. R package version 0.6.0. https:// cran.r‑ proje ct. org/ packa ge ggsig nif 33. Wickham H (2016) ggplot2: elegant graphics for data analysis. Springer 34. Auguie B (2017) gridExtra: miscellaneous functions for “Grid” graphics. R = package version 2.3 Weiße et al. Vet Res (2021) 52:112 Page 18 of 18 35. Lê S, Josse J, Husson F (2008) FactoMineR: an R package for multivariate analysis. J Stat Soft 25:1–18. https:// doi. org/ 10. 18637/ jss. v025.i. 01 36. Kassambra A, Mundt F (2020) factoextra: extract and visualize the results of multivariate data analyses. R package version 1.0.7 37. Slowikowski K (2020) ggrepel: automatically position non‑overlapping text labels with ‘ggplot2’. R package version 0.8.2 pathogen and zoonotic agent Streptococcus suis. Future Microbiol 7:259–279 52. Gómez‑Gascón L, Luque I, Tarradas C, Olaya‑Abril A, Astorga RJ, Huerta B, Rodríguez‑Ortega MJ (2018) Comparative immunosecretome analysis of prevalent Streptococcus suis serotypes. Comp Immunol Microbiol Infect Dis 57:55–61 38. Pedersen TL (2020) patchwork: the composer of plots. R package version 53. Seele J, Hillermann LM, Beineke A, Seitz M, von Pawel‑Rammingen U, 1.0.1 39. Bunk B, Jakóbczak B, Florian V, Dittmar D, Mäder U, Jarek M, Häußler S, Baums CG, Völker U, Michalik S (2021) Complete genome sequences of Streptococcus suis pig‑pathogenic strains 10, 13‑00283‑02, and 16085/3b. Microbiol Resour Announc 10:e01137‑e1220 40. Aranda J, Teixidó L, Fittipaldi N, Cortés P, Llagostera M, Gottschalk M, Barbé J (2012) Inactivation of the gene encoding zinc‑binding lipoprotein 103 impairs the infectivity of Streptococcus suis. Can J Vet Res 76:72–76 41. Wichgers Schreur PJ, Rebel JMJ, Smits MA, van Putten JPM, Smith HE (2011) Troa of Streptococcus suis is required for manganese acquisition and full virulence. J Bacteriol 193:5073–5080 42. Bonifait L, Vaillancourt K, Gottschalk M, Frenette M, Grenier D (2011) Purifi‑ cation and characterization of the subtilisin‑like protease of Streptococcus suis that contributes to its virulence. Vet Microbiol 148:333–340 43. Xu B, Zhang P, Li W, Liu R, Tang J, Fan H (2017) hsdS, belonging to the type I restriction‑modification system, contributes to the Streptococcus suis serotype 2 survival ability in phagocytes. Front Microbiol 8:1524 44. Zhang W, Lu CP (2007) Immunoproteomic assay of membrane‑associ‑ ated proteins of Streptococcus suis type 2 China vaccine strain HA9801. Zoonoses Public Health 54:253–259 45. Geng H, Zhu L, Yuan Y, Zhang W, Li W, Wang J, Zheng Y, Wei K, Cao W, Wang H, Jiang Y (2008) Identification and characterization of novel immunogenic proteins of Streptococcus suis serotype 2. J Proteome Res 7:4132–4142 46. Li Y, Martinez G, Gottschalk M, Lacouture S, Willson P, Dubreuil JD, Jacques M, Harel J (2006) Identification of a surface protein of Streptococcus suis and evaluation of its immunogenic and protective capacity in pigs. Infect Immun 74:305–312 47. Wisselink HJ, Vecht U, Stockhofe‑Zurwieden N, Smith HE (2001) Protec‑ tion of pigs against challenge with virulent Streptococcus suis serotype 2 strains by a muramidase‑released protein and extracellular factor vaccine. Vet Rec 148:473–477 48. Baums CG, Kock C, Beineke A, Bennecke K, Goethe R, Schröder C, Waldmann K‑H, Valentin‑Weigand P (2009) Streptococcus suis bacterin and subunit vaccine immunogenicities and protective efficacies against serotypes 2 and 9. Clin Vaccine Immunol 16:200–208 49. Li Y, Gottschalk M, Esgleas M, Lacouture S, Dubreuil JD, Willson P, Harel J (2007) Immunization with recombinant Sao protein confers protection against Streptococcus suis infection. Clin Vaccine Immunol 14:937–943 50. Gottschalk M, Segura M (2000) The pathogenesis of the meningitis caused by Streptococcus suis: the unresolved questions. Vet Microbiol 76:259–272 51. Fittipaldi N, Segura M, Grenier D, Gottschalk M (2012) Virulence factors involved in the pathogenesis of the infection caused by the swine Valentin‑Weigand P, Baums CG (2015) The immunoglobulin M‑degrading enzyme of Streptococcus suis, IdeSsuis, is a highly protective antigen against serotype 2. Vaccine 33:2207–2212 54. Baums CG, Valentin‑Weigand P (2009) Surface‑associated and secreted factors of Streptococcus suis in epidemiology, pathogenesis and vaccine development. Anim Health Res Rev 10:65–83 55. Van Calsteren MR, Gagnon F, Calzas C, Goyette‑Desjardins G, Okura M, Takamatsu D, Gottschalk M, Segura M (2013) Structure determination of Streptococcus suis serotype 14 capsular polysaccharide. Biochem Cell Biol 91:49–58 56. Van Calsteren M‑R, Gagnon F, Lacouture S, Fittipaldi N, Gottschalk M (2010) Structure determination of Streptococcus suis serotype 2 capsular polysaccharide. Biochem Cell Biol 88:513–525 57. Van Calsteren MR, Goyette‑Desjardins G, Gagnon F, Okura M, Takamatsu D, Roy R, Gottschalk M, Segura M (2016) Explaining the serological char‑ acteristics of Streptococcus suis serotypes 1 and 1/2 from their capsular polysaccharide structure and biosynthesis. J Biol Chem 291:8387–8398 58. Zheng B, Zhang Q, Gao J, Han H, Li M, Zhang J, Qi J, Yan J, Gao GF (2011) Insight into the interaction of metal ions with TroA from Streptococcus suis. PLoS One 6:e19510 59. Bonifait L, Grenier D (2011) The SspA subtilisin‑like protease of Streptococ- cus suis triggers a pro‑inflammatory response in macrophages through a non‑proteolytic mechanism. BMC Microbiol 11:47 60. Bonifait L, de la Cruz Dominguez‑Punaro M, Vaillancourt K, Bart C, Slater J, Frenette M, Gottschalk M, Grenier D (2010) The cell envelope subtilisin‑ like proteinase is a virulence determinant for Streptococcus suis. BMC Microbiol 10:42 61. Wu Z, Shao J, Ren H, Tang H, Zhou M, Dai J, Lai L, Yao H, Fan H, Chen D, Zong J, Lu C (2016) A Streptococcus suis LysM domain surface protein contributes to bacterial virulence. Vet Microbiol 187:64–69 62. Ge J, Feng Y, Ji H, Zhang H, Zheng F, Wang C, Yin Z, Pan X, Tang J (2009) Inactivation of dipeptidyl peptidase IV attenuates the virulence of Strep- tococcus suis serotype 2 that causes streptococcal toxic shock syndrome. Curr Microbiol 59:248–255 63. Dumesnil A, Martelet L, Grenier D, Auger JP, Harel J, Nadeau E, Gottschalk M (2019) Enolase and dipeptidyl peptidase IV protein sub‑unit vaccines are not protective against a lethal Streptococcus suis serotype 2 challenge in a mouse model of infection. BMC Vet Res 15:448 Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub‑ lished maps and institutional affiliations. • fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld• rapid publication on acceptance• support for research data, including large and complex data types• gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year • At BMC, research is always in progress.Learn more biomedcentral.com/submissionsReady to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:
10.1371_journal.pcbi.1008853	RESEARCH ARTICLE Interactions between all pairs of neighboring trees in 16 forests worldwide reveal details of unique ecological processes in each forest, and provide windows into their evolutionary histories 4,5, Yi Jin6, 7, Jill ThompsonID 12, Sara GermainID 7, Heming Liu13, Joseph SmokeyID 1☯, Bin Wang2☯, Shuai Fang3, Yunquan WangID 8, Kyle E. Harms9, Sandeep Pulla10,11, Christopher WillsID James LutzID Bonifacio PasionID Hsin Su15, Nathalie ButtID 20, H. S. DattarajaID YangID Shameema EsufaliID Chang-Fu Hsieh27, Fangliang He18, Stephen Hubbell28, Zhanqing Hao3‡, Akira Itoh29, 30, Buhang Li18, Xiankun Li2, Keping Ma5, Michael MorecroftID David KenfackID Xiangcheng Mi5, Yadvinder Malhi32, Perry Ong33†‡, Lillian Jennifer RodriguezID 35, Raman SukumarID 10,34, I Fang SunID S. SureshID 13, Xugao Wang3, T. L. YaoID Maria Uriarte38, Xihua WangID 16,17, Chengjin Chu18, George ChuyongID 23, 21, Stuart Davies22, Sisira EdiriweeraID 31, 33, H. 14, Sheng- 19, Chia-Hao Chang- 10, Sylvester Tan36‡, Duncan Thomas37, 25, Jess Zimmermann39 24, Christine Dawn Fletcher25, Nimal Gunatilleke26, Savi Gunatilleke26, a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Wills C, Wang B, Fang S, Wang Y, Jin Y, Lutz J, et al. (2021) Interactions between all pairs of neighboring trees in 16 forests worldwide reveal details of unique ecological processes in each forest, and provide windows into their evolutionary histories. PLoS Comput Biol 17(4): e1008853. https://doi.org/10.1371/journal.pcbi.1008853 Editor: Mercedes Pascual, University of Chicago, UNITED STATES Received: May 28, 2020 Accepted: March 3, 2021 Published: April 29, 2021 Copyright: This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication. Data Availability Statement: BCI FDP data are available from https://repository.si.edu/handle/ 10088/11. The authors do not own the data and are unable to share it in a public repository, as the majority of the data is held by various government agencies. However interested researchers are able to access all the data through the Smithsonian Institution’s ForestGeo project (https://forestgeo.si. edu/explore-data) The EAA analysis programs, written in R, along with their documentation, are 1 Division of Biological Sciences, University of California, San Diego, La Jolla, California, United States of America, 2 Guangxi Key Laboratory of Plant Conservation and Restoration Ecology in Karst Terrain, Guangxi Institute of Botany, Guangxi Zhuang Autonomous Region and Chinese Academy of Sciences, Guilin, 3 Key Laboratory of Forest Ecology and Management, Institute of Applied Ecology, Chinese Academy of Sciences, Shenyang, 4 College of Chemistry and Life Sciences, Zhejiang Normal University, Jinhua, 5 State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, 20 Nanxincun, Xiangshan, Beijing, 6 College of Life Sciences, Zhejiang University, Hangzhou, 7 Department of Wildland Resources, Utah State University, Logan, Utah, United States of America, 8 Center for Ecology & Hydrology, Penicuik, Midlothian, Scotland, 9 Department of Biological Sciences, Louisiana State University, Baton Rouge, Los Angeles, United States of America, 10 Divecha Centre for Climate Change, Indian Institute of Science, Bengaluru, India, 11 National Centre for Biological Sciences, GKVK Campus, Bengaluru, India, 12 Center for Integrative Conservation, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Menglun, Mengla, Yunnan, 13 Zhejiang Tiantong Forest Ecosystem National Observation and Research Station, School of Ecological and Environmental Sciences, East China Normal University, Shanghai, 14 Department of Biology, Memorial University of Newfoundland, Newfoundland, Canada, 15 Institute of Ecology and Evolutionary Biology, National Taiwan University, Taipei, 16 School of Geography and the Environment, University of Oxford, Oxford, United Kingdom, 17 School of Biological Sciences, The University of Queensland, St. Lucia, Queensland, Australia, 18 Department of Ecology, State Key Laboratory of Biocontrol and School of Life Sciences, Sun Yat-sen University, Guangzhou, 19 Department of Botany and Plant Physiology, University of Buea, Cameroon, 20 Department of Biological Sciences, National Sun Yat- sen University, Kaohsiung, 21 National Centre for Biological Sciences, Bengaluru, India, 22 Center for Tropical Forest Science, Smithsonian Institution, Washington, DC, United States of America, 23 Faculty of Science and Technology, Uva Wellassa University, Badulla, Sri Lanka, 24 Department of Botany, University of Peradeniya, Peradeniya Sri Lanka, 25 Forest Research Institute Malaysia, Kepong Selangor, Malaysia, 26 Dept. of Botany, Faculty of Science, University of Peradeniya, Peradeniya Sri Lanka, 27 Taiwan Forestry Research Institute, Taipei, 28 Department of Ecology and Evolutionary Biology, University of California, Los Angeles, Los Angeles, California, United States of America, 29 Graduate School of Science, Osaka City University, Sumiyoshi Ku, Osaka, Japan, 30 Center for Tropical Forest Science–Forest Global Earth Observatory (CTFS-ForestGEO), Smithsonian Tropical Research Institute, NMNH—MRC, Washington, DC, United States of America, 31 Natural England Mail Hub, County Hall, Worcester, United Kingdom, 32 School of Geography and the Environment, Oxford University Centre for the Environment, University of Oxford, Oxford, United Kingdom, 33 Institute of Biology, College of Science, University of the Philippines Diliman, Diliman, Quezon City, Philippines, 34 Centre for Ecological Sciences, Indian Institute of Science, Bengaluru, India, 35 Department of Natural Resources and Environmental Studies, National Dong Hwa University, Hualien, 36 Forest Department Sarawak, Bangunan Wisma Sumber Alam, Jalan Stadium, Petra Jaya, PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 1 / 33 freely available from https://github.com/ wangbinzjcc/EAAr. Funding: The author(s) received no specific funding for this work. Competing interests: The authors have declared that no competing interests exist. Authors Perry Ong, Zhanqing Hao and Sylvester Tan were unable to confirm their authorship contributions. On their behalf, the corresponding author has reported their contributions to the best of their knowledge. Each of 16 forests shows a unique pattern of between-tree interactions Kuching, Sarawak, Malaysia, 37 Department of Biology, Washington State University, Vancouver, Washington State, United States of America, 38 Department of Ecology, Evolution, and Environmental Biology, Columbia University, New York city, New York, United States of America, 39 Dept of Environmental Sciences, University of Puerto Rico, Rio Piedras, San Juan, PR, United States of America ☯ These authors contributed equally to this work. † Deceased. ‡ Unavailable. cwills@ucsd.edu Abstract When Darwin visited the Galapagos archipelago, he observed that, in spite of the islands’ physical similarity, members of species that had dispersed to them recently were beginning to diverge from each other. He postulated that these divergences must have resulted pri- marily from interactions with sets of other species that had also diverged across these other- wise similar islands. By extrapolation, if Darwin is correct, such complex interactions must be driving species divergences across all ecosystems. However, many current general eco- logical theories that predict observed distributions of species in ecosystems do not take the details of between-species interactions into account. Here we quantify, in sixteen forest diversity plots (FDPs) worldwide, highly significant negative density-dependent (NDD) com- ponents of both conspecific and heterospecific between-tree interactions that affect the trees’ distributions, growth, recruitment, and mortality. These interactions decline smoothly in significance with increasing physical distance between trees. They also tend to decline in significance with increasing phylogenetic distance between the trees, but each FDP exhibits its own unique pattern of exceptions to this overall decline. Unique patterns of between-spe- cies interactions in ecosystems, of the general type that Darwin postulated, are likely to have contributed to the exceptions. We test the power of our null-model method by using a deliberately modified data set, and show that the method easily identifies the modifications. We examine how some of the exceptions, at the Wind River (USA) FDP, reveal new details of a known allelopathic effect of one of the Wind River gymnosperm species. Finally, we explore how similar analyses can be used to investigate details of many types of interactions in these complex ecosystems, and can provide clues to the evolution of these interactions. Author summary Worldwide, ecosystems are collapsing or in danger of collapse, but the precise causes of these collapses are often unknown. Observational and experimental evidence shows that all ecosystems are characterized by strong interactions between and among species, and that these webs of interactions can be important contributors to the preservation of eco- system diversity. But many of the interactions–such as those involving pathogenic micro- organisms and the chemical defenses that are mounted by their prey–are not easily identified and analyzed in ecosystems that may have hundreds or thousands of species. Here we use our equal-area-annulus analytical method to examine census data from over three million trees in forest plots from around the world. We show how the method can be used to flag pairs and groups of species that exhibit unusual levels of interaction and PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 2 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions that are likely on further investigation to yield information about their causative mecha- nisms. We give a detailed example showing how some of these interactions can be traced to defense mechanisms that are possessed by one of the tree species. We explore how our method can be used to identify the between-species interactions that play the largest roles in the maintenance of ecosystems and their diversity. Introduction The] inhabitants of each separate [Galapagos] island, though mostly distinct, are related in an incomparably closer degree to each other than to the inhabitants of any other part of the world‥‥ [Dissimilarities] between the endemic inhabitants of the islands may be used as an argument against my views; for it may be asked, how has it happened in the several islands situated within sight of each other, having the same geological nature, the same height, cli- mate, &c., that many of the immigrants should have been differently modified, though only in a small degree. This long appeared to me a great difficulty: but it arises in chief part from the deeply-seated error of considering the physical conditions of a country as the most important for its inhabitants; whereas it cannot, I think, be disputed that the nature of the other inhabitants, with which each has to compete, is at least as important, and generally a far more important element of success. Charles Darwin, The Origin of Species, 1st ed. 1859, p. 400. In this passage from the Origin, Darwin effectively founded the field of evolutionary ecology. He was faced with the difficulty of explaining recent adaptive radiations that sometimes resulted in distinct species on the different islands of the Gala´pagos archipelago, even though the islands have similar physical properties. The solution, he suggested, must lie in these evolv- ing populations’ interactions with other species, the mix of which should differ among the individual islands. (And those other species, by his reasoning, would simultaneously be evolv- ing in their own unique directions as a result of their own sets of between-species interactions.) But his claim that the importance of such between-species interactions "cannot be disputed" was far from being demonstrated at the time. In the century and a half since the Origin, ecologists and evolutionary biologists have explored the many interactions among species that share the same ecological community, in ever-greater detail and with ever-more-sophisticated tools [1]. Modeling has pointed the way [2–6]. Even so, such interactions must involve many more species, occupying a variety of dif- ferent trophic levels, than those that can be examined in a typical study. Host-pathogen inter- actions were early postulated to be important in the maintenance of species diversity [7], and were soon realized to have a high likelihood of contributing to negative density-dependent (NDD) interactions between host species [8, 9]. Such interactions have been detected in the relatively simple ecosystems of the Gala´pagos [10] and in complex ecosystems such as tropical forests [11–13], coral reefs [14], and lacustrine fish communities [15]. The classic Lotka-Volterra model, based on competition coefficients, examines species that compete directly for resources, and shows that the species can coexist if each has resources that other species cannot access regardless of their numbers [2]. A second important group of mod- els involves NDD interactions, in which a selective advantage to species that are locally rare switches to a selective disadvantage when those species become locally common. NDD can lead to multiple stable internal density equilibria that permit numerous species to occupy the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 3 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions same ecosystem [3–6]. Possible mechanisms for NDD effects can include species interactions with both physical and biological factors. Many general ecological theories seeking to explain the distributions of species that occupy the same or similar trophic levels have tended to gloss over such complexities. In 2010, McGill [16] surveyed six “unified theories of biodiversity,” all of which had shown success in predict- ing observed species abundance distributions and species-area curves at scales of 100 m and above. He showed that all these theories employ three assumptions: intraspecific clumping, intraspecific variation in global abundance, and—most importantly for the present study— spatial independence of the distributions of different coexisting species. Darwin had postulated that species populating a multitude of trophic levels in an ecosystem are continuously interacting, and that these interactions contribute to evolutionary divergence. Given the growing evidence for such interactions (see [17] for an extreme example), it is sur- prising that apparently successful general ecological theories can be constructed using the assumption that species-species interactions are irrelevant to the overall structure of communi- ties. General unified theories of ecosystems may indeed be congruent with the distributions of component species that happen to be easily countable, provided that the scale is 100 m and above. But they incorporate no information about the existence of biotic and abiotic interac- tions at smaller scales, which is where most between-species interactions are likely to take place. How common, how complex, and how significant in their effects are the fine-structure between-species interactions that Darwin postulated? Can an understanding of these interac- tions lead to more complete theories that underlie ecosystem structures and their evolutionary trajectories? Here we test the spatial independence assumption that McGill shows is basic to the most general ecological distribution theories. We examine sixteen multiply-censused forest diversity plots (FDPs) that are scattered over a wide variety of biogeographic regions (Table 1), and find that the assumption does not hold at the scale of meters. We also show that the pat- tern of departures from independence can reveal new information about between-species interactions. We use the Equal-Area Annulus (EAA) [18] point-pattern method to visualize and quantify non-random patterns of tree clustering, distributions of tree recruitment and mortality, and the influence of surrounding trees on tree growth. We show details of how these interactions occur not only between conspecifics, where they are well-known [5, 13, 19, 20], but also between species that are separated across a wide range of phylogenetic distances. We show that, although the interactions decrease smoothly in significance with increasing physical dis- tance between trees, they exhibit complex relationships with phylogenetic distance that are unique to each of the study’s FDPs. Such complexities are not dealt with in the global theories examined by McGill. Because of the many differences between EAA and the commonly-used regression-based methods that are used to detect NDD effects, and because of the many modifications that have been made to EAA since its original publication (subheads 1–8 in the Modifications to the Original Method section), we have chosen to place the extensive Materials and Methods sec- tion immediately following this introduction. We emphasize how the EAA method avoids the statistical biases [21] that are inherent in regression-based methods. The "regression dilution" issue flagged by that paper is not a problem in our analyses because a bias towards zero makes the EAA method less—not more—likely to detect NDD (i.e. errors in predictors would reduce the statistical power of the method instead of increasing the Type I error rate). The EAA method also incorporates a number of desirable features that ideally should be exhibited by methods designed to detect density-dependent effects, as discussed in a recent review [22]. A method should (1) measure the relative magnitudes of conspecific NDD and heterospecific NDD and how they vary among different between-species interactions, (2) PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 4 / 33 PLOS COMPUTATIONAL BIOLOGYTable 1. Some characteristics of the FDPs examined in this paper, arranged by latitude. FDP Dim (m) No. of census intervals Total of species recorded Avg. tree density (trees/m2) & no. annuli used Annual Rainfall (mm) Latitude/ Longitude Min/max or avg. temp (oC) Each of 16 forests shows a unique pattern of between-tree interactions Pasoh (Peninsular Malaysia) Lambir (Sarawak, Malaysia) Korup (Cameroon) 1000 X 500 1040x500 1000 x 500 Sinharaja (Sri Lanka) 500 x 500 Barro Colorado Island (Panama) Mudumalai (India) 1000 x 500 1000 x 500 Palanan (Philippines) 400 x 400 Luquillo (Puerto Rico) 320x500 Nonggang 500 x 300 Heishiding Fushan 1000 x 500 500 x 500 Gutianshan 600 x 400 Tiantong 500 x 400 Changbaishan 500 x 500 Wind River (US) 800 x 340 Wytham Woods (UK) 300x600 5 2 2 2 6 898 1180 494 239 320 2 (4-yr intervals) 76 3 4 1 1 1 2 1 2 1 2 319 163 217 245 111 159 154 52 26 24 https://doi.org/10.1371/journal.pcbi.1008853.t001 0.671 20 0.665 10 0.656 10 0.829 10 0.470 20 0.035 10 0.210 10 0.289 10 0.453 10 0.546 10 0.463 10 0.586 10 0.604 10 0.155 10 0.116 10 0.112 10 2000 2700 2.98N/102.3E 25.8/28.3 4.2N/114E 31.4/22.1 5500 (seasonal) 5.1N/8.9E 22.7/30.6 5000 6.4N/80.4E 20.4/24.7 2600 (seasonal) 9.15N/79.85W 23/32 1300 (seasonal) 11.6N/76.5E 16.4/27.4 3200 (typhoons) 17.0N/122.4E 26.1 3500 (hurricanes) 18.3N/65.8W 23.0 1300 (seasonal) 22.4N/107.0E 19.0/27.2 1700 (seasonal) 23.3N/111.5E 10.6/28.4 4300 (typhoons) 24.8N/121.6E 18.2 2000 29.1N/118.1E 4.3/27.9 5000 (some typhoons) 700 29.8N/121.8E 16.2 42.4N/128.1E 3.6 2300 (seasonal) 45.8N/122W -2/27 700 51.8N/1.34W 10 evaluate the relative roles of conspecific and heterospecific NDD in the maintenance of eco- logical diversity, (3) remove the biases inherent in statistical methods that do not compare the actual data with appropriate null models, (4) distinguish the relative sizes of the contributions to NDD of species with different abundances and life histories, and the contributions of bio- geographic factors such as latitude and rainfall, (5) follow the contributions of organisms at different stages in their life histories, (6) provide a route for further examination of the details of the NDD mechanisms themselves and the long-term evolutionary implications of these mechanisms. Unlike regression-based methods, EAA uses null models that isolate the variables being examined while leaving all the other properties of these extensive data sets unchanged. This enables us to present details of the species-species interactions with high statistical confidence. For clarity, we provide a step-by-step illustration of an EAA analysis (Fig 1). We also provide an example showing that the method is highly sensitive to small deliberately-introduced changes in the FDP data (subhead 11 of the Modifications to the Original Method section and Fig 2). In the Results section, we present EAA analyses of all sixteen FDPs (Figs 3–7), and examine in detail the causes of the heterospecific interaction peaks and valleys that are observed at the temperate Wind River (Washington State, USA) FDP (Fig 8). We show how EAA analysis PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 5 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Fig 1. An overview of a typical EAA analysis. At top is a diagram of a large focal tree in the Lambir (Malaysia) FDP, surrounded by 10 concentric annuli each of area 50 m2. For simplicity, the trees shown in the diagram are only a sample of some of the annular surviving trees (S), recruits (R), and trees that die (D). In the diagram the trees shown are sampled from among the trees that lie at zero or at 185 Ma (mega-annum) phylogenetic distances from their LCA with the focal tree, although of course all the trees in the annuli are used in the entire EAA analysis. Generalized Additive Model (GAM) fits to patterns of clustering of surviving annular trees, using data from the closest and the furthest annulus, are shown in the two-dimensional graphs on the right of the diagram. In this analysis, the observed annular clustering is presented as deviations (z-values) from a null model expectation for four quantiles of focal tree diameter. The null model is generated by repeated shuffling of the focal tree diameters within species, so that any positive z-values for some of the focal-annular size quantiles must be balanced by negative values for others. Such positive-negative balances are expected in analyses of recruitment, clustering and mortality, but not in analyses of growth (Materials and Methods). The 95% confidence intervals of the GAM curves are shown in gray. Brown horizontal lines show the 95% confidence intervals around zero z-values. To help orient the viewer, gray arrows connect some of the closely-related and distantly-related survivors in the diagram to the places at which their data contributes to the largest-quantile focal tree lines (red) on the two-dimensional graphs. Each data point in the 2D graphs represents the difference between the actual and the null-model data for all focal trees in a given census period that have annular trees within a specific phylogenetic range. The null-model data have been generated by repeated shuffling of focal tree properties (size or growth rate) within species. A new point is generated for each of the ten replicates of the actual-null comparisons and for each of the census periods at the FDP. The points in the graphs form clusters because, with each replicate, species pairs separated by similar phylogenetic distances are shuffled at random in order to fill each of the phylogenetic distance quantiles. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 6 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Thus, each of the gray arrows that shows the contribution of an individual tree simply shows where the tiny amount of information that is contributed by that tree’s focal-annular interactions ends up in the data points in the graphs. The three-dimensional graphs show GAM fits of the largest-quantile (graph A) and smallest-quantile (graph B) focal tree size data across all ten annuli. Regions of the surfaces that lie within the range of non-significant z-values along the z-axis are gray; those that lie outside this range, and that therefore represent significant z-values, are colored. The colors start with green and shade through blue as the significance of the positive or negative z-values increases. The orientations of the three-dimensional graphs presented here sometimes differ, in order to reveal details of the surfaces. As with the lines on the two-dimensional graphs, the 3D surfaces themselves have confidence intervals, but the confidence intervals are not shown here. Typical confidence intervals on the 3D surfaces, which tend to be small, are visualized more easily if these three-dimensional graphs can be rotated by the viewer. A sampling of such rotatable graphs is presented as html files in S3–S11 Figs. https://doi.org/10.1371/journal.pcbi.1008853.g001 provides new details of the role played by strong allelopathic effects of a gymnosperm, the western hemlock Tsuga heterophylla, on some but not all of the angiosperms in the FDP to which it is very distantly related [23]. This example demonstrates the potential of EAA to examine the roles of heterospecific interactions that may involve a wide variety of tree charac- teristics and microenvironmental factors. EAA provides a tool to measure the sizes of the con- tributions of these variables to species distributions and life history patterns. Our preliminary findings show that EAA results and the experiments that they suggest will help to pinpoint unusually significant interactions that can be investigated further through field observations and experiments. These findings will in turn enable us to unravel the true complexity and the evolutionary histories of the many between-species interactions that, as Darwin had believed, “cannot be disputed.” Materials and methods FDP data This paper surveys demographic and tree-distribution data from 16 repeatedly-censused forest dynamics plots (FDPs) (https://forestgeo.si.edu/sites-all). The FDPs have been established in locations ranging from tropical to high-temperate latitudes, and encompass a wide range of seasonal and non-seasonal rainfall patterns (Table 1). Repeated censuses of the FDPs include all trees present during each census that have a diameter of 1 cm or greater at a height of 1.3 m. The EAA method The EAA method [18] examines interactions between "focal" trees, made up of all the surviving trees during a census period in the FDP, and the "annular" trees that occupy successive concen- tric annuli of equal area around the focal trees. The use of these successive annuli, which each consist of similar amounts of data that can be analyzed with the same statistical power, permits unbiased statistical comparisons of the significance of interactions over a range of physical focal-annular tree distances. EAA draws on many previous studies that have employed quadrat or point pattern analysis, coupled with a null modeling or Monte Carlo approach to generating control distributions in which specific components of the data have been randomized [11, 24–26]. The EAA method is an extension of neighborhood density functions such as the O-ring spatial statistic [26] and the Dx index [27]. EAA is similar to bivariate mark correlation analyses [26, 28]. The r-mark func- tion is a non-parametric estimator of the response of the growth of small trees to the presence of a large tree at distance r. Point pattern methods are not based on regression analysis, like many of the methods that have been used to search for positive or negative density-dependent effects in forest data. These regression-based methods have recently been criticized as suscepti- ble to over- or underestimation of the magnitudes of the effects that are being searched for [21]. EAA, in contrast, compares the distributions of the actual data with distributions in null models in which only the variable or variables of interest are repeatedly randomized and all other PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 7 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions parameters of the FDPs are left untouched. Each iteration of the null model is analyzed in the same way as the real data, and the entire distribution of these null model results is used to test the difference between the real and randomized data. In each of these iterated replicates of the null-model data, any heteroskedasticity of the distributions of within-species focal tree growth rates and focal tree sizes is left unaltered and therefore cannot bias the results. A diagrammatic example of EAA analysis Fig 1 shows, in diagrammatic form, the steps of a typical EAA analysis, in this case the relation- ship between focal tree size and clustering of annular trees in the Lambir FDP (Sarawak, Bor- nean Malaysia). Overview of the EAA tests Table 2 lists the current EAA tests, the focal and annular tree properties that each test exam- ines, and the expected results if focal-annular NDD interactions are present. Each of the tests is carried out as illustrated in Fig 1. Additional details of each test, including details of the null models used, are given below. Modifications of the original EAA method Overview of the modifications The EAA method has been redesigned since its initial publication. In addition to the use of equal-area annuli surrounding focal trees, the method now divides sets of FDP data into equal Table 2. Summary of the properties of, and expectations for, the EAA analyses that are used in this paper. The expected results for each test are for NDD focal-annu- lar interactions; the PDD expectation is the opposite. Focal tree properties Annular tree properties Null Model Comparison Expected Results Focal-annular properties employed, and expectations shared by all tests: Species, position, diameter, growth rate, recruitment and mortality for all trees in each census period For each annulus: species, basal area, distance from focal tree in Ma, whether trees were recruited or died during census period A single focal-tree attribute is repeatedly shuffled within species to serve as a control Focal-annular differences from null model should decline in significance as either physical or phylogenetic focal-annular distance increases Properties examined and expect results for each null model comparison test, assuming NDD focal-annular interactions: Test 1) Relationship between focal survivor sizes and their annular survivor summed basal area Focal survivor size Summed basal area of annular tree survivors that fall within a given quantile of phylogenetic distance from focal tree Shuffle focal tree survivor sizes within species Negative relationship between focal tree size and annular survivor summed basal area Test 2) Relationship between focal survivor sizes and their annular recruit fraction Focal survivor size Fraction of trees in the annulus and phylogenetic distance quantile that recruit Shuffle focal tree survivor sizes within species Negative relationship between focal tree size and annular recruit fraction Test 3) Relationship between focal survivor sizes and their annular mortality fraction Focal survivor size Fraction of trees in the annulus and phylogenetic distance quantile that die Shuffle focal tree survivor sizes within species Positive relationship between focal tree size and fraction of annular trees that die Test 4) Relationship between a focal tree’s growth rate (normalized within species) and its annular tree basal area Focal survivor growth rate Summed annular survivor basal area in the phylogenetic distance quantile Shuffle normalized focal tree survivor growth rates within species Negative relationship between focal tree growth rate and annular tree summed basal area Test 5) Differences between focal trees that do and do not recruit and their annular recruit fractions Focal recruits vs. other focal trees Fraction of trees in the annulus and phylogenetic distance quantile that recruit Shuffle properties of all focal trees within species Higher fraction of annular recruits around focal recruits Test 6) Differences between focal trees that do and do not die and their annular mortality fractions Focal trees that die (separated into small and large) vs. other focal trees Fraction of trees in the annulus and phylogenetic distance quantile that die Shuffle properties of all focal trees within species Higher fraction of annular trees that die around focal trees that die https://doi.org/10.1371/journal.pcbi.1008853.t002 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 8 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions quantiles, so that the statistical power of the analyses of all the subdivisions of the data in a given FDP are equivalent. Two- and three-dimensional GAM curves are now fit to the data. These curves reveal details and significance of effects traceable to phylogenetic distances between species. A uniform method of estimating phylogenetic distances between species is applied to all the FDPs. Equations that quantify the analyses of focal-annular NDD-influenced effects are derived in [18]. Significances of the effects, compared to the mean of 1,000 iterations of the null models, are estimated using z-scores. The GAM analyses we employ here use spline-based smooth terms [29]. GAM fits of curves to the data use the formula y ~s(x, k = k-value), where s estab- lishes the parameters of spline-based smooth terms and the k-value is the number of smooth terms employed. The optimal k-value is determined using gam.check [30]. Edge-effect correc- tions have been applied to all the FDPs [31]. The z-values obtained by all the analyses are adjusted for the discovery of false positives, using the Benjamini-Hochberg correction for independent statistics [32]. The data used to generate the GAM graphs are given in Supporting Information compressed data files S1 and S2 Datas. Division of the data into quantiles Because we are comparing different FDPs that exhibit a wide range of species numbers, tree size distributions and densities, plot sizes, and distributions of phylogenetic distances between species, we have introduced standardized protocols for subdividing the data. Our goal is to ensure that each of the subdivisions of a set of data are of approximately equal size, so that the statistical power of the EAA tests remains the same across successive concentric annuli, phylo- genetic distance intervals, and subdivisions of the focal and annular trees. Annuli vary from 5 to 20 in number in the different FDPs, depending on overall tree density, but the total area around each focal tree encompassed by the annuli is 500 m2 in all FDPs. Focal tree diameters at the start of each census period for each species are divided into four equal quantiles. Totals of annular tree biomasses in each of the annuli (approxi- mated by the sum of the areas at standardized height) are divided into five equal quantiles. Phylogenetic distances between species are also divided into quantiles, but this division poses special problems. First, the amount of data varies among FDPs. Therefore, in order to ensure that there is sufficient data for analysis, we have used different numbers of phyloge- netic distance quantiles in different FDPs. We have been able to use as many as 20 quantiles in large, species-rich FDPs such as BCI and Pasoh, but have been limited to as few as 5 quantiles in smaller, less speciose FDPs such as Wind River and Wytham Woods. Second, because the distribution of pairwise phylogenetic distances between species is different in each FDP, and these distributions are far from uniform, subdivision of quantile differences often means that many species pairs that are separated by the same or similar phylogenetic distances will fall into different quantiles. Each analysis for each census period, therefore, is repeated ten times, each with 100 iterations of the null model. With each repetition, the order within each set of focal-annular pairs that share the same phylogenetic distance value is shuffled. This ensures that if large numbers of species pairs in an FDP share the same phy- logenetic distance, subdivision into quantiles in the replicated analyses will have placed dif- ferent random mixes of these pairs in adjacent quantiles. Measurement of effect of annular trees on focal tree growth rates We define focal survivors as trees that are present at the beginning and end of a five-year cen- sus interval. For each data graph, data from all focal trees of all species are pooled. Normalized focal tree growth rates are calculated for each FDP as standard deviations from the mean of a PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 9 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions given decile of diameters of the focal trees of a given species within a census period. This approach avoids the confounding effects of tree size differences, species differences, and secu- lar trends over time on the growth rates of focal trees. Summed annular tree basal areas are estimated as summed area at "breast height," the sum of the trees’ cross-sectional areas at a height of 1.30 m. The areas of multi-stemmed trees are summed. Basal areas and focal trees are each divided into five size quantiles. In the data pre- sented in this paper, only the annular tree growth effects on the smallest size quintile of focal trees are examined, though as reported earlier there is a smaller but often significant negative effect of annular trees’ summed basal area on the growth of larger focal trees [18]. In these analyses, as in the analyses of recruitment, mortality, and annular tree clustering, each annulus is examined separately. Thus, an analysis of fifth-annulus annular tree effects on focal tree growth begins by examining all focal trees of the smallest diameter quintile. These focal trees either have, or do not have, annular trees in their fifth annulus that fall within a given quantile of Ma values to their last common ancestor (LCA) with the focal tree. Focal trees that have no such annular trees in their fifth annulus, regardless of whether or not they have such trees in their other annuli, form the comparison group. The focal trees that have such annular trees in their fifth annulus are divided into the five quintiles of summed annular tree basal areas and their growth rates are compared to those of the controls. The null model used for comparison is generated by repeated randomization, within spe- cies, of the growth rates of the focal trees. Annulus number and the number of phylogenetic distance quantiles used in each FDP analysis are adjusted to ensure that within each of the concentric annuli there will be a substan- tial number of such control annuli. Focal-annular phylogenetic distances We estimated DNA-based divergences times between focal and annular species (last com- mon ancestor (LCA) in mega-anna (Ma)) using the S.PhyloMaker program written by YJ (available at https://github.com/jinyizju/S.PhyloMaker). Table 3 presents the proportion of species in each FDP that are present in S.PhyloMaker’s Phytophylo DNA dataset. These proportions vary from 100% at Wind River to 14% at Sinharaja. The majority of phyloge- netic distances between species must therefore be estimated at the genus rather than the species level in most of the FDPs. In this study the estimation was made by using Scenario 2 of S.PhyloMaker. For the species for which only genus-level information is known, this scenario picks uniformly-distributed distances from the interval from the present back to its genus’ LCA. There is unavoidably some noise in the phylogenetic distances, especially at FDPs such as Lambir and Pasoh where few species have been characterized genetically (Table 3). Fur- ther, and unavoidably, we are forced to add more noise because we divide the pairwise dis- tances to the LCA into Ma interval quantiles that vary in number according to the amount of information in the FDP. This division has the advantage that it equalizes the amount of information in each quantile, but the disadvantage that it can add further noise to focal and annular species that are separated by pairwise distances that fall in a sparsely population region of the range of pairwise distances at the FDP. The noise problem can be overcome to some extent because we repeatedly sample the pairwise distances during the analysis. We have concluded that the advantage of having equal statistical power in each of the pairwise distance quantiles outweighs the disadvantage of introduced noise. This is because, when we have exhaustively analyzed FDPs more than once using this methodology, the results are essentially indistinguishable. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 10 / 33 PLOS COMPUTATIONAL BIOLOGYTable 3. Numbers and fraction of species in each of the FDPs in this study that are found in Phytophylo. Each of 16 forests shows a unique pattern of between-tree interactions Species Found in Phytophylo Species Not Found in Phytophylo Fraction of Total Species found in Phytophyo FDP BCI Changbaishan Fushan Gutianshan Heishiding Korup Lambir Luquillo Mudumalai Nonggang Palanan Pasoh Sinharaja Tiantong Windriver Wytham https://doi.org/10.1371/journal.pcbi.1008853.t003 266 26 45 117 132 117 209 130 33 82 49 190 33 105 17 19 54 26 66 42 166 390 1127 32 50 135 267 708 206 51 0 5 0.8313 0.5 0.4054 0.7358 0.443 0.2308 0.1564 0.8025 0.3976 0.3779 0.1551 0.2116 0.1381 0.6731 1 0.7917 Null models for clustering, recruitment and mortality In order to isolate the influence of focal tree diameter on annular tree properties, the observed distribution of clustering, recruitment or mortality of annular trees around differ- ent diameter classes of focal trees is compared with 1,000 iterations in which focal tree diameters are randomized within species within census intervals in an FDP which is other- wise identical to the original FDP. Thus, these null models leave all other properties of the FDPs intact, including the positions and species identifications of all of the annular trees and the distributions of sizes of each species of focal tree. The only real-data components of clustering, recruitment and mortality that are measured are in the form of z-values of differ- ences between the real data and the means of the Monte Carlo randomizations of focal tree diameters within species within census periods. This avoids the introduction of possible unknown variables, which is a problem with the regression analyses that are commonly used to search for density-dependent effects on recruitment and mortality [21]. Regression analyses search for differences in the properties of trees that surround trees that recruit or those that die, compared with those surrounding survivors, but the resulting regressions may have many sources traceable to the distributions of tree positions and tree properties that will vary across species. Null models and the influence of storage effects There is a built-in “delay” in NDD-influenced factors that affect recruitment, mortality and clustering. This delay results from spatial and temporal storage effects [4, 6], allowing relatively dense clusters of trees of the same or phylogenetically related species to become established in regions where physical resources are initially plentiful and where species-specific pathogens, browsers, and seed-predators are initially few. As the trees in the clusters grow older and larger, their species-specific pathogens begin to accumulate, browsers and seed-predators become increasingly attracted to the area, and species-specific physical resources become lim- iting [33]. As a consequence, the trees in the clusters thin out over time, so that new clusters of saplings of the same species as those in the clusters can only be established elsewhere. By PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 11 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions combining the original Janzen-Connell theory with spatial and temporal storage theory, it has been possible to explain the apparently contradictions between the Janzen-Connell model and the fact that many species of tree in forests are clustered rather than overdispersed [4, 6, 11]. Thus, null models that randomize only focal tree size and leave annular tree clustering or over- dispersal intact are central to the EAA analyses, because these storage effects are the same in both the real and the randomized data. NDD- influenced focal-annular interactions Many focal-annular interactions can be examined by the EAA method. We chose the interac- tions, listed below, that were shown in our previous study [18] to have an NDD component that remains significant across a wide range of focal-annular phylogenetic distances. The effect of focal tree size on annular tree recruitment. NDD predicts that recruitment of annular trees at any phylogenetic distance from the focal tree should tend to be highest around small focal trees and diminish as the focal trees increase in size. Conditions favoring annular recruitment result from the accumulation of NDD effects of pathogens and parasites shared between focal and annular trees [8, 9], and from the depletion of shared physical and biologically-generated resources (niche-complementarity) [34, 35]. As focal trees grow, as shared pathogens accumulate, and as shared resources become scarcer, annular recruitment should diminish [3, 8, 9, 19]. Seedling data are not available for these data sets, and we there- fore use the fraction of annular trees that have achieved a diameter of 1 cm during a census period as a proxy for recruitment rates. The effect of focal tree size on annular tree mortality. The same NDD processes should result in low mortality among the annular trees that surround small focal trees and high mor- tality among annular trees that surround large focal trees [8, 9, 36]. The effect of focal tree size on annular tree clustering. The combination of NDD effects influencing recruitment and mortality should over time result in high summed basal area of annular trees around small focal trees and lower summed basal area of annular trees around large focal trees, again across a wide range of focal-annular phylogenetic separations. The effect of annular tree basal area on focal tree growth. If a tree’s growth rate is slo- wed by the effects of competition for physical resources with nearby conspecific or phylogenet- ically related annular trees, or through the sharing of pathogens and predators with these annular trees, there may be a negative impact on its fitness [37]. Trees growing in regions that have few related trees nearby exhibit a fitness advantage over those growing in regions where there are many related trees nearby [24]. As noted above, and as in the original EAA analyses [18], focal tree growth rates are nor- malized within species, censuses, and focal tree size classes. Additional analyses Other analyses based on comparisons of the actual data with randomized null models may be applied to these data. For example, the focal trees that recruit or die during a census period can be examined to determine the fraction of their annular trees that are also recruits or trees that die (tests 5 and 6 of Table 2). The null models for these tests determine the same ratios after the positions of the focal trees that recruit, survive and die are repeatedly shuffled within spe- cies and the properties of the annular trees are left untouched. These tests, however, only examine how interactions between focal trees that recruit or die and their annular trees that recruit or die, which are small fractions of all the focal-annular interactions, differ from those of the remainder of the focal-annular interactions. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 12 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Interpretation of the graphical results In interpreting the graphs of tests for these classes of focal-annular interactions, recall that in EAA analyses the z-values for the annular tree properties express differences from null-model data sets that are randomized only with respect to focal tree size. The z-values from the four quan- tiles of focal tree sizes are therefore symmetrically distributed around the mean of all the z-values. Unlike the results for clustering, recruitment and mortality, the growth results are not sym- metrically distributed around zero. Instead, each line shows the difference in z-scores between (a) the normalized growth rates of small focal trees that have, in a specified annulus, a specified summed basal area of annular trees within a specified quantile of phylogenetic distances from the focal tree, and (b) the normalized growth rates of small focal trees with no such annular trees in the specified annulus. In general, throughout the FDPs, the larger the summed basal area of the specified subset of annular trees, and the closer to the focal tree they are in either physical or phylogenetic distance, the greater the expected negative effect of the annular trees on normalized focal tree growth. A test for a possible relationship between tree size and patterns of tree mortality in this study Mortality varies across life cycle, with NDD effects being most pronounced in the smallest trees [38]. NDD patterns of seedling mortality are primarily mediated by fungal pathogens [20, 39]. Mammals, foliar herbivores and foliar pathogens tend to contribute little to mortality at these early stages [40]. Phylogenetic distance is known to play an important role in the inci- dence of seedling mortality, which decreases as the phylogenetic distance between focal and surrounding trees increases [41–43]. The smallest trees for which mortality is measured in the present study have a dbh of 1 cm. We hypothesized that such small trees are more likely to die if they are near large focal trees, and less likely to die if they are near small focal trees. Larger trees that die might show a weaker association with focal tree size, because there might be a greater influence on the mortality of large trees of factors such as wind, fire and large herbivores that may have a small NDD com- ponent. To test this possibility, we divided annular trees of each species that died into two equal-sized groups, designated small and large. A control manipulation of the BCI data to check the sensitivity of the EAA analyses Fig 2 below shows that focal-annular phylogenetic curves undergo the expected changes when FDP data are deliberately manipulated. EAA analyses are therefore highly sensitive to small differences in the data sets. A pronounced reduction in significance in the BCI FDP data (marked with a circled numeral 1 at slightly over 100 Ma focal-annular phylogenetic distance in Fig 2 below) indicates that focal-annular species pairs separated by this phylogenetic distance are only interacting at low levels. This “valley” in significance values is seen clearly in focal-annular clustering, recruitment and the mortality of large annular trees. It is not apparent in mortality of small annular trees, or in focal tree growth. To investigate the validity of this signal, we determined the set of phylogenetic distances between pairs of species at this FDP that lie between 108 and 117 Ma, and selected the interac- tions involving the three commonest species that are separated by these distances for our sensi- tivity test. Our reasoning was that, because these species are common in the plot, they are likely to contribute disproportionately to between-species interactions. The three, in PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 13 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Fig 2. Test of the sensitivity of EAA analysis. The circled 1 and 2 mark unusual valleys and peaks respectively in the significance of focal-annular interactions. Asterisks mark the appearance of a new “valley” in significance levels after deliberate manipulation of the data (see text). Other legends as in Fig 1. https://doi.org/10.1371/journal.pcbi.1008853.g002 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 14 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions descending order of abundance, are Trichilia tuberculata (Meliaceae) (5.0% of the stems), Mouriri myrtilloides (Melastomataceae) (3.4%), and Tetragastris panamensis (Burseraceae) (2.5%). The focal-annular phylogenetic distances to their LCA that fall in this window are T. tuberculata with M. myrtilloides (111 Ma) and T. panamensis with M. myrtilloides (111 Ma), while T. tuberculata with T. panamensis (72 Ma) falls outside the window. We modified each of the focal-annular phylogenetic distances separating these three species to 50 Ma, and repeated the EAA analysis. The right-hand set of graphs in Fig 2 show the appearance of a new region of lowered significance (indicated with an asterisk) close to the 50 Ma mark for the clustering, recruitment and large-tree mortality graphs. The new lowered-sig- nificance signal does not appear precisely at 50 Ma, because these altered between-species dis- tances now join a large group of other distances that contribute to the phylogenetic distance quantile that includes the 50 Ma distance. The original “valley” in significance values remains, though somewhat reduced in size, demonstrating that additional species pairs contribute to it. The growth graph shows no new signal, however, suggesting that the causes of the reduction in between-species interactions do not influence focal tree growth. Selective removals of spe- cies or focal-annular data from the data sets, such as those which were carried out at Wind River that are presented below, are therefore capable of providing consistent and detailed information about the extent of individual between-species interactions. Results Each of the 16 forest diversity plots (FDPs) examined in this study exhibits smooth declines across increasing physical focal-annular distance (in m) in the significance of each of the four focal-annular interactions tested. There is one exception: recruitment at Mudumalai shows com- plexity across physical distance, possibly because of recent influence of elephants and/or fires (S1 Fig). Clustering at Mudumalai, which measures the results of longer-term processes, shares with the other FDPs the common pattern of a smooth decline with increasing physical distance. These interactions also tend to show an overall decline in significance with increasing phy- logenetic distances (in Ma) between the species, but there are many localized exceptions to this decline that result in unique patterns of “peaks” and “valleys” in the magnitude of significance along the phylogenetic distance axis for each FDP. Table 4 summarizes the expected (from Table 2) and the observed results from the four EAA tests employed in this paper. In general, the results agree with NDD expectation, but there are many localized departures from a smooth decrease in significance with phylogenetic distance. In addition, there is a puzzling weak PDD signal seen for small annular trees that die and that warrants further investigation. In Fig 3 we present the results, for each of the first four NDD-influenced focal-annular interactions that are listed in Table 2, at the BCI (Panama) FDP. The results are presented as two- and three-dimensional graphs generated from generalized additive model (GAM) analy- ses. The three-dimensional graphs in the figure present data from GAM analyses along both the physical (focal-annular distance in meters) and phylogenetic (Ma back to the last common ancestor (LCA)) axes, permitting a comparison of the effects of physical and phylogenetic dis- tance in a single graph. Note that in these 3D graphs the physical distance decline remains smooth across all concentric annuli, while the irregularities in the phylogenetic distance curve are preserved across all concentric annuli. The smooth and gradual decline in significance with increasing physical focal-annular distance is clearly distinguishable from the complex fluctuations in significance that are seen across the range of phylogenetic distances. Figs S1 and S2 show that these patterns are seen in 3D analyses across FDPs, with the exception of recruitment at Mudumalai that was noted earlier. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 15 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Table 4. Expected and observed results for the EAA tests, assuming NDD focal-annular interactions. Expected (from Table 2) Observed Results expected and observed in all tests: Focal-annular differences from null model should decline in significance as either physical or phylogenetic focal-annular distance increases Smooth decline in significance with increasing physical distance. A general decline as phylogenetic distance increases, but many FDPs show significant peaks and valleys Results expected and observed in each test, given NDD expectation: Test 1) Relationship between focal survivor sizes and their annular survivor summed basal area Negative relationship between focal tree size and annular survivor summed basal area Generally negative, as predicted, but there are some localized switches into PDD (positive) territory in some FDPs Test 2) Relationship between focal survivor sizes and their annular recruit fraction Negative relationship between focal tree size and annular recruit numbers Generally negative, as predicted. Peaks and valleys along the phylogenetic axis often match annular survivor curves Test 3) Relationship between focal survivor sizes and their annular mortality fraction Positive relationship between focal tree size and numbers of annular trees that die Positive, as predicted for large annular trees that die, but weakly negative for small annular trees that die Test 4) Relationship between a focal tree’s growth rate (normalized within species) and its annular tree basal area Negative relationship between focal tree growth rate and annular tree summed basal area Negative relationship, as predicted. Significance falls with decreasing annular summed basal area, with many significant peaks and valleys along the phylogenetic axis Test 5) Relationship between focal trees that do and that do not recruit. and their annular recruit fractions Higher fraction of annular recruits around focal recruits At BCI, a higher fraction, as predicted. Peaks and valleys along the phylogenetic axis do not match those for surviving focal trees Test 6) Relationship between focal trees that do and do not die and their annular mortality fractions Higher fraction of annular trees that die around focal trees that die At BCI, a higher fraction, as predicted, for both small and large focal trees that die https://doi.org/10.1371/journal.pcbi.1008853.t004 Fig 4 shows phylogenetic distance results from a preliminary application of Tests 5 and 6 to the data from BCI. The results are highly significant and in agreement with NDD expectation (Table 4, Tests 5 and 6). As with the other analyses, there is a smooth decline with increasing focal-annular physical distance (not shown), and a complex decline with increasing phylogenetic distance. The curves for mortality are in agreement with those shown for large annular tree mortality in Fig 3. How- ever, the phylogenetic significance decline for recruitment appears to show a different pattern of hills and valleys from the equivalent analysis in Fig 3, suggesting that focal-annular interactions of focal trees that recruit or die with annular trees that recruit or die may be different from those of focal survivors. In addition, the phylogenetic peaks and valleys for recruitment are less complex than those found in the Fig 3 analysis, possibly because in Test 5 data are being examined from a smaller fraction of the focal trees. As these and other tests are explored further, they will provide additional windows of opportunity for investigation of detailed focal-annular interactions. First-annulus two-dimensional graphs for Tests 1–4 in the remaining fifteen FDPs, which illustrate each FDP’s unique phylogenetic distance patterns, are shown in Figs 5, 6 and 7. NDD-influenced focal-annular interactions for clustering and recruitment are present in all of the FDPs, and NDD-influenced focal tree growth interactions with the summed basal areas of annular trees are also found in 15 of the FDPs. Three-dimensional GAM graphs for clustering, recruitment and focal growth in all of the FDPs are shown in S1and S2 Figs. These three-dimensional graphs show that, with the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 16 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Fig 3. A comparison of two- and three-dimensional graphs of the z-values for NDD-influenced focal-annular patterns at the BCI (Panama) FDP. The 2D graphs show first-annulus z-values of these patterns for all subdivisions of focal tree sizes and annular tree biomasses, compared to null models (Materials and Methods). The 3D graphs show the surfaces formed by the z-values across both physical and phylogenetic distances between focal and annular trees. The surfaces shown are for the largest focal trees or annular biomasses (red lines in the 2D graphs) and the smallest PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 17 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions focal trees or annular biomasses (blue lines in the 2D graphs). Circled numerals 1 and 2 mark unusually low or high z- values that are found at certain focal-annular phylogenetic distances. The k-values are the optimized number of smooth terms used in the GAM analyses (Materials and Methods). The 95% confidence limits for the expected 2D z- values of zero are shown as brown horizontal lines, and the 95% confidence intervals of the 2D lines are shown in gray. The regions of significant z-values on the 3D surfaces are heat-map colored as in Fig 1, and the non-significant regions are gray. https://doi.org/10.1371/journal.pcbi.1008853.g003 exception of recruitment at Mudumalai, the significance of NDD-related effects declines smoothly rather than irregularly with increasing physical distance, while the patterns of peaks and valleys along the phylogenetic axis that are unique to each FDP are preserved across the range of physical focal-annular distances. These idiosyncratic phylogenetic distance patterns show that the species in each of the FDPs have in the past been shaped by distinct evolutionary trajectories that have led to this wide variety of patterns of species-species interactions. Some of the exceptions to a smooth decline in significance with increasing focal-annular phylo- genetic distance are observed at the same phylogenetic distances of an FDP’s recruitment, cluster- ing, and/or focal tree growth analyses. In Figs 3 and 5–7, which present first-annulus 2D graphs for all sixteen FDPs, the most pronounced of these exceptions are marked with circled numbers 1 or 2 to mark significant local reductions or increases in significance respectively. Fig 6 shows a par- ticularly striking example in the subtropical Luquillo FDP. At this FDP a valley and a peak, cen- tered at 100 and 150 Ma back to the LCA respectively, are seen in clustering, recruitment and growth. If such exceptions are found at the same phylogenetic distance in more than one type of focal-annular interaction, they may have underlying causes in common (see the detailed analysis of the Wind River patterns below). Note that these resemblances are not the result of correlations in the numbers of recruits and survivors in the annuli, because these correlations are preserved in the null-model data to which the real data are compared (Materials and Methods). Mortality results show a more complex and difficult-to-interpret pattern Thirteen of the sixteen FDPs show significant deviations from the null model in the pattern of small annular tree mortality. The pattern, however, is the opposite of that seen in previous studies of seedling mortality [20, 39]. More small annular trees than expected die around small focal trees, and fewer die than expected around large focal trees. This pattern, which is consis- tent with positive rather than negative density dependence, is largely confined to conspecific annular trees, except at the Nonggang, Tiantong and Wind River FDPs at which the PDD effect extends to some heterospecifics. This test is different from the BCI analysis of Fig 4 (d), which shows that more annular trees than expected die near the focal trees that die regardless of their size. The Fig 4 (d) analyses are consistent with NDD effects, in which mortality—espe- cially mortality among closely-related trees—is expected to be spatially clustered. In Figs 3, 5, 6 and 7 the mortality pattern exhibited by large annular trees that die is the reverse of that seen in the small annular trees, and like the Fig 4 (c) analysis is consistent with NDD effects. In fifteen of the FDPs, with the exception of the dry tropical forest plot at Mudu- malai, there is a deficiency of mortality in large annular trees around small focal trees and an excess around large focal trees, as NDD would predict. We explore some possible reasons for the different outcomes of these different mortality tests in the Discussion. Known between-species interactions account for some of the variation in significance levels along the phylogenetic distance axis at the Wind River FDP At the Wind River (Washington State, USA) FDP, which has only 26 tree species, EAA analysis reveals an unusually large exception to declines of NDD effects as focal-annular phylogenetic PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 18 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Fig 4. Tests at the BCI FDP of the proportions of recruits or trees that die in Annulus 1 around focal recruits or focal trees that die, compared with the proportions expected from a null model in which the properties (recruits, survivors, died) of the focal trees are shuffled repeatedly within species. https://doi.org/10.1371/journal.pcbi.1008853.g004 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 19 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Fig 5. This figure and the following two figures show two-dimensional GAM analyses of NDD-influenced focal-annular interactions for the remaining fifteen FDPs in the study (excluding BCI, which is presented in Fig 3). This figure shows results from the lowest-latitude FDPs. Legends as in Figs 1 and 3. The k-values are the optimized number of smooth terms. Circled numbers 1 and 2 represent local reductions or increases respectively in the significance of the effects along the phylogenetic distance axis. https://doi.org/10.1371/journal.pcbi.1008853.g005 distance increases. At the largest focal-annular phylogenetic distances in this FDP, there are large increases in the significance of deviations of annular clustering and recruitment, and a large negative effect on focal tree growth rate. First-annulus data are used in the analysis PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 20 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Fig 6. Continuation of the first-annulus two-dimensional phylogenetic distance analyses of all the FDPs, showing FDPs at intermediate latitudes. Legends as in Figs 1 and 3. https://doi.org/10.1371/journal.pcbi.1008853.g006 presented in Fig 7, but as with the other FDPs these unique patterns of phylogenetic peaks and valleys at Wind River are preserved across more distant annuli. A likely factor contributing to the unusual EAA pattern found in this FDP may be allelo- pathic inhibition. Large-diameter trees of the western hemlock Tsuga heterophylla have dense canopies and have been shown to have allelopathic needles [23]. These properties reduce the PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 21 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Fig 7. The highest-latitude set of the first-annulus two-dimensional phylogenetic distance analyses of all the FDPs. Legends as in Figs 1 and 3. https://doi.org/10.1371/journal.pcbi.1008853.g007 growth rate of trees of other species when T. heterophylla are nearby, and also reduce cluster- ing and recruitment of trees of a range of species around large T. heterophylla [23, 43]. The effects of removing different combinations of species from the Wind River EAA analy- sis are shown in Fig 8. Trees of the commonest species, the vine maple Acer circinatum, and those of the second most common species, Tsuga heterophylla, are close to each other in num- bers. Together, these two species make up two-thirds of the stems, and those of all the other PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 22 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions species make up the remainder. We therefore divided the species into three categories: A. circi- natum, T. heterophylla, and a third category made up of all the other species. We examined the effects of removing different combinations of focal or annular trees, or both, that fall into these three categories. The row of graphs (1) at the top of Fig 8 show (A) annular clustering, (B) annular recruit- ment, and (C) focal tree growth GAM-generated curves for all the Wind River first-annulus data. All three of these focal-annular interactions show increases in significance among species separated by 150 Ma or more from their LCA, in the direction expected from NDD models. This is in striking contrast to the patterns of declining significance with increasing focal-annu- lar phylogenetic distance that would be predicted if the strength of focal-annular interactions decreases with increasing phylogenetic distance between them. For focal growth, however, a group of species pairs that are separated by a little more than 100 Ma shows strong positive effects of annular trees on the growth of focal trees, consistent with PDD. The next row of graphs (2) show results from only focal T. heterophylla and all annular spe- cies except for A. circinatum. The anomalous results at the most distant phylogenetic intervals are retained for clustering and recruitment, but disappear for focal tree growth (as does the equally anomalous PDD peak at more intermediate distances). This is the pattern to be expected if focal T. heterophylla are suppressing recruitment of distantly-related annular species (pre- dominantly Angiosperms), but if the presence of these distantly-related annular species is not affecting the growth of the focal T. heterophylla in either a positive or a negative direction. Graphs in row (3) show the results for only focal A. circinatum and all annular species except for T. heterophylla. These graphs present only the interactions between focal trees of the commonest species in the plot and annular species that are not known to have allelopathic effects. These interactions show a general decline of NDD-influenced effects with increasing phylogenetic distance across all phylogenetic distances. Thus, at Wind River, such a pattern— seen, though with many localized exceptions, at most of the FDPs—is revealed when only the interactions between focal trees of the commonest species A. circinatum and all annular species except for T. heterophylla are analyzed. When the entire data set is examined, however, this pattern is masked because of the strong allelopathic effects of T. heterophylla. Row (4) conditions are the same as in Row (3), except that T. heterophylla has been added back to the annular trees. Clustering and recruitment patterns around focal A circinatum are little changed from Row (3), showing a lack of interactions between annular T. heterophylla and focal A. circinatum. There may be a small negative effect of T. heterophylla on the growth of focal A. circinatum, but it is at the margin of significance. The graphs in Row (5) show results from the dataset consisting of all focal species except for T. heterophylla and all annular species except for A. circinatum. These results are the mirror image of Row (2). Clustering and recruitment of annular species are not influenced by allelop- athy from focal trees other than T. heterophylla, and therefore do not show enhanced effects at extreme phylogenetic distances. But growth of distantly related focal species is slowed, as expected, by the allelopathic effects of annular T. heterophylla. And the striking peak in growth of intermediate-distance focal trees is again apparent, strongly indicating that annular T. het- erophylla may be responsible for this effect as well. Taken together, these results show that T. heterophylla influences at least some distantly related species negatively, and may influence trees at intermediate phylogenetic distances posi- tively, but it does not influence the commonest species A. circinatum. Removal of other combi- nations of focal and annular species yields results that are consistent with this interpretation (not shown). Previously-published measurements of the allelopathic effects of T. heterophylla had not detected A. circinatum’s immunity to T. heterophylla’s effects [23, 43]. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 23 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Fig 8. EAA two-dimensional first-annulus analysis of allelopathic effects of Tsuga heterophylla on other species in the Wind River FDP. Three-dimensional analyses (not shown) show smooth declines in significance with increasing physical distance but preservation of localized phylogenetic distance features across annuli. See text for interpretation. Legends and confidence intervals as in Fig 1. https://doi.org/10.1371/journal.pcbi.1008853.g008 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 24 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions The positive effects on focal growth seen in rows (1) and (4) of Fig 8 suggest that T. hetero- phylla has positive density-dependent effects on the growth of focal trees separated from it by intermediate phylogenetic distances. This interesting result warrants further investigation. Discussion The FDPs examined here span a wide range of environments, from tropical to northern tem- perate. Across this diversity of ecosystems, the EAA results reinforce the growing body of evi- dence that NDD interactions between even distantly related species are common [44]. Here we provide a brief summary of conclusions that can be drawn and questions that can be raised from EAA analyses. Peaks and valleys in the significance of focal-annular species interactions along the phylogenetic axes for each FDP may result from unusual rates of evolutionary divergence and convergence between pairs or groups of species For example, a valley in significance of annular tree effects on focal tree growth might result from the convergent evolution of distantly-related species on the ability to obtain resources from the environment [45]. A peak in significance in the NDD component of recruitment of distantly-related annular trees around focal trees might result from convergent evolution that has led to susceptibilities to similar pathogens [41] or browsers [46], while a valley in these sig- nificance levels between distantly-related focal and annular trees might result from divergent evolution in these susceptibilities. Evolution of allelopathy and of other offensive or defensive mechanisms in particular species, as in the Wind River FDP (Fig 8), must also play an impor- tant role in generating peaks and valleys. We emphasize that the evolutionary changes leading to the peaks and valleys that we observe are not likely to have originated in the FDPs being examined, but must have had a much longer history of natural selection that took place in the various ecosystems that were inhabited by these species’ ancestors. Regardless of these anomalies’ precise origins, EAA can detect the between-species interactions that are most significant and that are therefore most likely with further study to yield useful information about the species’ evolutionary histories. Mortality interactions that have an NDD component may be confounded with effects that do not have an NDD component Mortality of seedlings is known to have a strong NDD component [41, 42]. On the assumption that factors unrelated to NDD effects might play an important role in the mortality of large trees (such as large herbivores, strangler figs, windstorms, etc.), we divided the focal trees of each species that died in each FDP into two equally numerous groups (small and large) accord- ing to their diameters. When Test 3 of Table 2 was applied, the mortality pattern differences in these two groups were striking and unexpected. For small annular trees that die, most of the FDPs show patterns consistent with positive density-dependence. There is low mortality around the largest focal trees and high mortality around the smallest focal trees. The FDPs Changbaishan and Palanan show no significant effects. In contrast, when the larger annular trees that die are examined, a positive density- dependent pattern is only seen at Changbaishan. Instead, a pattern that is consistent with NDD effects is seen at BCI, Fushan, Gutianshan, Heishiding, Korup, Lambir, Nonggang, Sin- haraja, Tiantong, Wytham Wood and possibly Wind River. Examination of the causes of small-tree mortality may reveal the source or sources of the conspecific PDD effects. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 25 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Storm-damaged FDPs show a variety of EAA patterns The FDPs Palanan, Fushan and Luquillo are repeatedly damaged by cyclonic wind storms. These three FDPs have recently been compared with each other and with BCI, and were shown to have significant differences in mortality, growth and recruitment [47]. The present study found that all four FDPs also show differences in NDD patterns of mortality, growth and recruitment (Fig 6). Palanan is a species-rich tropical forest in the Philippines that has been battered by three supertyphoons and numerous other storms during the 18 years that it has been censused. (It was hit with another supertyphoon in 2018, with effects yet to be measured.) This FDP shows weak focal-annular interactions that do not decline with increasing phylogenetic distance. Although Luquillo is at the same latitude as Palanan and was damaged by two severe hurri- canes during the period covered by this study, it shows a different set of focal-annular interac- tions. Luquillo, like Palanan, exhibits weak clustering and recruitment NDD effects that persist across most phylogenetic distances, but unlike at Palanan these effects diminish markedly at the largest distances. Fushan, at a higher latitude, is a similarly storm-battered submontane FDP. It shows an interaction pattern more typical of the majority of FDPs, with strong interactions among con- specifics and an overall decline (with some dramatic exceptions) with increasing focal-annular phylogenetic distance. Palanan and Luquillo have lower tree densities than Fushan, even though Fushan lies fur- ther north. In part this is the result of the more severe effects of the storms at Palanan and Luquillo, which often rip away the tops of canopy trees. These events open up the areas around large trees to higher levels of successful recruitment of all species [48, 49]. Enhanced recruit- ment may help explain why NDD-associated recruitment and clustering patterns are weakly significant and often change little over most phylogenetic distance in these FDPs. But at Luquillo, the reduction in the significance of clustering and recruitment at extreme phyloge- netic distances is more consistent with the majority of FDPs. The three-dimensional GAM graphs show the persistence of peaks and valleys in species-species interactions across a range of physical distances The unique shapes of the three-dimensional surfaces along their phylogenetic distance axis are retained across annuli at each FDP (Fig 3 and S1 and S2 Figs). They are retained even in distant annuli in which the means of the z-values that the surfaces represent are not themselves signifi- cantly different from zero (gray regions of the 3D surfaces). The surfaces can retain their phy- logenetic-distance shapes, even in the gray regions, because the 95% confidence limits on the surfaces are small. As noted above, the rotatable three-dimensional GAM graphs from Fushan, Lambir and Pasoh that are presented in S3–S11 Figs allow the viewer to assess the relationships between the surfaces and their confidence intervals. In each case these intervals are small com- pared to the confidence intervals of the z-values themselves. Thus, the GAM analyses of the EAA data are able to detect the details of focal-annular interactions, even if the z-values them- selves may be below the level of significance. Species with different properties do not on average contribute disproportionately to the results In the first EAA paper [18] it was shown for the BCI FDP that between-species NDD interac- tions are of approximately equal strength across a variety of groupings of species into subsets that have different phenotypic and ecological properties. In that analysis it was decided not to PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 26 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions divide species into phenotypic classes that would have led to unequal-sized subdivisions, because this would have made the tests for NDD-influenced patterns less directly comparable to each other. To preserve the statistical power of the tests, and to make them comparable to each other, the BCI species were subdivided using two different criteria. The first criterion sorted the species according to their abundance, and the second sorted them according to the CV of their diameters across all the stems of the species. After ranking of the species according to each of these criteria, the pooled trees of the ranked species were then divided into thirds, each third consisting of equal numbers of individual trees. EAA analyses of each of these six subdivisions showed that all six exhibited the same EAA NDD patterns as the BCI FDP as a whole, though as expected the subdivision results were less significant than they were in the total data. Thus, in this FDP, EAA analysis of species with a wide range of properties yields similar results. Detailed examinations of NDD interactions between conspecifics of individual species of differ- ent abundances at BCI, however, have suggested substantial differences in the strength of NDD- influenced patterns, to the point that the least-influenced species may be in danger of local extinc- tion [50]. EAA can be used to investigate the effects of the removal of information about individual species that have similar abundances in an FDP, in order to test this observation further. EAA analyses can be used to investigate ecological-evolutionary processes in detail It is possible to use EAA to examine interactions between focal-annular species pairs that are based, not only on the phylogenetic distance between them, but also on other quantifiable characteristics: differences in the species’ physical and biochemical phenotypes, in their defen- sive and allelopathic mechanisms, in their shared interactions with different classes of patho- gens, herbivores and parasites, and in their associations with the plots’ topographies and soil types. The only requirement for such an analysis is the ability to arrange focal-annular species pairs on a scale of numerical values for the character, from least divergent to most divergent. The EAA approach can then be modified to "sieve out" focal-annular species combinations that show the greatest discordance between phylogenetic distance and such scalable pheno- typic and environmental characteristics. A large focal-annular phylogenetic distance, coupled with a small focal-annular distance in a simultaneously measured phenotypic or niche-related character, would suggest evolutionary convergence in the character being measured, while the reverse situation would suggest an unusually high rate of evolutionary divergence. This approach will permit the isolation of the characteristics that are most likely to be associated with cases of unusual focal-annular divergence or convergence in FDPs and in similar complex ecosystems. The nature of these interactions can be explored further by experimental manipu- lations in the field or in greenhouse experiments. Most of the thousands of species in these plots are rare. Their aggregate contributions to NDD effects may be substantial, but most of these contributions are unlikely to be detectable at the species level. EAA provides us with a tool for finding the species that are likely to repay further study, while still applying the rigorous EAA standard of examining the effects of only one variable at a time. The EAA approach may also be sensitive enough to detect small changes over time in the focal-annular dynamics of multiply-censused FDPs such as BCI and Pasoh that may be corre- lated with climatic change. EAA and Darwin’s hypothesis EAA analyses can be used to test an important ecological-evolutionary prediction that follows from Darwin’s observations and from his speculations that were quoted at the beginning of PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 27 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions this paper. When species have lived in close physical proximity to each other within an ecologi- cal community for some time, some of these species’ evolutionary changes should have been driven by the direct and indirect interactions among them. Even if particular pairs of tree spe- cies are not physically close in the ecosystem, they will share symbionts, pollinators, pathogens, parasites, predators, and herbivores that often have high levels of dispersal. These shared bio- logical agents must also have undergone evolutionary change as a result of their interactions with their hosts. It is likely that the variety of evolutionary trajectories in the NDD-influenced focal-annular effects that are seen in the FDPs of this study stems in part from such biotic interactions. EAA can detect the pairs of tree species in an FDP that are likely candidates for detailed studies of these interactions, which will in turn help towards eventual clarification of their true causes and their evolutionary histories. At the same time, such extended studies will reveal the proportion of these interactions that are responses to physical factors in the species’ environment, and provide firm evidence for or against Darwin’s hypothesis that biological fac- tors play a large role in ecosystem evolution. EAA and ecosystem preservation In order to preserve threatened ecosystems, we must understand the mechanisms that main- tain their diversity. EAA can be used to flag between-species interactions that are unusually strong or weak and are therefore likely to yield significant results when subjected to experi- mental manipulation. Each such case that is understood in depth will increase our understand- ing of the kinds of interactions that must be preserved in order to maintain the overall structure of both intact and endangered ecosystems. General ecological models that ignore these interactions are of little help in understanding which aspects of ecosystems are important in their long-term preservation. Supporting information S1 Fig. Three-dimensional graphs showing the differences between the physical (m) and phylogenetic (Ma) axes for z-values that measure the significance of the NDD-influenced component or annular tree clustering and recruitment for the sixteen FDPs. The graphs show the patterns seen around the largest quantile of focal tree sizes. Legends and surface col- ors as in Figs 1 and 3. Note that levels of significance decrease smoothly with increasing focal- annular physical distance and irregularly with increasing focal-annular phylogenetic distance at each FDP. In addition, the shapes of the phylogenetic distance curves for clustering and recruitment often resemble each other, for reasons discussed in the Results section of the main paper. (TIF) S2 Fig. Three-dimensional graphs showing the effect of the largest quintile of annular trees on the growth of focal trees for all 16 FDPs. Legend as in Figs 1 and 3. (TIFF) S3 Fig. bci_focal_growth_largest_annular_biomass_type_factor.html. An interactive rotat- able graph of a three-dimensional GAM analysis of focal growth data from the BCI FDP. The graph allows the viewer to examine the surfaces formed by the data from all vantages, in order to visualize the differences between the physical and the phylogenetic distance axes. The areas of the surface on the graph that lie in regions of significant z-values are shown in color. The areas that lie in regions of non-significant z-values are in gray. Because of the size of the graph, it is possible to show clearly the magnitude of the 95% confidence intervals of the surface itself, which tend to be small. The surface with its confidence interval also shows that irregular PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 28 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions features along the surface’s phylogenetic distance axis retain their shapes and their significance even in parts of the surface that lie within regions where the individual z-values themselves are non-significant. (HTML) S4 Fig. bci_annular_clust_largest_focal_trees_type_factor.html. An interactive rotatable graph of a three-dimensional GAM analysis of clustering data from the BCI FDP. Legend as in S3 Fig. (HTML) S5 Fig. bci_annular_born_largest_focal_trees_type_factor.html. An interactive rotatable graph of a three-dimensional GAM analysis of recruitment data from the BCI FDP. Legend as in S3 Fig. (HTML) S6 Fig. fushan_focal_growth_largest_annular_biomass_type_factor.html. An interactive rotatable graph of a three-dimensional GAM analysis of focal growth data from the Fushan FDP. Legend as in S3 Fig. (HTML) S7 Fig. fushan_annular_clust_largest_focal_trees_type_factor.html. An interactive rotat- able graph of a three-dimensional GAM analysis of clustering data from the Fushan FDP. Leg- end as in S3 Fig. (HTML) S8 Fig. fushan_annular_born_largest_focal_trees_type_factor.html. An interactive rotat- able graph of a three-dimensional GAM analysis of recruitment data from the Fushan FDP. Legend as in S3 Fig. (HTML) S9 Fig. luquillo_focal_growth_largest_annular_biomass_type_factor.html. An interactive rotatable graph of a three-dimensional GAM analysis of focal growth data from the Luquillo FDP. Legend as in S3 Fig. (HTML) S10 Fig. luquillo_annular_clust_largest_focal_trees_type_factor.html. An interactive rotat- able graph of a three-dimensional GAM analysis of clustering data from the Luquillo FDP. Legend as in S3 Fig. (HTML) S11 Fig. Luquillo_annular_born_largest_focal_trees_type_factor.html. An interactive rotatable graph of a three-dimensional GAM analysis of recruitment data from the Luquillo FDP. Legend as in S3 Fig. (HTML) S1 Data. Data for FDP’s Mudumalai, Wytham Woods, Heishiding, Nonggang and Pala- nan. (ZIP) S2 Data. Data for remainder of FDP’s in the study. These zipped files contain all the data used to plot the figure graphs, including information needed to generate the errors on the graphs, in .csv format. (ZIP) PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 29 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Acknowledgments We are most grateful for valuable advice on the paper provided by Scott Rifkin, Joshua Kohn, Bret Elderd, Jim Dalling and Will Pearse. We also gratefully acknowledge help in data analysis from Mahidhar Tatineni, Jerry Greenberg, Ron Hawkins and Paul Rodriguez of the San Diego Supercomputer Center, and assistance from Sandra L. Yap and Edwino S. Fernando. Dedication Dedicated to the memories of Abdul Rahman bin Kassim (1963–2018) (Pasoh FDP) and of Perry S. Ong (1960–2019) (Palanan FDP), who made many valued contributions to these and to many other forest plot studies. Author Contributions Conceptualization: Christopher Wills, Shuai Fang, Kyle E. Harms. Data curation: Shuai Fang, Yunquan Wang, James Lutz, Jill Thompson, Sandeep Pulla, Boni- facio Pasion, Sara Germain, Heming Liu, Joseph Smokey, Sheng-Hsin Su, Nathalie Butt, Chengjin Chu, George Chuyong, Chia-Hao Chang-Yang, H. S. Dattaraja, Stuart Davies, Sisira Ediriweera, Shameema Esufali, Christine Dawn Fletcher, Nimal Gunatilleke, Savi Gunatilleke, Chang-Fu Hsieh, Fangliang He, Stephen Hubbell, Zhanqing Hao, Akira Itoh, David Kenfack, Buhang Li, Xiankun Li, Keping Ma, Michael Morecroft, Xiangcheng Mi, Yadvinder Malhi, Perry Ong, Lillian Jennifer Rodriguez, H. S. Suresh, I Fang Sun, Raman Sukumar, Sylvester Tan, Duncan Thomas, Maria Uriarte, Xihua Wang, Xugao Wang, T. L. Yao, Jess Zimmermann. Formal analysis: Christopher Wills, Bin Wang, Yunquan Wang, Yi Jin, James Lutz, Jill Thompson, Kyle E. Harms, Sandeep Pulla, Bonifacio Pasion, Sara Germain, Heming Liu, Joseph Smokey, Nathalie Butt, Chengjin Chu, Chia-Hao Chang-Yang, Sisira Ediriweera, Shameema Esufali, Christine Dawn Fletcher, Nimal Gunatilleke, Savi Gunatilleke, Chang- Fu Hsieh, Fangliang He, Zhanqing Hao, Akira Itoh, Buhang Li, Xiankun Li, Keping Ma, Xiangcheng Mi, Raman Sukumar, Sylvester Tan. Funding acquisition: Heming Liu, Sheng-Hsin Su, George Chuyong, Chia-Hao Chang-Yang, Stuart Davies, Shameema Esufali, Christine Dawn Fletcher, Nimal Gunatilleke, Savi Guna- tilleke, Chang-Fu Hsieh, Fangliang He, Stephen Hubbell, Zhanqing Hao, Akira Itoh, David Kenfack, Buhang Li, Yadvinder Malhi, Perry Ong, Lillian Jennifer Rodriguez, H. S. Suresh, I Fang Sun, Raman Sukumar, Duncan Thomas, Xihua Wang, Xugao Wang, T. L. Yao, Jess Zimmermann. Investigation: Bin Wang, Shuai Fang, James Lutz, Joseph Smokey, Chia-Hao Chang-Yang, H. S. Dattaraja, Savi Gunatilleke, Stephen Hubbell, H. S. Suresh. Methodology: Shuai Fang, Yunquan Wang, Yi Jin, Sandeep Pulla, Maria Uriarte. Project administration: Yunquan Wang, Nathalie Butt, Chengjin Chu, George Chuyong, H. S. Dattaraja, Stuart Davies, Sisira Ediriweera, Shameema Esufali, Christine Dawn Fletcher, Nimal Gunatilleke, Savi Gunatilleke, Chang-Fu Hsieh, Fangliang He, Stephen Hubbell, Zhanqing Hao, Akira Itoh, David Kenfack, Buhang Li, Xiankun Li, Keping Ma, Michael Morecroft, Xiangcheng Mi, Yadvinder Malhi, Perry Ong, Lillian Jennifer Rodriguez, H. S. Suresh, I Fang Sun, Raman Sukumar, Sylvester Tan, Duncan Thomas, Maria Uriarte, Xihua Wang, Xugao Wang, T. L. Yao, Jess Zimmermann. Resources: Shuai Fang, Sheng-Hsin Su, Nathalie Butt, Chengjin Chu, Chia-Hao Chang-Yang, H. S. Dattaraja, Stuart Davies, Sisira Ediriweera, Keping Ma, Michael Morecroft. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 30 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions Software: Christopher Wills, Bin Wang, Yi Jin. Supervision: Christopher Wills, Bonifacio Pasion. Validation: Jill Thompson, Heming Liu. Visualization: Bin Wang. Writing – original draft: Christopher Wills. Writing – review & editing: Christopher Wills, Bin Wang, Shuai Fang, James Lutz, Jill Thompson, Kyle E. Harms, Sandeep Pulla, Bonifacio Pasion, Sara Germain. References 1. Cavender-Bares J, Kozak KH, Fine PVA, Kembel SW. The merging of community ecology and phyloge- netic biology. 2009; 12(7):693–715. https://doi.org/10.1111/j.1461-0248.2009.01314.x PMID: 19473217 2. Broekman MJE, Muller-Landau HC, Visser MD, Jongejans E, Wright SJ, de Kroon H. Signs of stabilisa- tion and stable coexistence. Ecol Lett. 2019; 22(11):1957–75. https://doi.org/10.1111/ele.13349 PubMed PMID: WOS:000477122900001. PMID: 31328414 3. Bagchi R, Henrys PA, Brown PE, Burslem D, Diggle PJ, Gunatilleke CVS, et al. Spatial patterns reveal negative density dependence and habitat associations in tropical trees. Ecology. 2011; 92(9):1723–9. https://doi.org/10.1890/11-0335.1 PubMed PMID: WOS:000294152800003. PMID: 21939068 4. Chesson P. General theory of competitive coexistence in spatially-varying environments. Theoretical Population Biology. 2000; 58(3):211–37. https://doi.org/10.1006/tpbi.2000.1486 PubMed PMID: WOS:000166247900004. PMID: 11120650 5. Marden JH, Mangan SA, Peterson MP, Wafula E, Fescemyer HW, Der JP, et al. Ecological genomics of tropical trees: how local population size and allelic diversity of resistance genes relate to immune responses, cosusceptibility to pathogens, and negative density dependence. Mol Ecol. 2017; 26 (9):2498–513. https://doi.org/10.1111/mec.13999 PubMed PMID: WOS:000400335200008. PMID: 28042895 6. Chesson P. Updates on mechanisms of maintenance of species diversity. Journal of Ecology. 2018; 106(5):1773–94. https://doi.org/10.1111/1365-2745.13035 PubMed PMID: WOS:000452178600002. 7. Gillett JB. Pest pressure, an underestimated factor in evolution. In: Nichols D, editor. Taxonomy and Geography, a Symposium. London: London Systematics Association; 1962. p. 37–46. PMID: 13884603 8. Janzen DH. Herbivores and the number of tree species in tropical forests. Am Nat. 1970; 104:501–29. 9. Connell JH. On the role of natural enemies in preventing competitive exclusion in some marine animals and in rain forest trees. In: Den Boer PJ, Gradwell G, editors. Dynamics of Populations. New York: PUDOC; 1971. p. 298–312. 10. Lindstrom KM, Foufopoulos J, Parn H, Wikelski M. Immunological investments reflect parasite abun- dance in island populations of Darwin’s finches. Proc R Soc B-Biol Sci. 2004; 271(1547):1513–9. https://doi.org/10.1098/rspb.2004.2752 PubMed PMID: WOS:000222600600011. PMID: 15306324 11. Wills C, Condit R, Foster R, Hubbell S. Strong density- and diversity-related effects help to maintain tree species diversity in a neotropical forest. Proceedings of the National Academy of Sciences. 1997; 94(4):1252–7. 12. Wills C, Condit R. Similar non-random processes maintain diversity in two tropical rainforests. Proc R Soc B-Biol Sci. 1999; 266(1427):1445–52. https://doi.org/10.1098/rspb.1999.0799 PubMed PMID: WOS:000081833200006. PMID: 10457617 13. Comita LS, Queenborough SA, Murphy SJ, Eck JL, Xu KY, Krishnadas M, et al. Testing predictions of the Janzen-Connell hypothesis: a meta-analysis of experimental evidence for distance- and density- dependent seed and seedling survival. Journal of Ecology. 2014; 102(4):845–56. https://doi.org/10. 1111/1365-2745.12232 PubMed PMID: WOS:000338027500003. PMID: 25253908 14. Marhaver KL, Vermeij MJA, Rohwer F, Sandin SA. Janzen-Connell effects in a broadcast-spawning Caribbean coral: distance-dependent survival of larvae and settlers. Ecology. 2013; 94(1):146–60. PubMed PMID: WOS:000316186900017. https://doi.org/10.1890/12-0985.1 PMID: 23600249 15. Karvonen A, Wagner CE, Selz OM, Seehausen O. Divergent parasite infections in sympatric cichlid species in Lake Victoria. J Evol Biol. 2018; 31(9):1313–29. Epub 2018/06/27. https://doi.org/10.1111/ jeb.13304 PMID: 29944770. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 31 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions 16. McGill BJ. Towards a unification of unified theories of biodiversity. Ecol Lett. 2010; 13(5):627–42. https://doi.org/10.1111/j.1461-0248.2010.01449.x PubMed PMID: WOS:000276790200011. PMID: 20337695 17. Kaliszewicz A, Uchmanski J. A cross-phyla response to Daphnia chemical alarm substances by an aquatic oligochaete. Ecological Research. 2009; 24(2):461–6. https://doi.org/10.1007/s11284-008- 0522-0 PubMed PMID: WOS:000263504100022. 18. Wills C, Harms KE, Wiegand T, Punchi-Manage R, Gilbert GS, Erickson D, et al. Persistence of Neigh- borhood Demographic Influences over Long Phylogenetic Distances May Help Drive Post-Speciation Adaptation in Tropical Forests. PLOS ONE. 2016; 11(6):e0156913. https://doi.org/10.1371/journal. pone.0156913 PMID: 27305092 19. Harms KE, Wright SJ, Calderon O, Hernandez A, Herre EA. Pervasive density-dependent recruitment enhances seedling diversity in a tropical forest. Nature. 2000; 404(6777):493–5. https://doi.org/10. 1038/35006630 PubMed PMID: WOS:000086257700047. PMID: 10761916 20. Bagchi R, Gallery RE, Gripenberg S, Gurr SJ, Narayan L, Addis CE, et al. Pathogens and insect herbi- vores drive rainforest plant diversity and composition. Nature. 2014; 506(7486):85–8. https://doi.org/10. 1038/nature12911 http://www.nature.com/nature/journal/v506/n7486/abs/nature12911. html#supplementary-information. PMID: 24463522 21. Detto M, Visser MD, Wright SJ, Pacala SW. Bias in the detection of negative density dependence in plant communities. Ecol Lett. 2019; 22(11):1923–39. https://doi.org/10.1111/ele.13372 PubMed PMID: WOS:000486708600001. PMID: 31523913 22. Hu¨lsmann L, Chisholm RA, Hartig F. Is Variation in Conspecific Negative Density Dependence Driving Tree Diversity Patterns at Large Scales? Trends Ecol Evol. 2020. https://doi.org/10.1016/j.tree.2020. 10.003 PMID: 33589047 23. Lutz JA, Larson AJ, Furniss TJ, Donato DC, Freund JA, Swanson ME, et al. Spatially nonrandom tree mortality and ingrowth maintain equilibrium pattern in an old-growth Pseudotsuga-Tsuga forest. Ecol- ogy. 2014; 95(8):2047–54. https://doi.org/10.1890/14-0157.1 PubMed PMID: WOS:000340215900003. PMID: 25230456 24. Wills C, Harms KE, Condit R, King D, Thompson J, He F, et al. Nonrandom Processes Maintain Diver- sity in Tropical Forests. Science. 2006; 311(5760):527–31. https://doi.org/10.1126/science.1117715 PMID: 16439661 25. Wiegand T, Gunatilleke S, Gunatilleke N. Species associations in a heterogeneous Sri lankan diptero- carp forest. Am Nat. 2007; 170(4):E77–E95. https://doi.org/10.1086/521240 PubMed PMID: WOS:000249307700001. PMID: 17891727 26. Wiegand T, Moloney KA. Handbook of Spatial Point-Pattern Analysis in Ecology. Ch Hall Crc App Env. 2014:1–495. PubMed PMID: WOS:000364179900008. 27. Stoyan D, Bertram U, Wendrock H. Estimation Variances for Estimators of Product Densities and Pair Correlation-Functions of Planar Point-Processes. Ann I Stat Math. 1993; 45(2):211–21. https://doi.org/ 10.1007/Bf00775808 PubMed PMID: WOS:A1993LL84100002. 28. Illian J, Penttinen A, Stoyan H, Stoyan D. Statistical Analysis and Modelling of Spatial Point Patterns. Chichester, UK: Wiley; 2008. 29. Wood SN. Generalized additive models: an introduction with R. Second edition. ed. Boca Raton: CRC Press/Taylor & Francis Group; 2017. xx, 476 pages p. 30. Team RC. R: A language and environment for statistical computing https://www.R-project.org/: Founda- tion for Statistical Computing, Vienna, Austria; 2019. 31. Goreaud F, Pelissier R. On explicit formulas of edge effect correction for Ripley’s K-function. Journal of Vegetation Science. 1999; 10(3):433–8. https://doi.org/10.2307/3237072 PubMed PMID: WOS:000082205100014. 32. Benjamini Y, Hochberg Y. On the Adaptive Control of the False Discovery Rate in Multiple Testing With Independent Statistics. 2000; 25(1):60–83. https://doi.org/10.1016/s0165-0270(99)00194-6 PMID: 10720676 33. Fang S, Munoz F, Ye J, Lin F, Yuan ZQ, Kuang X, et al. Deterministic processes drive functional and phylogenetic temporal changes of woody species in temperate forests in Northeast China. Annals of Forest Science. 2019; 76(2). doi: ARTN 42 https://doi.org/10.1007/s13595-019-0830-2 PubMed PMID: WOS:000465547100002. 34. Hector A, Schmid B, Beierkuhnlein C, Caldeira MC, Diemer M, Dimitrakopoulos PG, et al. Plant diversity and productivity experiments in European grasslands. Science. 1999; 286(5442):1123–7. https://doi. org/10.1126/science.286.5442.1123 PubMed PMID: WOS:000083534200033. PMID: 10550043 35. Loreau M. Biodiversity and ecosystem functioning: recent theoretical advances. Oikos. 2000; 91(1):3– 17. https://doi.org/10.1034/j.1600-0706.2000.910101.x PubMed PMID: WOS:000165237400001. PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 32 / 33 PLOS COMPUTATIONAL BIOLOGYEach of 16 forests shows a unique pattern of between-tree interactions 36. Paine CET, Norden N, Chave J, Forget P-M, Fortunel C, Dexter KG, et al. Phylogenetic density depen- dence and environmental filtering predict seedling mortality in a tropical forest. Ecol Lett. 2012; 15 (1):34–41. https://doi.org/10.1111/j.1461-0248.2011.01705.x PMID: 22004454 37. Davi H, Cailleret M, Restoux G, Amm A, Pichot C, Fady B. Disentangling the factors driving tree repro- duction. Ecosphere. 2016; 7(9). doi: UNSP e01389 https://doi.org/10.1002/ecs2.1389 PubMed PMID: WOS:000387210100001. 38. Green PT, Harms KE, Connell JH. Nonrandom, diversifying processes are disproportionately strong in the smallest size classes of a tropical forest. Proc Natl Acad Sci U S A. 2014; 111(52):18649–54. https://doi.org/10.1073/pnas.1321892112 PubMed PMID: WOS:000347444400061. PMID: 25512498 39. Johnson DJ, Bourg NA, Howe R, McShea WJ, Wolf A, Clay K. Conspecific negative density-dependent mortality and the structure of temperate forests. Ecology. 2014; 95(9):2493–503. PubMed PMID: WOS:000342340200012. 40. Mangan SA, Schnitzer SA, Herre EA, Mack KML, Valencia MC, Sanchez EI, et al. Negative plant–soil feedback predicts tree-species relative abundance in a tropical forest. Nature. 2010; 466(7307):752–5. https://doi.org/10.1038/nature09273 PMID: 20581819 41. Gilbert GS, Webb CO. Phylogenetic signal in plant pathogen-host range. Proc Natl Acad Sci U S A. 2007; 104(12):4979–83. https://doi.org/10.1073/pnas.0607968104 PubMed PMID: WOS:000245256700040. PMID: 17360396 42. Parker IM, Saunders M, Bontrager M, Weitz AP, Hendricks R, Magarey R, et al. Phylogenetic structure and host abundance drive disease pressure in communities. Nature. 2015; 520(7548):542–4. https:// doi.org/10.1038/nature14372 http://www.nature.com/nature/journal/v520/n7548/abs/nature14372. html#supplementary-information. PMID: 25903634 43. Das AJ, Larson AJ, Lutz JA. Individual species-area relationships in temperate coniferous forests. Jour- nal of Vegetation Science. 2018; 29(2):317–24. https://doi.org/10.1111/jvs.12611 PubMed PMID: WOS:000431503000019. 44. LaManna JA, Mangan SA, Alonso A, Bourg NA, Brockelman WY, Bunyavejchewin S, et al. Plant diver- sity increases with the strength of negative density dependence at the global scale. Science. 2017; 356 (6345):1389–92. https://doi.org/10.1126/science.aam5678 PubMed PMID: WOS:000404351500039. PMID: 28663501 45. Hubbell SP. Neutral theory and the evolution of ecological equivalence. Ecology. 2006; 87(6):1387–98. https://doi.org/10.1890/0012-9658(2006)87[1387:ntateo]2.0.co;2 PubMed PMID: WOS:000238658400006. PMID: 16869413 46. Forrister DL, Endara MJ, Younkin GC, Coley PD, Kursar TA. Herbivores as drivers of negative density dependence in tropical forest saplings. Science. 2019; 363(6432):1213-+. https://doi.org/10.1126/ science.aau9460 PubMed PMID: WOS:000461329000048. PMID: 30872524 47. Hogan J, Zimmerman J, Thompson J, Uriarte M, Swenson N, Condit R, et al. The Frequency of Cyclonic Wind Storms Shapes Tropical Forest Dynamism and Functional Trait Dispersion. 2018; 9 (7):404. https://doi.org/10.3390/f9070404 48. Hubbell SP, Foster RB, O’Brien ST, Harms KE, Condit R, Wechsler B, et al. Light-gap disturbances, recruitment limitation, and tree diversity in a neotropical forest. Science. 1999; 283(5401):554–7. https://doi.org/10.1126/science.283.5401.554 PubMed PMID: WOS:000078203300042. PMID: 9915706 49. Comita LS, Uriarte M, Thompson J, Jonckheere I, Canham CD, Zimmerman JK. Abiotic and biotic driv- ers of seedling survival in a hurricane-impacted tropical forest. Journal of Ecology. 2009; 97(6):1346– 59. https://doi.org/10.1111/j.1365-2745.2009.01551.x PubMed PMID: WOS:000270787100025. 50. Stump SM, Comita LS. Interspecific variation in conspecific negative density dependence can make species less likely to coexist. Ecol Lett. 2018; 21(10):1541–51. https://doi.org/10.1111/ele.13135 PubMed PMID: WOS:000444413300009. PMID: 30129216 PLOS Computational Biology \| https://doi.org/10.1371/journal.pcbi.1008853 April 29, 2021 33 / 33 PLOS COMPUTATIONAL BIOLOGY
10.1371_journal.pclm.0000173	RESEARCH ARTICLE Discounting the future: The effect of collective motivation on investment decisions and acceptance of policies for renewable energy Fabian MarderID Martin QuaasID 1,2, Torsten MassonID 3 1,2, Immo FritscheID 3, Julian Sagebiel1,2, Christina MartiniID 1,2, 1 German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Leipzig, Germany, 2 Department of Economics, Leipzig University, Leipzig, Germany, 3 Department of Social Psychology, Leipzig University, Leipzig, Germany * fabian.marder@idiv.de (FM); torsten.masson@uni-leipzig.de (TM) Abstract Climate protection is a collective project. However, most previous research on people’s pro- climate behavior ignores the collective dimension, looking at personal private-sphere behav- ior and considering personal cost-benefit predictors only. The present paper transcends this individualistic perspective by addressing behaviors that target collective transformation (i.e., financial investments in renewable energy projects and the acceptance of renewable energy policies) and predictors of collective cognition and motivation (i.e., social identity). Combin- ing insights and methods from economics and psychology, the current research investigates if collective pro-environmental motivation (e.g., pro-environmental ingroup norms, collective climate efficacy beliefs) can add to the explanation of investment decisions and the accep- tance of policies for renewable energies, also beyond personal psychological and economic factors. Results from a multi-country survey (31 European countries, N = 18,037), including a discrete choice experiment, showed that collective pro-environmental motivation was pos- itively correlated with the acceptance of green energy policies and negatively correlated with discounting of future benefits (money discount rate) in investment decisions for renew- able energies. Importantly, collective pro-environmental motivation remained a significant predictor of policy acceptance and the discount rate after controlling for personal pro-envi- ronmental motivation. Furthermore, the associations between collective pro-environmental motivation and our outcome measures were stronger for respondents who highly identified with their group compared to low identifiers. Our (correlational) results are one of the first to show that collective psychological factors are a unique predictor of green investment behav- ior and acceptance of green policies. From an applied perspective, our findings suggest that interventions should target agentic social identities with norms supporting pro-environmen- tal behavior to increase acceptance of and participation in the transformation towards car- bon neutrality, particularly for persons with low personal pro-environmental motivation. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Marder F, Masson T, Sagebiel J, Martini C, Quaas M, Fritsche I (2023) Discounting the future: The effect of collective motivation on investment decisions and acceptance of policies for renewable energy. PLOS Clim 2(6): e0000173. https://doi.org/10.1371/journal.pclm.0000173 Editor: Abdul Rehman, Henan Agricultural University, CHINA Received: February 6, 2023 Accepted: April 24, 2023 Published: June 5, 2023 Peer Review History: PLOS recognizes the benefits of transparency in the peer review process; therefore, we enable the publication of all of the content of peer review and author responses alongside final, published articles. The editorial history of this article is available here: https://doi.org/10.1371/journal.pclm.0000173 Copyright: © 2023 Marder et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: The data that support the findings of this study are openly available in Zenodo at http://doi.org/10.5281/zenodo.3524917. PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 1 / 22 PLOS CLIMATEThe codes can be retrieved from: https://git.idiv.de/ fm58hufi/discounting-the-future.git. Funding: The authors received no specific funding for this work. Competing interests: The authors have declared that no competing interests exist. Discounting the future Introduction Scientific forecasts show that the ecological, social and economic consequences of continued global warming will be dramatic [1]. Previous calls to action to stop global warming were inef- fective or insufficient. Why is this the case? Perhaps, the wrong actions and the wrong levers of action were addressed: Fighting a large-scale social problem, such as climate change requires collectives to act and societies to transform. However, ignoring the collective and transforma- tive nature of climate action, environmental behavioral sciences and interventions have long been focusing on explaining and changing private (consumption) behavior, instead of behav- iors supporting collective ecological system change. Also, they addressed climate action as the result of a personal decision of individuals, instead of considering the impact of collective cog- nition and motivation (i.e., social identity) [2–4]. Before we show how, in the present research, we introduced the collective dimension in climate action research, we explain why ignoring the collective dimension might have been wrong-headed and insufficient. First, the urgency and scale of global environmental degradation require the immediate transformation of societies’ production and consumption systems. Specifically, dramatic changes in the infrastructural, economic, and legal boundary conditions of individuals’ behavior are needed to enable large-scale changes in private environmentalism across different societal milieus and groups. This is because current structures often discourage or disable pro-environ- mental behavior options as ecologically sustainable products or services are not offered or only at high personal costs in terms of money, effort, or safety (e.g., biking is often perceived as dan- gerous in car-crowded cities, and frequent public transport connections are often missing in rural areas). At the same time, dynamics of free-riding and commons dilemma situations [5–7] require regulations and prohibitions to induce people to make personally costly contributions to the common (environmental) good [8,9]. As a consequence, understanding and changing individuals’ environmental behavior needs a focus on structural changes [10]. This does not mean, however, that investigating and supporting pro-environmental action in individuals is not important. The opposite is true. It is just pivotal to look at the relevant types of action. Thus, instead of limiting the focus to private consumption, behavioral sciences urgently need to understand when, how, and why individuals support or oppose societal and economic transi- tion processes. These actions may include the passive acceptance of green policy measures (e.g. increased taxes on fossil fuels), but also more active behaviors like participation in collective environmental projects, such as investment in renewable energy sites. In the realm of economic behavior, much more than through individual pro-environmental consumption, a person might be able to effectively support the transformation towards carbon-neutrality by investing money in green businesses. In other words, behavioral sciences are now needed to explain indi- viduals’ actions that are directed on changing the system, and not just their personal environ- mental behaviors. This is why the present research seeks to explain the psychological and economic drivers of both the acceptance of environmental policies and personally costly invest- ment decisions in green businesses, such as financial investments in renewable energy projects. There is a second reason why the current focus on personal behavior decisions is insuffi- cient. It refers to an inaccurate conception of individuals’ behavior as a solely personal decision that is driven by personal cost-benefit analyses, personal morals, and personal capabilities. If environmental action would be a solely personal decision, probably, people would never start to act. This is because the current large-scale environmental crises that burden people [11,12] did not emerge, and cannot be solved by, an individual’s action alone. In the global North, it even does not threaten most individuals’ current personal well-being, but that of the many generations of people to come. Obviously, environmental crises such as climate change, are solely collective, but not personal, problems. So, why should people act? We propose that they PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 2 / 22 PLOS CLIMATEDiscounting the future do so, nevertheless, because their basic psychic design implies that humans think and act as group members instead of idiosyncratic and isolated persons [13,14]. That is, people act upon collective problems on the ground of their identification with, and their perception of, a collec- tive they categorize themselves as [15]. Collectives may refer to groups from different levels of inclusiveness, ranging from small activist groups to very inclusive social categories (e.g. gener- ational or national groups; [16]). Then, group members’ environmental cognition and action depend on whether they consider their group as being in favor of pro-environmental action and as having the capabilities to significantly affect environmental crises [17]. Recently, such theorizing on collective pro-environmental motivation has been introduced to the study of pro-environmental behavior [15,18,19]. Building on the Social Identity Approach [13], this work indicates that collective pro-environmental motivation may be an important, but some- times overlooked factor in transition processes towards carbon neutrality [20]. The present research aims to shed light on the question, of how the human capacity to think and act as social group members uniquely shape people’s efforts to mitigate large-scale environmental crises. Extending previous work, we target environmental behaviors that are more directly related to structural changes, namely acceptance of environmental policies and the subjective discount rate in investment decisions for renewable energies. The discount rate is an important factor to consider in investment behavior as it represents the time preference for consumption and reflects the opportunity cost of a specific investment, such as an invest- ment in a renewable energy project. A high discount rate would result in a lower present value of future benefits from the investment, making it less attractive to private or public investors. In contrast, a low discount rate would increase the present value of future benefits and make the investment more appealing. The subjective discount rate can have a significant impact on the pace and success of the transformation towards a carbon-neutral future, as it determines the perceived value and feasibility of investments in green businesses. Economic research on (subjective discount rates in) investment in renewable energy projects has mainly focused on the role of markets and incentive-based policies, for example how to design feed-in tariffs to induce efficient investments into renewable electricity generation [21,22]. However, less is known about the effects of collective psychological factors on investment decisions. Bringing together economic and psychological research, the present work aims to provide novel and interdisciplinary insights into how collective pro-environmental motivation may affect the investment behavior and the acceptance of policies for renewable energies and—as a conse- quence—may increase private engagement for the transformation towards carbon neutrality. Social identity and pro-environmental behavior Psychological research investigating the cognitive and motivational drivers of people’s pro- environmental behavior has tended to focus on personal beliefs and motivation, such as per- sonal environmental attitudes, perceived personal behavior costs or (personal) self-efficacy beliefs. However, we need to consider collective cognition and motivation as well, i.e. the switch from the personal ‘I’ to the collective ‘we’, if we aim to understand and support people’s pro-environmental behavior [15,19,23,24]. Recently, environmental psychology has started to investigate the effects of collective motivation on pro-environmental conduct. In line with the Social Identity Approach [13], this work proposes that–if certain conditions are met–individu- als think and act in terms of their group membership (social identity) when appraising and responding to environmental problems. This self-categorization as a group member increases the importance of collective motivation for pro-environmental behavior. But how exactly does group membership affect environmental appraisal and behavior? Models of collective pro-environmental action, such as the Social Identity Model of Pro- PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 3 / 22 PLOS CLIMATEDiscounting the future Environmental Action (SIMPEA; [15]), describe three key factors that influence how group members respond to perceived environmental crisis: ingroup norms and goals, collective effi- cacy beliefs, ingroup identification. Specifically, SIMPEA proposes that individuals are more likely to act in a pro-environmental manner if the norms and goals of their group support such behavior, particularly for members who are highly identified with their group. Similarly, col- lective environmental efficacy beliefs, i.e. the perception that the ingroup is capable (or not) to achieve its pro-environmental goals, should affect pro-environmental action. If the group is perceived as agentic and capable to achieve its pro-environmental goals, group members, espe- cially high identifiers, should be more motivated to engage in pro-environmental action. How- ever, collective factors may also influence how individuals appraise environmental issues. For example, social identities may increase or decrease acceptance of anthropogenic climate change, depending on whether (or not) climate change denial is perceived as prototypical for the salient group [25]. A growing body of research has shown that collective pro-environmental motivation can foster people’s pro-environmental behavior, albeit less work has been carried out regarding the effects of collective motivation on appraisal processes (see [17,18], for recent reviews). For example, increasing the salience of their political identity reduced acceptance of anthropogenic climate change and climate action intentions among self-identified political right-wingers [26]. Similarly, environmental ingroup norms, i.e. norms supportive or not supportive of pro- environmental behavior, were found to affect pro-environmental action intentions across dif- ferent behavioral domains, including mobility behavior, energy-saving behavior, recycling or sustainable food choice [27–29]. Importantly, the effects of ingroup norms on action inten- tions were stronger for individuals highly identified with their group compared to low identifi- ers [30,31]. Corroborating these findings, meta-analytic results indicated that a stronger endorsement of a social identity with clear climate-protective norms was associated with higher behavioral intentions to fight climate change or self-reported climate-protective behav- ior [32]. Finally, strong beliefs about the ingroup’s capability to mitigate climate change increased climate-protective private consumption behavior as well as climate activist behavior [33–35]. Notably, the effects of collective pro-environmental motivation on pro-environmen- tal action are not limited to groups inherently related to environmental issues (e.g. environ- mental activist groups) but were also observed for broader social categories (e.g., community identification; [36]). This suggests that social identities may provide a point of entry for inter- ventions to foster pro-environmental action across different social contexts. The majority of the studies on collective pro-environmental motivation and pro-environmental behavior, however, have targeted private consumption behaviors or activist behavior (Fritsche et al., 2018). In contrast, fewer studies have investigated the effects of collective pro-environmental motivation on economic behavior, such as decisions about investment in green businesses or acceptance of green, but relatively costly policy measures [37]. Applying the social identity per- spective to the study of green investment behavior may be a timely endeavor, as raising invest- ment in green businesses can be considered a key strategy to facilitate the transformation towards carbon neutrality. Economic research on investment behavior for renewable energies From the economics perspective, an investment into a renewable energy project is profitably if its present value exceeds the costs of the investment. This present value depends on the cash flow of the project. A large literature asks how to design economic instruments that increase the cash flow in order to set the correct investment incentives (reviewed in [21,38]). In addi- tion, the present value of a renewable energy project depends on the discount rate applied to PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 4 / 22 PLOS CLIMATEDiscounting the future future payments. In more psychological terms, a subjective discount rate represents the (reduced) present value people assign to investment outcomes they expect only for the future, not for today. As an example, imagine the choice between receiving €100 today or €100 in one year. If the discount rate is 5%, the €100 somebody receives in one year is worth less to this person today, or in other words, €100 in one year is equivalent to €95.23 today (€100/(1+0.05) = €95.23). The larger the subjective discount rate, the less favorable an investment becomes. Higher discount rates thus make investments with long-term payouts or benefits, such as ben- efits for future generations, substantially less attractive. As a consequence, the subjective dis- count rate could be a crucial factor influencing support for private and public investments for the transformation towards carbon neutrality. While there is a growing body of literature showing that individual discount rates are shaped by personal and contextual circumstances [39,40], much less is known about how social identities and collective motivation affect dis- count rates and, hence, investment strategies. Present research The present research investigates the effects of personal and collective pro-environmental motivation on efforts to support the transformation towards carbon neutrality. Previous work in environmental psychology has often focused on private consumption behaviors (e.g., recy- cling, private mobility behavior) and personal-level variables when predicting pro-environ- mental behavior (e.g., personal attitudes; [41]). The present research extends these studies by testing how collective pro-environmental motivation (e.g., perceived ingroup norms support- ive of pro-environmental action, collective environmental efficacy beliefs) may influence behaviors that are more directly related to changes in our production and consumption pat- terns. Specifically, we examine if collective pro-environmental motivation can uniquely add to the explanation of investment decisions and acceptance of policies for renewable energies. We use investment in renewable energy projects as a key possibility for individuals to contribute to the transformation towards carbon neutrality. The key parameter for private or public deci- sion-making in such climate-related investments is the subjective discount rate [42–45] which converts future payoffs into a present-day equivalent value. The subjective discount rate thereby takes into account the time value of money and other factors. Using data from a multi-country survey in 31 European countries (N = 18,037), we test if personal pro-environmental motivation (H1a) and collective pro-environmental motivation (H1b) are negatively associated with subjective money discount rate in a choice experiment on investment in renewable energy projects and positively associated with acceptance of green energy policies (personal motivation: H2a, collective motivation: H2b). In line with social iden- tity theory, we also examine if the effects of collective pro-environmental motivation on dis- count rate (H3a) and policy acceptance (H3b) are stronger for participants with a strong identification with their group compared to low identifiers. Although the primary focus of the present research is on collective pro-environmental motivation, we explore if the expected cor- relations between collective pro-environmental motivation and our two outcome variables remain significant after controlling for personal pro-environmental motivation. In other words, we examine if collective motivation can uniquely add to the explanation of investment decisions and policy acceptance. Our research is unique as it is the first study that combines choice exper- iments and psychological research methods to investigate investments in renewable energy projects with a focus on time preferences. There exist only few studies that use choice experi- ments to investigate time preferences [46,47] and, to our knowledge, there is no study linking discount rates to collective and individual motivation items. Our study is thus novel, being the first intertemporal choice experiment, to explain time preferences with psychological items. PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 5 / 22 PLOS CLIMATEDiscounting the future Materials and methods Survey and participants We use data from an online multi-country survey collected in the ECHOES Horizon 2020 project (echoes-project.eu; [48]). The survey region covered 31 European countries (EU 27, Norway, Switzerland, Turkey, UK) and the online questionnaire was administered by a market research company. All survey materials were presented to the participants in their native lan- guage and monetary values were translated from Euros into an equivalent value of the national currency, where applicable. About 600 respondents were recruited in each target country using quota sampling methods to ensure that the samples were representative concerning income, age and gender. The total sample amounted to 18,037 completed questionnaires. Par- ticipants received a compensation of €5 after completing the questionnaire. Table 1 presents a summary of the socio-demographic indicators of the survey sample. Questionnaire and measurement of psychological variables The questionnaire included information on respondents’ socio-demographic situation, their decisions in a choice experiment to invest in renewable energy projects, as well as items on respondents’ pro-environmental and energy-related attitudes, beliefs, personal norms and behaviors (and behavioral intentions). Participants were also asked to answer a number of group-related items on energy norms, efficacy beliefs and behaviors as well as their social iden- tification for different social ingroups (see [48], for the full survey). For this, participants were randomly assigned to respond to group-based questions that referred to one out of three social ingroups: their municipality (N = 5919), their country (N = 6007), or Europe (N = 6111). For the current research, we use items on personal pro-environmental motivation and group pro- environmental motivation as predictor variables. Our central outcome measures are the subjec- tive money discount rate (see description of the choice experiment below) and the acceptance of green energy technologies. If not indicated otherwise, all items were measured on five-point scales, ranging from 1 = “strongly disagree” to 5 = “strongly agree”. Acceptance of green technologies was assessed with one item (‘I would accept energy policies that protect the environment even when these induce higher costs, e.g., policies that increase the prices of fossil fuels.’). This variable will henceforth be called Acceptance. Personal pro-envi- ronmental motivation includes two items on personal norms to save energy and to support the energy transition (example item: ‘I feel a personal obligation to support energy policies that support the energy transition.’), a single item on environmental self-identity (‘Acting pro- environmentally is an important part of who I am.’) as well as a graphical measure of inclusion Table 1. Respondent socio-demographic characteristics. Characteristic Description Mean Median Min Max Age 18–34 Age 35–44 Age 45–54 Age 55+ Male respondent age 18–34 respondent age 35–44 respondent age 45–54 respondent age 55+ = 1 if respondent identifies as male Household size number of residents in the household Kids Employed University Income = 1 if there are children under age 14 in the household = 1 if a person is full or part-time employed = 1 if a respondent has a university or equivalent degree estimated net monthly income based on income tranches in 1000’s https://doi.org/10.1371/journal.pclm.0000173.t001 0.35 0.23 0.20 0.23 0.51 2.74 0.60 0.62 0.48 2.02 0 0 0 0 1 3 1 1 0 0 0 0 0 0 1 0 0 0 1 1 1 1 1 6 1 1 1 1.5 0.02 8.18 PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 6 / 22 PLOS CLIMATEDiscounting the future of nature in self (adapted from [49]), a single item on self-efficacy beliefs to support the energy transition (‘As an individual, I can do a lot to support the energy transition.’) and two items on climate change beliefs (‘Most scientists say that the world’s temperature has slowly been rising over the past 100 years. Do you think this has been happening?’, ranging from 1 = “No, defi- nitely not” to 5 = “Yes, definitely”; ‘Assuming that the world’s temperature is rising, do you think this is caused mostly by natural causes, about equally by natural causes and human activ- ity, or mostly by human activity?‘, ranging from 1 = “Mostly by natural causes”to 3 = “Mostly by human activity“). We z-standardized all eight items and combined them into a single mea- sure of personal pro-environmental motivation (Cronbachs α = .80), henceforth called per- sonal motivation index (PMI). Items measuring collective pro-environmental motivation refer to the salient ingroup (municipality, national, or EU). Collective pro-environmental motivation includes two items on perceived injunctive ingroup norms to save energy and to support the energy tran- sition (example item: ‘Many people in [my municipality, the country I live in, the EU] would support it if I used less energy, e.g., using public transport instead of a personal car, turning off lights when leaving the room, using technical appliances which help to save energy.’), two items on perceived descriptive ingroup norms to save energy and to support the energy transition (example item: ‘A growing number of people in [my municipality, the country I live in, the EU] try to save energy, e.g., using public transport instead of a personal car, turning off lights when leaving the room, using technical appliances which help to save energy.’), and a single item on collective efficacy beliefs to support the energy transition (‘We as people in [my municipality, the country I live in, the EU] can act together to achieve the energy transition.’). We z-standardized all items and averaged them into a single mea- sure of collective pro-environmental motivation (Cronbachs α = .79), henceforth called col- lective motivation index (CMI). Finally, social identification, i.e. identification with the salient ingroup, was assessed with one item (‘How much do you see yourself as a citizen of [your municipality, the country you live in, Europe]?’, ranging from 1 = “not at all” to 5 = “very much”). This variable will henceforth be called ID. The summary of these variables are depicted in Table 2. The choice experiment The ECHOES survey incorporated a discrete choice experiment (DCE) to examine preferences for community renewable energy (CRE) projects. A DCE is a research method used to study the preferences of individuals. It is a type of stated preference study, which is used to measure how individuals would choose among different options. The method involves presenting respondents with a series of hypothetical choices between two or more options, where each option is defined by a set of attributes. The respondents are asked to indicate which option they would choose in each scenario. Within the ECHOES’ DCE, the respondents were presented with two hypothetical invest- ment opportunities in eight different scenarios. In each scenario, respondents could choose to Table 2. Summary statistics of the respondent-specific variables. Variable Description PMI CMI ID Personal pro-environmental motivation index Collective pro-environmental motivation index How much do you see yourself as a citizen of [your municipality, the country you live in, Europe]? N Mean St. Dev. Min Max 18,037 0.000 0.643 18,037 0.000 0.763 18,037 3.372 1.019 −2.789 1.171 −2.563 1.482 1 5 Acceptance I would accept energy policies that protect the environment even when these induce higher costs (e.g., policies 18,037 3.298 1.130 1 5 that increase the prices of fossil fuels). https://doi.org/10.1371/journal.pclm.0000173.t002 PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 7 / 22 PLOS CLIMATEDiscounting the future Fig 1. Example choice card (Source: [48]). https://doi.org/10.1371/journal.pclm.0000173.g001 invest in a wind park or solar farm, with the investment levels, holding time and other attri- butes of the options varying between scenarios. A third ‘opt-out’ option was also provided in each scenario, allowing respondents not to invest. The order of the scenarios was randomized, and the survey included three blocks of eight scenarios for a total of 24 choice scenarios. An example choice card is depicted in Fig 1. The experimental design uses the D-efficiency criteria with Bayesian priors for creating choice sets. More information about the statistical design of the DCE can be found in [50]. The levels of the holding periods vary between 5, 10 and 15 years. To calculate the profit we use the profit rate (0%, 5%, 10%, 20% or 50%) and the investment level which were randomly assigned. The investment levels—€100, €500, €1000, €2000, or €5000 —, were not varied between the scenarios in order to simplify the choice tasks for the respondents. In Table 3 we describe all attributes and list their levels. Further, the survey included a treatment that told respondents that a local government, national government, or EU official had endorsed the investment opportunities. Each treatment was shown to one-quarter of the respondents in each country, with the remaining respondents seeing only a briefing explaining the investment opportunities. PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 8 / 22 PLOS CLIMATEDiscounting the future Table 3. Attribute levels and description. Attribute Profit rate Holding Period Visibility Description The percent of money you get on top of your initial investment. Levels 0%, 5%, 10%, 20%, 50% The number of years until you get your money back, including any profits. 5, 10, 15 years If the proposed wind or solar park is visible from your home. visible or not visible Administrator The group that handles your investment and is in charge of building and running the power plant. community organization, utility company or government entity https://doi.org/10.1371/journal.pclm.0000173.t003 Econometric model: Empirical model based on random utility theory Our model assumes that people maximize utility over time [51]. Utility in a broad sense depends on individual-level factors, both tangible economic variables, such as the amount and timing of monetary payoffs, and personal behavior of self-efficacy beliefs. It further includes variables that capture collective cognition and motivation relevant to the decision-making situ- ation. Specifically, utility is a function of observable characteristics of the investment alterna- tives, in particular the profit rate, the project length, the investment volume, the visibility of the renewable energy project, and the administrator of the project, as specified in the choice experiment. Moreover, the parameters of the utility function are modeled as functions of observed individual and collective motivations of the respondents. The main aim of the paper is to analyze how the respondents’ preferences are shaped by these latter variables. In denotes the investment, which is independent of the choice alternative j but varies with respondent n, with In2{100,500,1000,2000,5000} Euros. The profit rate is πj2{0,0.05,0.10,0.20,0.50}, and is one of the attributes changing with choice alternatives. After the specified holding period for the choice alternative, Tj2{5,10,15} years have passed, the project delivers the cash flow In(1+πj). The utility from cash flow and other characteristics of the renewable energy investment Xnj at the end of the investment period Tj is described by the following utility function Vnj ¼ e(cid:0) dnTjðIn ð1 þ pjÞÞaðeXnjÞb; ð1Þ with α>0, and where the β is a vector of parameters indicating the marginal utility of other investment-specific characteristics Xj. As these are categorical variables, they enter linearly in the log of utility. The utility derived from the investment accrues Tj periods into the future, whereas the decision is made at present. Thus utility is expressed as a present value, which is obtained by applying the subjective (annual) utility discount rate δ. Taking logs and adding an independently and identically distributed random component εnj, we obtain the model Unjp ¼ a lnðIn ð1 þ pjÞÞ þ b Xnj (cid:0) dn Tj þ εnj: ð2Þ Applying the model to the data from the choice experiment allows us to identify the model parameters. We model the utility discount rate as a function of individual and collective moti- vation indicator variables, which we summarize in the variable Yn. The utility discount rate becomes: dn ¼ d0 þ d1 Yn: ð3Þ The effects of the social and psychological variables Yn on the utility discount rate δn are empirically identified by the estimated parameters for the interaction between these variables PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 9 / 22 PLOS CLIMATEDiscounting the future and the holding period Tn, leading to Unjp ¼ a lnðIn ð1 þ pjÞÞ þ bXj (cid:0) d0Tj (cid:0) d1Tj � Yn þ εnj: To facilitate interpretation, we convert the utility discount rate δn into a money discount rate, dividing it by the estimated coefficient for log profit, α. We thus obtain the money dis- count rate rn ¼ d0 þ d1Yn a : ð4Þ ð5Þ In the choice experiment, respondents choose repeatedly between two hypothetical invest- ment alternatives. We assume that the alternatives are mutually exclusive and the respondent chooses either one of the two investment alternatives j2{1,2} or chooses not to invest (opt out) j = 0. In this setting, the parameters from this utility function can be estimated using a condi- tional logit model. Assuming that εnj is Extreme Value Type I (Gumbel) distributed, we obtain the logit probability � � Pr ynj ¼ i ¼ PJ expðUniÞ j¼1 expðUnjÞ ð6Þ As only differences in utility matter, the model can only be identified if the error variance is normalized. The normalization implies that the estimated parameters are confounded with the scale of the error variance so that the parameters have arbitrary values which cannot be directly interpreted. However, by dividing the subjective utility discount rate by the coefficient of the log profit α, the scale parameters drop out and we can interpret money discount in units of % of profit per year. We are particularly interested in the subjective money discount rate, ρn, which has been identified to be a key variable in decision-making related to climate change, as pointed out in the introduction. Results In Models 1–5 (Table 4) we estimate Conditional Logit models using the DCE data. The dependent variable is the choice made by the respondents. The models include alternative-spe- cific constants (ASC_A and ASC_B), which show the preferences for investment options A and B (i.e. respondent decides to invest in the energy project) over the opt-out alternative (i.e. respondent decides not to invest in the energy project). We also entered alternative-specific variables (Profit, Holding period, Visible installation, Community admin, Utility admin) and respondent-specific variables in the analysis (personal motivation index, collective motivation index, ID, group assignment: municipality, country, EU). We are in particular interested in the ratio of coefficients of the variable Holding period and ln(Profit) which we can interpret as the money discount rate, i.e. one of our central outcomes. Specifically, we aim to examine the impact of respondent-specific variables on the money discount rate, by analyzing interaction effects between the variable Holding period and the respondent-specific variables (personal motivation index, collective motivation index, ID, group assignment). For testing our hypoth- eses, we included the two-way interaction term of Holding period and personal motivation index in Model 1 (H1a), the two-way interaction term of Holding period and collective moti- vation index in Model 2 (H1b), as well as all two-way and three-way interaction terms of Hold- ing period, collective motivation index and ingroup identification (ID) in Model 4 (H3). For exploring if collective pro-environmental motivation uniquely predicts the money discount PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 10 / 22 PLOS CLIMATETable 4. Results of the conditional logit models. Discounting the future ASC_A ASC_B ln(Profit) Hold Visible Community Utility HoldPMI HoldCMI HoldID HoldCMIID HoldMun HoldCMIMun HoldIDMun HoldCMIIDMun HoldCountry HoldCMICountry HoldIDCountry HoldCMIIDCountry No Observations No Respondents LL (Null) LL (Converged) Model 1 -0.275 (0.021) -0.146 * (0.021) 4.720 * (0.037) 0.095 * (0.001) -0.021 * (0.008) 0.051 * (0.010) -0.122 * (0.010) -0.043 * (0.002) Model 2 -0.279 * (0.021) -0.150 * (0.021) 4.714 * (0.037) 0.094 * (0.001) -0.020 * (0.008) 0.052 * (0.010) -0.122 * (0.010) -0.033 * (0.002) Model 3 -0.271 * (0.021) -0.142 * (0.021) 4.725 * (0.037) 0.095 * (0.001) -0.022 * (0.008) 0.051 * (0.011) -0.122 * (0.010) -0.030 * (0.002) -0.018 * (0.002) Model 4 -0.279 * (0.021) -0.150 * (0.021) 4.715 * (0.037) 0.104 * (0.004) -0.020 * (0.008) 0.051 * (0.010) -0.122 * (0.010) -0.019 * (0.005) -0.003 * (0.001) -0.004 * (0.001) 144088 18037 -158297 -137454 144088 18037 -158297 -137677 144088 18037 -158297 -1372256 144088 18037 -158297 -137649 Model 5 -0.277 * (0.024) -0.145 * (0.024) 4.701 * (0.042) 0.107 * (0.008) -0.026 * (0.009) 0.055 * (0.012) -0.116 * (0.012) -0.013 * (0.009) -0.004 (0.002) -0.005 (0.003) -0.003 (0.011) 0.005 (0.013) 0.002 (0.003) -0.001 (0.004) -0.004 (0.011) -0.018 * (0.013) 0.002 (0.003) 0.005 (0.004) 108248 13552 -118923 -103438 Notes: (i) This table presents the Multinomial Logit (MNL) estimation derived from Eq (4).(ii) The dependent variable is the utility obtained from making investment decisions. (iii) Dummy variables ASC_A and ASC_B represent investment alternatives A and B, respectively. The variable ln(Profit) is the natural logarithm of the profit, Hold represents the duration of the holding period in years, and Visible is a dummy variable that equals 1 if the energy project is visible from the respondent’s home, and 0 otherwise. Dummy variables Community and Utility indicate the administrator of the power plant, with the government admin as the base category. PMI denotes personal motivation index, CMI denotes collective motivation index, and ID denotes ingroup identification. Municipality (Mun) and Country represent the group assignment, with the EU as the base category. (iv) Robust standard errors are reported in parentheses. Statistical significance is denoted as * for p < 0.01, for p < 0.05, and * for p < 0.1. https://doi.org/10.1371/journal.pclm.0000173.t004 PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 11 / 22 PLOS CLIMATEDiscounting the future rate, we included personal motivation index, collective motivation index and their two-way interaction terms with Holding period in Model 3. Probability to invest in energy project First, we analyzed respondents’ choices to invest or not invest in the proposed energy project. We observe a consistent preference for the opt-out alternative over an investment in the proj- ect A and a consistent preference for the opt-out alternative over an investment in the project B, ceteris paribus, evidenced by the significant negative regression coefficients for the variables ASC_A and ASC_B (Models 1–5). Overall, 27% of the choices were project A, 30% project B and 43% opt-out. We also find that higher profit rates, non-visible installation (vs. visible installation) and community-based administration (vs. administration by a utility company or public authority) of the energy site increased probability of investment in the energy project. These results are in line with previous findings on private investments in renewable energy projects [50]. (Money) discount rate From the coefficients of the variable Holding period and ln(Profit) in Models 1–5, we can derive the money discount rates by applying Eq (5). We expect that the discount rate is nega- tively associated with personal pro-environmental motivation (H1a) and collective pro-envi- ronmental motivation (H2a). The results of Models 1 and 2 support our assumptions. Specifically, we find a negative interaction effect of Holding period and personal motivation index (coefficient of HoldPMI) in Model 1, indicating that higher levels of personal pro-envi- ronmental motivation are associated with a lower money discount rate (see Fig 2A). The ratio of the coefficient of HoldPMI and the coefficient of ln(Profit) describes the impact of an Fig 2. A: Money discount rate and personal motivation index. B: Money discount rate and collective motivation index. (Shaded areas indicate the 95% confidence intervals). https://doi.org/10.1371/journal.pclm.0000173.g002 PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 12 / 22 PLOS CLIMATETable 5. Calculated money discount rates. Discount rate Discount rate PMI CMI CMI ID –1 SD ID +1 SD ID –1 SD ID +1 SD Discounting the future Model 1 Model 2 Model 3 Model 4 Level of money discount rate in % 2.00 * (0.030) 1.99 * (0.030) 2.01 * (0.030) Change in money discount rate in % with a one-point increase of -0.91 * (0.039) -0.70 * (0.036) -0.65 * (0.046) -0.38 * (0.038) 2.07 * (0.039) 1.95 * (0.039) -0.60 * (0.044) -0.76 * (0.045) Notes: (i) This table presents money discount rates in % derived from Eq (5). (ii) In row one the mean money discount rates are represented. (iii) Row 2 represents the discount rates depending on the ingroup identification (ID) (only relevant in model 4). (iv) The third and fourth rows indicate the changes in the discount rates that occur due to a one-point increase in the personal motivation index (PMI) and collective motivation index (CMI), respectively. (v) Row five indicate the changes in the discount rates that occur due to a one-point increase in the collective motivation index (CMI) depending on the ingroup identification (ID) (only relevant in model 4). (vi) Due to insignificant results, model 5 is omitted in this table. (vii) Robust standard errors are reported in parentheses. Statistical significance is denoted as * for p < 0.01, ** for p < 0.05, and * for p < 0.1. https://doi.org/10.1371/journal.pclm.0000173.t005 increase in the personal motivation index by one unit on the money discount rate (seeTable 5). Given that the mean value of the personal motivation index is zero, the mean money discount rate across all respondents is 2.00% per year. In other words: €100 in one year is equivalent to €98.00 today (€100/(1+0.02) = €98.00). Further, increasing the personal motivation index by one unit decreases the mean money discount rate by 0.91%. Similarly, results also reveal a neg- ative interaction effect of Holding period and collective motivation index in Model 2 (coeffi- cient of HoldCMI), showing that a stronger collective pro-environmental motivation is related to a lower money discount rate (see Fig 2B). The mean money discount rate here is 1.99% and decreases by 0.7% with an increase of the collective motivation index by one unit. Next, we explored if the negative association between collective pro-environmental motiva- tion and the money discount rate will remain stable after controlling for the effects of personal pro-environmental motivation. Results of Model 3 indicate that including the interaction effect of personal motivation index and Holding period (HoldPMI) did not change the inter- action effect of collective motivation index and Holding period (see Fig 3). Put differently, the negative relationship between collective pro-environmental motivation and the money dis- count rate remained robust after controlling for personal pro-environmental motivation. The results of Models 2 and 3 support our assumption that a stronger collective pro-environmental motivation is associated with a lower money discount rate. Building on the Social Identity Approach, we expect that the negative relationship between collective pro-environmental motivation and the money discount rate is stronger for participants who are highly identified with their group compared to low identifiers (H3a). The results of Model 4 support this assumption, revealing a statistically significant three-way interaction effect of Holding period, PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 13 / 22 PLOS CLIMATEDiscounting the future Fig 3. Money discount rate, personal motivation and collective motivation index. (Shaded areas indicate the 95% confidence intervals). https://doi.org/10.1371/journal.pclm.0000173.g003 collective motivation index and ID (coefficient of HoldCMIID). Inspection of the simple slopes (see Fig 4) showed that the negative association between collective pro-environmental motivation and the money discount rate was stronger for high identifiers (+1SD) than for respondents with low levels of ID (-1SD). Specifically, high identifiers exhibited a lower money discount rate compared to low identifiers when collective pro-environmental motivation was high. However, we found no difference in money discount rate between high and low identifi- ers for low levels of collective pro-environmental motivation. Finally, we also tested if the nega- tive correlation between money discount rate and collective pro-environmental motivation changed for different salient ingroups (municipality, country, EU). Results of Model 5 showed no significant interaction effects of Holding period, collective motivation index and the dummy variables for the type of salient identity (coefficients of HoldCMIMunicipal and HoldCMICountry). This suggests that the negative relationship between collective motiva- tion and money discount rate can be generalized across different forms of collectives. Acceptance of green energy policies Table 6 presents the results of a linear mixed model to investigate the relationships between policy acceptance, our second outcome measure, and the respondent-specific variables. The fixed effects in this model are represented by the coefficients of the independent variables per- sonal motivation index, collective motivation index, and ID, as well as the interaction term of collective motivation index and ID. These coefficients represent the average effect of each vari- able on policy acceptance across all groups. The random effect in this model is represented by the Survey country variable. This variable accounts for the fact that the data was collected from multiple groups (countries) and that the variation within each group may be different from the variation across groups. The inclusion of random effects in this model helps to account for the PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 14 / 22 PLOS CLIMATEDiscounting the future Fig 4. Money discount rate, collective motivation index and group identification. (Shaded areas indicate the 95% confidence intervals). https://doi.org/10.1371/journal.pclm.0000173.g004 Table 6. Results of the linear mixed model. (Intercept) PMI CMI ID CMIID AIC BIC LL Num. obs. Num. groups: Survey country Model 6 3.292 (0.034) 0.793 * (0.022) 0.134 * (0.031) 0.000 (0.007) 0.037 * (0.008) 48136 48230 -24056 18037 31 Notes: (i) This table presents the result of a linear mixed model. (ii) The dependent variable is the acceptance of green energy policies, which is measured one-to-five Likert scale. (iii) PMI denotes personal motivation index, CMI denotes collective motivation index, and ID denotes ingroup identification. (iv) Standard errors are reported in parentheses and are robust. Statistical significance is denoted as * for p < 0.01, for p < 0.05, and * for p < 0.1. https://doi.org/10.1371/journal.pclm.0000173.t006 PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 15 / 22 PLOS CLIMATEDiscounting the future Fig 5. Policy acceptance, collective motivation index and group identification. (Shaded areas indicate the 95% confidence intervals). https://doi.org/10.1371/journal.pclm.0000173.g005 non-independence of observations within groups and leads to more accurate estimates of the fixed effects of our independent variables. We expected that policy acceptance is positively associated with personal pro-environmental motivation (H1b) and collective pro-environ- mental motivation (H2b). We also expect that the correlation between policy acceptance and collective motivation is stronger for high identifiers compared to low identifiers (H3b). In line with H1b and H2b, the results of Model 6 (Table 6) indicate significant positive relationships between personal motivation index and acceptance of green energy policies (coefficient of PMI) as well as between collective motivation index and policy acceptance (coefficient of CMI). Although the correlation between personal motivation index and policy acceptance is stronger, collective pro-environmental motivation can uniquely add to the explanation of pol- icy acceptance. Furthermore, we found a significant interaction effect of collective motivation index and ID (coefficient of CMI*ID). Inspection of the simple effects (see Fig 5) revealed that the correlation between collective motivation index and policy acceptance is stronger when ID is high (+1SD) than for low levels of ID (-1SD). Results of Model 6 thus support H3b. Discussion Given the urgency of the ecological transformation of whole societies, it is important to deter- mine when and why citizens are ready to support systemic changes by accepting green policies and by investing their money in green businesses. The collective nature of effectively coping with large-scale environmental crises suggests that such support cannot be fully explained as a personal decision people make on the ground of their perceived personal costs, benefits, and capabilities [15,23]. Instead, support for a green transformation might be better understood as an individual’s expression of collective action. That is, people support–personally costly–sys- temic changes towards ecological sustainability when they define themselves as a member of a collective that has collectively shared pro-environmental norms and goals and appears to be PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 16 / 22 PLOS CLIMATEDiscounting the future agentic in initiating collective action and effectively contributing to fighting environmental crises [17]. The current research supports this novel look at individuals’ pro-environmental action: Collective motivation to protect the environment, indicated by people’s perception of pro-environmental collective norms and collective efficacy, predicted both people’s acceptance of green energy policies and lower discounting of future gains in hypothetical green energy investment decisions. While personal motivation (sense of personal obligation to protect the environment and personal pro-environmental identity) predicted these pro-environmental behaviors as well (as suggested by previous results; e.g., [32,41,52]) the effects of collective motivation remained present when controlling for the effect of personal motivation. That is, collective motivation predicted support of the transformation independent of personal motiva- tion. At the same time, controlling for personal motivation effects reduced the effects of collec- tive motivation. This suggests that part of the collective motivation effect could be mediated via people’s personal sense of pro-environmental obligation and identity. In other words, per- ceived collective norms and efficacy might affect people’s pro-environmental behavior by changing the personal attitudes that then drive pro-environmental action. Our results are in line with other studies showing that collective motivation can foster pro-environmental behav- ior, either directly [30] or through changes in personal pro-environmental motivation [35,53]. For example, previous results indicate that a strong sense of identification with an energy com- munity initiative was positively associated with sustainable energy behavior and behavior intentions [54]. As a further indication that the effects of norms and collective efficacy are also truly collec- tive, we found that the effects were stronger in people who indicated higher identification with their salient ingroup. Obviously, it needs identified group members to make collective motiva- tion factors work. Groups may not just have the power and magnitude to bring about signifi- cant pro-environmental change through societal transformation but they also provide identified members with a sense of agency in the face of collective problems causing personal helplessness, and they validate their actions as being appropriate. This is why, in our study across 31 different European countries, not just very large and highly powerful collective iden- tities, such as “EU Citizens”, had the observed motivating effects, but also smaller groups, such as the people in one’s own country or municipality. Obviously, just thinking about the self in terms of some collective strengthens people’s motivation to support pro-environmental sys- temic change (see [55] for similar results in the context of energy community initiatives). This transcends earlier research showing that ingroup norms affect group members’ environmental behavior only when they were highly identified with their group [30,31,56]. The current results show that ingroup identification is a crucial boundary condition for collective motivation fac- tors more broadly, including collective efficacy beliefs, to affect people’s pro-environmental behavior, as proposed in the social identity model of pro-environmental action [15]. Economic analysis usually takes preferences as given. This is true in particular for the dis- count rate, which is often assumed to be a constant, independent of time and circumstances also in the analysis of climate change mitigation policies [57]. Our study provides evidence that “personal circumstances” affect the discount rate. Specifically, personal and collective pro- environmental motivations influenced the discount rate people applied to renewable energy investment decisions in a choice experiment. Our study thus may help to inform the analysis of climate policies and renewable energy transition with endogenously changing preferences [58]. To our knowledge, there exists only limited work on individual discount rates in the con- text of climate policies [47] combined a DCE on energy efficiency investments with a method to identify individual discount rates. Similar to our results, they find a strong influence of indi- vidual discount rates on willingness to pay (in their case for energy-efficient home appliances). They also find several socio-demographic variables to influence the individual discount rate PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 17 / 22 PLOS CLIMATEDiscounting the future but do not investigate psychological factors [59] investigate psychological factors affecting the willingness to invest in socially responsible investments (which includes an environmental component) but do not investigate individual discount rates. They find strong links between the willingness to invest and attitudinal variables (warm glow, social norms). To that end, we can conclude that other studies find similar patterns between investment decisions, individual discount rates and psychological factors. Yet, our study is unique in linking individual dis- count rates directly to group motivational factors. To increase investments in renewable energy projects and other environmental projects, policymakers can initiate marketing campaigns that aim to increase collective and individual motivations and strengthen collective experiences. Through this channel, the average social discount rate may reduce, facilitating the willingness to invest. Our study has a few limitations: First, there is a debate about statistical challenges from inte- grating indicator variables such as self-reported measures of perceptions and attitudes in dis- crete choice models. There are two reasons to expect endogeneity in these variables [60]. First, indicator variables are often measured on Likert scales, which can lead to measurement error. Second, the dependent variable in a choice experiment and an indicator variable can be simul- taneously caused by a third unobserved variable, leading to a correlation between the error term and the indicator variable. Such a setting can lead to bias in the estimated parameters. The literature suggests different approaches to deal with such endogeneity issues. The first issue, measurement error could be handled by hybrid choice models, which model the endoge- nous variables as explanatory variables for a latent variable [61]. Yet, these models cannot cap- ture the bias caused by simultaneity. Other approaches used to deal with endogeneity in choice models are control function approaches and instrumental variable regressions [62]. However, all these approaches can cause new issues, which may be more severe than the biases caused by measurement error and simultaneity. We, therefore, decided not to use them in this paper. Second, our experiment does not allow us to separate the discount rate for environmental benefits from the discount rate of money because we do not vary the timing of the environ- mental benefit. In a future study, it would be interesting to investigate time preferences for money vis-à-vis time preferences for environmental benefits. For example, a choice experi- ment could include an attribute describing after how many years the investment will create positive externalities in terms of CO2 mitigation. Third, the generalizability of our results needs to be done with care. On the one hand, our sample is not representative for the 30 European countries. For example, our samples have higher education degrees than the country averages. On the other hand, the choices made in the choice experiment are hypothetical and may suffer from hypothetical bias [63]. Respon- dents likely overstate their willingness to invest, i.e. they would be less likely to do so in real life. Conclusion Pursuing rapid societal transformation towards ecological sustainability requires citizens’ sup- port. Obviously, environmentalism has definitely entered the stage where it is no longer suffi- cient to consider private consumption and lifestyle behavior as the individuals’ contribution to saving the environment. Instead, now this is about supporting systemic, collective changes. This further illustrates that pro-environmental action is basically collective in nature and is motivated on the ground of collective cognition. The present study provides evidence for the crucial role of collective motivation in explaining individuals’ support of an ecological trans- formation of societies, although the correlational nature of our data requires conceptual repli- cations in experimental or longitudinal studies to provide clear causal evidence. On the more PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 18 / 22 PLOS CLIMATEDiscounting the future methodological side, our study shows that insights from psychology can meaningfully contrib- ute to our understanding of economic decision-making, thus opening up a new perspective for fruitful interdisciplinary collaboration. Supporting information S1 Fig. Histograms of personal (A) and collective (B) motivaion indices. (TIF) S1 Table. Items of personal motivation index. (DOCX) S2 Table. Items of collective motivation index. (DOCX) Acknowledgments We thank participants of the Breathing Nature Conference 2022 in Leipzig for helpful discussion. Author Contributions Conceptualization: Fabian Marder, Torsten Masson, Julian Sagebiel, Christina Martini, Martin Quaas, Immo Fritsche. Formal analysis: Fabian Marder, Torsten Masson, Martin Quaas, Immo Fritsche. Investigation: Fabian Marder, Torsten Masson, Christina Martini. Methodology: Fabian Marder, Torsten Masson, Julian Sagebiel, Martin Quaas, Immo Fritsche. Software: Fabian Marder, Julian Sagebiel. Supervision: Martin Quaas, Immo Fritsche. Visualization: Fabian Marder. Writing – original draft: Fabian Marder, Torsten Masson, Martin Quaas, Immo Fritsche. Writing – review & editing: Fabian Marder, Torsten Masson, Julian Sagebiel, Martin Quaas, Immo Fritsche. References 1. IPCC. Climate change 2022: Impacts, adaptation and vulnerability: summary for policymakers. Geneva: IPCC; 27 February 2022. 2. Bamberg S, Mo¨ser G. Twenty years after Hines, Hungerford, and Tomera: A new meta-analysis of psy- cho-social determinants of pro-environmental behaviour. Journal of Environmental Psychology. 2007; 27:14–25. https://doi.org/10.1016/j.jenvp.2006.12.002 3. Hamann K, Masson T. Kollektives nachhaltiges Handeln und Psychologie. In: Genkova P, editor. Hand- buch Globale Kompetenz: Grundlagen–Herausforderungen–Krisen. Wiesbaden: Springer Fachmedien Wiesbaden; Imprint Springer; 2020. p. 1–16. https://doi.org/10.1007/978-3-658-30684-7_35–1 4. Lange F, Dewitte S. Measuring pro-environmental behavior: Review and recommendations. Journal of Environmental Psychology. 2019; 63:92–100. https://doi.org/10.1016/j.jenvp.2019.04.009 5. Gordon HS. The Economic Theory of a Common-Property Resource: The Fishery. Journal of Political Economy. 1954; 62:124–42. https://doi.org/10.1086/257497 6. Hardin G. The tragedy of the commons. Science. 1968; 162:1243–8. https://doi.org/10.1126/science. 162.3859.1243 PMID: 17756331 PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 19 / 22 PLOS CLIMATEDiscounting the future 7. Stavins RN. The Problem of the Commons: Still Unsettled after 100 Years. American Economic Review. 2011; 101:81–108. https://doi.org/10.1257/aer.101.1.81 8. Bronnmann J, Stoeven MT, Quaas M, Asche F. Measuring Motivations for Choosing Ecolabeled Sea- food: Environmental Concerns and Warm Glow. Land Economics. 2021; 97:641–54. https://doi.org/10. 3368/wple.97.3.101119-0147R 9. Quaas MF, Meya JN, Schenk H, Bos B, Drupp MA, Requate T. The social cost of contacts: Theory and evidence for the first wave of the COVID-19 pandemic in Germany. PLoS One. 2021; 16:e0248288. https://doi.org/10.1371/journal.pone.0248288 PMID: 33740007 10. Becker E, editor. Soziale O¨ kologie: Grundzu¨ ge einer Wissenschaft von den gesellschaftlichen Natur- verha¨ ltnissen. Frankfurt: Campus Frankfurt / New York; 2006. 11. Clayton S, Karazsia BT. Development and validation of a measure of climate change anxiety. Journal of Environmental Psychology. 2020; 69:101434. https://doi.org/10.1016/j.jenvp.2020.101434 12. Salomon E, Preston JL, Tannenbaum MB. Climate change helplessness and the (de)moralization of individual energy behavior. J Exp Psychol Appl. 2017; 23:15–28. https://doi.org/10.1037/xap0000105 PMID: 28165276 13. Reicher S, Spears R, Haslam SA. The Social Identity Approach in Social Psychology. In: Wetherell M, Mohanty C, Mohanty CT, editors. The SAGE handbook of identities. Los Angeles, Calif., London: SAGE; 2010. p. 45–62. https://doi.org/10.4135/9781446200889.n4 14. 15. 16. Turner JC. Rediscovering the social group: A self-categorization theory. Oxford: Basil Blackwell; 1987. Fritsche I, Barth M, Jugert P, Masson T, Reese G. A social identity model of pro-environmental action (SIMPEA). Psychol Rev. 2018; 125:245–69. https://doi.org/10.1037/rev0000090 PMID: 29265852 Lickel B, Hamilton DL, Wieczorkowska G, Lewis A, Sherman SJ, Uhles AN. Varieties of groups and the perception of group entitativity. J Pers Soc Psychol. 2000; 78:223–46. https://doi.org/10.1037//0022- 3514.78.2.223 PMID: 10707331 17. Masson T, Fritsche I. We need climate change mitigation and climate change mitigation needs the ‘We’: a state-of-the-art review of social identity effects motivating climate change action. Current Opinion in Behavioral Sciences. 2021; 42:89–96. https://doi.org/10.1016/j.cobeha.2021.04.006 18. 19. Fritsche I, Masson T. Collective climate action: When do people turn into collective environmental agents? Curr Opin Psychol. 2021; 42:114–9. https://doi.org/10.1016/j.copsyc.2021.05.001 PMID: 34130199 Fielding KS, Hornsey MJ. A Social Identity Analysis of Climate Change and Environmental Attitudes and Behaviors: Insights and Opportunities. Front Psychol. 2016; 7:121. https://doi.org/10.3389/fpsyg. 2016.00121 PMID: 26903924 20. Bamberg S, Rees JH, Schulte M. Environmental protection through societal change. In: Clayton SD, Manning C, editors. Psychology and climate change: Human perceptions, impacts, and responses. London, San Diego, Cambridge, Oxford: Academic Press; 2018. p. 185–213. https://doi.org/10.1016/ B978-0-12-813130-5.00008–4 21. Requate T. Green tradable certificates versus feed-in tariffs in the promotion of renewable energy shares. Environ Econ Policy Stud. 2015; 17:211–39. https://doi.org/10.1007/s10018-014-0096-8 22. Lancker K, Quaas MF. Increasing marginal costs and the efficiency of differentiated feed-in tariffs. Energy Economics. 2019; 83:104–18. https://doi.org/10.1016/j.eneco.2019.06.017 23. Barth M, Masson T, Fritsche I, Fielding KS, Smith JR. Collective responses to global challenges: The social psychology of pro-environmental action. Journal of Environmental Psychology. 2021; 74:101562. 24. Schulte M, Bamberg S, Rees J, Rollin P. Social identity as a key concept for connecting transformative societal change with individual environmental activism. Journal of Environmental Psychology. 2020; 72:101525. https://doi.org/10.1016/j.jenvp.2020.101525 25. Bliuc A-M, McGarty C, Thomas EF, Lala G, Berndsen M, Misajon R. Public division about climate change rooted in conflicting socio-political identities. Nature Clim Change. 2015; 5:226–9. https://doi. org/10.1038/nclimate2507 26. Unsworth KL, Fielding KS. It’s political: How the salience of one’s political identity changes climate change beliefs and policy support. Global Environmental Change. 2014; 27:131–7. https://doi.org/10. 1016/j.gloenvcha.2014.05.002 27. Bamberg S, Rollin P, Schulte M. Local mobility culture as injunctive normative beliefs–A theoretical approach and a related measurement instrument. Journal of Environmental Psychology. 2020; 71:101465. https://doi.org/10.1016/j.jenvp.2020.101465 28. Barth M, Jugert P, Fritsche I. Still underdetected–Social norms and collective efficacy predict the accep- tance of electric vehicles in Germany. Transportation Research Part F: Traffic Psychology and Behav- iour. 2016; 37:64–77. https://doi.org/10.1016/j.trf.2015.11.011 PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 20 / 22 PLOS CLIMATEDiscounting the future 29. Masson T, Jugert P, Fritsche I. Collective self-fulfilling prophecies: group identification biases percep- tions of environmental group norms among high identifiers. Social Influence. 2016; 11:185–98. https:// doi.org/10.1080/15534510.2016.1216890 30. Masson T, Fritsche I. Adherence to climate change-related ingroup norms: Do dimensions of group identification matter? Eur. J. Soc. Psychol. 2014; 44:455–65. https://doi.org/10.1002/ejsp.2036 31. Terry DJ, Hogg MA, White KM. The theory of planned behaviour: self-identity, social identity and group norms. Br J Soc Psychol. 1999; 38 (Pt 3):225–44. https://doi.org/10.1348/014466699164149 PMID: 10520477 32. Vesely S, Masson T, Chokrai P, Becker AM, Fritsche I, Klo¨ ckner CA, et al. Climate change action as a project of identity: Eight meta-analyses. Global Environmental Change. 2021; 70:102322. https://doi. org/10.1016/j.gloenvcha.2021.102322 33. Hamann KRS, Reese G. My Influence on the World (of Others): Goal Efficacy Beliefs and Efficacy Affect Predict Private, Public, and Activist Pro-environmental Behavior. Journal of Social Issues. 2020; 76:35–53. https://doi.org/10.1111/josi.12369 34. Hornsey MJ, Fielding KS. A cautionary note about messages of hope: Focusing on progress in reducing carbon emissions weakens mitigation motivation. Global Environmental Change. 2016; 39:26–34. https://doi.org/10.1016/j.gloenvcha.2016.04.003 35. Jugert P, Greenaway KH, Barth M, Bu¨ chner R, Eisentraut S, Fritsche I. Collective efficacy increases pro-environmental intentions through increasing self-efficacy. Journal of Environmental Psychology. 2016; 48:12–23. https://doi.org/10.1016/j.jenvp.2016.08.003 36. Goedkoop F, Sloot D, Jans L, Dijkstra J, Flache A, Steg L. The Role of Community in Understanding Involvement in Community Energy Initiatives. Front Psychol. 2021; 12:775752. https://doi.org/10.3389/ fpsyg.2021.775752 PMID: 35222148 37. Berneiser JM, Becker AC, Loy LS. Give Up Flights? Psychological Predictors of Intentions and Policy Support to Reduce Air Travel. Front Psychol. 2022; 13:926639. https://doi.org/10.3389/fpsyg.2022. 926639 PMID: 35992431 38. Requate T. Dynamic incentives by environmental policy instruments—a survey. Ecological Economics. 2005; 54:175–95. https://doi.org/10.1016/j.ecolecon.2004.12.028 39. Voors MJ, Nillesen EEM, Verwimp P, Bulte EH, Lensink R, van Soest DP. Violent Conflict and Behavior: A Field Experiment in Burundi. American Economic Review. 2012; 102:941–64. https://doi.org/10. 1257/aer.102.2.941 40. Matousek J, Havranek T, Irsova Z. Individual discount rates: a meta-analysis of experimental evidence. Exp Econ. 2022; 25:318–58. https://doi.org/10.1007/s10683-021-09716-9 41. Klo¨ckner CA. A comprehensive model of the psychology of environmental behaviour—A meta-analysis. Global Environmental Change. 2013; 23:1028–38. https://doi.org/10.1016/j.gloenvcha.2013.05.014 42. Stern N. The Economics of Climate Change: Cambridge University Press; 2014. 43. Nordhaus WD. A Review of the Stern Review on the Economics of Climate Change. Journal of Eco- nomic Literature. 2007; 45:686–702. https://doi.org/10.1257/jel.45.3.686 44. Quaas MF, van Soest D, Baumga¨ rtner S. Complementarity, impatience, and the resilience of natural- resource-dependent economies. Journal of Environmental Economics and Management. 2013; 66:15– 32. https://doi.org/10.1016/j.jeem.2013.02.001 45. Drupp MA, Freeman MC, Groom B, Nesje F. Discounting Disentangled. American Economic Journal: Economic Policy. 2018; 10:109–34. https://doi.org/10.1257/pol.20160240 46. Cubitt RP, Read D. Can intertemporal choice experiments elicit time preferences for consumption? Exp Econ. 2007; 10:369–89. https://doi.org/10.1007/s10683-006-9140-2 47. Newell RG, Siikama¨ ki J. Individual Time Preferences and Energy Efficiency. American Economic Review. 2015; 105:196–200. https://doi.org/10.1257/aer.p20151010 48. Reichl J, Cohen J, Kollmann A, Azarova V, Klo¨ckner C, Royrvik J, et al. International survey of the ECH- OES project. Zenodo. 2019; 1. 49. Psychology of sustainable development. New York, NY: Springer Science + Business Media; 2012. 50. Cohen JJ, Azarova V, Kollmann A, Reichl J. Preferences for community renewable energy investments in Europe. Energy Economics. 2021; 100:105386. https://doi.org/10.1016/j.eneco.2021.105386 51. McFadden D. Conditional logit analysis of qualitative choice behavior. In: Frontiers in econometrics. New York [u.a.]: Academic Press, 1974; 1974. 52. Udall AM, Groot JIM de, Jong SB de, Shankar A How I See Me-A Meta-Analysis Investigating the Asso- ciation Between Identities and Pro-environmental Behaviour. Front Psychol. 2021; 12:582421. https:// doi.org/10.3389/fpsyg.2021.582421 PMID: 33796041 PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 21 / 22 PLOS CLIMATEDiscounting the future 53. Hoppe A, Fritsche I, Chokrai P. The “I” and the “We” in Nature Conservation—Investigating Personal and Collective Motives to Protect One’s Regional and Global Nature. Sustainability. 2023; 15:4694. https://doi.org/10.3390/su15054694 54. Sloot D, Jans L, Steg L. Can community energy initiatives motivate sustainable energy behaviours? The role of initiative involvement and personal pro-environmental motivation. Journal of Environmental Psychology. 2018; 57:99–106. https://doi.org/10.1016/j.jenvp.2018.06.007 55. Sloot D, Jans L, Steg L. In it for the money, the environment, or the community? Motives for being involved in community energy initiatives. Global Environmental Change. 2019; 57:101936. https://doi. org/10.1016/j.gloenvcha.2019.101936 56. Fielding KS, Terry DJ, Masser BM, Hogg MA. Integrating social identity theory and the theory of planned behaviour to explain decisions to engage in sustainable agricultural practices. Br J Soc Psychol. 2008; 47:23–48. https://doi.org/10.1348/014466607X206792 PMID: 17535461 57. Ha¨nsel MC, Drupp MA, Johansson DJA, Nesje F, Azar C, Freeman MC, et al. Climate economics sup- port for the UN climate targets. Nature Clim Change. 2020; 10:781–9. https://doi.org/10.1038/s41558- 020-0833-x 58. Mattauch L, Hepburn C, Spuler F, Stern N. The economics of climate change with endogenous prefer- ences. Resource and Energy Economics. 2022; 69:101312. https://doi.org/10.1016/j.reseneeco.2022. 101312 59. Gutsche G, Ziegler A. Which private investors are willing to pay for sustainable investments? Empirical evidence from stated choice experiments. Journal of Banking & Finance. 2019; 102:193–214. https:// doi.org/10.1016/j.jbankfin.2019.03.007 60. Budziński W, Czajkowski M. Endogeneity and Measurement Bias of the Indicator Variables in Hybrid Choice Models: A Monte Carlo Investigation. Environ Resource Econ. 2022; 83:605–29. https://doi.org/ 10.1007/s10640-022-00702-0 61. Ben-Akiva M, McFadden D, Train K, Walker J, Bhat C, Bierlaire M, et al. Hybrid Choice Models: Prog- ress and Challenges. Marketing Letters. 2002; 13:163–75. https://doi.org/10.1023/A:1020254301302. 62. Guevara CA. Critical assessment of five methods to correct for endogeneity in discrete-choice models. Transportation Research Part A: Policy and Practice. 2015; 82:240–54. https://doi.org/10.1016/j.tra. 2015.10.005 63. Haghani M, Bliemer MCJ, Rose JM, Oppewal H, Lancsar E. Hypothetical bias in stated choice experi- ments: Part I. Macro-scale analysis of literature and integrative synthesis of empirical evidence from applied economics, experimental psychology and neuroimaging. Journal of Choice Modelling. 2021; 41:100309. https://doi.org/10.1016/j.jocm.2021.100309 PLOS Climate \| https://doi.org/10.1371/journal.pclm.0000173 June 5, 2023 22 / 22 PLOS CLIMATE
10.1186_s13071-020-04140-z	Xavier et al. Parasites Vectors (2020) 13:278 https://doi.org/10.1186/s13071-020-04140-z Parasites & Vectors RESEARCH Open Access H1-antihistamines as antischistosomal drugs: in vitro and in vivo studies Rogério P. Xavier1, Ana C. Mengarda1, Marcos P. Silva1, Daniel B. Roquini1, Maria C. Salvadori2, Fernanda S. Teixeira2, Pedro L. Pinto3, Thiago R. Morais1, Leonardo L. G. Ferreira4, Adriano D. Andricopulo4 and Josué de Moraes1* Abstract Background: Schistosomiasis is a socioeconomically devastating parasitic infection afflicting hundreds of millions of people and animals worldwide. It is the most important helminth infection, and its treatment relies solely on the drug praziquantel. Oral H1-antihistamines are available worldwide, and these agents are among the most widely used of all medications in children and adults. Given the importance of the drug repositioning strategy, we evaluated the antischistosomal properties of the H1-antihistamine drugs commonly used in clinical practices. Methods: Twenty-one antihistamine drugs were initially screened against adult schistosomes ex vivo. Subsequently, we investigated the anthelmintic properties of these antihistamines in a murine model of schistosomiasis for both early and chronic S. mansoni infections at oral dosages of 400 mg/kg single dose or 100 mg/kg daily for five consecu- tive days. We also demonstrated and described the ability of three antihistamines to induce tegumental damage in schistosomes through the use of scanning electron microscopy. Results: From phenotypic screening, we found that desloratadine, rupatadine, promethazine, and cinnarizine kill adult S. mansoni in vitro at low concentrations (5–15 µM). These results were further supported by scanning electron microscopy analysis. In an animal model, rupatadine and cinnarizine revealed moderate worm burden reductions in mice harboring either early or chronic S. mansoni infection. Egg production, a key mechanism for both transmission and pathogenesis, was also markedly inhibited by rupatadine and cinnarizine, and a significant reduction in hepa- tomegaly and splenomegaly was recorded. Although less effective, desloratadine also revealed significant activity against the adult and juvenile parasites. Conclusions: Although the worm burden reductions achieved are all only moderate, comparatively, treatment with any of the three antihistamines is more effective in early infection than praziquantel. On the other hand, the clinical use of H1-antihistamines for the treatment of schistosomiasis is highly unlikely. Keywords: Schistosomiasis, Antischistosomal, Schistosoma, Antihistamines, Drug repositioning Background Infection with trematodes (blood flukes) of the genus Schis- tosoma, the causative agents responsible for schistoso- miasis, causes chronic and debilitating disease in millions Correspondence: moraesnpdn@gmail.com 1 Núcleo de Pesquisa em Doenças Negligenciadas, Universidade Guarulhos, Guarulhos, SP, Brazil Full list of author information is available at the end of the article of people and animals worldwide [1]. Although not com- monly fatal, schistosomiasis significantly contributes to a huge economic burden associated with low productivity and the perpetuation of the poverty cycle, as well as impos- ing a large burden on healthcare costs. Schistosomiasis is among the most prevalent parasitic diseases worldwide, and it is the most important human helminth infection in terms of global mortality and morbidity [2]. Approximately 800 million people may be at risk of infection worldwide, © The Author(s) 2020. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/ zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Xavier et al. Parasites Vectors (2020) 13:278 Page 2 of 12 and almost 240 million are infected [3]. The combination of the global healthcare burden, the prevalence of these helminths, and limited treatment options have led to the inclusion of schistosomiasis on the World Health Organi- zation’s list of neglected tropical diseases [4]. Schistosoma mansoni is the prevalent species in Africa, the Middle East, South America, and the Caribbean, in regions where the intermediate snail host, a freshwater snail of the genus Biomphalaria, is present. This parasite has a lifespan of several years and female schistosomes continuously produce eggs (each S. mansoni female worm can produce up to 300 eggs/day), which are able to pass through the intestinal lumen to be finally excreted with feces. Some of the eggs can be trapped in the tissues of the mammalian host instead of being excreted in the feces. Most, if not all, of the pathology in a schistosome infection results from the deposition of eggs in the tis- sues and the host’s response to them. The result is local, and systemic pathological effects include impaired cogni- tion, anemia and growth stunting, as well as organ-spe- cific effects, leading eventually to severe pathology such as hepatosplenomegaly and even death [5]. Praziquantel is a broadly effective trematocide and cestocide widely employed in veterinary and human medicine, and it is the only drug available to treat schis- tosomiasis. Praziquantel is very effective against adult worms (patent infection), but it is, unfortunately, poorly active against juvenile stages (prepatent infection), mean- ing that praziquantel must be given periodically for effec- tive treatment and control [6]. In addition, widespread use of praziquantel in both humans and domestic ani- mals [7, 8], along with the identification of laboratory and field isolates with reduced susceptibility to praziquantel [9, 10] raise serious concerns about the risk of selection of drug-resistant strains. Thus, new antischistosomal agents are needed, especially those targeting multiple stages of the parasite. For this reason, efforts to discover and develop novel antischistosomal agents have been intensified in recent years (for a review see [10, 11]). On the other hand, drug discovery is a lengthy and arduous process that inevitably struggles to deliver new therapies in a timely manner. Since the disease mainly affects poor people living in developing countries, pharmaceutical companies have little interest in developing new drugs. Thus, drug repurposing, the process of identifying new uses for existing drugs, is a promising strategy that has been used in recent years [12]. In schistosomes, and other flatworms, histamine is an important neuroactive substance [13] and G protein- coupled receptors (GPCRs) responsive to histamine have been described in S. mansoni [14, 15]. Due to their involvement in diverse biological and physiological pro- cesses, their pharmacological importance and potential as biological target, GPCRs are promising targets for new anthelmintic agents [16]. We have previously shown that promethazine, an old H1-antihistamine drug, had antischistosomal properties against S. mansoni adult worms ex vivo and in an animal model of schistosomiasis [17]. In view of these studies and in an attempt to explore drug repositioning strategy, here we evaluated the antiparasitic effect of a set of 21 H1-antihistamines com- monly used in clinical practice. In this context, from phe- notypic screening we found four antihistamine drugs that effectively killed S. mansoni adult worms ex vivo. Sub- sequently, these drugs were tested in vivo using an early and a chronic S. mansoni infection in a murine model. We also demonstrated the ability of these antihistamines to induce tegumental damage in adult worms through the use of scanning electron microscopy. Methods Drugs and reagents All H1-antihistamines were purchased from Cayman Chemical (Ann Arbor, MI, USA), Sigma-Aldrich (St. Louis, MO, USA) and Toronto Research Chemicals (Toronto, Ontario, Canada). Praziquantel was kindly pro- vided by Ecovet (São Paulo, SP, Brazil). The structures of all tested H1-antihistamines are shown in Additional file 1: Table S1. RPMI 1640 culture medium, penicillin G/streptomy- cin sulfate, and inactivated fetal bovine serum (FBS) were purchased from Vitrocell (Campinas, SP, Brazil). HEPES buffer, glutaraldehyde solution, and dimethyl sulfoxide (DMSO) were purchased from Sigma-Aldrich. Maintenance of the S. mansoni life‑cycle The life-cycle of S. mansoni (BH strain) is maintained by passage through snails (Biomphalaria glabrata) and a mice (Mus musculus) as described by de Moraes [18]. The host snails were exposed to light (60 W incandescent light bulbs) for up to 3 h and subsequently cercariae of S. mansoni were harvested. Female Swiss mice, 3 weeks-old (purchased from Anilab, São Paulo, Brazil) were infected subcutaneously with approximately 150 cercariae. Both snails and mice were kept under environmentally con- trolled conditions (25 °C; humidity of 50%), with free access to water and food. In vitro anthelmintic assay An in vitro anthelmintic assay was performed as previ- ously described [19, 20]. Briefly, adult schistosomes (49 day-old) were collected from the portal system and mes- enteric veins from infected mice (parasite ex vivo). Next, schistosomes were placed in RPMI 1640 culture medium supplemented with 10% FBS, containing 100 μg·ml−1 streptomycin 100 IU·ml−1 penicillin, and incubated in Xavier et al. Parasites Vectors (2020) 13:278 Page 3 of 12 a 24-well culture plate (Corning, New York, NY, USA). Drugs were dissolved in DMSO to obtain stock solu- tions of 10 mM and then were tested at a concentra- tion of 50 μM (one pair of parasites per well). Each drug was assessed in five replicates. Helminths were kept for 72 h (37 °C, 5% CO2) and their viability was monitored microscopically. The compounds that produced an effect superior to 90% after 72 h post-exposure underwent determination of their half maximum lethal concentra- tion (LC50) using 1:2 serial dilutions from 0.78 to 50 µM [17, 21]. Each concentration was tested in triplicate, and experiments were repeated once. Negative control (using the highest concentration of DMSO) and positive control (praziquantel 2 µM) were included [22]. Microscopy analysis During the in vitro experiments, parasites were moni- tored using a light microscope (Leica Microsystems EZ4E, Wetzlar, Germany). In addition, schistosomes were visualized using a scanning electron microscope (JEOL SM-6460LV; JEOL Tokyo, Japan) whose experi- mental protocols were previously published [23]. Briefly, adult worms (control and treated groups) were fixed in 2.5% glutaraldehyde and mounted specimens were metal- ized with gold (Desk II sputter coater; Denton Vacuum LLC, Moorestown, NJ, USA) before observation under scanning electron microscopy. Studies in an animal model of schistosomiasis Considering the in vitro results, we progressed cinnar- izine, desloratadine and rupatadine to in vivo studies in both early and chronic S. mansoni-murine models as pre- viously described [24]. Eighty Swiss mice, 3 weeks-old, were infected subcutaneously with 80 S. mansoni cercar- iae each. Animals were randomly divided into 16 groups (5 mice per group) and drugs were administered 21 days (early infection) or 42 days (chronic infection) post-infec- tion by oral gavage. For treatment, drugs were dissolved in 2% ethanol in water (v/v) and tested at a single dose of 400 mg/kg or a dose of 100 mg/kg/day for five successive days. Groups of S. mansoni-infected control were given a corresponding amount of vehicle on the same timetable. At 56 days post-infection, all animals were euthanized by the CO2 method and dissected; parasites were then col- lected, sexed, and counted [25, 26]. Therapeutic activity was also based on the technique of qualitative and quan- titative oograms in intestine, as well as the Kato-Katz method for quantitative fecal examination [27]. Randomization and blinding Animal studies are reported in compliance with the National Centre for the Replacement and Refinement & Reduction of Animals in Research (NC3Rs) ARRIVE guidelines. The mice were randomly assigned to the experimental groups, and pharmacological treatments were also performed randomly. The mice were eutha- nized in a random manner inside a group. All parameters (worm counts, measurement of the mass of the organs, quantitative and qualitative oogram, and quantitative fecal examination) were performed by different peo- ple (at least by two different investigators). Therefore, to eliminate bias in interpretation, manipulators of the experiments were not the same as the data analysts. Statistical analysis Statistical analyses were performed using GraphPad Prism version 7 (GraphPad Software, San Diego, CA) in accordance with the recommendations in the pharma- cology field [24]. All data from the in vitro anthelmintic experiments are presented as the mean ± standard devia- tion (SD) of at least three independent assays. LC50 val- ues were calculated using sigmoid dose-response curves and 95% confidence intervals [28]. Kaplan-Meier survival analyses were also used to compare in vitro survival data, and P-values calculated using the log-rank (Mantel-Cox) test. For experimental analysis of animal studies, a para- metric Dunnett’s test was applied to compare the control group with the treated group. The level of statistical sig- nificance was set to P < 0.05 [17]. Molecular and physicochemical properties Molecular and physicochemical properties were calcu- lated using the default parameters in the ADME/QSAR models of StarDrop version 6.6 (Optibrium, Cambridge, UK). The heat maps were performed using the same platform. Results In vitro efficacy of H1‑antihistamine drugs against adult schistosomes Effect of H1‑antihistamines on parasite viability The 21 H1-antihistamine drugs were initially screened against adult schistosomes ex vivo at 50 µM. Of all H1-antihistamines tested, four (desloratadine, rupata- dine, cinnarizine and promethazine) showed antischisto- somal properties after 72 h, and these H1-antihistamines were further tested at a range of concentrations for their LC50 determination. Out of these, compared with the control group, cinnarizine, desloratadine, and prometh- azine achieved an LC50 below 10 μM, whereas rupatadine had a LC50 value of ~15 μM (Fig. 1). Comparison of LC50 values revealed that the order of potency was prometh- azine (Mantel-Cox signed-rank test: χ2 = 29.09, df = 6, P < 0.0001), cinnarizine (Mantel-Cox signed-rank test: Xavier et al. Parasites Vectors (2020) 13:278 Page 4 of 12 χ2 = 41.03, df = 6, P < 0.0001), desloratadine (Mantel-Cox signed-rank test: χ2 = 28.68, df = 6, P < 0.0001), rupata- dine (Mantel-Cox signed-rank test: χ2 = 23.19, df = 5, P < 0.001). The calculation of the molecular and physico- chemical properties of the tested drugs (Additional file 1: Table S1) suggests that the antischistosomal activity may be correlated with the polar surface area of the com- pounds. The topological polar surface area (TPSA) val- ues, which have been demonstrated to correlate well with passive transmembrane transport, were, in general, lower for the active drugs compared with the inactive com- pounds [29]. The active drugs had an average TPSA value of 15.70, while the inactive antihistaminic agents had the significantly higher TPSA value of 42.43. This indicates that the active drugs may be better diffused across para- site membranes. Heat maps constructed for the active compounds (Fig. 2) illustrate the contribution of the dif- ferent portions of the molecules to TPSA. The temporal effects of different concentrations of H1-antihistamine drugs on adult schistosomes are depicted in Fig. 3. Control parasites remained viable over the entire observation period of 72 h. Cinnarizine and promethazine (TPSA values of 6.48) were able to kill all schistosomes within 24 h of contact at a concen- tration of 50 μM. A slightly slower onset of action was observed when parasites were incubated with deslorata- dine or rupatadine (TPSA values of 24.92 and 29.02, respectively). This time-dependence is consistent with the TPSA values calculated for the active drugs. All adult schistosomes died within 48 h. In contrast, praziquantel had a very fast onset of action on schistosomes. Effect of H1‑antihistamines on parasite tegument Since we have previously shown morphological changes in the tegument of the adult worms induced by pro- methazine [30], we conducted further studies with cin- narizine, desloratadine, and rupatadine using scanning electron microscopy. Figure 3a shows the tegument of a male parasite (control) depicting ridges and tubercles covered by spines that are somewhat uniformly distrib- uted. By 24 h after incubation with 50 μM of cinnarizine (Fig. 3b), desloratadine (Fig. 3c), or rupatadine (Fig. 3d), extensive destruction was visible on the entire tegument of all adult worms analyzed. For example, rupture of the tegument along the whole dorsal body surface, including blebbing, shrinking and sloughing was visible. Moreo- ver, tubercles had lost their spines. Similar morphologi- cal observations were made when the tegument of the schistosomes was evaluated after 48 h of exposure to cin- narizine 25 μM (Fig. 3e), desloratadine 50 μM (Fig. 3f ) and rupatadine 50 μM (Fig. 3g). After incubation for 72 h, shrinking and swelling of the tegument was seen on all parasites exposed to cinnarizine 12.5 μM (Fig. 3h), as well as desloratadine at 25 μM (Fig. 3i) and 12.5 μM (Fig. 3j). Interestingly, massive bubbles were observed on all worms exposed to rupatadine at 25 μM after 72 h of incubation (Fig. 3k). The positive control (praziquantel 2 μM) caused massive shrinking and swelling of the tegu- ment (Fig. 3l). The efficacy of H1‑antihistamine drugs in mice harboring either early or chronic S. mansoni infection Since promethazine was already tested in vivo and results published [17], we investigated the antischistosomal effect of cinnarizine, desloratadine, and rupatadine in mice harboring either early or chronic S. mansoni infec- tion. Results were compared to the control infected but untreated animal harboring either early or chronic infec- tion. Of note, all drugs were well tolerated, and all mice survived until the end of the experimental work. Effect of H1‑antihistamines on worm burden Figure 4 summarizes the antischistosomal activity of cin- narizine, desloratadine, and rupatadine given in single or multiple oral doses in both early and chronic infection, compared to control S. mansoni-infected animals. In early infection, using a single oral dose (400 mg/kg), cinnarizine achieved the highest worm burden reduc- tion (55.1%; ANOVA: F(13, 56) = 15.06, P = 0.0026). In the experiments where cinnarizine, rupatadine, and deslorat- adine were administered daily for 5 days (100 mg/kg), a decrease in total worm burden of 66.9% (ANOVA: F(13, 56) = 15.06, P = 0.0003), 66.5% (ANOVA: F(13, 56) = 15.06, P = 0.0004), and 50.7% (ANOVA: F(13, 56) = 15.06, P = 0.0052), respectively, was observed. In chronic infection, using a single oral dose (400 mg/ kg), the H1-antihistamine drugs caused a total worm burden reduction ranging from 50.1% (ANOVA: F(13, 56) = 15.06, P = 0.0059) to 55.6% (ANOVA: F(13, 56) = 15.06, P = 0.0021). In the treatment using multiple oral doses (5 × 100 mg/kg), cinnarizine and rupatadine achieved high total worm burden reductions of 73.6% (ANOVA: F(13, 56) = 15.06, P (ANOVA: F(13, 56) = 15.06, P = 0.0052, P < 0.0001), respectively. Lower but significant worm burden reduction values were obtained for desloratadine (59.2%; ANOVA: F(13, 56) = 15.06, P = 0.0008). < 0.0001) and 75.4% Effect of H1‑antihistamines on egg burden The egg load was evaluated using the oogram tech- nique (immature, mature and dead worms in the intes- tine) and the Kato-Katz technique for quantitative fecal examination. Regarding the oogram, in early infection, multiple oral doses of any of the three antihistamines led to a significant reduction in the number of immature eggs Xavier et al. Parasites Vectors (2020) 13:278 Page 5 of 12 ◂ Fig. 1 Viability of adult S. mansoni parasites ex vivo following exposure to H1-antihistamine drugs. Adult parasites were collected from the hepatic portal and mesenteric veins of mice and placed on plates containing the indicated concentrations of H1-antihistamines. Parasites were monitored for up to 72 h using a microscope and results are expressed as the percent mortality recorded by Kaplan-Meier survival curves. Mean values of viability were derived from a minimum of three experiments (n was performed with five replicates. LC50 values were determined at 72 h. Control (dashed line): RPMI 1640 praziquantel at 2 μM 3), and each experiment 0.5% DMSO. PZQ, + = (ANOVA: F(7, 35) = 8.43, P = 0.026), whereas drugs admin- istered in a single dose showed a non-significant reduc- tion in the number of eggs when compared to control infected mice. In contrast, the number of immature eggs was highly reduced in mice harboring a chronic S. mansoni infection, especially when any of the three drugs were administered in multiple doses (ANOVA: F(7, 35) = 8.43, P = 0.0006). The percentages of immature, mature and dead eggs are summarized in Fig. 5. With respect to fecal examination, cinnarizine and rupatadine administered daily for 5 days to mice harbor- ing chronic infection greatly reduced the number of eggs in feces by 73.8% (ANOVA: F(7, 35) = 8.43, P < 0.0001) and 80.1% (ANOVA: F(7, 35) = 8.43, P < 0.0001), respectively. Under the same drug regimen, desloratadine showed moderate but significant reductions in egg burden (56.5%; ANOVA: F(7, 35) = 8.43, P < 0.0001). In the experi- ments where antihistamines were administered with a single dose (400 mg/kg), a lower percentage reduction in the number of eggs in fecal samples relative to control infected mice was observed, especially in animals with early infection (Fig. 6). Effect of H1‑antihistamines on hepato‑ and splenomegaly Treatment of S. mansoni-infected animals with antihis- tamines also achieved a significant reduction of hepato- and-splenomegaly, as measured by weight, compared to control infected rodents (Fig. 7). In a chronic infection model, cinnarizine or rupatadine reduced liver mass by 22.3% (ANOVA: F(4, 7) = 7.16, P = 0.039) to 27.4% (ANOVA: F(4, 7) = 7.16, P = 0.034) (Fig. 7a) and spleen mass by 26.4% (ANOVA: F(3, 9) = 7.94, P = 0.038) to 32.7% (ANOVA: F(3, 9) = 7.94, P = 0.012) (Fig. 7b), whereas a moderate but significant reduction in the liver by 13.6% (ANOVA: F(4, 7) = 7.16, P = 0.0086) to 18.5% (ANOVA: F(4, 7) = 7.16, P = 0.034) and spleen by 20.1% (ANOVA: F(3, 9) = 7.94, P = 0.0041) to 24.9% (ANOVA: F(3, 9) = 7.94, P = 0.0006) was observed with desloratadine. On the other hand, hepatomegaly and splenomegaly were reduced more slightly in early schistosome infection. Xavier et al. Parasites Vectors (2020) 13:278 Page 6 of 12 Fig. 2 Heat maps for topological polar surface area (TPSA). The yellow regions contribute to increasing the value of TPSA, whereas the green regions have no influence Discussion Parasitic flatworm infections are treated by a limited number of drugs and, in most cases, control is reliant upon praziquantel monotherapy. However, praziquantel’s lack of efficacy against immature worms and the emer- gence of resistance against praziquantel cast a shadow on the global effort to control helminthiasis, as both treat- ment and control rely significantly on this drug. Since new drugs take a decade or longer to develop, and cost millions of dollars, drug repurposing is a promising approach. Phenotypic screening has successfully iden- tified praziquantel and other anthelmintic agents (e.g. ivermectin and albendazole) that are in veterinary and medical use [31]. In this study, from a screening of 21 H1-antihistamines, we found four drugs which affect the viability of S. mansoni. In vitro results showed that two first-generation anti- histamines (cinnarizine and promethazine) and two second-generation antihistamines (desloratadine and rupatadine) are highly active against adult schistosomes, with LC50 values of 5.8–15.4 µM, whereas the other anti- histamines were found to be inactive when screened at 50 μM. Although less potent than praziquantel, which had an LC50 value of approximately 0.1 µM [32, 33], cinnar- izine, promethazine, desloratadine and rupatadine are more potent than most antischistosomal compounds described so far (for review see [10, 34]). In vitro results of this study with cinnarizine are inconsistent with previ- ously reported results, which showed the lack of in vitro anti-parasitic activity against the larval [35] and adult [36] stages of S. mansoni. Interestingly, assessing the activity profile of an FDA-approved compound library against S. mansoni [37], tested promethazine and cinnarizine against adult parasites even at a high concentration of 33 µM and did not see any antischistosomal activity. These inconsistencies are likely a combination of differences in drug concentrations and life stages tested. In addition, it may also be possible that strain differences result in dif- fering drug susceptibilities. For example, Sarhan et al. [36] and Panic et al. [37] used an Egyptian and Liberian strain, respectively, whereas we used a Brazilian strain. inflammation, and Histamine has an important role as a chemical mes- senger in physiological responses, neurotransmission, allergic immunomodulation. Its receptors (named H1, H2, H3 and H4) are traditional GPCRs of extensive therapeutic interest [30, 38]. As the target of 33% of all small-molecule drugs, GPCRs are an important class of proteins in drug discovery [39]. Although GPCRs have been described in schistosomes [14, 18], the exact mechanism by which desloratadine, rupatadine, cinnarizine, and promethazine exert their anthelmintic action on schistosomes is still not clear. From a structural point of view, desloratadine, rupata- dine and loratadine are similar, but loratadine was inactive in vitro against adult schistosomes. Rupata- dine contains a 5-methylpyridin-3-yl group connected through a methylene to the basic amine of deslorata- dine. Interestingly, all four active H1-antihistamines had marked effects on the tegument of S. mansoni. However, it is not possible to distinguish causative from consequent action with regard to tegument damage; the drugs may induce it as part of their mechanism of action, or it may be a consequence of parasite death from another mechanism. Unlike nematodes, which are protected by a cuticle, Schistosoma species are covered by a living syncytium, called the tegument. This tissue Xavier et al. Parasites Vectors (2020) 13:278 Page 7 of 12 Fig. 3 Scanning electron micrographs of S. mansoni after exposure to H1-antihistamines drugs. Adult parasites were collected from the hepatic portal and mesenteric veins of mice and placed on plates containing various concentrations of H1-antihistamines. Parasites were monitored at different times up to 72 h and micrographs of the mid-body region of male worms were obtained using a scanning electron microscope. a Control showing tubercles (T) and spines on the surface (arrow). b–d Twenty-four hours after incubation of cinnarizine 50 μM (b), desloratadine 50 μM (c) and rupatadine 50 μM (d). e–g Forty-eight hours after incubation of cinnarizine 25 μM (e), desloratadine 50 μM (f) and rupatadine 50 μM (g). h–k Seventy-two hours after incubation of cinnarizine 12.5 μM (h), desloratadine 25 μM (i), desloratadine 12.5 μM (j) and rupatadine 25 μM (k). l Praziquantel 2 μM. In figures b–l, the dorsal tegumental surface shows roughening (ro), disintegration (di), bubbles (bu) and shrinking (sh). Images were captured using a JEOL SM-6460LV scanning electron microscope. Scale-bars: 10 μm Xavier et al. Parasites Vectors (2020) 13:278 Page 8 of 12 Fig. 4 Effect of H1-antihistamine drugs on the parasite burden in a preclinical mouse model. Drugs were administered orally using a single dose of 400 mg/kg or 100 mg/kg for five consecutive days to mice harboring either early or chronic S. mansoni infection. On day 56 post-infection, all animals were humanely euthanized and parasite burdens were determined by sex (male and female parasites). Points represent data from individual mice (n control by Dunnett’s test. Abbreviations: WBR, worm burden reduction; CNZ, cinnarizine; DES, desloratadine; RUP, rupatadine 5 per group). Horizontal bars represent median values. P < 0.05, P < 0.01, P < 0.001 compared with infected untreated = is bounded at its basal surface by a usual invaginated plasma membrane, whereas its apical surface has an atypical heptalaminate appearance [40]. This heptala- mellar layer forms many surface pits that substantially enlarge the surface area of the schistosomes. Antihis- tamines are heterogeneous groups of compounds, with markedly different chemical structures. Comparing the physicochemical properties, the active H1-anti- histamines have lower values of TPSA; this membrane permeability parameter may be important to facilitate the permeation of the drugs through the parasite’s tegument and, consequently, the interaction with their molecular target(s). Furthermore, cinnarizine is also a calcium channel blocker, and the possibility of action on the helminth’s calcium channels cannot be excluded. Further studies are needed to elucidate the mechanism of action of the H1-antihistamines in schistosomes. Cinnarizine, desloratadine, and rupatadine were evalu- ated in both early and chronic S. mansoni infection mod- els in mice. The oral doses of H1-antihistamines that were chosen (single dose of 400 mg/kg and 100 mg/kg daily for 5 days) followed the protocol recommended for a mouse model of schistosomiasis (e.g. [24, 41]). In addition, these drug regimens (single dose or daily, once a day) are in tune with those recommended for the treatment of aller- gic symptoms. Of note, most H1-antihistamines have an extended duration of clinical activity which allows once‐ daily administration. In this study, the treatment with any of the three H1-antihistamines, mainly using 100 mg/kg daily, revealed significant worm burden reductions in ani- mals harboring either chronic or early S. mansoni infection. It should be noted that praziquantel treatment exerts high cure rates of 70–90% [42], but it is concerning that some infections in humans and in various other species of ani- mals appear to be refractory to treatment [43, 44]. Impor- tantly, praziquantel has low efficacy against immature parasites (early infection) [45]. Comparatively, oral treat- ment with cinnarizine, desloratadine, or rupatadine is more effective in early infection than praziquantel. In contrast, praziquantel is more effective in chronic infection that the Xavier et al. Parasites Vectors (2020) 13:278 Page 9 of 12 Fig. 6 Effect of H1-antihistamine drugs on the egg burden in the feces in a preclinical mouse model. Drugs were administered orally using a single dose of 400 mg/kg or 100 mg/kg for five consecutive days to mice harboring either early or chronic S. mansoni infection. On day 56 post-infection, all animals were humanely euthanized and egg burden in feces was measured by Kato-Katz technique. Data are presented as the mean 5 per group). The numbers SD (n represent the percentages of egg reduction vs infected untreated control. P < 0.05, P < 0.01, P < 0.001 compared with infected untreated control groups by Dunnett’s test. Abbreviations: CNZ, cinnarizine; DES, desloratadine; RUP, rupatadine ± = and splenomegaly was also recorded. This result could be attributed to a decrease in the number of parasites as a result of treatment with antihistamines and/or reduction of egg-laying by female helminths. A significant decrease in egg-laying in the intestine or feces has been recently reported with other anthelmintic agents [17, 29, 46]. More- over, the pathology normally associated with the parasite eggs in the liver and spleen was ameliorated, mainly when antihistamines were given for five days. This finding could be assigned to a decrease in the number of worms and egg- laying. Additionally, it is well established that, in addition to their effects on H1 receptors, antihistamines also possess anti-inflammatory properties and, thus, H1-antihistamine therapy may have contributed to reducing hepatomegaly and splenomegaly in S. mansoni-infected mice. In tandem, rupatadine, cinnarizine, and deslotaradine revealed a moderate reduction in worm and egg burden in mice harboring either early or chronic S. mansoni infection. Clinically, a typical dose of these H1-antihis- tamines is a single 5–20 mg tablet. Even allowing for pharmacokinetic (PK) differences between mice and humans (see dose translation from animal to human studies described by Reagan-Shaw et al. [47]), this dose is much less than 100 mg/kg (let alone 400 mg/kg). Simi- larly, the maximum serum concentration (Cmax) for these drugs in humans is < 10 ng/ml (< 0.1 µM), far lower than the concentrations needed to kill schistosomes in Fig. 5 Effect of H1-antihistamine drugs on the egg developmental stage in a preclinical mouse model. Drugs were administered orally using a single dose of 400 mg/kg or 100 mg/kg for five consecutive days to mice harboring either early or chronic S. mansoni infection. On day 56 post-infection, all animals were humanely euthanized and egg burdens were determined by counting eggs in the intestine (quantitative and qualitative oogram technique). Data are presented as the mean the percentages of egg reduction vs infected untreated control. P < 0.05, P < 0.01, P < 0.001 compared with infected untreated control groups. Abbreviations: CNZ, cinnarizine; DES, desloratadine; RUP, rupatadine 5 per group). The numbers represent SD (n = ± three H1-antihistamines. Collectively, this finding high- lights the advantage of using cinnarizine, rupatadine and desloratadine instead of praziquantel in immature schisto- some stages. In vivo results of this work with cinnarizine in part mirrored the in vivo studies mentioned earlier [36], in that cinnarizine was effective in reducing the worm burden in early infection, surpassing praziquantel. Egg production, a key mechanism for both transmis- sion and pathogenesis, was also markedly inhibited by antihistamines, and a mitigation effect on hepatomegaly Xavier et al. Parasites Vectors (2020) 13:278 Page 10 of 12 Conclusions In conclusion, of all the H1-antihistamines tested, pro- methazine, cinnarizine, desloratadine, and rupatadine are schistosomicidal agents in vitro, which is consist- ent with the extensive structural damage caused by these compounds. In a rodent model of schistosomiasis, desloratadine and mainly rupatadine and cinnarizine greatly reduced worm burden, egg production, and hepa- tomegaly and splenomegaly. Although the worm and egg burden reductions achieved were all only moderate, com- paratively, treatment with any of the three antihistamines is more effective in early infection than praziquantel. On the other hand, the clinical use of H1-antihistamines for the treatment of schistosomiasis is highly unlikely. Finally, the exact mechanism by which these H1-antihis- tamines exert their anthelmintic effect is still not clear, and further investigation of this property and identifica- tion of parasitic-selective ligands that convey this effect are warranted because this could lead to a directed medicinal chemistry effort to identify schistosome-selec- tive compounds. Supplementary information Supplementary information accompanies this paper at https ://doi. org/10.1186/s1307 1-020-04140 -z. Additional file 1. Table S1. Molecular properties of H1-antihistamine drugs. Abbreviations Cmax: Maximum serum concentration; DMSO: Dimethyl sulfoxide; FBS: Fetal bovine serum; GPCRs: G protein-coupled receptors; LC50: 50% Lethal concen- tration; PK: Pharmacokinetic; SD: Standard deviation; TPSA: Topological polar surface area. Acknowledgements We would like to thank Anderson B. Gonçalves, Yone C. Xavier and Silvia G. Chiodelli for technical support (Núcleo de Enteroparasitas, Instituto Adolfo Lutz, SP, Brazil). We are also grateful to Ramon M. Cogo, Bianca C. Silva and Cris- tiane S. Morais for support during the animal experiments (Núcleo de Pesquisa em Doenças Negligenciadas, Universidade Guarulhos, SP, Brazil). Authors’ contributions ACM, MPS and DBR conducted the in vitro and in vivo experiments and assessed the data. MCS and FST performed microscopy procedures. PLP provided adult S. mansoni parasites and analyzed the data from the in vitro experiments. TRM provided support with chemical data. LLGF and ADA analyzed the drugs’ pharmacokinetic parameters. JM performed the statistical analysis. RPX, TRM and JM analyzed the data and wrote the manuscript. All authors contributed intellectually to the article. All authors read and approved the final manuscript. Funding This study was supported by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), Brazil (grant # 2016/22488-3). ACM, MPS and DBR were supported by a fellowship from the Coordenação de Aperfeiçoamento de Pes- soal de Nível Superior (CAPES), Brazil. TRM received a postdoctoral fellowship (PNPD-CAPES). ADA was supported by FAPESP-CIBFar (grant # 2013/07600- 3). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Fig. 7 Effect of H1-antihistamine drugs on the organ pathology in a preclinical mouse model. a Liver weight. b Spleen weight. Drugs were administered orally using a single dose of 400 mg/kg or 100 mg/kg for five consecutive days to mice harboring either early or chronic S. mansoni infection. On day 56 post-infection, all animals were humanely euthanized and organ pathology was determined by liver and spleen weights. Data are presented as the mean 5 per group). The numbers represent the percentages of egg reduction vs infected untreated control. P < 0.05, P < 0.01, *P < 0.001 compared with infected untreated control groups by Dunnett’s test. Abbreviations: OWR, organ weight reduction; CNZ, cinnarizine; DES, desloratadine; RUP, rupatadine SD (n ± = culture. Therefore, although these drugs are quite safe, that difference is highly unlikely to support clinical use for schistosomiasis. Xavier et al. Parasites Vectors (2020) 13:278 Page 11 of 12 Availability of data and materials The dataset supporting the conclusions of this article is included within the article and its additional file. Ethics approval and consent to participate This study was conducted following the guidelines of the Animal Ethics Com- mittee (Universidade Guarulhos, Guarulhos, SP, Brazil) according to Brazilian law. All experimental protocols were approved by the Universidade Guarulhos (Approval ID 31/2017). Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1 Núcleo de Pesquisa em Doenças Negligenciadas, Universidade Guarulhos, Guarulhos, SP, Brazil. 2 Instituto de Física, Universidade de São Paulo, São Paulo, SP, Brazil. 3 Núcleo de Enteroparasitas, Instituto Adolfo Lutz, São Paulo, SP, Brazil. 4 Laboratório de Química Medicinal e Computacional, Instituto de Física de São Carlos, Universidade de São Paulo, São Paulo, SP, Brazil. Received: 22 February 2020 Accepted: 21 May 2020 References 1. Colley DG, Bustinduy AL, Secor WE, King CH. Human schistosomiasis. Lancet. 2014;383:2253–64. 2. McManus DP, Dunne DW, Sacko M, Utzinger J, Vennervald BJ, Zhou XN. Schistosomiasis. Nat Rev Dis Primers. 2018;4:13. 3. WHO. Schistosomiasis. Geneva: World Health Organization. 2019. https :// www.who.int/news-room/fact-sheet s/detai l/schis tosom iasis . Accessed 17 Mar 2020. 4. WHO. Schistosomiasis. Geneva: World Health Organization. 2019. https :// www.who.int/gho/negle cted_disea ses/schis tosom iasis /en/. Accessed 17 Mar 2020. 5. Vale N, Gouveia MJ, Rinaldi G, Brindley PJ, Gärtner F, Correia da Costa JM. Praziquantel for schistosomiasis: single-drug metabolism revisited, mode of action, and resistance. Antimicrob Agents Chemother. 2017;61:e02582. 6. Melman SD, Steinauer ML, Cunningham C, Kubatko LS, Mwangi IN, Wynn NB, et al. Reduced susceptibility to praziquantel among naturally occurring Kenyan isolates of Schistosoma mansoni. PLoS Negl Trop Dis. 2009;3:e504. 8. 7. Mwangi IN, Sanchez MC, Mkoji GM, Agola LE, Runo SM, Cupit PM, et al. Praziquantel sensitivity of Kenyan Schistosoma mansoni isolates and the generation of a laboratory strain with reduced susceptibility to the drug. Int J Parasitol Drugs Drug Res. 2014;4:296–300. Fallon PG, Doenhoff MJ. Drug-resistant schistosomiasis: resistance to praziquantel and oxamniquine induced in Schistosoma mansoni in mice is drug specific. Am J Trop Med Hyg. 1994;51:83–8. Sabra AN, Botros SS. Response of Schistosoma mansoni isolates having different drug sensitivity to praziquantel over several life cycle passages with and without therapeutic pressure. J Parasitol. 2008;94:537–41. 9. 10. Lago EM, Xavier R, Teixeira T, Silva LM, da Silva Filho AA, de Moraes J. Antischistosomal agents: state of art and perspectives. Future Med Chem. 2018;10:89–120. 11. Mafud AC, Ferreira LG, Mascarenhas YP, Andricopulo AD, de Moraes J. Discovery of novel antischistosomal agents by molecular modeling approaches. Trends Parasitol. 2016;32:874–86. 12. Lago EM, Silva MP, Queiroz TG, Mazloum SF, Rodrigues VC, Carnaúba PU, et al. Phenotypic screening of nonsteroidal anti-inflammatory drugs identified mefenamic acid as a drug for the treatment of schistosomiasis. EBioMedicine. 2019;43:370–9. 13. Ribeiro P, El-Shehabi F, Patocka N. Classical transmitters and their recep- tors in flatworms. Parasitology. 2005;131(Suppl. 1):S19–40. 14. Hamdan FF, Abramovitz M, Mousa A, Xie J, Durocher Y, Ribeiro P. A novel Schistosoma mansoni G protein-coupled receptor is responsive to hista- mine. Mol Biochem Parasitol. 2002;119:75–86. 15. El-Shehabi F, Ribeiro P. Histamine signalling in Schistosoma mansoni: immunolocalisation and characterisation of a new histamine-responsive receptor (smgpr-2). Int J Parasitol. 2010;40:1395–406. 16. Hahnel S, Wheeler N, Lu Z, Wangwiwatsin A, McVeigh P, Maule A, et al. Tissue-specific transcriptome analyses provide new insights into GPCR signalling in adult Schistosoma mansoni. PLoS Pathog. 2018;14:e1006718. 17. Roquini DB, Cogo RM, Mengarda AC, Mazloum SF, Morais CS, Xavier RP, et al. Promethazine exhibits antiparasitic properties in vitro and reduces worm burden, egg production, hepato-, and splenomegaly in a schistosomiasis animal model. Antimicrob Agents Chemother. 2019;63:e01208–19. 18. de Moraes J. Antischistosomal natural compounds: present challenges for new drug screens. In: Rodriguez-Morales AJ, editor. Current topics in tropical medicine. Rijeka: InTech Open; 2012. p. 333–58. 19. de Castro CC, Costa PS, Laktin GT, de Carvalho PH, Geraldo RB, de Moraes J, et al. Cardamonin, a schistosomicidal chalcone from Piper aduncum L. (Piperaceae) that inhibits Schistosoma mansoni ATP diphosphohydrolase. Phytomedicine. 2015;22:921–8. 20. Mafud AC, Silva MP, Monteiro DC, Oliveira MF, Resende JG, Coelho ML, et al. Structural parameters, molecular properties, and biological evalua- tion of some terpenes targeting Schistosoma mansoni parasite. Chem Biol Interact. 2016;244:129–39. 21. Silva AP, Silva MP, Oliveira CG, Monteiro DC, Pinto PL, Mendonça RZ, et al. Garcinielliptone FC: antiparasitic activity without cytotoxicity to mam- malian cells. Toxicol In Vitro. 2015;29:681–7. 22. Quelemes PV, Perfeito ML, Guimarães MA, dos Santos RC, Lima DF, Nascimento C, et al. Effect of neem (Azadirachta indica A. Juss) leaf extract on resistant Staphylococcus aureus biofilm formation and Schistosoma mansoni worms. J Ethnopharmacol. 2015;175:287–94. 23. Guimarães MA, de Oliveira RN, Véras LM, Lima DF, Campelo YD, Cam- pos SA, et al. Anthelmintic activity in vivo of epiisopiloturine against juvenile and adult worms of Schistosoma mansoni. PLoS Negl Trop Dis. 2015;9:e0003656. 24. Guerra AR, Silva MP, Silva TC, Salvadori MC, Teixeira FS, de Oliveira RN, et al. Spironolactone as an antischistosomal drug capable of clinical repurposing: in vitro and in vivo studies. Antimicrob Agents Chemother. 2019;63:e01722. 25. Silva MP, de Oliveira RN, Mengarda AC, Roquini DB, Allegretti SM, Sal- vadori MC, et al. Antiparasitic activity of nerolidol in a mouse model of schistosomiasis. Int J Antimicrob Agents. 2017;50:467–72. 26. Lima LI, Py-Daniel KR, Guimarães MA, Muehlmann LA, Mafud AC, Mascar- enhas YP, et al. Self-nanoemulsifying drug-delivery systems improve oral absorption and antischistosomal activity of epiisopiloturine. Nanomedi- cine. 2018;13:689–702. 27. de Moraes J, de Oliveira RN, Costa JP, Junior AL, de Sousa DP, Freitas RM, et al. Phytol, a diterpene alcohol from chlorophyll, as a drug against neglected tropical disease schistosomiasis mansoni. PLoS Negl Trop Dis. 2014;8:e2617. 28. de Moraes J, Dario BS, Couto RA, Pinto PL, da Costa Ferreira AM. Antisch- istosomal activity of oxindolimine-metal complexes. Antimicrob Agents Chemother. 2015;59:6648–52. 29. Ertl P, Rohde B, Selzer P. Fast calculation of molecular polar surface area as a sum of fragment-based contributions and its application to the predic- tion of drug transport properties. J Med Chem. 2000;43:3714–7. 30. Parsons ME, Ganellin CR. Histamine and its receptors. Br J Pharmacol. 2006;147(Suppl. 1):S127–35. 31. Woods DJ, Lauret C, Geary T. Anthelmintic discovery and development in the animal health industry. Expert Opin Drug Discov. 2007;2(Suppl. 1):S25–33. 32. de Moraes J, Keiser J, Ingram K, Nascimento C, Yamaguchi LF, Bittencourt CR, et al. In vitro synergistic interaction between amide piplartine and antimicrobial peptide dermaseptin against Schistosoma mansoni schisto- somula and adult worms. Curr Med Chem. 2013;20:301–9. 33. Campelo YDM, Mafud AC, Véras LMC, Guimarães MA, Yamaguchi LF, Lima DF, et al. Synergistic effects of in vitro combinations of piplartine, epiisopiloturine and praziquantel against Schistosoma mansoni. Biomed Pharmacother. 2017;88:488–99. 34. de Moraes J. Natural products with antischistosomal activity. Future Med Chem. 2015;7:801–20. 35. Abdulla MH, Ruelas DS, Wolff B, Snedecor J, Lim KC, Xu F, et al. Drug dis- covery for schistosomiasis: hit and lead compounds identified in a library Xavier et al. Parasites Vectors (2020) 13:278 Page 12 of 12 of known drugs by medium-throughput phenotypic screening. PLoS Negl Trop Dis. 2009;3:e478. 36. Sarhan R, Thabet H, Nazeer J, William S. The impact of cinnarizine and griseofulvin on juvenile and adult stages of Schistosoma mansoni. J Helminthol. 2020;94:E41. 37. Panic G, Vargas M, Scandale I, Keiser J. Activity profile of an FDA-approved compound library against Schistosoma mansoni. PLoS Negl Trop Dis. 2015;9:e0003962. 38. Thangam EB, Jemima EA, Singh H, Baig MS, Khan M, Mathias CB, et al. The role of histamine and histamine receptors in mast cell-mediated allergy and inflammation: the hunt for new therapeutic targets. Front Immunol. 2018;9:1873. 39. Santos R, Ursu O, Gaulton A, Bento AP, Donadi RS, Bologa CG, et al. A comprehensive map of molecular drug targets. Nat Rev Drug Discov. 2017;16:19–34. 40. Skelly PJ, Alan Wilson R. Making sense of the schistosome surface. Adv Parasitol. 2006;63:185–284. 41. Mengarda AC, Mendonça PS, Morais CS, Cogo RM, Mazloum SF, Salvadori MC, et al. Antiparasitic activity of piplartine (piperlongumine) in a mouse model of schistosomiasis. Acta Trop. 2020;205:105350. 42. Coulibaly JT, Panic G, Silue KD, Kovac J, Hattendorf J, Keiser J. Efficacy and safety of praziquantel in preschool-aged and school-aged children infected with Schistosoma mansoni: a randomised controlled, parallel- group, dose-ranging, phase 2 trial. Lancet Glob Health. 2017;5:e688–98. 43. Ismail M, Metwally A, Farghaly A, Bruce J, Tao LF, Bennett JL. Characteri- zation of isolates of Schistosoma mansoni from Egyptian villagers that tolerate high doses of praziquantel. Am J Trop Med Hyg. 1996;55:214–8. 44. Jesudoss Chelladurai J, Kifleyohannes T, Scott J, Brewer MT. Praziquantel resistance in the zoonotic cestode Dipylidium caninum. Am J Trop Med Hyg. 2018;99:1201–5. 45. Bergquist R, Elmorshedy H. Artemether and praziquantel: origin, mode of action, impact, and suggested application for effective control of human schistosomiasis. Trop Med Infect Dis. 2018;1:118–25. 46. Guimarães MA, de Oliveira RN, de Almeida RL, Mafud AC, Sarkis ALV, Ganassin R, et al. Epiisopilosine alkaloid has activity against Schistosoma mansoni in mice without acute toxicity. PLoS ONE. 2018;13:e0196667. 47. Reagan-Shaw S, Nihal M, Ahmad N. Dose translation from animal to human studies revisited. FASEB J. 2008;22:659–61. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.